Synchronized Spontaneous Ca2+ Transients in Acute Anterior Pituitary Slices*

We investigated the organization of spontaneous rises in cytosolic free Ca2+ concentration ([Ca2+] i ) due to electrical activity in acute pituitary slices. Real time confocal imaging revealed that 73% of the cells generated fast peaking spontaneous [Ca2+] i transients. Strikingly, groups of apposing cells enhanced their [Ca2+] i in synchrony with a speed of coactivation >1,000 μm/s. Single-cell injection of Neurobiotin or Lucifer yellow labeled clusters of cells, which corresponded to coactive cells. Halothane, a gap junction blocker, markedly reduced the spread of tracers. Coupling between excitable cells was mainly homologous in nature, with a prevalence of growth hormone-containing cells. We conclude that spontaneously active endocrine cells are either single units or arranged in synchronized gap junction-coupled assemblies scattered throughout the anterior pituitary gland. Synchrony between spontaneously excitable cells may help shape the patterns of basal secretion.

Endocrine pituitary cells that release hormones from large dense core vesicles (LDCV) 1 by calcium-mediated exocytosis exhibit spontaneous firing of action potentials. In cultured cell preparations, individual cells present asynchronous activity with different firing patterns (pace-making or bursting mode). When electrical recordings are combined with fluorescent monitoring of cytosolic free Ca 2ϩ concentration ([Ca 2ϩ ] i ), single spontaneous spikes trigger transient rises in [Ca 2ϩ ] i with characteristic features; a time to peak of less than 1 s and a return within a couple of seconds (1). The lag between the onset of the [Ca 2ϩ ] i rise and exocytosis is also within a subsecond range (2,3), sharing common features of stimulus-secretion coupling with neuroendocrine cells (e.g. chromaffin cells) and neurons, which release peptides from LDCV (4 -6).
Although electrical activity has already been detected in the anterior pituitary gland, both in vivo and in tissue preparations (7)(8)(9), the dynamics and organization of [Ca 2ϩ ] i rises associated with spontaneous action potentials have not yet been investigated at the tissue level. Based on the heterogeneous distribution of the five secretory cell types throughout the tissue (10), endocrine cells would display asynchronous firing so that the overall activity of each secretory type would simply reflect the average of single cell events. Cell regulation should mainly depend on the input of hypothalamic clocks, such as sequential release of growth hormone-releasing factor and somatostatin, which have been shown to pace growth hormone (GH) release (11). However, the gland disconnected from the hypothalamic inputs still shows pulsatile GH release (12), suggesting a synchronization of cellular signals within the tissue. With regard to the mechanisms accounting for synchronization in other tissues (13)(14)(15)(16)(17)(18), two sources of cell-to-cell communication, not mutually exclusive, could be proposed. First, both endocrine and non-endocrine (folliculostellate) pituitary cells release various products (ATP, dopamine, and so forth), which locally act on neighboring cells (19 -21). Second, gap junctions present in the anterior pituitary (9,(22)(23)(24) may allow both metabolic and electrical coupling between connected cells.
To study the behavior of spontaneously active cells within the adenohypophysis, we measured the multicellular patterns of spontaneous [Ca 2ϩ ] i rises in acute slices of guinea pig pituitary, which preserved tissue structure (25). Real time confocal laser microscopy with the Ca 2ϩ -sensitive fluorescent dye fluo-3 offers a sensitive method for optical recording of the fast peaking [Ca 2ϩ ] i transients due to spontaneous action potentials. By visualizing the multicellular profiles of [Ca 2ϩ ] i activity in these slices, we detected clusters of spontaneously coactive endocrine cells, which were scattered throughout the anterior pituitary. An abstract of a preliminary account of these results has already been presented (51).

