Report

Daily Electrical Silencing in the Mammalian Circadian Clock

See allHide authors and affiliations

Science  09 Oct 2009:
Vol. 326, Issue 5950, pp. 281-284
DOI: 10.1126/science.1169657

Abstract

Neurons in the brain’s suprachiasmatic nuclei (SCNs), which control the timing of daily rhythms, are thought to encode time of day by changing their firing frequency, with high rates during the day and lower rates at night. Some SCN neurons express a key clock gene, period 1 (per1). We found that during the day, neurons containing per1 sustain an electrically excited state and do not fire, whereas non-per1 neurons show the previously reported daily variation in firing activity. Using a combined experimental and theoretical approach, we explain how ionic currents lead to the unusual electrophysiological behaviors of per1 cells, which unlike other mammalian brain cells can survive and function at depolarized states.

In mammals, behavior and physiology are regulated on a daily basis by the brain’s master circadian (~24-hour) clock in the suprachiasmatic nuclei (SCNs). The period 1 (per1) gene is a key component of the molecular mechanism of this clock (1); its expression in the SCN peaks during the day and is low at night, and can be used as a marker of clock-containing SCN neurons and their circadian phase (2, 3). SCN neurons are also thought to express time of day by changing their firing frequency, with high rates during the day and lower rates at night (46). A fundamental question in circadian biology is how the intracellular molecular circadian clock regulates the electrophysiology of SCN neurons. A Hodgkin-Huxley–type model of SCN neurons shows that circadian changes in ionic conductances can account for the circadian variation in firing rate (7). This model also predicts that the molecular clock can drive SCN neurons to exhibit an unusual depolarized rest state not previously reported experimentally. Because not all SCN neurons appear to contain the molecular clock machinery (8), a further prediction of this model is that SCN neurons that express the key clock genes (such as per1) will have different electrical properties than those that do not express these genes.

To measure time-dependent changes in the membrane properties and excitability of SCN neurons in vitro, and to investigate potential differences among these neurons, we made targeted whole-cell recordings from SCN of mice expressing an enhanced green fluorescent protein (EGFP) reporter of per1. We sampled from 309 “per1” neurons showing detectable EGFP and 109 neurons in which EGFP could not be detected (“non-per1”). We found a marked difference in the electrophysiological behavior during the day of these two populations.

Sampling from per1 and non-per1 neurons throughout the SCN across the projected day-night cycle revealed large circadian variation in resting membrane potential (RMP), membrane input resistance (Rinput), and overall electrophysiological behaviors (Figs. 1 and 2). In the morning [Zeitgeber Time (ZT) 2 to 4.25, where ZT0 represents lights-on and ZT12 represents lights-off], per1 neurons were at moderate RMP (Vm = –44 ± 0.6 mV; n = 47 neurons) and generated action potentials (APs) (4.3 ± 0.1 Hz) (Figs. 1 and 2). In the afternoon (ZT5.25 to 10.75), per1 neurons were spontaneously depolarized and either showed low-amplitude 2- to 7-Hz oscillations in membrane potential (MP) (mean Vm = –35 ± 0.7 mV; n = 50 neurons) or were completely silent (Vm = –27 ± 0.4 mV; n = 50 neurons) (Figs. 1 and 2). Around dusk (ZT10.75 to 12.75), these cells were again at moderate RMP (–46 ± 1 mV; n = 45 neurons) and generated APs, but at lower frequency than in the morning (1 ± 0.1 Hz) (Figs. 1 and 2). This progression from moderate RMP in the morning to depolarized states (–25 to –34 mV) in the afternoon and back to moderate RMP at dusk was specific to per1 neurons; non-per1 cells remained at moderate RMP (Vm = –46 ± 0.9 mV; n = 48 neurons) and continued to generate APs until dusk (Figs. 1 and 2). In contrast, during the early night (ZT13 to 15.25) both per1 and non-per1 neurons became robustly hyperpolarized and showed no APs (per1, –68 ± 0.5 mV; n = 40 neurons; non-per1, –66 ± 1.1 mV; n = 24 neurons), but later in the night (ZT16.75 to 23) both per1 and non-per1 neurons were at or close to moderate RMP and generated APs (Figs. 1 and 2).

