The suprachiasmatic nucleus (SCN) is the master clock in mammals governing the daily physiological and behavioral rhythms. It is composed of thousands of clock cells with their own intrinsic periods varying over a wide range (20–28 h). Despite this heterogeneity, an intact SCN maintains a coherent 24 h periodic rhythm through some cell-to-cell coupling mechanisms. This study examined how the clock cells are connected to each other and how their phases are organized in space by monitoring the cytosolic free calcium ion concentration ([Ca2+]c) of clock cells using the calcium-binding fluorescent protein, cameleon. Extensive analysis of 18 different organotypic slice cultures of the SCN showed that the SCN calcium dynamics is coordinated by phase-synchronizing networks of long-range neurites as well as by diffusively propagating phase waves. The networks appear quite extensive and far-reaching, and the clock cells connected by them exhibit heterogeneous responses in their amplitudes and periods of oscillation to tetrodotoxin treatments. Taken together, our study suggests that the network of long-range cellular connectivity has an important role for the SCN in achieving its phase and period coherence.
The suprachiasmatic nucleus (SCN) is the mammalian circadian clock orchestrating a range of physiological and behavioral circadian rhythms (Moore & Eichler, 1972; Stephan & Zucker, 1972). Many earlier studies reported that the SCN circadian rhythm stems from individual SCN clock cells (Klein et al., 1991; Welsh et al., 1995; Liu et al., 2007) and that the intracellular circadian oscillations are the result of transcription–translation feedback loops arising from multiple ‘clock genes’, such as period1 (Per1), period2 (Per2) and clock (Reppert & Weaver, 2002; Kohsaka & Bass, 2006). Interestingly, populations of SCN clock cells are quite heterogeneous, with their intrinsic periods of oscillation varying significantly from 20 to 28 h (Welsh et al., 1995; Honma et al., 2004). Therefore, from the viewpoint of dynamical system theory, the SCN is a coupled network of heterogeneous populations of non-linear oscillators. In order for such a system to achieve a coherent circadian rhythm overall, there must be some cell-to-cell communication resulting in frequency locking (Low-Zeddies & Takahashi, 2001; Liu et al., 2007).
In addition to the phenomenon of frequency locking, populations of coupled non-linear oscillators generically exhibit some form of phase synchronization when the coupling strength of the interaction becomes larger than a critical value (Pikovsky et al., 2003). Therefore, some organized spatiotemporal phase dynamics can be expected to occur in the SCN. Indeed, recent studies imaging period1-luciferase (Yamaguchi et al., 2003), PERIOD2::Luciferase reporters (Yan et al., 2007; Evans et al., 2011; Foley et al., 2011), or period1-green fluorescent protein (Per1-GFP) (Quintero et al., 2003; Hughes et al., 2008) suggest that the phases of different SCN clock cells are not synchronized but advance progressively in space. However, there is no definite picture of the spatiotemporal organization of the SCN circadian rhythm, and quantifying the phase dynamics and its biological implications is currently a large challenge in conjunction with the anatomical structure of the SCN (Butler & Silver, 2009; Welsh et al., 2010; Mohawk & Takahashi, 2011).
Intracellular free calcium ions are often viewed as a second messenger mediating intercellular communication and as an intracellular agent coupling the clock gene oscillation and action potential (AP) firing rhythm in SCN neurons (Pennartz et al., 2002; Honma & Honma, 2003; Park et al., 2003; Ikeda, 2004). This article reports detailed analysis on the spatiotemporal phase dynamics of cytosolic free calcium ion concentration ([Ca2+]c) in rat SCNs that are organotypically cultured. The calcium-binding fluorescent protein yellow cameleon (YC) 2.1 was used to obtain [Ca2+]c signals continuously from individual clock cells. The phases of the [Ca2+]c circadian rhythms appear to be quite random in space. However, based on a careful analysis, we find that the SCN circadian dynamics is governed by two different mechanisms working together simultaneously; first, diffusive local wave conduction, and second, long-range direct cell-to-cell coupling through neurites. Therefore, from the viewpoint of dynamical systems, the SCN resembles the cardiac ventricles that support progressive excitable waves, the conduction speed of which is facilitated by an extensive network of Purkinje fibers.