EXPERIMENTAL PROCEDURES
Tissue Slice Preparation-Acute pituitary slices were prepared according to previously reported methods (26). Briefly, the pituitary gland was removed from 4 -8-week-old female guinea pigs (OCF-DH albinos) that had been killed by decapitation after pentobarbital anesthesia. After keeping the gland in ice-cold saline for 2 min, it was glued onto an agarose cube and transferred to the stage of a vibratome (Microslicer®, DTK-1000, D.S.K, Dosaka EM Co. Ltd., Kyoto, Japan). Coronal slices of 150-m thickness were then cut with a razor blade and transferred to a storage chamber thermostated at 32°C, containing Ringer's saline (in mM): 125 NaCl, 2.5 KCl, 2 CaCl 2 , 1 MgCl 2 , 1.25 NaH 2 PO 4 , 26 NaHCO 3 , 12 glucose, and buffered to pH 7.4. The saline was continuously bubbled with carbogen (95% O 2 , 5% C0 2 ). As reported for the slices of the intermediate lobe (26), slices of the anterior lobe were suitable for patch-clamp recordings and Ca 2ϩ signal measurements immediately after cutting. Slices were viable up to 8 h, as seen by the presence of spontaneous [Ca 2ϩ ] i elevations. To achieve optical and/or electrophysiological recordings, pituitary slices were transferred to a recording chamber attached to the stage of an upright microscope fitted with differential interference contrast optics (Axioskop FS, Zeiss, Le Pecq, France) and continuously superfused with Ringer's saline at 30°C.
Confocal Microscopy-Fast spontaneous [Ca 2ϩ ] i transients were routinely measured by a real time (30 -480 frames/s) confocal laser scanning microscope equipped with an Ar/Kr laser (Odyssey XL with In-terVision 1.4.1 software, Noran Instruments Inc., Middleton, WI). Cells were viewed with a 63 ϫ 0.9 numerical aperture achroplan water immersion objective lens (Zeiss). Various thicknesses of confocal images was obtained by selecting different detection slits. The larger slit (100 m) was used for [Ca 2ϩ ] i signals, giving bright images with a 3.1-m axial resolution. When cells were subsequently loaded with Lucifer yellow (see below), confocal images were acquired with a 25-m slit, which provided an axial resolution of ϳ1.3 m. Slices were loaded with the Ca 2ϩ -sensitive fluorescent probe fluo-3 by exposure to 10 M fluo-3 acetoxymethyl ester (fluo-3/AM, Molecular Probes, Eugene, OR) for 20 -30 min at 32°C (27). Fluo-3 was excited through a 488-nm band pass filter, and the emitted fluorescence was collected through a 515-nm barrier filter. Transmitted images were acquired using the longer wavelength of the laser beam (647 nm), which penetrated deeply into pituitary slices. To follow the time course of fluo-3 emission changes, the bright over time tool of the InterVision 1.4.1 software package was applied to areas that surrounded cells either on live images or following capture of sequential images into memory of an Indy R4600SC/133 MHz Silicon Graphics station equipped with a Cosmo compress JPEG board. To ensure that the image rate acquisition was adequate to resolve the time delay between [Ca 2ϩ ] i changes measured in cells of the same field, the fluorescence changes were also acquired in the line scanning mode at a rate of 4.4 ms/line (6400 ns sample time, 480 lines). In some experiments, the line trigger (TTL signal) from the Odyssey XL was used to trigger a voltage pulse that helped synchronize its timing with the line 100. A voltage stimulator delivered these depolarizing pulses (2 ms, 6 V) to a patch pipette filled with Ringer's saline and positioned on the cell (28). Because fluo-3 is a single-wavelength dye, its emission is a function of both intracellular Ca 2ϩ and dye concentrations. [Ca 2ϩ ] i changes were therefore expressed as the F/F min ratio where F min was the minimum fluorescence intensity measured during the recording (27). No detectable difference was noted between slices used just after cutting or after spending several hours in the storage chamber. Acquired data were then processed for analysis using either the Indy station (InterVision 1.4.1, two-dimensional analysis module) or a PowerPC 8100/100 MHz (NIH Image 1.6.0, Adobe Photoshop 3.0.5 or Igor Pro 2.03).
Electrophysiology-Membrane potential was recorded in the wholecell configuration of the patch-clamp technique (29). Patch pipettes were pulled to a resistance of 4 -8 megohms from borosilicate glass (1.5-mm outer diameter, 1.17-mm inner diameter) and filled with the following internal solution (in mM): 140 potassium gluconate, 10 KCl, 2 MgCl 2 , 1.1 EGTA, 5 HEPES, that was titrated to pH 7.2 with KOH. For simultaneous recordings of membrane potential and [Ca 2ϩ ] i , the perforated whole-cell patch-clamp technique was preferred to the conventional patch-clamp technique. In this case, the internal pipette solution was composed of (in mM): 10 KCl, 10 NaCl, 70 K 2 SO 4 , 7 MgCl 2 , 5 HEPES, and 100 g/ml nystatin (Sigma). Nystatin was added to the electrode solution before filling the patch pipettes. Perforation was usually achieved within 10 min after seal formation. Cells with an access resistance Ͼ30 megohms were discarded. Membrane potential was recorded under current-clamp conditions using a List EPC-9 patchclamp amplifier (HEKA Electronik, Lambrecht/Pfalz, Germany) and filtered at 3 kHz. Patch-clamp signals were acquired and analyzed using Pulse ϩ PulseFit softwares (version 7.86, HEKA Electronik) on a PowerPC 8100/100.
