Sleep spindles are generated in the absence of T-type calcium channel-mediated low-threshold burst firing of thalamocortical neurons.

Transcription

1 Sleep spindles are generated in the absence of T-type calcium channel-mediated low-threshold burst firing of thalamocortical neurons Jungryun Lee a,b, Kiyeong Song c, Kyoobin Lee b,1, Joohyeon Hong c, Hyojung Lee c, Sangmi Chae b, Eunji Cheong c,2, and Hee-Sup Shin a,2 a Center for Cognition and Sociality, Institute for Basic Science, Daejeon , Korea; b Center for Neural Science, Korea Institute of Science and Technology, Seoul , Korea; and c Department of Biotechnology, Translational Research Center for Protein Function Control, College of Life Science and Biotechnology, Yonsei University, Seoul , Korea Contributed by Hee-Sup Shin, November 6, 2013 (sent for review August 25, 2013) T-type Ca 2+ channels in thalamocortical (TC) neurons have long been considered to play a critical role in the genesis of sleep spindles, one of several TC oscillations. A classical model for TC oscillations states that reciprocal interaction between synaptically connected GABAergic thalamic reticular nucleus (TRN) neurons and glutamatergic TC neurons generates oscillations through T- type channel-mediated low-threshold burst firings of neurons in the two nuclei. These oscillations are then transmitted from TC neurons to cortical neurons, contributing to the network of TC oscillations. Unexpectedly, however, we found that both WT and KO mice for Ca V 3.1, the gene for T-type Ca 2+ channels in TC neurons, exhibit typical waxing-and-waning sleep spindle waves at a similar occurrence and with similar amplitudes and episode durations during non-rapid eye movement sleep. Single-unit recording in parallel with electroencephalography in vivo confirmed a complete lack of burst firing in the mutant TC neurons. Of particular interest, the tonic spike frequency in TC neurons was significantly increased during spindle periods compared with nonspindle periods in both genotypes. In contrast, no significant change in burst firing frequency between spindle and nonspindle periods was noted in the WT mice. Furthermore, spindle-like oscillations were readily generated within intrathalamic circuits composed solely of TRN and TC neurons in vitro in both the KO mutant and WT mice. Our findings call into question the essential role of low-threshold burst firings in TC neurons and suggest that tonic firing is important for the generation and propagation of spindle oscillations in the TC circuit. Sleep spindles are one type of several rhythmic brain waves detected by electroencephalography (EEG) during normal non-rapid eye movement (NREM) sleep. A spindle consists of characteristic waxing-and-waning field potentials grouped into 7- to 14-Hz oscillations that last for 1 3 s and recur once every 5 10 s in the thalamus and the cortex (1 3). Spindles are also visible under anesthesia, particularly with barbiturates but also with ketamine-xylazine combinations (4, 5). These oscillations are generated in the thalamus as a result of synaptic interactions between inhibitory [i.e., thalamic reticular nucleus (TRN)] neurons and excitatory thalamocortical (TC) neurons, and are propagated to the cortex. Corticothalamic projections back to the thalamus complete the cortico-thalamo-cortical loop. In vivo data suggest that TRN neurons are spindle pacemakers, because spindles can be generated in deafferented TRN neurons (6) but disappear in TC regions after disconnection from TRN neurons (7). However, in vitro data suggest that an intact TC-TRN network is a necessity, because spindles are abolished after disconnection of TC and TRN neurons (8). Two distinct firing patterns, tonic and burst, are displayed by both TRN and TC neurons. Burst firing is mediated by lowthreshold T-type Ca 2+ channels (9). Of the three subtypes of T-type channels, Ca V 3.1 is expressed exclusively in TC regions, whereas Ca V 3.2 and Ca V 3.3 are abundant in TRN regions (10). T-type channels in TC neurons have been proposed to be a critical component in the generation of physiological and pathological TC oscillations, such as sleep rhythms (1, 11, 12) and the spikewave discharges (SWDs) of absence seizures (11, 13, 14). One generally accepted hypothesis proposes that inhibitory inputs from TRN neurons de-inactivate T-type Ca 2+ channels in TC neurons, leading to induction of burst firings in TC neurons, which in return excite reciprocally connected TRN neurons. These thalamic oscillations are then transmitted from TC neurons to cortical neurons. This model proposes that T-type channel-mediated