Materials and methods
This study was approved by our institutional review board for animal research, the Korea University Animal Care and Use Committee (KUIACUC-2011-197).
Suprachiasmatic nucleus organotypic slice cultures and transfection of yellow cameleon (YC2.1) vectors
Organotypic SCN slice cultures were prepared as described previously (Hong et al., 2010). Briefly, 2–3-day-old Sprague-Dawley rats [Crl:CD(SD), Orient Bio., South Korea] that had been bred under a 12 : 12 h light/dark cycle (lights on 07 : 00–19 : 00 h) at room temperature (22–24 °C) were decapitated with a surgical tool. Coronal hypothalamic slices (400 μm in thickness) containing a pair of SCN nuclei were cut using a vibroslicer (NVSL; World Precision Instrument, FL, USA) on a clean bench. The diameter of each SCN nucleus was about 500 μm, and generally we cut out three coronal slices and used the medial part only for the slice cultures. Considering the sample-to-sample size variation and the uncertainty associated with identifying the locations and orientations of SCN nuclei, our SCN slice cultures could contain some portion of the rostral and caudal part. The slices were transferred to a 0.40 μm filter cup (Millicell-CM, Millipore, MA, USA), placed in a standard six-well plate, and cultured with 1 mL of medium that consisted of 50% Eagle’s basal medium, 25% Earle’s balanced salt solution, and 25% heat-inactivated horse serum, supplemented with 5 mg/mL glucose and 1 : 100 Glutamax (Invitrogen, Carlsbad, CA, USA). The cultures were maintained in a CO2 incubator at 36 °C and with 5% CO2. The culture medium was changed every 3–4 days. Only the slice culture that clearly showed the rostrocaudal center of the SCN was used for imaging.
The YC genes (YC2.1) linked to a neuron-specific enolase promoter (pNSE/YC2.1) were transfected to rat SCN slice cultures using a Gene Gun system (Helios™; Bio-Rad Laboratories, CA, USA). Small gold particles (0.6 μm diameter, 5 mg) coated with vectors, carrying pNSE/YC2.1 genes (20 μg), were blasted into the slice cultures at day 7 or 8 in vitro, and the samples were incubated for another 7–10 days prior to use. Finally, a cameleon-expressing SCN slice was cut out from the filter cup and the upper surface of the slice was plated down onto a dish coated with pig collagen (Cellmatrix, Nitta Gelatin, Osaka, Japan). A volume (200 μL) of the culture medium was changed every day.
Cameleon fluorescence resonance energy transfer image acquisition
An organotypic slice culture sample was placed in a small home-made incubator chamber for live-cell imaging. The incubator chamber housing was machined out of an aluminium block, temperature controlled (36 °C, SDM9000; San-up Electronics, South Korea), humidity saturated, and continuously perfused with a mixed gas (5% CO2, 95% air). The culture medium was changed every day. The images were acquired using an electron multiplying charge-coupled device (EMCCD) camera (iXon DV887; Andor, Northern Ireland, UK) mounted on an inverted microscope (IX71; Olympus, Tokyo, Japan) that was equipped with a 10 × objective lens (LUCPlanFL; NA0.30, Olympus), an optical module (Dual-View; Optical-Insights, AZ, USA), and a 440 nm LED lamp (pE; CoolLED, Brimingham, UK). A filter set (OI-05-Ex; Chroma Technology, VT, USA) that included an excitation bandpass filter (436 ± 20 nm) and a dichroic mirror (455DCLP; Chroma Technology) was used for excitation. The Dual-View system separated the mixed fluorescent emission of cyan fluorescent protein (CFP) and yellow fluorescent protein (YFP) with a dichroic mirror (505DCXR, Chroma Technology) and two emission bandpass filters (480 ± 30 and 535 ± 40 nm), and mapped the separated CFP and YFP images to the left and right half of the charge-coupled device (CCD) plate, respectively. An electromagnetic shutter (Vincent Associates, NY, USA) was placed in front of the LED lamp housing. The shutter control, image acquisition and online image processing were coordinated by custom-built software based on Andor software development kit sub-routine packages. The fluorescent images were collected every 10 min, typically for more than 10 days. With our current optical arrangement using 10 × objective lens, only one SCN nucleus can be viewed.