Cell-to-Cell Communication and Dye Coupling-The fluorescent dye Lucifer yellow (LY, 4% in 150 mM LiCl) was introduced into cells through a sharp microelectrode. The cells were impaled and filled for a few minutes, before image acquisition with the confocal microscope. When LY was injected into cells not subjected to [Ca 2ϩ ] i imaging, the cells were selected on the basis of their round or oval shape and the presence of dense vesicles, which resembled hormone-containing vesicles. In addition, the hormonal content of impaled cells was routinely characterized by immunofluorescence after formaldehyde fixation of slices.
Cells were also loaded with the Neurobiotin TM (N-(2-aminoethyl)biotinamide hydrochloride) tracer (1% in the internal solution) by diffusion through the patch pipette for 10 -30 min. To increase the rate of Neurobiotin dialysis, depolarizing pulses (500-ms duration, 0.5 Hz) were periodically applied. In some cells, the presence of gap junctional communication was assessed by using the size exclusion properties of dextran conjugates (9). Cells were dialyzed with a patch pipette containing 1% Neurobiotin plus 2 mg/ml dextran Texas Red (lysine fixable, M r 3,000) for at least 10 min. After fixation (see below), slices were rinsed in phosphate-buffered saline (PBS, pH 7.4) plus 0.1% bovine serum albumin (BSA) and then incubated in PBS ϩ BSA ϩ 0.8% saponin for 1 h at room temperature. After several rinses in PBS ϩ BSA, Neurobiotin staining was revealed using fluorescein-or Texas Red-labeled avidin D (1:200 dilution, room temperature, 5 or 1 h, respectively).
Slices were incubated overnight at 4°C in the presence of antibodies raised against human GH (developed in rabbit, used at 1:2500 dilution) with 0.05% BSA in TBS. The antiserum was kindly donated by Dr. Tillet (INRA, Nouzilly, France) and the NIDDK-human GH-B-1 was used as an antigen. Incubation was followed by several washes with TBS plus 0.05% BSA. The primary antibodies were localized by a 1-hr incubation with a Cy5-or fluorescein-conjugated anti-rabbit IgG developed in donkey (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA), used at 1:250 dilution in TBS plus 0.05% BSA. After washes, slices were postfixed in formaldehyde 3% (PBS, pH 7.4) and rinsed in 50 mM NH 4 Cl before mounting in Slowfade TM Light (Molecular Probes). Images were acquired with the confocal microscope as described above. Cells were viewed with a 63 ϫ 1.4 numerical aperture plan-apochromat with a 15-m detection slit (ϳ0.65-m axial resolution).
Test Substances-Drugs were pressure-ejected from an extracellular micropipette (tip diameter 2-5 m), the tip of which was positioned in the vicinity of the recorded cells. The concentration reported are those in the pressure pipette. Lucifer yellow, somatostatin-14, and the calcium channel blocker CdCl 2 , were purchased from Sigma. The gap junction blocker halothane (2-bromo-2-chloro-1,1,1-trifluoroethane) was from Fluka. To obtain Ca 2ϩ -free solution, CaCl 2 was omitted from, and 5 mM EGTA was added to modified Ringer's saline. The Neurobiotin tracer was purchased from Vector Laboratories (Biosys S.A., Compiègne, France).
Statistics-Numerical data are expressed as the mean Ϯ S.E. Student's t test was used to compare means when appropriate. Differences between groups were assessed by using the non-parametric Mann-Whitney U test. Differences with p Ͻ 0.05 were considered significant.