burst firing in TC neurons underlies sleep spindles and other sleep rhythms within TC circuits (1, 11). There have long been doubts regarding the extent to which TC T-type Ca 2+ channels contribute to the heterogeneity of TC oscillations during NREM sleep, which consists of multiple EEG components including slow waves (<1 Hz), delta waves (1 4 Hz), and sleep spindles (7 14 Hz). T-type channels have received particular attention in the genesis of spindles and delta waves, both of which are thought to originate from thalamic neurons (8), although cortically generated delta waves also have been found in cats with thalamic lesions (15). A role for TC T-type channels in sleep has been demonstrated in two studies, one using mice with a global deletion (16) and the other using mice with a thalamus-restricted deletion (17) of Ca V 3.1 T-type channels. Both mice exhibited reduced delta waves with intact slow waves (16, 17). Fragmented sleep was observed in both mice, indicating that this sleep phenotype in the global Ca V 3.1 / mice Significance This study addresses one of the most fundamental issues in sleep rhythm generation. The theory that low-threshold burst firing mediated by T-type calcium channels in thalamocortical neurons is the key component for sleep spindles, has been accepted as dogma and appears throughout the literature. In this study, however, in vivo and in vitro evidence shows that sleep spindles are generated normally in the absence of T-type channels and burst firing in thalamocortical neurons. Furthermore, our data indicate a potentially important role of tonic firing in this rhythm generation. This study advances the knowledge of sleep and vigilance control to another level of understanding. Author contributions: E.C. and H.-S.S. designed research; J.L., K.S., K.L., J.H., S.C., and E.C. performed research; J.L., K.S., K.L., J.H., H.L., and E.C. analyzed data; and J.L., E.C., and H.-S.S. wrote the paper. The authors declare no conflict of interest. 1 Present address: Samsung Advanced Institute of Technology, Kiheung , Korea. 2 To whom correspondence may be addressed. eunjicheong@yonsei.ac.kr or shin@ ibs.re.kr. This article contains supporting information online at /pnas /-/DCSupplemental PNAS December 10, 2013 vol. 110 no. 50

2 (Fig. 1B). These findings appear to indicate that sleep spindles are not altered in Ca V 3.1 / mice. Decreased Duration of Barbiturate-Induced Spindle Episodes in Ca v 3.1 / Mice. We next examined the characteristics of spindle oscillations induced by barbiturate injection. In agreement with the literature (20), we found that barbiturate-induced spindles were characterized by large-amplitude oscillations of increased duration compared with sleep spindles (Fig. 2A). The mean duration of barbiturate-induced spindle episodes was almost doubled compared with the spindle episode duration during NREM sleep in the Ca V 3.1 +/+ mice, but not in the Ca V 3.1 / mice. Compared with the Ca V 3.1 +/+ mice, the Ca V 3.1 / mice exhibited shorter barbiturate-induced spindle episodes (*P < 0.05; Fig. 2B), as well as smaller peak-to-peak amplitudes (*P < 0.05), whereas there was no difference in the peak frequency of spindles between the two genotypes (Fig. 2B). Of note, the spindle episodes in the Ca V 3.1 / mice were composed of shorter episodes compared with those seen in the WT mice (Fig. 2C). Taken together, these results suggest that low-threshold spikes (LTSs) mediated by Ca V 3.1 T-type Ca 2+ channels in TC neurons are not essential for spindle generation, but may play a role in increasing the duration of spindles episodes induced by barbiturates. Spike Discharges in TC Neurons During in Vivo Spindle Oscillations. Our findings regarding natural sleep spindles and barbiturate- Fig. 1. Sleep spindles during NREM sleep in Ca V 3.1 +/+ (WT) and Ca V 3.1 / (KO) mice. (A) Sample traces show the raw (upper trace) and filtered (lower trace) EEG signals recorded during NREM sleep. Bandpass-filtered (6 15 Hz) EEG signals clearly show spindle events (arrowheads) in both genotypes. (B) There were no differences in the mean length of each spindle episode, number of episodes, mean peak-to-peak amplitude, and peak frequency between Ca V 3.1 +/+ and Ca V 3.1 / mice. is related to a defect in TC neurons. The effect of the mutation on spindle rhythms was unclear, however (16). In the present study, we examined the role of low-threshold burst firing in sleep spindles expressed in TC neurons using mice lacking Ca V 3.1 T-type Ca 2+ channels. We observed intact sleep spindles in Ca V 3.1 / mice during NREM sleep. Our findings suggest