We used tetrodotoxin (TTX) to block the intercellular communication by synaptic coupling and to see how that affected the phase relationships among SCN cells. Stock solution (TTX; 1 mm; Sigma) was made with phosphate-buffered saline and working solutions (either 0.25 or 0.50 μm) were prepared by diluting the stock solution with SCN culture medium.
Image data analysis
From the acquired YFP/CFP image stacks, fluorescence resonance energy transfer (FRET) signals were obtained from the individual neurons expressing cameleon based on the following procedure. First, the overall background fluorescence level, which is the mean gray value of a sample disc area with a 60 μm diameter not containing cameleon expression cells, was subtracted from each image file to remove the long-term change associated with the light source. Second, a cell-tracking algorithm was applied because the cells within the slice cultures moved a small but significant distance (approximately one cell diameter over several days) during the observation time span. For each active clock cell, a region of interest was defined as either a form of disc or ellipsoid that essentially fitted into the cell body. The centroid and center of mass of the defined area were calculated and compared. The centroid position assumed a uniform density (i.e. gray scale), whereas the center of mass reflected the density variation over the space. The (centroid) position of the region of interest was moved to its center of mass for the next image frame. By repeating this process iteratively, the movement of a cell could be tracked effectively. Third, the space-averaged YFP/CFP signals were obtained from the individual cells in their moving coordinates. Finally, the ratio of the space-averaged YFP value to that of the CFP value was considered to be a raw FRET signal measuring [Ca2+]c. This set of processes was carried out using a custom-made plugin (SCN FRET) for ImageJ (NIH).
The raw FRET time series contained high-frequency fluctuations as well as very slow drifting components over several days. As neither of them was a focus of this study, the time series was low-pass filtered by fitting successive data points locally to a third-order polynomial function (sliding window size: 1 day) and a high-pass filter was also applied to detrend the slow component. Then, in order to extract the phase from the processed FRET time series, we applied a Hilbert transformation that extracted the time-varying phase of a nearly sinusoidal signal in a mathematically rigorous way (Yamazaki et al., 1998; Phillips et al., 2010). The peak positions of the FRET time series matched the phase value of 0 quite closely. The Zeitgeber time (ZT) values corresponding to the peak positions were identified and used to construct the plot of ΔZT vs. distance. ZT 0 matches the light-on time of 07 : 00 h.
Averages are expressed as mean ± SDs. For Student’s t-test, a directional test was conducted and significance was accepted when P < 0.05. Estimating the conduction velocity of SCN [Ca2+]c waves (positive side: 179.9 μm/24 h; negative side: 189.7 μm/24 h) was based on the analysis of 23 different SCN nuclei (Fig. 2C). The analyses of two independent sub-groups (12 and 11 nuclei) gave rise to similar values (positive side: 168.3 μm/24 h and 173.2 μm/24 h; negative side: 136.2 μm/24 h and 201.5 μm/24 h, respectively). The result of the TTX treatment shown in Fig. 8 was reproduced more than three times, all with P < 0.05.