RESULTS
Experiments were performed in coronal slices (150-m thickness) from pituitary of 4 -8-week-old female guinea pigs. Slices were loaded with fluo-3/AM by bath application of the Ca 2ϩ indicator, which produced widespread staining of the first and second layers of cells on the slice surface. When visualized with epifluorescence under the upright microscope, numerous cells within each field exhibited spontaneous changes in fluo-3 emission, reflecting rises in [Ca 2ϩ ] i (31). Time-lapse optical sequences of the cells showing spontaneous [Ca 2ϩ ] i rises were then recorded with fast scanning confocal imaging (120 images/s with averaging 4 frames) during different experimental protocols.
Presence of Spontaneously Coactive Cells in Acute Pituitary Slices-In most slices, one to several groups of adjacent active cells which fired synchronously were observed, as visualized at first using the epifluorescent port of the microscope. Real time optical imaging revealed then that these clusters of synchronous cells coexisted with asynchronous neighboring cells as illustrated in Fig. 1A. The top left image shows a field in which four fluo-3-loaded cells could be observed in the same optical section. The montage of consecutive optical slices depicts a time series of fluo-3 emission frames encoded in pseudocolors (from blue to red with [Ca 2ϩ ] i increasing). The three bottom cells fired spontaneous fast-peaking [Ca 2ϩ ] i transients. The plots of relative fluo-3 emission changes show that two cells paced their [Ca 2ϩ ] i in synchrony, whereas the third one had its own rhythm. With only the first pair of synchronized [Ca 2ϩ ] i transients, it seems likely that the red-circled cell became active before the green-circled one while the lag between the following pairs of [Ca 2ϩ ] i transients was indistinguishable under the time resolution used in these experiments.
Both synchronous and asynchronous cells presented spontaneous [Ca 2ϩ ] i transients with roughly similar kinetics (time to peak ϭ 150.6 Ϯ 6.8 ms, n ϭ 141), patterns (pace-maker or bursts) and firing frequency (0.54 Ϯ 0.07 Hz, range from 0.1 to 2.4 Hz, n ϭ 43). No difference was noted with regard to the location of synchronized cell clusters within coronal slices. In addition, spontaneously coactive cells were also detected in sagittal slices (n ϭ 4) suggesting a lack of preferential orientation for the spread of coactivation among spontaneously active cells. Confocal imaging routinely enabled us to observe doublets of coactive cells in single optical sections. However, multiplets with three to four coactive cells were occasionally encountered (Fig. 1B).
The kinetic features of spontaneous [Ca 2ϩ ] i transients observed in acute pituitary slices along with previous [Ca 2ϩ ] i data obtained in cultured pituitary cells (1,32) suggest that these [Ca 2ϩ ] i fluctuations resulted from voltage-gated Ca 2ϩ entry. We therefore investigated their extracellular Ca 2ϩ dependence. Local application of a Ca 2ϩ -deprived solution reversibly blocked synchronous [Ca 2ϩ ] i transients (Fig. 1C, n ϭ 5) as well as asynchronous [Ca 2ϩ ] i fluctuations (n ϭ 24). A Cd 2ϩ -containing Ringer's saline (0.5 mM CdCl 2 ) ejected in the vicinity of cells had a similar effect on both synchronous (n ϭ 3) and asynchronous [Ca 2ϩ ] i transients (n ϭ 8, data not illustrated). Interestingly, the first [Ca 2ϩ ] i transients still occurred synchronously after washout (Fig. 1C, inset), suggesting that blockade of Ca 2ϩ entry did not alter the coactivation spread. When [Ca 2ϩ ] i monitoring was combined simultaneously with recordings of membrane potential using the perforated patch-clamp technique, the triggering of a single action potential was sufficient to elicit a rapid increase in [Ca 2ϩ ] i with kinetics (time to peak ϭ 182.4 Ϯ 25.1 ms, n ϭ 18) similar to those of spontaneous [Ca 2ϩ ] i transients recorded in cells not subjected to patchclamping (p Ͼ 0.05, data not shown).