that the classical view of the roles of T-type channels and burst firing in TC neurons with respect to the generation of spindle oscillations may need to be revised. Results Sleep Spindles Are Not Altered in Ca v 3.1 / Mice During Natural NREM Sleep. First, in an attempt to elucidate the role of lowthreshold burst firing mediated by TC T-type channels in generating sleep spindles, we examined EEGs recorded during a state of natural sleep in mice lacking the Ca V 3.1 T-type channels that are highly expressed in TC neurons. EEG/electromyography (EMG) recordings were acquired with a telemetry system over a continuous period of 48 h in mice housed in their home cages. These data were analyzed to assign the sleep categories of awake, NREM, and rapid eye movement (REM) states based on standard criteria for the analysis of rodent sleep (18). The sleep spindle episodes in both Ca V 3.1 +/+ and Ca V 3.1 / mice exhibited a characteristic waxing-and-waning pattern and typical recurring sequences (Fig. 1A) that met the stereotypical criteria for sleep spindles (1, 8, 19). There were no significant differences in the number of spindle episodes, mean duration of episodes, peak-to-peak amplitudes, or peak frequencies of spindle oscillations between the Ca V 3.1 +/+ and Ca V 3.1 / mice Fig. 2. The duration of spindle episodes induced by barbiturate injection was diminished in Ca V 3.1 / mice compared with Ca V 3.1 +/+ mice. (A) Sample traces illustrate EEG signals with a systemic injection of barbiturate (20 mg/ kg i.p.). The upper trace is a raw EEG trace, and the lower trace shows bandpass-filtered (6 15 Hz) EEG signals. Arrowheads denote spindle events. (B) Ca V 3.1 / mice had a shorter mean spindle episode duration and smaller peak-to-peak spindle amplitude compared with WT mice, but similar number of episodes and peak frequency. (C) Barbiturate-induced spindle sequences were categorized according to the duration of each event. Spindle events in Ca V 3.1 / mice were of short duration. NEUROSCIENCE Lee et al. PNAS December 10, 2013 vol. 110 no

3 induced spindles in Ca V 3.1 / mice prompted us to investigate the properties of TC neuron spike discharges during spindle episodes in the absence of low-threshold burst firing. To do this, we performed in vivo extracellular single-unit recordings in TC neurons while simultaneously monitoring the cortical EEG under barbiturate anesthesia. Both tonic and burst spikes were observed in Ca V 3.1 +/+ TC cells, whereas only tonic spikes were detected in Ca V 3.1 / TC cells (Fig. 3 A C), confirming previous in vitro data (13). Bursts were identified according to the conventional criteria (21), with a burst defined as a cluster of two or more spikes with interspike intervals of 4 ms, in which the first spike in the burst has a preceding interspike interval of >100 ms (Fig. S1). Spikes recorded over 15 consecutive spindle sequences (SP1 SP15) were aligned with the initiation time of each cortical EEG spindle event. The perievent histograms showed a large number of TC spikes co-occurring within cortical spindle events (Fig. 3B), as reported previously (22, 23). Of note, we found the spike frequency was significantly higher during spindle periods compared with nonspindle periods in both WT mice (1.71 ± 0.23 Hz vs ± 0.16 Hz) and KO mice (1.94 ± 0.43 Hz vs ± 0.04 Hz) (Fig. 3C, a). Of particular importance, there was no difference in the total spiking rate (across tonic and burst activity) within spindle events between two genotype TC neurons (P = 0.42; Fig. 3C, a) with event correlation (Fig. S2), whereas the nonspindle spike frequency was significantly lower in Ca V 3.1 / neurons compared with Ca V 3.1 +/+ neurons (Fig. 3C, a). Further analysis of the firing mode of TC neurons revealed a significantly higher tonic spike frequency in both WT and mutant mice during spindle periods than during nonspindle periods (Fig. 3C, b). During spindle episodes, the tonic spike frequency was significantly higher in mutant KO mice compared with WT mice, perhaps explaining why the total spiking rate of mutant TC neurons within cortical spindle events was as high as that in WT neurons, even though the mutants are devoid of burst spikes. In contrast, there was no significant difference in burst firing between spindle and nonspindle periods (Fig. 3C, b), either in the frequency of burst spikes (0.58 ± 0.15 Hz vs ± 0.09 Hz) or in the occurrence of burst events (0.28 ± 0.09 Hz vs ± 0.06 Hz; P = 0.27). In other words, the majority of Ca V 3.1 +/+ TC cells discharged burst spikes without preference for spindle episodes; just 2 of 13 fired bursts preferentially