Heterogeneous phase distribution of circadian [Ca2+]c oscillators
The number of cameleon-transfected cells was typically < 50 per SCN nucleus, and they were randomly distributed, covering the entire SCN with no particular regional preference (Fig. 1A). The [Ca2+]c level was monitored continuously over 10 days using a FRET imaging system. With no extrinsic manipulation, approximately 63.1 ± 2.5% (mean ± SD) of the SCN neuronal population expressing cameleon (402 out of 637 neurons; number of nuclei, 23; number of slices, 18) exhibited a well-defined circadian [Ca2+]c oscillation (Fig. 1B) with a period of 24.0 ± 1.3 h (population mean ± SD). The remaining 36.9% of the population (‘x’ symbols in Fig. 1C) was silent; in other words, they did not show any oscillatory behavior. These values were very close to those obtained with mice (64%, 23.8 h) (Ikeda et al., 2003).
Figure 1B illustrates the [Ca2+]c time series taken from four different cells, as marked in Fig. 1A. Typically, the [Ca2+]c time traces were not in phase. Indeed, their phases were distributed quite widely over the entire 24 h circadian range, as shown in the phase distribution of Fig. 1C, which maps the ZT corresponding to the peak of the [Ca2+]c time series in color. Quite often, two neighboring cells located in close proximity to each other exhibited very different phases, like the pairs of cells in the two boxed areas in Fig. 1C. However, some cells located far away from each other can share the same phase as in the case of cells ‘c’ and ‘d’. Overall, no large-scale coherent pattern was easily discernible, and this was a general trend observed in all 18 SCN slice culture samples that were monitored. Knowing that previous studies monitoring the circadian rhythmic activity of Per1 (Quintero et al., 2003; Yamaguchi et al., 2003) and Per2 (Yan et al., 2007; Evans et al., 2011; Foley et al., 2011) gene expression in SCN slices showed some spatially organized coherent waves, the seemingly random, spatially incoherent [Ca2+]c oscillations were quite unexpected.
Local [Ca2+]c phase wave conduction in the suprachiasmatic nucleus
As a first step towards understanding the complex phase distribution, Fig. 2A plots ZT differences (ΔZT) vs. spatial distances for all possible combinations of cell pairs in an SCN slice culture. All [Ca2+]c active cells were identified within a given slice, two of them were chosen, and the ZT difference and physical distance between the two were measured. The same process was repeated for all different combinations of pairs and the data were plotted. As expected, the data points were scattered over the entire dynamic range of ΔZT (−12–12 h), and no functional relationship seemed to exist among them. However, two features were notable: first, there were relatively fewer data points in the regime where the distance was 25–100 μm and |ΔZT| was 8–12 h; and second, the data points were populated in the range of distance of 100–300 μm and |ΔZT| < 6 h.
The experiment was repeated with 23 different SCN nuclei, the same analysis was carried out, and then all the data points were plotted on the single graph in Fig. 2B, searching for some general features to emerge. The overall characteristics of Fig. 2B are essentially similar to those of Fig. 2A. Again, the data points cover most of the plane. However, the density map (Fig. 2C) of Fig. 2B clearly reveals (i) a wedge-shaped steep bound (connected black dots on red contour) at the left-hand side and (ii) a region of high density (‘hot spot’) in the middle. In the regime of physical distances of < 100 μm, the density map fell off rather quickly across the wedge-shaped bound. Both the positive and negative sides of the contour fit very well on a straight line [white solid lines; the slopes were 179.9 μm/24 h (R2 = 0.95) and 189.7 μm/24 h (R2 = 0.91) for the positive and negative sides, respectively]. The very existence of this wedge-shaped contour was indirect evidence of the existence of waves propagating locally. Indeed, the two slopes of the wedge were nothing but the inverse of the velocity of phase waves in the direction normal to the wave front. A similar wedge-shaped bound existed, even if we normalized the count of each bin with respect to the total count of each band of distance (Fig. S1). The new slope estimated with the different normalization was 166.9 μm/24 h (R2 = 0.9) and 167.6 μm/24 h (R2 = 0.3), which was very close to the earlier values obtained based on the density map of Fig. 2C.
