M. Yamashita, Department of Physiology I, Nara Medical University, Shijo-cho 840, Kashihara 634-8521, Japan Fax: +81 744 29 0306 Tel: +81 744 29 8827 E-mail: email@example.com
Synchronous Ca2+ oscillation occurs in various cell types to regulate cellular functions. However, the mechanism for synchronization of Ca2+ increases between cells remains unclear. Recently, synchronous oscillatory changes in the membrane potential of internal Ca2+ stores were recorded using an organelle-specific voltage-sensitive dye [Yamashita et al. (2006) FEBS J273, 3585–3597], and an electrical coupling model of the synchronization of store potentials and Ca2+ releases has been proposed [Yamashita (2006) FEBS Lett580, 4979–4983]. This model is based on capacitative coupling, by which transient voltage changes can be synchronized, but oscillatory slow potentials cannot be communicated. Another candidate mechanism is synchronization of action potentials and ensuing Ca2+ influx through voltage-dependent Ca channels. The present study addresses the question of whether Ca2+ increases are synchronized by action potentials, and how oscillatory store potentials are synchronized across the cells. Electrophysiological and Ca2+-sensitive fluorescence measurements in early embryonic chick retina showed that synchronous Ca2+ oscillation was caused by releases of Ca2+ from Ca2+ stores without any evidence of action potentials in retinal neuroepithelial cells or newborn neurons. High-speed fluorescence measurement of store membrane potential surprisingly revealed that the synchronous oscillatory changes in the store potential were periodic repeats of a burst of high-frequency voltage fluctuations. The burst coincided with a Ca2+ increase. The present study suggests that synchronization of Ca2+ release is mediated by the high-frequency fluctuation in the store potential. Close apposition of the store membrane and plasma membrane in an epithelial structure would allow capacitative coupling across the cells.
Oscillatory Ca2+ increases are synchronized across multiple cells in epithelium-like structures such as confluent monolayers of endothelial cells [1,2], Madin–Darby canine kidney (MDCK) cells [3,4], connected hepatocytes , and in polarized epithelium of cholangiocytes . The synchronous Ca2+ oscillation also occurs in the ventricular zone of neocortex [7,8] and in the developing retina [9–14]. Intercellular traveling waves of Ca2+ increase may be explained by the release of a transmitter , but it is unclear how the oscillatory Ca2+ increases are synchronized across cells. As gap junction blockade disrupts the synchrony of Ca2+ oscillation [4–6], it is likely that gap junction coupling mediates the synchronization of Ca2+ increases. However, passive diffusion of a second messenger or Ca2+ ions through gap junctions cannot account for the synchronization of Ca2+ increases, because agonist-induced inositol 1,4,5-trisphosphate (InsP3)-mediated Ca2+ release does not cause synchronous Ca2+ increases even in cells that are in contact  or in closely positioned cells that show synchronous Ca2+ oscillation . Thus the exact mechanism for synchronization of oscillatory Ca2+ releases remains unknown. Synchronized Ca2+ increases may be caused by synchronous action potentials and ensuing Ca2+ influx through voltage-dependent Ca channels. However, this is impossible for non-excitable cells.
As the release of Ca2+ is regulated by an electrochemical driving force for Ca2+ efflux from the store, changes in the luminal potential of the Ca2+ store will affect the release of Ca2+, and vice versa [14,16–18]. Recently, an electrical coupling hypothesis has been proposed as an explanation for the synchronization of oscillatory Ca2+ releases . Electrical coupling of Ca2+ stores can be achieved by close contact of cells, in which the outer membrane of a Ca2+ store is closely apposed to the plasma membrane (fig. 2 in ). A rapid change in the luminal potential produces a capacitative current, which passes through the plasma membranes to change the luminal potential of the Ca2+ store in the neighboring cell. However, such capacitative coupling (AC coupling) occurs only when the luminal potential changes transiently. AC coupling cannot mediate transmission of steady levels of luminal potential (DC potentials) or slow changes in the luminal potential.