during the spindle period. Also of interest is our finding of no difference in the overall spike frequency of TC neurons during spindle events between the two genotypes. Of note, WT TC neurons exhibited a higher tonic spike frequency than burst spike frequency during spindle events. More importantly, the tonic spike frequency in WT mice was significantly ( 3.5-fold) higher during spindle periods than during nonspindle periods, whereas the burst spike frequency did not differ in the two periods. The tonic spike frequency observed in Ca V 3.1 / TC cells coincided even more strongly with spindles ( 10-fold higher). In addition, spike discharges recorded in both Ca V 3.1 +/+ and Ca V 3.1 / TRN neurons were more frequent during spindle episodes (Fig. S3), in parallel with the findings in TC neurons. Spindle-Like Oscillations Elicited in Intrathalamic Circuits in Ca v 3.1 / Mice. Our finding that robust generation of cortical spindles is accompanied by increased TC spike discharges in Ca V 3.1 / mice prompted us to further investigate whether spindle oscillations can be generated within an intrathalamic circuit composed only of TRN and TC neurons in the presence or absence of LTSs in TC neurons. For this purpose, we assessed thalamic network Fig. 3. TC neurons discharge tonic spikes abundantly during spindle events in both Ca V 3.1 +/+ and Ca V 3.1 / mice in vivo. (A) Sample traces showing cortical EEG signals over the ipsilateral frontal lobe (upper trace) and extracellular single-unit recordings from TC regions (lower trace) in Ca V 3.1 +/+ and Ca V 3.1 / mice under barbiturate anesthesia. Burst and tonic spikes are represented by vertical red and blue lines, respectively. (B) Perievent histograms depicting the number of spikes in a 200-ms bin for 15 cortical spindle episodes (SP1 SP15). Red, blue, and turquiose bars represent the spiking frequency of burst, tonic, and mixed spikes, respectively. The mixed spikes describe the events in which both tonic and burst firing were detected within a 250-ms bin. The duration of each spindle episode is shown as a horizontal gray line. (C, a) Total spike frequency during spindle and nonspindle events between WT (solid bar) and Ca V 3.1 / (open bar with diagonal stripes) mice. Burst spikes are in red, and tonic spikes are in blue. (C, b) Tonic and burst firing frequencies during spindle and nonspindle periods in WT and KO mutant mice. Values are mean ± SEM. *P < 0.05; **P < Lee et al.

4 activity in TC slices with and without attached cortex (24). Previous studies have demonstrated that oscillations within the range observed for sleep spindles in vivo (22) can be evoked in thalamic slices by electrical stimulation of the internal capsule (IC) (25). These are thus referred to as spindle-like oscillations (SLOs) (26). We adopted this paradigm using the multielectrode array system as described in SI Materials and Methods. Thalamic oscillations were evoked by a single 20- to 100-μV, 60- to 80-μs electrical stimulation in the IC in TC slices. Clusters of spikes composed of multiple units with various amplitudes were detected simultaneously in TRN and TC nuclei (Fig. 4A). SLOs were readily generated in these thalamic slices from both Ca V 3.1 +/+ and Ca V 3.1 / mice in the presence and absence of cortex (Fig. 4A). There were no significant differences in the durations of evoked oscillatory activities between Ca V 3.1 +/+ and Ca V 3.1 / thalamic slices with cortex (7 slices from 7 WT mice vs. 8 slices from 7 mutant mice) or without cortex (13 slices from 8 WT mice vs. 12 slice from 9 mutant mice) (Fig. 4B), although the total TC spike numbers in Ca V 3.1 / thalamic slices were lower than those in Ca V 3.1 +/+ slices in the absence of cortex (P = 0.047; Fig. 4C). The foregoing results provide direct evidence contrary to the classical view that postinhibitory low-threshold burst firing mediated by T-type Ca 2+ channels in TC neurons is essential for the generation of oscillations within the TRN-TC circuit. They suggest that instead, sleep spindles observed in Ca V 3.1 / mice may be generated within the TRN-TC circuit in the absence of TC bursts. Excitability of Ca V 3.1 / TC Neurons in Vitro. Our next step was to examine whether generation of spindle oscillations within the Ca V 3.1 / intrathalamic circuit might be related to changes in the intrinsic excitability of TC neurons rather than to a loss of lowthreshold burst firing by deletion of Ca V 3.1 T-type channels. To do this, we used the whole-cell patch-clamp technique to measure the excitability of TC neurons. We found no differences between Ca V 3.1 +/+ (n = 12) and