To elucidate this point more clearly, this study considered a toy model of a sinusoidal planar wave passing through a disc area, approximately the size of an SCN nucleus, along the vertical direction, as illustrated in Fig. 3A (top). Simply, it was an analytical solution of a linear wave equation in two-dimensional space. We assumed its period to be 24 h and velocity to be 500 μm/24 h. The pseudo-color represented the phase and the small circles distributed randomly represented the individual cells that were sampled for the analysis. Similar to Fig. 2A, aΔZT vs. distance graph [Fig. 3A (middle)] was plotted based on all possible combinations of the model cell pairs and Fig. 3A (bottom) shows the corresponding density map of the data points. Figure 3A (middle and bottom) clearly shows a wedge-shaped bound on the left-hand side. Basically, this linear bound was formed by pairs of cells located more or less along the same vertical line (i.e. in the normal direction of wave propagation, e.g. OA, OG, OH, and OD). Therefore, the |1/slope| was nothing but the normal conduction speed of the assumed planar wave. The wedge-shaped linear bound also existed for a sinusoidal circular wave, as shown in Fig. 4, and was a universal property of any phase waves having a constant conduction speed.
Synchronization of circadian [Ca2+]c oscillations mediated by long-range direct connections
There were a few puzzles regarding the result shown in Fig. 2. First, the two slopes of the bounding lines were too steep, or alternatively, the normal wave conduction speed (about 7.2 μm/h) was rather small for the wave to cover the entire SCN nucleus, which was approximately 500 μm in diameter, within 1 day. Second, the density of data points in the central area of Fig. 2C was quite high, in contrast with that of the toy model in Fig. 3A (bottom). The ‘hot spot’ in Fig. 2C meant that there were many pairs of cells that were separated far away from each other but with little phase difference. Waves with a simple geometrical shape (i.e. planar, circular, elliptical, parabolic) cannot result in such a hot spot [Fig. 3A (bottom) and Fig. 4A (bottom)]. Third, simple wave forms cannot explain the fact that there were many pairs of closely neighboring cells with a large phase difference (points in the yellow-shaded areas in Fig. 2B).
A close-up examination of the images of cameleon-transfected SCN slice cultures provided a clear answer to these puzzles. Figure 5A shows a typical YFP image of a whole SCN nucleus. The cameleon transfection rendered some networks of long-range neurite connections visible. The cells connected physically by neurites had their [Ca2+]c oscillations in phase, e.g. the three cells marked by circles in Fig. 5A. Even the neurites (marked by small rectangles) connecting them also oscillated in phase, as illustrated in Fig. 5B. This is clear evidence showing that oscillatory clock cells, which are located far apart but connected directly to each other, can have their [Ca2+]c oscillations synchronized. The intercellular networks appeared random, but were quite extensive throughout the entire SCN nucleus and far-reaching.
To emulate the experimental observation, the toy model in Fig. 3A was reconsidered assuming that 25% of the whole cell population formed three different synchronizing sub-networks, as shown in Fig. 3B. As a consequence, some data points are now placed in the yellow-shaded area and the point density increases in the central area near ΔZT = 0. At the same time, the wedge-shaped bound becomes noisier. If the percentage is increased further to 50% (with five sub-networks) as in Fig. 3C, the hot spot becomes more pronounced in the central region and the wedge bound gets more smeared. The corresponding density map [Fig. 3C (bottom)] closely resembles the experimental results shown in Fig. 2C. The hot spot in Fig. 2C is peaked sharply in the separation distance of 100–200 μm, and it may be the typical extent of the direct network connections.