In the present study, the mechanism for the synchronization of Ca2+ increases was investigated in early embryonic chick retina with electrophysiological methods and high-speed fluorescence measurements of the membrane potential of Ca2+ stores. Synchronous Ca2+ oscillations are observed in the chick retina from embryonic day 3 (E3)  and at E5–E6 [10,11]. Synchronous Ca2+ increases and simultaneous bursts of action potential occur at E13–E18, when retinal axons refine their connections . In the present study, the early stage of retinal development was studied and the possibility of synchronization of action potentials was examined first. However, this was found not to be the case at the early stage of neurodevelopment. Instead, it was surprisingly revealed that synchronous voltage fluctuations burst periodically with a Ca2+ increase. The capacitative electrical coupling model is discussed on the basis of this unexpected finding.
Figure 1A illustrates the early development of embryonic chick retina in the vertical plane. An E3 retina is composed of neuroepithelial cells (soma diameter 4–6 μm, Fig. 1B) undergoing interkinetic nuclear migration during the cell cycle [20,21]. Retinal ganglion cells, the first type of retinal neurons, develop at E4–E6. The early embryonic retina was so thin (approximately 40 μm at E3) that it was difficult to prepare a slice for conventional patch–clamp recording. Instead, intracellular recording was performed using a fine microelectrode in the inner layer of a flat-mounted retina. In the E3 retina, recordings with resting membrane potentials more positive than −60 mV were discarded to ensure the reliability of data. Although the E3 cell showed quite stable resting membrane potentials (−72.0 ±2.1 mV, n =12, mean ± SE), the input resistance was rather low (28.6 ± 5.2 MΩ, n =12, measured using ± 0.2 nA current pulses). These cells did not fire even with 1 nA current injection (Fig. 1C). The low input resistance might be due to the microelectrode penetration. However, the leak resistance was estimated to be quite high (discussed in supplementary Doc. S1). Another possible reason for the low input resistance was gap junction coupling; the neuroepithelial cells adhere to each other through gap junctions immediately below the outer limiting membrane (Fig. 1A) . To address this issue, carbenoxolone (CBX, a gap junction channel blocker) was bath-applied during intracellular recording and dramatically increased the input resistance (Fig. 1D): 10 μm of CBX increased the input resistance 2-4-fold (n =2) and 100 μm of CBX increased it 3- to 8.5-fold (5.2 ± 0.5-fold, 132.0 ± 15.9 MΩ, n =10, Fig. 1E). It was also noted that the resting membrane potential became unstable and shifted in the depolarizing direction in the presence of CBX (Fig. 1D). This depolarization appears to be due to the decoupling of gap junctions and a current leaking into the impaled cell (supplementary Doc. S1). These results suggest that gap junction coupling stabilizes the resting membrane potential of neuroepithelial cells by lowering the input resistance. The gap junction coupling was verified by intracellular injection of Alexa Fluor 488 in five cells in five E3 retinae (Fig. 1F and supplementary Fig. S1).