Fig. 1

ZT-dependent changes in RMP of per1 (green) and non-per1 (red) neurons. (A and A1) Scatter plot of RMP versus ZT for per1 (green dots) and non-per1 (red dots) cells. (B and B1) Morning per1 and non-per1 cells at moderate RMP. (C) Late-morning per1 cell in transitional phase displaying bistability. (D) Afternoon to late-afternoon per1 cell displaying low-amplitude oscillations in MP [group A (GA)] or silenced [group B (GB)] as determined by their Rinput and electrophysiological behaviors (Fig. 2). (C1) Non-per1 neurons do not display bistability or depolarized RMP, but remained at moderate RMP, generating APs. (E and D1) Dusk per1 and non-per1 cells at moderate RMP around the time of lights-off. (F and E1) Early night per1 and non-per1 cells hyperpolarized and silenced, receiving excitatory postsynaptic potentials (EPSPs). (G and F1) Late-night per1 and non-per1 neurons at moderate RMP.

Fig. 2

Changes in RMP and associated changes in Rinput and firing frequency of per1 and non-per1 cells across a day-night cycle. Per1 cells in GA and GB at ZT5.25 to 10.75 were determined by their electrophysiological behaviors (Fig. 1) and Rinput (P < 0.01) (Fig. 2B). GA cells displayed low-amplitude 2- to 7-Hz oscillations in MP, whereas GB cells were silent. (A) Changes in RMP of per1 cells and (B) associated changes in their Rinput. (C) Scatter plot of firing frequency versus ZT for per1 cells, showing per1 cells generating APs in the morning, at dusk, and during late night; however, per1 cells were completely silent during the afternoon and early night. (D) Changes in RMP of non-per1 cells, and (E) associated changes in their Rinput. (F) Scatter plots of firing frequency versus ZT for non-per1 cells, showing the previously reported daily variation in firing frequency of these cells. *P < 0.05; **P < 0.001; ***P < 1 × 10−5. Numerical data represent ±SEM

In many studies, unusually depolarized neurons are considered unhealthy, so we were careful to establish that such per1 neurons were physiologically sound. Indeed, depolarized states were only observed in cells recorded during the afternoon and never in cells tested in morning, dusk, or night. Further, silent per1 cells characteristic of the state attained in the afternoon had extremely high Rinput and, by application of steady-state hyperpolarizing currents, could be induced to fire regular APs like those of cells measured in the morning (Rinput at –25 mV = 2.9 ± 0.1 gigohm, n = 38 neurons) (fig. S1). Alternatively, steady-state depolarizing currents reversibly silenced regular AP-firing per1 cells in the morning, rendering them indistinguishable from neurons measured in the afternoon portion of the light cycle (n = 26 neurons) (fig. S1). Depolarization of non-per1 neurons in a similar manner compromises their electrophysiological properties. Thus, per1 neurons appear to be healthy and functioning within normal whole-cell electrophysiological parameters.

We used mathematical modeling to determine whether these observed electrical behaviors could be explained by existing data on the ionic currents measured within SCN neurons. To simulate per1 neurons at various circadian phases, we first adapted a model of unidentified SCN neurons (7) to account for the enhanced electrical excitability of per1 neurons (fig. S2). We then incorporated into the model circadian variations in K+ and Ca2+ conductances (fig. S3), and intracellular Ca2+ concentration (fig. S4), from published experimental observations (911). Although a circadian variation in only one ionic current could cause transitions from a hyperpolarized steady state to repetitive firing, and also from repetitive firing to a depolarized steady state (figs. S5 and S6), circadian rhythms in both K+ and Ca2+ currents were required to faithfully reproduce the fine details of these transitions (Fig. 1, B to G, and Fig. 3).

Fig. 3

Mathematical simulations predict the electrophysiological behavior of SCN clock neurons observed throughout a day-night cycle. These simulations use a revised version of the Sim-Forger model (7) specific for per1 neurons and incorporate randomly generated postsynaptic potentials (fig. S2). (A to F) Panels are similar to experimental data in Fig. 1, B to G, and show that known rhythms in ionic currents (fig. S3) can explain the experimentally observed behaviors.