Effect of blocking action potentials on the circadian [Ca2+]c oscillations of suprachiasmatic nucleus clock cells
To check the importance of cell-to-cell electrical coupling for the maintenance of SCN circadian oscillations, this study examined the effect of TTX on the circadian [Ca2+]c oscillations of SCN clock cells, particularly with regard to their amplitudes and periods. For approximately 54% of the [Ca2+]c-active clock cell population, 3 days of 0.50 μm TTX treatment had no significant effect except for some increase in the amplitude of the [Ca2+]c oscillation (Fig. 6, top). However, for approximately 46% of the active population, the circadian [Ca2+]c oscillation disappeared completely in the presence of TTX. The oscillation either reappeared after being washed (29%; Fig. 6, bottom) or was permanently extinct (17%). Surprisingly, some of the originally silent sub-population of clock cells newly gained a circadian [Ca2+]c oscillation in the presence of TTX. This oscillation often persisted, even after the TTX was washed away (53%; Fig. 6, middle). Even when the sample was treated with 0.25 μm TTX, the heterogeneous responses were similar. Table 1 lists the results.
Table 1. Heterogeneous responses to 0.25 and 0.5 μm TTX
0.25 μm TTX (7 days)
0.5 μm TTX (3 days)
The ‘gain’ denotes the cases in which an initially silent cell newly acquires a circadian oscillation, whereas the ‘lost’ stands for the case in which an initially oscillatory cell loses its rhythm during the TTX treatment. The term ‘extinct’ refers to the cases in which the initial rhythm is removed by the TTX treatment and is never recovered, even after the wash. The percentage values are based on data collected from four different SCN slices for each different concentration.
The change in the period was another important consequence of the TTX treatment. Figure 7A (left frame) is a YFP image of a cameleon-transfected SCN nucleus, which again reveals clock cells with many neurite connections. Figure 7B illustrates two different time traces of [Ca2+]c oscillation taken at two different cells (‘a’ and ‘b’ as marked in Fig. 7A), which appear to be physically connected to each other. During 3 days of the TTX treatment, the mean period of cell ‘a’ was lengthened from 25.7 to 26.8 h, whereas that of cell ‘b’ was shortened rather dramatically from 23.9 to 19.9 h (Fig. 7B). The removal of electrical coupling appeared to bring the cells back to their independent oscillatory states with a different period. As a consequence of the period change during the TTX treatment, the clock cells acquired a large phase shift.
The significance of the phase shift is well reflected in the return maps of Fig. 7C and D. A return map is a useful tool for revealing any complex-periodic nature of a discrete sequence of data, but here it is used to illustrate the significance of the phase shift caused by the application of TTX, compared with the degree of the intrinsic day-to-day fluctuation during the control. In Fig. 7C, the red points are ZTi vs. ZTi+1 (i.e. first return map of ZT) of cell ‘1’ before the TTX treatment, where ZTi represents the ZT value of the ith peak of the [Ca2+]c oscillation. In comparison, the blue point plots (<ZTbefore>, <ZTafter>). During the control, the [Ca2+]c oscillations peaked at near ZT ∼ 18 ± 2 h but the peak had moved to ZT ∼ 2 h after the treatment, which is significant. Figure 7D shows a similar return map plotting the same for 22 different cells together. The majority of red points fall within the shaded ± 2 h diagonal band, whereas the blue points are located significantly off the diagonal band. This is clear proof that the phases of circadian [Ca2+]c oscillations were altered significantly by the TTX treatment. In addition, the red points were well spread along the entire diagonal band, again confirming that the SCN clock cells were not phase-synchronized in general. The decrease in the mean period, caused by the TTX treatment, was approximately 1 h (Fig. 8; 24.0 ± 1.5–22.9 ± 1.8 h, t85 = 4.87, P < 0.001), but the amount of change varied from one cell to another, as shown in Fig. 7B. The various responses of the cells following a TTX treatment are well reflected in the sequence of ZT map images given in Fig. 7A.
Phase inhomogeneity and wave nature of suprachiasmatic nucleus circadian oscillations
Revealing the exact shape and direction of propagation of the SCN [Ca2+]c waves based on pNSE/YC2.1 was impossible for some intrinsic and technical reasons. First of all, the number of clock cells expressing pNSE/YC2.1 and actually showing [Ca2+]c oscillations was rather small. More importantly, the rat SCN slices contained non-trivial phase-synchronizing network structures that ‘masked’ the [Ca2+]c waves.