The first morphological sign of ganglion cell differentiation is the detachment of the ventricular (outer) process from the junctional complex at the ventricular surface . Thus, a newborn ganglion cell loses the gap junction coupling at the outer surface. In fact, the ganglion cell in E5 chick retina shows very few gap junctions . Intracellular injection of Alexa Fluor 488 did not show dye coupling in the E5 ganglion cell (Fig. 2A, n =7). In accordance with this morphological change, presumptive ganglion cells in the inner layer of E4–E6 retinae showed distinctly high input resistance (112.1 ± 4.4 MΩ; mean resting membrane potential −56.7 ± 1.1 mV, n =38) and showed voltage decay in response to depolarizing current injection (Fig. 2B). This appeared as a steep downward drop (Fig. 2B, asterisk) or a slow decay at a more positive potential (Fig. 2B, arrow). The decay became steeper with negative shifts in the holding potential (Fig. 2C). A hyperpolarizing response was observed upon termination of depolarizing current (Fig. 2D, arrow). These results indicate that hyperpolarizing conductance was activated by Ca2+ influx through voltage-dependent (e.g. T-type) Ca channels. To address this issue, 10 μm of mibefradil (a selective T-type Ca channel blocker ) was puff-applied near the recorded cell, and this abolished the voltage decay (Fig. 2E,F, n =10). The voltage decay was also abolished by puff application of 2 μm of paxilline (a blocker for BK-type Ca2+-activated potassium channels ) (Fig. 2G,H, n =9). These results suggest that BK channels (big/maxi K channels) were activated by Ca2+ influx through T-type Ca channels, although other types of potassium channels or Ca channels may also be activated. Puff application of tetrodotoxin (TTX, 1 μm) did not produce any detectable change (n =4). TTX-sensitive action potentials were observed later at E11 (supplementary Fig. S2).
To verify the Ca2+ influx through voltage-dependent Ca channels, a high-K+ (50 mm) solution was bath-applied to an E5 retina loaded with a fluorescent Ca2+ indicator (Oregon Green 488 BAPTA-1 acetoxymethyl ester, cell-permeant). This caused a Ca2+ increase in a ganglion cell (soma diameter ≥ 10 μm ), and this Ca2+ increase was abolished in the absence of extracellular Ca2+ (Fig. 3A–F). Ni2+ and mibefradil inhibited the high-K+-induced Ca2+ increase (supplementary Fig. S3). Immunolabeling with an antibody against CaV3.1 supported the expression of T-type Ca channels (supplementary Fig. S4) .
Ca2+ oscillations (frequency approximately 3–10 min−1) occurred at rather high temperatures (28–38 °C). If Ca2+ oscillations are caused by the activation of T-type Ca channels, the oscillatory Ca2+ increase may be attenuated by the exogenous depolarization due to the inactivation of T-type Ca2+ channels and a decrease in the driving force for Ca2+ influx across the plasma membrane. However, synchronous Ca2+ oscillation occurred almost independently of the high-K+-induced Ca2+ increase (Fig. 3G–J). It even occurred in the absence of extracellular Ca2+ (Fig. 3K). These facts strongly suggested that the Ca2+ influx through voltage-dependent Ca channels is not essential for the generation of synchronous Ca2+ oscillation [7,8].
Another way to cause a Ca2+ increase is the release of Ca2+ from intracellular Ca2+ stores. The Ca2+ efflux changes the membrane potential of Ca2+ store. This voltage change could mediate the synchronization of Ca2+ releases if the stores are electrically coupled across the cells . To test this hypothesis, DiOC5(3) (3,3′-dipentyloxacarbocyanine iodide), a voltage-sensitive fluorescent probe for the endoplasmic reticulum and nuclear envelope , was applied to an E5 retina. The DiOC5(3) fluorescence oscillated synchronously across the cells at 6–7 s intervals (Fig. 4A–C). By injecting a cell-impermeant Ca2+ indicator (Oregon Green 488 BAPTA-1 hexapotassium salt), it was shown that the increase in DiOC5(3) fluorescence, which indicates a negative shift in the luminal potential , coincided with the Ca2+ increase (Fig. 4D,E). In order to observe the changes in DiOC5(3) fluorescence intensity at higher time resolution, the intensity was measured using a photomultiplier from the half of the field of view through the eyepiece of the spinning disc confocal scanner simultaneously with an intensified charge coupled device (ICCD) video camera. The recording using the photomultiplier surprisingly revealed that the voltage oscillation comprised repeats of a burst of high-frequency voltage fluctuations (Fig. 4F, background noise is shown in supplementary Fig. S5). The low time resolution of the ICCD video camera (sampling rate of 66 ms for one frame) covered the high-frequency voltage fluctuation (Fig. 4H); the smoothed waveform of the record obtained using the photomultiplier (sampling rate of 1 kHz, 67 points smoothed, Fig. 4G) appeared similar to the record obtained using the video camera (Fig. 4C). To determine whether the burst of high-frequency voltage fluctuations was associated with activation of InsP3 receptor channels, a muscarinic agonist (carbamylcholine, 100 μm) was bath-applied, and clearly increased the frequency of bursts (Fig. 4I).