Mathematical analysis revealed that as the day progresses, transitions in the behavior of per1 neurons occur when a quiescent state gains or loses stability through Hopf bifurcations (fig. S3). This common mathematical structure has been found in many other neural systems and corresponds to Type II excitability (spiking emerging with nonzero frequency) in Hodgkin’s original classification system (12). The changes in stability of the hyperpolarized rest state we observed at night in per1 neurons are similar in character to those seen in other neurons (13). However, per1 neurons are different in that a second Hopf bifurcation occurs during the day, when an unusually depolarized rest state becomes stable or unstable. The physiological importance of these bifurcations is that minor, molecular clock–driven changes in certain ionic conductances can have a major effect on the bioelectrical output of the neuron. Near bifurcation points, neurons can exhibit the types of behaviors seen in per1 neurons, such as low-amplitude oscillations in MP or noise-induced transitions between oscillatory and quiescent states (12, 14). These behaviors do not appear to depend on the initial state of the neuron (fig. S3) and can be achieved with a variety of parameter choices (figs. S4, S5, S6, and S7).

In both per1 and non-per1 cells, the cyclical changes in RMP were closely associated and in phase with alterations in Rinput (Fig. 2). For per1 cells, the amplitude of the circadian variation in RMP and Rinput were 42 mV [day, –27 ± 0.4 mV; night, –68 ± 0.5 mV; P < 1 × 10−5; all P values were determined by means of analysis of variance (ANOVA) and post-hoc Bonferroni test unless otherwise stated] and 2.1 gigohm (day, 2.9 ± 0.1 gigohm; night, 0.8 ± 0.03 gigohm; P < 1 × 10−5), respectively. The measured membrane time constant was also significantly larger during the day than at night (afternoon, 48.3 ms; early night, 23.3 ms; P < 0.001; n = 20 neurons). These differences between day and night per1 cells were not associated with any change in cell capacitance (day, 9.4 ± 0.04 pF; night, 9.3 ± 0.05 pF; n = 10 neurons), and thus are not a result of changes in cell size. In non-per1 cells, the differences in RMP and Rinput between the day and night were 20 mV (day, –46 ± 0.9 mV; night, –66 ± 1.1 mV; P < 0.001) and 0.9 gigohm (day, 1.6 ± 0.1 gigohm; night, 0.7 ± 0.1 gigohm; P < 0.001), respectively.

Our measurements and modeling indicate that a clock-controlled reduction in K+ channel conductance causes depolarization in per1 cells in the afternoon, and an increase in K+ channel conductance at night causes the hyperpolarization. We also reasoned that L-type Ca2+ channels may contribute to determining the RMP of these cells (9, 15, 16). We therefore bath-applied tetraethylammonium (TEA; a broad spectrum K+ channel blocker) or nimodipine (an L-type Ca2+ channel blocker) during the night or the morning.

TEA (30 mM) caused a significant depolarization in per1 cells recorded in the morning (from –43 ± 1.2 mV to –28 ± 0.9 mV; P < 0.001, n = 8 neurons, ANOVA and post hoc test), and significantly increased their Rinput (1.5 ± 0.03 to 2.9 ± 0.2; P < 0.001, ANOVA and post hoc test) (fig. S8). Nimodipine (2 μM) depolarized per1 cells recorded in the morning so that their cellular behaviors and RMP were similar to those per1 cells tested in the afternoon (fig. S8). Nimodipine also caused cells whose RMP was near –33 mV to depolarize to –25 mV (fig. S8). Antagonizing L-type Ca2+ channels in SCN neurons inhibits Ca2+-dependent K+ channels (Kca) (15), and in many neuronal systems inhibition of Kca channels causes depolarization (17). So, we tested whether the large- (BKCa) and/or the small-conductance (SKCa) Kca are involved in setting the RMP of per1 neurons. Iberiotoxin (IbTX; BKCa-selective receptor antagonist) (100 nM) causes small depolarization (2 to 3 mV) in early- and late-morning per1 cells. However, concomitant application of IbTX and apamin (SKCa selective receptor antagonist) (100 to 200 nM) during early and late morning caused significant and irreversible depolarization (13 to 15 mV) in per1 cells, and significantly increased their Rinput (1.6 ± 0.03 to 3 ± 0.2; P < 0.001) to values that were indistinguishable from those of per1 cells measured in the afternoon (Figs. 2 and 4). During the apamin- and IbTX-induced depolarization, morning per1 cells displayed behaviors and membrane properties that were typical of afternoon per1 cells (figs. S9 and S10). Thus, spontaneous depolarization and associated change in Rinput observed in per1 neurons between dawn and dusk result from a clock-controlled reduction in BKCa and SKCa conductances. IbTX and apamin had little effect on the RMP of non-per1 cells during the day (fig. S11). Although TEA (n = 6 neurons) dose-dependently depolarized cells in the early-night state and increased their Rinput to values similar to those of cells measured in the morning (Fig. 2 and figs. S8 and S12), these night effects were not seen with nimodipine, IbTX, or apamin (n = 7 neurons) (fig. S12). This suggests that BKCa, SKCa, and L-type Ca2+ channels minimally influence the RMP of per1 neurons during the night.