Nevertheless, the existence of [Ca2+]c phase waves could be confirmed and their conduction velocity in the normal direction of propagation could be estimated accurately by repeating the same experiment many times and considering the graph of ΔZT vs. distance with the collection of data. According to the toy models described in Figs 3 and 4, the existence of a wedge-shaped bound in the plot of ΔZT vs. distance is good supporting evidence for the existence of locally conducting waves. Figure 2B clearly reveals such a wedge-shaped bound.
Estimated value of the conduction velocity of suprachiasmatic nucleus [Ca2+]c waves
The estimation of the normal conduction speed from Fig. 2C was approximately 7.2 μm/h, which was too small for a single planar wave to cover a distance of approximately 500 μm (about the diameter of an SCN nucleus) within a circadian period of 24 h. Indeed, the measured value was approximately five to eight times smaller than those that we estimated from Yamaguchi et al. (2003) and Welsh et al. (2010). The reason for the large discrepancy is unclear, but there are a few possibilities. First, the shape of the propagating wave front would matter because the actual distance that the wave front needs to travel during a circadian cycle will depend on the shape. For example, a circular wave front moving towards the core from the outer shell needs to cover approximately half of the diameter of the SCN during a 24 h cycle. The topology of [Ca2+]c waves in our SCN slice cultures could possibly have many small-scale features so that the waves needed to travel only a short distance (i.e. a small fraction of the SCN nucleus) during a circadian cycle. Second, our estimate of the normal conduction velocity from Fig. 2 excluded the effect of long-range connections, which can effectively speed up the propagation. However, those that we estimated from the previous reports (Yamaguchi et al., 2003; Welsh et al., 2010) were based on a bioluminescence signal that is diffusive. In other words, the effect of direct long-range connections could have already been reflected in the acquired signals and in the estimated values of the conduction velocity. In cardiac physiology, it is well known that Purkinje fibers, which wrap around the ventricles, expedite the AP wave conduction throughout ventricular tissues. Likewise, it is possible that the long-range cell-to-cell networks in the SCN play a similar role.
Network structures revealed by cameleon transfection
The low success rate of cameleon transfection only guarantees the time series of [Ca2+]c specifically from individual cells. More importantly, some intricate network structures connecting spatially distributed cells are observed, as shown in Figs 5 and 7. We often find that clock cells connected by neurites share a very similar circadian phase, and their circadian [Ca2+]c signals are disrupted when the connecting neurites are physically cut off by laser ablation. The existence of a hot spot (−4 h < ΔZT < 4 h; 100 μm < distance < 200 μm) in Fig. 2C is also good supporting evidence for the existence of synchronizing networks. Subsequently, this study proposes that there are sub-networks, and that the cells belonging to two different sub-networks generally show a different circadian phase, as shown in Fig. 3. The hypothesis easily explains the fact that two immediately neighboring cells, as marked in Fig. 1, can have rather different circadian phases. At this point, it is noteworthy that earlier studies, monitoring SCN clock gene expression, also found many phase-synchronized cells that were scattered widely in the medial part of the SCN (Yamaguchi et al., 2003; Evans et al., 2011) in the presence of an overall macroscopic wave.
It is generally accepted that the SCN nucleus is divided into two compartments, a region referred to as the ‘shell’, which contains neurons expressing arginine vasopressin, and another region referred to as the ‘core’, which contains the neurons immunoreactive for vasoactive intestinal polypeptide and gastrin-releasing peptide (Abrahamson & Moore, 2001). The shell region is delineated by rhythmic Per mRNA and arginine vasopressin expression (Jin et al., 1999), whereas the core region, taking afferent terminals from the retina, either lacks detectable rhythmicity or shows only a very small change in clock gene expression (Antle & Silver, 2005). However, the SCN neurons expressing cameleon are spread widely over the entire SCN nucleus (Fig. S2), and the morphology of the cell-to-cell networks appears random and dense without any region-specific feature.