The present results indicate that the synchronous Ca2+ oscillation is not caused by synchronous action potentials or Ca2+ influx through voltage-dependent Ca channels at the early stage of retinal development. Instead, it was generated by the release of Ca2+ from internal stores. The Ca2+ release may be caused by opening of InsP3 receptor channels, as muscarinic acetylcholine receptor activation induces and is essential for Ca2+ oscillation in developing retina and cerebral cortex [10,13,28]. In the E3 chick retina, stimulation of muscarinic receptors, P2Y purinoceptors and lysophosphatidic acid receptors induces a robust Ca2+ release [29–31]. The agonist-induced Ca2+ increases, however, start asynchronously despite the presence of gap junctions, and the frequency of synchronous Ca2+ oscillation then increases . This implies that diffusion of InsP3 or released Ca2+ itself through gap junctions is not sufficient for synchronization of Ca2+ increases. The synchronous Ca2+ increases in newborn ganglion cells, which lack gap junctions, also suggest that another mechanism underlies the synchronization.
It was unexpectedly found in the present study that the store potential fluctuated at high frequencies during synchronous Ca2+ oscillation. The high-frequency voltage fluctuation burst periodically in a synchronous manner across the cells. This might reflect periodic changes in noise of the photomultiplier. However, the mean level of DiOC5(3) fluorescence intensity shifted upwards during the burst. This result is in accordance with the recording using the ICCD video camera, and indicates a hyperpolarizing shift in the store potential. It is very likely that the burst of voltage fluctuations is initiated by Ca2+ efflux from Ca2+ stores. The Ca2+ releasing channel responsible for the Ca2+ efflux may be the InsP3 receptor channel, which is activated by a constant concentration of InsP3  or its increase . In the present study, application of a muscarinic agonist clearly increased the frequency of bursts of the high-frequency voltage fluctuation. The high-frequency voltage fluctuation may be primarily due to the kinetic properties of InsP3 receptor channels (mean open time of 2.9–6 ms [34–36]).
How does the high frequency voltage fluctuation mediate intercellular communication? At early stages of retinal development, nuclei are surrounded by a thin cytoplasmic rim and intercellular spaces are virtually absent, although the intercellular space becomes larger as ganglion cells differentiate . As the outer membrane of the nuclear envelope, which forms part of a Ca2+ store [35,36], is closely apposed to the plasma membrane (supplementary Fig. S6) [14,27], and the intercellular space is very narrow (supplementary Fig. S7), the store potentials may affect each other by AC coupling through the capacitance of the outer nuclear membrane and the plasma membrane (supplementary Doc. S2A) . The high-frequency fluctuation in the luminal potential of a Ca2+ store could be transmitted to the adjacent Ca2+ store in the neighboring cell through series capacitance of the outer nuclear membrane and the plasma membrane. Thus the luminal potentials of Ca2+ stores would be synchronized.
As the driving force for Ca2+ efflux depends on the store membrane potential, it is plausible that the AC-coupled synchronous voltage fluctuation could entrain releases of Ca2+. The store potential would fluctuate by ≥ 20 mV in amplitude, assuming that a 1 mV change produces a 1.3% change in the DiOC5(3) fluorescence intensity . As the luminal potential fluctuates in both positive and negative directions alternately, Ca2+ ions may leave and enter the store. However, Ca2+ ions cannot enter the store through Ca2+-permeable channels due to the very large difference in [Ca2+] between the inside and outside of the store. Ca2+ ions will be released from the store when the luminal potential shifts in the positive (depolarizing) direction during the burst of voltage fluctuations. The Ca2+ efflux in turn shifts the luminal potential in the negative (hyperpolarizing) direction. Then K+ ions will enter the store to depolarize the luminal potential again. Thus Ca2+ ions could be released from the store during the burst of voltage fluctuations with the co-operation of a counter-influx of K+, and the mean luminal potential shifts in the negative (hyperpolarizing) direction due to the overwhelming Ca2+ efflux.