Ca V 3.1 / (n = 14) TC neurons in membrane capacitance (141.2 ± 15.6 pf vs ±10.5 pf), resting membrane potential ( ± 5.34 mv vs ± 6.31 mv), or membrane resistance (219.5 ± 48.2 MΩ vs ± 38.2 MΩ). We then measured hyperpolarization-activated cyclic nucleotide-gated channel-mediated current (I h ) in TC neurons at the tail current at 0.7 s after a test voltage pulse (Fig. 5A) to minimize the capacitative contribution (27) or non I h -active current (28). We observed an increase in the current density of I h in Ca V 3.1 / TC neurons compared with Ca V 3.1 +/+ TC neurons only when membrane potential was hyperpolarized to 100 mv (Fig. 5 A and B). We then measured the firing pattern of TC neurons. Low-threshold burst firing was ablated in Ca V 3.1 / TC neurons, whereas vigorous burst firing was induced in Ca V 3.1 +/+ TC neurons, with a membrane potential rebound to 60 mv after hyperpolarization to 70 to 120 mv (Fig. 5 C and D). Some Ca V 3.1 / TC neurons (6 of 10) produced a single spike when the membrane potential was returned to 60 mv after hyperpolarization to 100 mv (Fig. 5 C and D), possibly related to the enhanced I h current (Fig. 5 A and B). However, the enhanced I h current did not affect either the tonic or burst firing rate elicited by depolarizing current injections (Fig. 5 C E), and did not alter any other intrinsic membrane properties of Ca V 3.1 / TC neurons. Thus, we conclude that deletion of Ca V 3.1 T-type channel did not significantly alter the intrinsic excitability of TC neurons. The firing pattern of TRN neurons was not changed either (Fig. S4). Fig. 4. SLOs were elicited in the intrathalamic network in both Ca V 3.1 +/+ and Ca V 3.1 / thalamic slices in either the presence or absence of the cortex. (A) Sample traces of simultaneous multiunit recordings of SLOs from TRN (upper trace) and TC (lower trace) nuclei in Ca V 3.1 +/+ (Left) and Ca V 3.1 / (Right) thalamic slices demonstrate oscillatory activity in both nuclei. Network oscillations were evoked by single electrical shocks to the IC. Magnified traces show that multiple units with various amplitudes were detected in both nuclei in the thalamus. (B) Duration of SLOs in TRN and TC nuclei in Ca V 3.1 +/+ and Ca V 3.1 / thalamic slices in the presence or absence of cortex in the slice. (C) Total spike activity in TRN during the oscillatory activity evoked by IC stimulation in the presence or absence of cortex in the slice. Discussion We have demonstrated here that sleep spindles and intrathalamic SLOs are generated in the absence of Ca V 3.1 T-type Ca 2+ channels in TC neurons. This finding calls into question the validity of the long-standing hypothesis that LTSs mediated by T- type channels in TC neurons are essential for spindle oscillations. Ca V 3.1 T-Type Channels in Sleep Rhythms. Sleep spindles have been proposed to arise from the intrathalamic circuit composed of TRN and TC neurons (8, 29). Postinhibitory LTSs in TC neurons has been considered essential for the generation of intrathalamic oscillations and their propagation to the cortex (1). According to this hypothesis, spindle oscillations should be diminished, if not abolished, in cortical EEGs of Ca V 3.1 / mice lacking LTSs in TC neurons. In contrast to this expectation, we clearly observed sleep spindle waves in cortical EEG recordings from both Ca V 3.1 +/+ and Ca V 3.1 / mice during NREM sleep. We chose a method suitable for analyzing spindle waves in mice (19) used in previous studies (1 3). In a previous study on the sleep phenotype of Ca V 3.1 / mice, we filtered EEG data with a high-pass cutoff at 0.1 Hz to include and analyze all sleep rhythms. This did not exclude the possibility that a significant reduction in delta power in Ca V 3.1 / mice might exaggerate the difference in the EEG power density of the spindle component (16). In the present study, we used a 6- to 15-Hz bandpass filter to clearly reveal spindle oscillations and exclude delta and slow waves (Figs. 1A and 2A and Fig. S5), and found no difference in the number of sleep spindle episodes between Ca V 3.1 +/+ and Ca V 3.1 / mice. Neither the duration nor the amplitude of the spindle episodes was reduced in Ca V 3.1 / mice during natural NREM sleep, raising questions about the essential role of TC neuron T-type NEUROSCIENCE Lee et al. PNAS December 10, 2013 vol. 110 no