So, there is a puzzle: how are the SCN core cells able to show oscillations of Ca2+ but not of clock gene expression? Currently, we do not have a clear answer to this question but propose two different possibilities. First, we cannot exclude the possibility that some of the ‘core region’ in our slice cultures could have originated, at least partially, from the ‘shell (rostral or caudal) regions,’ as the slice thickness of 400 μm is large enough to contain those parts. An alternative possibility is that circadian [Ca2+]c oscillations are universal, whereas those of (some) clock gene expressions are region-specific. For that matter, we note that Sugiyama et al. (2004) have shown that, in the presence of antisense oligonucleotides, Per2 gene expression is suppressed, whereas the oscillations in the firing rate of APs and [Ca2+]c are not affected at all.
Effects of tetrodotoxin on the circadian [Ca2+]c oscillations of the suprachiasmatic nucleus
In an earlier study, Ikeda et al. (2003) reported that, even in the presence of TTX, the circadian [Ca2+]c oscillations of mouse SCN neurons were persistent, even though the AP firing activity of the SCN was blocked completely. In other words, the presence of APs was not mandatory for the generation of SCN [Ca2+]c circadian oscillations. However, in an independent study, Yamaguchi et al. (2003) reported that cell-to-cell phase coherence was disrupted by the application of TTX in an organotypic slice culture of the mouse SCN expressing period1-luciferase. They also showed that the original phase coherence was re-established later when TTX was removed. Therefore, APs appear to be an essential component for the synchronization of distributed SCN clock cells.
The effect of TTX on the [Ca2+]c dynamics of the rat SCN was investigated in this report. For most cells (> 50%), the circadian [Ca2+]c oscillations were persistent during the TTX treatment, just as reported by Ikeda et al. (2003). Interestingly, some neurons lost their [Ca2+]c oscillations, whereas others newly obtained circadian oscillations. This heterogeneous response of clock cells to the application of TTX is consistent with a recent report by Webb et al. (2009). SCN nuclei are comprised of several different types of neurons excreting various neurotransmitters (such as vasoactive intestinal polypeptide, gastrin-releasing peptide and γ-aminobutyric acid), and thus influencing each other. Moreover, some of them function as either excitatory or inhibitory interneurons, and thus any significant change in their neural activity can cause subsequent changes in their neighboring post-synaptic cells. Therefore, depending on how a particular SCN neuron is connected to other cells in the network, its response to the TTX application can be diverse.
The more interesting consequence of the TTX treatment was that, for many clock cells, the periods of oscillation changed significantly. The degree of period detuning was not uniform across the clock cell population, as shown in Fig. 7. Therefore, the relative phases among the different clock cells were significantly changed after the TTX treatment. The shape of the period distribution did not change very much during and after the treatment, but the mean period was reduced by approximately 1 h during the treatment (Fig. 8). This reduction was also observed by Webb et al. (2009) who used PERIOD2::Luciferase to monitor the circadian rhythms of mouse SCNs.
Our results show that the SCN calcium dynamics is coordinated by phase-synchronizing networks of long-range neurites as well as by diffusively propagating phase waves. The networks appear quite extensive and far-reaching, and have a role of expediting the conduction of circadian phase information throughout the SCN. Many questions need to be addressed in the future. The functional significance of both the phase waves and synchronizing networks is unknown. The wave dynamics and network structure may provide robustness to the SCN by generating a coherent circadian rhythm. Perhaps, the system is designed intricately in such a way that various region-specific output pathways channel a different phase (time) cue to a range of target areas.
The authors wish to thank Dr Masayuki Ikeda of Toyama University for helpful discussions and providing YC (pNSE/YC2.1). This study was supported by the Acceleration Project (R17-2007-017-01000-0) of the Korea Ministry of Science and Technology.