Termination of the synchronous burst of high-frequency voltage fluctuations may be explained by the quiescence of InsP3 receptor channels  if they close synchronously. Alternatively, a decrease in the conductance of the store membrane could account for the intermittence; closure of the BK channel in the membrane of the Ca2+ store will raise the store membrane resistance after the release of an amount of Ca2+, as the store BK channel is sensitive to the luminal [Ca2+] [14,38]. Closure of the store BK channel would decrease the counter-influx of K+ to terminate the Ca2+ release. The increase in store membrane resistance should lengthen the time constant of the store membrane. Lengthening of the membrane time constant interrupts propagation of high-frequency voltage fluctuations (supplementary Doc. S2B). In fact, membrane-permeant blockers for BK channels (quinidine and paxilline) inhibit the Ca2+ release and DiOC5(3) fluorescence oscillation . These results strongly suggest that activity of the store BK channel is essential for the release of Ca2+ and the generation of Ca2+ oscillation. The high-frequency voltage fluctuation could be transmitted to a newborn ganglion cell even after the loss of gap junctions (supplementary Doc. S2C). The AC-coupled Ca2+ oscillation model is an attractive example of the involvement of ‘pulse-coupled’ oscillators in synchronizing biological rhythms .
The synchronous Ca2+ oscillation may play essential roles in development of the retina. Increases in intracellular [Ca2+] regulate proliferation of retinal neuroepithelial cells [10,27,40,41] and are associated with the interkinetic nuclear movement . In addition to the regulation of proliferation, transient Ca2+ increases regulate neuronal migration, differentiation and neurite formation at early stages of neuronal development .
Preparation of retina
The neural retina was isolated from chick embryos incubated for three (E3) to six (E6) days at 38 °C. The optic cup was dissected from an E3 chick embryo, and the neural retina (the inner wall of the optic cup) was isolated. The diameter of the neural retina was approximately 400 μm. At E4–E6, the central part of the neural retina was trimmed so that the size of the retinal piece was 400 × 400 μm. The retina was positioned on the bottom of a recording chamber (0.2 mL volume) with the inner side up using a ring of nylon wire . The recording chamber was mounted on the fixed stage of an upright microscope (BX51WI; Olympus, Tokyo, Japan). The recording chamber was perfused at 2 mL·min−1 with normal bath solution (NBS) containing 137 mm NaCl, 5 mm KCl, 2.5 mm CaCl2, 1 mm MgCl2, 10 mm Hepes and 22 mm glucose, buffered to pH 7.3 by adding NaOH. Bath solutions were changed from NBS to the test solution using solenoid valves (General Valve Corp., Fairfield, NJ, USA). Recordings were made at 26–28 °C (or 28–38 °C for oscillations).