5 Fig. 5. Firing properties of Ca V 3.1 / TC neurons in vitro. (A) Hyperpolarization-activated cyclic nucleotide-gated channel current (I h ) responses (upper traces) to voltage steps (lower traces) recorded from Ca V 3.1 +/+ (Left) and Ca V 3.1 / (Right) TC neurons were measured at the tail current (arrowheads). The membrane potential was held at 60 mv, stepped to a hyperpolarizing potential between 120 and 70 mv to elicit the current, and then returned to 60 mv. (B) The current density of I h was larger in Ca V 3.1 / TC cells compared with Ca V 3.1 +/+ TC cells when membrane potential was hyperpolarized to 100 mv. (C) Representative traces from wholecell patch recordings show the firing pattern of Ca V 3.1 +/+ and Ca V 3.1 / TC neurons responding to hyperpolarizing or depolarizing current inputs. (D) The number of spikes per burst was counted in WT TC neurons (closed circles). No bursts were induced in mutants (open circles). (E) Number of tonic spikes per s induced by various depolarizing currents. channels in the generation of sleep spindles. However, the mean episode duration and amplitude of barbiturate-induced spindles were substantially reduced in Ca V 3.1 / mice, in accordance with our previous findings of a reduced EEG power density at delta and spindle frequencies in Ca V 3.1 / mice under high-dose urethane (2 mg/kg), another spindle-inducing anesthesia (16). These results suggest that T-type Ca 2+ channels and the resulting LTSs in TC neurons are not required for the generation of spindle oscillations in TC circuits, but may play a role in enhancing the strength of spindles, as observed in barbiturateinduced spindles. This view is quite different from the classical theory that the generation of sleep spindles requires low-threshold burst firing of TC neurons. A recent study in mice lacking the Ca V 3.3 T-type channels abundantly expressed in TRN found a selective reduction in spindle power before the onset of REM sleep (30). The same study also found that delta power was unaffected by the absence of Ca V 3.3 channels (30), whereas Ca V 3.1 / mice demonstrated a significantly reduced power density of the delta range (1 4 Hz) during NREM sleep (16). These findings suggest that Ca V 3.3 T- type Ca 2+ channels in TRN neurons contribute to spindles, whereas channels encoded by Ca V 3.1 genes in TC neurons contribute to the delta waves during NREM sleep. Firing Patterns During Spindle Oscillations in TC Neurons. We examined an intriguing question regarding the firing pattern of TC neurons that lack T-type channels in response to inhibitory inputs from the TRN neurons during spindle episodes. Only tonic spikes were observed in TC nuclei in Ca V 3.1 / mice, as expected (Fig. 3). Tonic spikes were more abundant than burst spikes during spindles in Ca V 3.1 +/+ TC neurons, whereas no difference in tonic and burst spike frequency was seen during nonspindle periods. With regard to the critical role of tonic firing in spindles, our most compelling observation may be that in WT TC neurons, the tonic spike frequency increases significantly during cortical spindle events compared with nonspindle periods. This pattern is clearly different from that seen for burst spike frequency in WT TC neurons, which is almost equal during spindle and nonspindle periods. Of interest, the overall spike frequency during spindle episodes in mutant TC neurons with no burst spikes did not differ from that in WT TC neurons, because the degree of increase in tonic spike frequency during spindle periods was much larger in mutant TC neurons. Nevertheless, the duration of barbiturate-induced spindles was shorter in the mutant mice compared with the WT mice (Fig. 2) even though the total spike frequency was similar in the two mice during barbiturate-induced spindles (Fig. 3), implying that TC burst firing may play a role in increasing the duration of barbiturate-induced spindle oscillations. The total spike frequency during nonspindle periods was lower in mutant mice than in the WT mice, a finding that remains to be explained. The explicit induction of oscillations within an isolated Ca V 3.1 / intrathalamic circuit convinced us that spindle oscillations may be generated within the TRN-TC network as suggested previously (31), but in the absence of LTSs in TC neurons. We also confirmed that deletion of Ca V 3.1 T-type channels had no effect on the intrinsic firing properties of TC neurons (Fig. 5). Taken together, these findings suggest that spindle oscillations can be generated and propagated via any type of TC spike activity as long as TC neurons discharge spikes at an appropriate frequency. This conclusion differs from previous observations on the critical role of TC T-type channels in sleep delta waves (16) and paroxysmal SWDs (13, 32, 33), and leads us to suggest that the various TC oscillations may be generated through diverse mechanisms Lee et al.