Dye Diffusion between Spontaneously Coactive Cells-Different communication mechanisms, either electrical or biochemical, could explain the synchronization of spontaneously active cells. Since coactivation always occurred between apposing cells of finite clusters, the synchronization signal should be restricted to coactive cell-cell boundaries, but not extensively diffused to neighboring asynchronous cells. Since gap junctions were described in the anterior pituitary (22), we carried out dye coupling experiments with a low molecular weight fluorescent dye LY (457 Da). The tracer was injected through a sharp microelectrode into single cells belonging to synchronized clusters. Fig. 2A illustrates an example of two neighboring cells exhibiting synchronized spontaneous [Ca 2ϩ ] i transients. After fluo-3 emission measurements, cell labeled 2 was impaled with LY (4% in 150 mM LiCl). A few seconds later, cell labeled 1 was also stained with LY, indicating the LY diffusion from the impaled cell to the coupled partner. No diffusion was observed in other adjacent cells (n ϭ 21). Moreover, LY diffusion was restricted to single impaled cells, which spontaneously fired with their own rhythm (Fig. 2B, n ϭ 14). These data strongly suggest that gap junctions could ensure the spread of coactivation between excitable pituitary cells. Since large molecules (Ն1,000 Da) usually do not permeate through gap junctions, further experiments were conducted with large molecular weight dextran conjugates. Cells were randomly loaded with Neurobiotin, a low molecular mass dye (323 Da, 1% in internal patch pipette solution) and dextran Texas Red (3,000 Da, 2 mg/ml), by diffusion for 10 -30 min. In two out of seven clusters, patched cells contained both markers, whereas coupled cells were labeled with Neurobiotin only (data not shown). In the others, no dye diffusion was observed.
When LY or Neurobiotin were injected into cells not subjected to [Ca 2ϩ ] i imaging, the tracers diffused from the impaled cell to neighboring cells in 47% of the clusters tested (n ϭ 57/121). The extent of dye coupling, however, did not usually exceed five to six cells, as seen by three-dimensional reconstruction of optical slices (data not shown). Similar experiments were further conducted in presence of the gap junction blocker halothane (33,34). In pituitary slices bathed for 15-60 min in Ringer's saline saturated with 3 mM halothane, the appearance of dye-coupled cells was reduced to 17.8% (Fig. 2C, p Ͻ 0.05). It was unlikely that the lowering of dye-coupling was linked to a decrease in [Ca 2ϩ ] i spiking activity since fast spontaneous [Ca 2ϩ ] i transients still occurred in the presence of the blocker. In most cells, halothane did not alter the shape (time to peak 157.5 Ϯ 7.8 ms, n ϭ 74 transients recorded in 14 different cells, p Ͼ 0.05) and frequency (0.51 Ϯ 0.09 Hz, n ϭ 14, p Ͼ 0.05) of spontaneous [Ca 2ϩ ] i transients. In the other cells, halothane, however, reduced [Ca 2ϩ ] i spiking. The reversibility of halothane effects was checked by placing treated slices in a halothane-free saline for 1-3 h. Under these conditions, the percentage of dye-coupled cells recovered to control level (Fig.  2C, p Ͼ 0.05). Fig. 2D shows the uncoupling effect of halothane (3 mM, 3-5 min bath application) on previously coactive cells (n ϭ 5). In addition, LY injection in one of these cells failed to label its neighboring cells. Taken together, these data strongly suggest that the spread of synchronization could be mediated by gap junctions.
Speed of Coactivation Is Higher than 1,000 m/s-Spread of coactivation through gap junctions can be due to either simple diffusion of Ca 2ϩ from a trigger cell to coupled cells or electrical coupling between these cells. Since these two mechanisms are associated with distinct speeds of propagation, x-t-series line scans were performed to estimate the speed of the wave of coactivation. A single horizontal line crossing synchronized cells was continuously scanned over 2,133 ms (4.4 ms/line, 480 lines). Fig. 3A shows each line displayed in time along the y axis. The time course of the fluorescence changes in synchronized cells revealed a 5.6-ms delay between the onset of spontaneous [Ca 2ϩ ] i transients. Taking into account the distance between cell centers, the average speed of the coactivation wave from the trigger cell to the other was 1,056 Ϯ 112 m/s (n ϭ 40, 5 different coactive clusters). Interestingly, the calculation of the delay between [Ca 2ϩ ] i transients within coactive cell assembly revealed that the trigger cell could become the responding cell and vice versa during recordings (four out of five clusters, Fig. 3B).
In parallel experiments, an external voltage stimulation (6 V, 2 ms) was delivered from a micropipette touching one cell at the slice surface. Such a stimulation induced a fast [Ca 2ϩ ] i rise which could propagate to one or two adjacent cells in the same focal plane (n ϭ 9, data not shown). The speed of propagation calculated from x-t-series data (1,639 Ϯ 613 m/s, n ϭ 4 clusters) did not differ from that found between spontaneous synchronized cells (p Ͼ 0.05).