Intracellular recording and dye injection
Microelectrodes were produced using a puller (P-97; Sutter Instrument Co., Novato, CA, USA) from a borosilicate glass capillary containing a filament (GC150F-10, Harvard Apparatus Ltd, Edenbridge, UK), and filled with 3 m KCl or 4 m potassium acetate . The electrode resistance (Re) was 140–180 MΩ. A conventional intracellular recording amplifier (MA-8504; Physio-Tech, Tokyo, Japan) was used. The output signal was low-pass filtered at 5 kHz and sampled at 10 kHz. The microelectrode was manipulated using a linear actuator (850F; Newport, Irvine, CA, USA) or a piezoelectric positioning system (PCS-1000; Burleigh Instruments Inc., Fishers, NY, USA). For dye injection, a microelectrode was filled with 200 mm KCl containing 1% Alexa Fluor 488 hydrazide (Molecular Probes Inc., Eugene, OR, USA; Re: 400–600 MΩ). When an impaled cell showed a resting membrane potential more negative than −40 mV, the dye was injected into the cell by passing hyperpolarizing current pulses (0.5–1.0 nA in amplitude, 400 ms in duration at 800 ms interval) for 0.5–3 min. Then the electrode was carefully pulled away after confirming that the cell still showed a resting membrane potential more negative than −20 mV. A period of at least 5 min was allowed for the dye to diffuse through gap junctions. The same electrode was used to inject the dye into the extracellular space using a break circuit, which connected the electrode to −120 V for about 1 s. Oregon Green 488 BAPTA-1 hexapotassium salt (1% in 200 mm KCl, Molecular Probes; Re: 530–690 MΩ) was injected for monitoring intracellular Ca2+.
Confocal fluorescence microscopy
Confocal fluorescence images were taken using a confocal scanner consisting of a Nipkow disc with microlens array (CSU10; Yokogawa, Tokyo, Japan) and a cooled ICCD video camera (Gen IV; Princeton Instruments and Nippon Roper, Chiba, Japan) through a water immersion objective (CFI Plan 100×W, numerical aperture 1.10, working distance 2.5 mm; Nikon, Tokyo, Japan or LUM Plan FL100XW, NA 1.00, WD 1.5 mm; Olympus). The images were recorded and analyzed using a fluorescence cytophotometric system (FC-500; Furusawa Labo, Kawagoe, Japan). Fluorescence was excited with a laser diode (473 nm, 30 mW, HK-5510; Shimadzu, Kyoto, Japan). The output power of the laser diode was regulated within ± 1% by an auto power control circuit. DiOC5(3) (2 μm; Molecular Probes) was bath-applied for 20 min [14,27]. The DiOC5(3) fluorescence intensity was also measured using a photomultiplier (H5784; Hamamatsu Photonics, Hamamatsu, Japan) from the eyepiece of the confocal scanner. The output signal of the photomultiplier was amplified and low-pass filtered at 5 kHz and sampled at 1 kHz using a data acquisition system (PowerLab 2/25; ADInstruments Japan Inc., Nagoya, Japan). To monitor intracellular Ca2+ increases, the retina was loaded with Oregon Green 488 BAPTA-1 acetoxymethyl ester (10 μm; Molecular Probes) for 1 h at 27 °C. Cremophor EL (0.5%; Sigma, St Louis, MO, USA) was used to dissolve the dye.
An E5 retina was fixed in 4% paraformaldehyde in NaCl/Pi for 15 min at room temperature. After washing with NaCl/Pi, the retina was incubated for 3 min in NaCl/Pi containing 0.2% Triton X-100. Nonspecific binding was blocked for 30 min at room temperature using 10% normal goat serum in NaCl/Pi containing 0.2% Triton X-100. The retina was then incubated with an antibody against CaV3.1 (ACC-021, diluted 1 : 130; Alomone Labs, Jerusalem, Israel) labeled with Alexa Fluor 488 (Zenon antibody labeling kits; Molecular Probes) in NaCl/Pi containing 0.2% Triton X-100 for 1 h at room temperature. After washing with NaCl/Pi, confocal fluorescence images were taken. A negative control was prepared following the same procedure without the antibody.
I thank Dr Kaoru Katoh for technical advice and Dr Yoshio Maruyama for discussions. This work was supported by the Japan Spina Bifida and Hydrocephalus Research Foundation (JSBHRF), the NISSAN Science Foundation, the Strategic Promotion System for Brain Science, Special Coordination Funds for Promoting Science and Technology (SCF), the Japan Society for the Promotion of Science (JSPS) and the Ministry of Education, Culture, Sports, Science and Technology.