6 Ca V 3.1 T-Type Channels in Spindles vs. SWDs. It has been proposed that TC oscillations occurring during sleep spindles and SWDs of absence seizures share common elements and mechanisms (11). Clinical observations of the temporal coincidence between sleep spindles and SWDs and experimental findings showing that penicillin, a weak GABA A R antagonist, induces the transformation of sleep spindles into SWDs gave rise to the view that SWDs may be a perverted form of sleep spindles (34). Previous studies have found that Ca V 3.1 / mice are resistant to GABA B receptor agonist-induced SWDs (13), and that deletion of Ca V 3.1 completely suppresses or markedly reduces SWDs in genetic mouse models of absence seizures (35), implying that Ca V 3.1 T-type Ca 2+ channels play a crucial role in the genesis of SWDs. In contrast, the increased T-type Ca 2+ currents in TC neurons occur alongside the generation of SWDs and absence seizure phenotypes in transgenic or mutant mice (32, 33). Our present findings support the hypothesis that Ca V 3.1 T- type Ca 2+ channels are not an essential component in the generation of sleep spindles. Overall, these results suggest that paroxysmal TC oscillations during SWDs require the postinhibitory low-threshold burst firing mediated by T-type channels in TC neurons, whereas physiological TC oscillations during sleep spindles do not. In conclusion, our findings suggest that T-type Ca 2+ channels in TC neurons might not be essential for the generation and propagation of spindle oscillations. This study provides valuable insights into the realm of T-type channels as a possible target for differentiation of various TC oscillations and implies that the TC oscillations generated under pathological conditions may be controlled by modulating T-type channels in TC neurons without affecting natural physiological sleep spindles. Our findings raise further questions, including what drives TC neurons to fire during spindles and what determines the spindle frequency. Future studies should be aimed at investigating the detailed mechanism through which each type of TC oscillation is generated. Materials and Methods Treatment of animals and all experiments were conducted in accordance with the ethical guidelines of the Institutional Animal Care and Use Committee of Yonsei University, Institute of Basic Science and the Korea Institute of Science and Technology. The mice and experiments for this study are described in detail in SI Materials and Methods. ACKNOWLEDGMENTS. We thank G. Ha for help with the PCR analysis. This work was supported by funding from the Institute for Basic Science (Grant HQ1301), the Translational Research Center for Protein Function Control (Grant NRF ), the Basic Science Research Program (Grant NRF ), and the Pioneer Research Center Program (Grant ), all of which are funded by the Korean Ministry of Science, Information and Communication Technology, and Future Planning. 1. Steriade M, McCormick DA, Sejnowski TJ (1993) Thalamocortical oscillations in the sleeping and aroused brain. Science 262(5134): Meeren HK, Veening JG, Möderscheim TA, Coenen AM, van Luijtelaar G (2009) Thalamic lesions in a genetic rat model of absence epilepsy: Dissociation between spike-wave discharges and sleep spindles. Exp Neurol 217(1): Schofield CM, Huguenard JR (2007) GABA affinity shapes IPSCs in thalamic nuclei. J Neurosci 27(30): Contreras D, Steriade M (1995) Cellular basis of EEG slow rhythms: A study of dynamic corticothalamic relationships. J Neurosci 15(1 Pt 2): Contreras D, Destexhe A, Steriade M (1997) Spindle oscillations during cortical spreading depression in naturally sleeping cats. Neuroscience 77(4): Steriade M, Domich L, Oakson G, Deschênes M (1987) The