Fig. 4

Role of large- (BKca) and small-conductance (SKca) calcium-activated K+ channels in determining the RMP of per1 neurons (n = 10 neurons). Inhibiting BKca channels alone by IbTX (100 nM) causes small depolarization (2 to 3 mV) of (A) early- and (B) late-morning per1 neurons. (C) However, bath-application of apamin (100 to 200 nM), a specific and irreversible antagonist for SKca channels, in the presence of IbTX, depolarized (13 to 15 mV) early- and late-morning per1 cells and significantly increased their Rinput. Although depolarized early-morning per1 neurons immediately displayed low-amplitude oscillations in MP [(A), inset 1], late morning per1 cells first show low-amplitude oscillations in MP during depolarization [(B), inset 2] and then became silent [(B), inset 3]. These behaviors were typical of afternoon per1 cells (Fig. 1 and figs. S1 and S8). **P < 0.001. Numerical data represent ±SEM.

Blocking of AP-dependent presynaptic potentials with TTX (200 nM to 1 μM; n = 8 neurons) or inhibition of postsynaptic receptors for glutamate and γ-aminobutyric acid, alone or in combination, had no effect on the RMP of per1 cells at any time (fig. S13). This suggests that the observed changes in RMP of per1 cells were cell autonomous and do not rely on the network properties of the SCN (18).

Both our experimental and modeling results establish that like invertebrate circadian clock neurons (1921), per1-containing SCN cells exhibit clock-controlled, high-amplitude circadian oscillations in key electrophysiological properties. A loss of BKCa and SKCa channel conductances underpin the depolarization of RMP seen in per1 neurons during the afternoon, whereas multiple K+ channels contribute to hyperpolarization of these cells during the night. In contrast, SCN neurons lacking detectable EGFP expression showed much smaller day-night differences in electrophysiological parameters but do exhibit the characteristic circadian differences in AP frequency (46). The level of per1-driven EGFP peaks about two hours before the day-night transition (4), with its morning and dusk phases showing strong association with AP generation. However, in the middle of the afternoon, neurons containing per1 are in an excited state but do not fire APs. This temporal dissociation between clock gene expression and AP generation, at this time of day, demonstrates that the relationships between per1 expression and neuronal membrane properties are much more complex than first supposed (2224). Our study also indicates that the neurophysiology of per1 neurons may not obey conventional electrophysiological principles.

Supporting Online Material

www.sciencemag.org/cgi/content/full/326/5950/281/DC1

Materials and Methods

Figs. S1 to S18

References

References and Notes

  1. We thank D. McMahon, Vanderbilt University, for providing us with the per1::2dGFP mice and R. Lucas and R. Baines, University of Manchester, C. Diniz Behn, University of Michigan, and D. Paydarfar, University of Massachusetts Medical School, for their critical reading of the manuscript. This work was funded by a project grant from the Biotechnology and Biological Sciences Research Council awarded to H.D.P. (BB/E00511X). M.D.C.B. is a Research Associate. C.O.D. is a NSF Graduate Research Fellow. D.B.F. is an Air Force Office of Scientific Research Young Investigator (FA 9550-08-01-0076).
View Abstract

Navigate This Article