Prevalence of Homologous Coupling between GH-containing Coactive Cells-Earlier studies carried out in pituitary cells isolated from the tissue have revealed that spontaneous electrical activity and ensuing [Ca 2ϩ ] i transients were mainly observed in cells containing either GH or prolactin (35). Since somatotrophs are the dominant secretory cell type in the anterior pituitary, we therefore investigated whether coactive cells could contain GH. To conduct this experiment, the hormonal content of cells was identified by immunohistofluorescence following formaldehyde fixation of slices. In 50% of cell clusters (14 out of 28), spontaneously coactive cells were GH-containing cells as illustrated in Fig. 4 (mid top frame). As expected from previously described results, the two synchronized cells were stained following LY injection (cells that turned green, second top frame). Interestingly, other apposing cells were also immunoreactive to GH but not dye-coupled to coactive cells. This suggests that homologous coupling could at least involve subsets of spontaneously active somatotrophs. Somatostatin, a native inhibitor of GH release, reversibly blocked spontaneous [Ca 2ϩ ] i transients issued from both synchronous and asynchronous somatotrophs (n ϭ 6 and 13, respectively, data not shown). It should be noted that clusters of synchronized cells composed of both GH-positive and -negative cells were also encountered (5 out of 28 clusters). The remainders were GHnegative (9 out of 28). DISCUSSION Our experiments describe for the first time the multicellular [Ca 2ϩ ] i rises, which spontaneously occur in acute slices from anterior pituitary. This spontaneous [Ca 2ϩ ] i activity is spatially organized within the tissue into small groups of excitable cells that pace their [Ca 2ϩ ] i synchronously. Combination of real time optical imaging with dye-coupling studies enabled us to observe a fast speed of coactivation (Ͼ1,000 m/s) that involves cell-to-cell communication via gap junctions. In contrast, other cells which spontaneously display asynchronous [Ca 2ϩ ] i transients are never dye-coupled to neighboring cells.
The discovery of coactivation of spontaneously active pituitary cells has marked a new step in our knowledge of how endocrine cells secreting from LDCV interact with one another in the absence of any stimulus. Although domains of spontaneously coactive cells have been extensively described, namely in the brain (16, 17, 36, 37), synchronization between excitable endocrine cells has only been reported in secretagogue-stimulated cells, such as beta cells from pancreatic islets exposed to high glucose levels. The latter are electrically silent at rest (i.e. at low glucose concentrations), while they display synchronized bursts of Ca 2ϩ -driven action potentials only in response to their fuel secretagogue (38,39).
In the brain, two distinct mechanisms support the spread of coactivation between spontaneously excitable cells. Synaptic transmission synchronizes neuronal activity in many brain areas (15), whereas coupling through gap junctions underlies local coactivation in, e.g. neocortex neuronal domains (16). Our results strongly suggest that gap junctions cause coupling of excitable pituitary cell subsets and thereby allow the synchronization of spontaneous [Ca 2ϩ ] i transients in these assemblies of cells. Two observations concur with this proposal. Cell injection of small tracers (LY or Neurobiotin) results in the selective labeling of coactive cells and halothane, a gap junction blocker, markedly lowers the appearance of dye coupling without affecting the time course of spontaneous [Ca 2ϩ ] i transients. Since all these studies have been done in slices, the relevance of these findings in vivo is yet unknown. Nevertheless, we and others have already observed LY transfer between unidentified pituitary cells in both neonatal rat pituitary slices maintained in long term (Ն1 month) organotypic culture (9) and rat hemipituitaries (24). Ultrastructural and immunohistofluorescent studies have also demonstrated the presence of gap junction plaques and the expression of two connexin types (Cx26 and Cx43) in the anterior pituitary (22,40,41). Altogether, this strongly suggests that cell-to-cell calcium signaling mediated by gap junction communication is indeed present in the gland and is not a side effect of the acute slice preparation.
Line scanning mode experiments reveal that the speed of coactivation is higher than 1,000 m/s. Although we do not rule out that Ca 2ϩ or a metabolite (e.g. inositol 1,4,5-trisphosphate) can slowly diffuse through gap junctions in the cell assemblies (13,14,42), cell coactivation underlying synchronized [Ca 2ϩ ] i transients is associated with a much faster and regenerative mechanism in the anterior pituitary. If single action potentials drive spontaneous [Ca 2ϩ ] i transients, they should therefore act as a trigger for coactivation between gap junction-coupled cells. However, the occurrence of synchronized [Ca 2ϩ ] i transients upon removing the blockade of Ca 2ϩ entry suggests that maintenance of connectivity is not due to Ca 2ϩ entry per se.
The mechanism driving the recruitment of coactive cells is tantalizing. What are the trigger cells? Given that external voltage stimulation causes a [Ca 2ϩ ] i rise, which can propagate to adjoining cells, we suggest that electrical coupling mediates intercellular communication between excitable cells. However, the entrainment is not likely to be associated with the firing of fast-spiking cells which "chatter" slower cells (17), since the frequencies of [Ca 2ϩ ] i transients are roughly similar in both synchronous and asynchronous cells. Interestingly, trigger cells can alternate with time within the coactive cell domain. The mechanisms that dictate the wide range of spiking patterns in the excitable pituitary cells remain hard to identify despite the fact that the voltage-gated channels that open during the action potential have been well characterized (35,43). One would therefore assume that the stochastic occurrence of action potentials would continuously determine which cell triggers spikes in neighboring cells via gap junctions. Paracrine interactions may also play a significant role in the selection of trigger cells. Since single action potentials seem to be efficient enough to trigger fast exocytosis in gland (adrenal) slices (44), a minute fraction of fast-acting factors (e.g. ATP, dopamine copacked with hormones in LDCV) (19,21) readily released upon single action potentials would quickly alter the firing frequency of any coactive cells (45) and thereby change the hierarchy of the propagation of electrical events. Finally, dyecoupled cells are not always within a single field at a single plane of focus suggesting that the cell phasing synchronization can be out of focus during optical recordings. A fine strategy for studying the organization of [Ca 2ϩ ] i events within coactive cell domains would therefore consist in applying three-dimensional imaging in real time.
The significance of spontaneously coactive cell domains in the anterior pituitary can be viewed in terms of [Ca 2ϩ ] i signal and hormone secretion. Cells containing GH in their LDCV prevail in coactive cell assemblies. Interestingly, coactive GHcontaining cells often coexist with nearby asynchronous GH cells. Hence, the patterns of GH release should depend on the integration of the exocytotic activities of both synchronous and asynchronous cells. Morphological studies have also suggested a polarized phenotype for GH cells in situ since the latter are mainly arranged in palisades alongside fenestrated capillaries (10). Thus, the concurrent level of secretory efficiency within the capillaries depends on the topographical distribution of the two distinct GH subsets within the columns of pituitary cells (so-called cell cords), which are separated by basal laminae, connective tissue and blood vessels (10).
An association between GH-positive and GH-negative cells was also encountered, suggesting that the intricate pattern of spontaneous coactivation may encode the trigger for releasing factors other than hormones. In our view, neurotransmitterlike factors found in LDCV (e.g. dopamine in somatotrophs and lactotrophs) (19) could represent putative candidates since their fast and evanescent actions on pituitary cells (46) could provide a fine tuning of nearby cell activities.
Besides a possible role for local control of secretion, coactivation of excitable endocrine cells may serve a more integrated function in the entire gland. This can require a substantial number of connections between synchronized groups. Can a mechanism account for close synchrony despite the apparent wide range of distances (and the presence of connective tissue) between spontaneously coactive areas in vivo? The episodic releases of hypothalamic secretagogues (e.g. GH-releasing factor) (11) might periodically alter the number of coupled cells by acting on gap junctions (47,48). This is certainly plausible, but these connections might not be dense enough to allow synchronization across very distant coactive areas. An alternative scenario is that other cells might pace collective rhythms within the anterior pituitary gland. Folliculostellate cells would be good candidates since they form a cell network, which extends throughout the anterior pituitary gland (49). These cells are coupled by gap junctions (24,49,50) and communicate with endocrine cells via paracrine interactions (20) and gap junctions (23,24). Extensive analysis of these two putative mechanisms would be of particular interest, insofar as long distance synchrony may be an important determinant for shaping the patterns of hormone release in the systemic circulation.