deafferented reticular thalamic nucleus generates spindle rhythmicity. J Neurophysiol 57(1): Steriade M, Deschênes M, Domich L, Mulle C (1985) Abolition of spindle oscillations in thalamic neurons disconnected from nucleus reticularis thalami. J Neurophysiol 54(6): von Krosigk M, Bal T, McCormick DA (1993) Cellular mechanisms of a synchronized oscillation in the thalamus. Science 261(5119): Huguenard JR (1996) Low-threshold calcium currents in central nervous system neurons. Annu Rev Physiol 58: Talley EM, et al. (1999) Differential distribution of three members of a gene family encoding low voltage-activated (T-type) calcium channels. J Neurosci 19(6): Beenhakker MP, Huguenard JR (2009) Neurons that fire together also conspire together: Is normal sleep circuitry hijacked to generate epilepsy? Neuron 62(5): Cheong E, Shin HS (2013) T-type Ca2 + channels in normal and abnormal brain functions. Physiol Rev 93(3): Kim D, et al. (2001) Lack of the burst firing of thalamocortical relay neurons and resistance to absence seizures in mice lacking alpha(1g) T-type Ca(2+) channels. Neuron 31(1): Cheong E, Shin HS (2013) T-type Ca²+ channels in absence epilepsy. Biochim Biophys Acta 1828(7): Steriade M, Nuñez A, Amzica F (1993) Intracellular analysis of relations between the slow (<1 Hz) neocortical oscillation and other sleep rhythms of the electroencephalogram. J Neurosci 13(8): Lee J, Kim D, Shin HS (2004) Lack of delta waves and sleep disturbances during nonrapid eye movement sleep in mice lacking alpha1g-subunit of T-type calcium channels. Proc Natl Acad Sci USA 101(52): Anderson MP, et al. (2005) Thalamic Cav3.1 T-type Ca2 + channel plays a crucial role in stabilizing sleep. Proc Natl Acad Sci USA 102(5): Radulovacki M, Virus RM, Djuricic-Nedelson M, Green RD (1984) Adenosine analogs and sleep in rats. J Pharmacol Exp Ther 228(2): Wang XJ, Liu Y, Sanchez-Vives MV, McCormick DA (2003) Adaptation and temporal decorrelation by single neurons in the primary visual cortex. J Neurophysiol 89(6): Fuentealba P, Timofeev I, Bazhenov M, Sejnowski TJ, Steriade M (2005) Membrane bistability in thalamic reticular neurons during spindle oscillations. J Neurophysiol 93(1): Sherman SM (1996) Dual response modes in lateral geniculate neurons: Mechanisms and functions. Vis Neurosci 13(2): Contreras D, Steriade M (1996) Spindle oscillation in cats: The role of corticothalamic feedback in a thalamically generated rhythm. J Physiol 490(Pt 1): Llinás RR, Steriade M (2006) Bursting of thalamic neurons and states of vigilance. J Neurophysiol 95(6): Agmon A, Connors BW (1991) Thalamocortical responses of mouse somatosensory (barrel) cortex in vitro. Neuroscience 41(2-3): Huntsman MM, Porcello DM, Homanics GE, DeLorey TM, Huguenard JR (1999) Reciprocal inhibitory connections and network synchrony in the mammalian thalamus. Science 283(5401): Jacobsen RB, Ulrich D, Huguenard JR (2001) GABA(B) and NMDA receptors contribute to spindle-like oscillations in rat thalamus in vitro. J Neurophysiol 86(3): Paz JT, et al. (2013) Closed-loop optogenetic control of thalamus as a tool for interrupting seizures after cortical injury. Nat Neurosci 16(1): McCormick DA, Pape HC (1990) Properties of a hyperpolarization-activated cation current and its role in rhythmic oscillation in thalamic relay neurones. J Physiol 431: Kim U, Bal T, McCormick DA (1995) Spindle waves are propagating synchronized oscillations in the ferret LGNd in vitro. J Neurophysiol 74(3): Astori S, et al. (2011) The Ca(V)3.3 calcium channel is the major sleep spindle pacemaker in thalamus. Proc Natl Acad Sci USA 108(33): Huguenard JR, McCormick DA (2007) Thalamic synchrony and dynamic regulation of global forebrain oscillations. Trends Neurosci 30(7): Ernst WL, Zhang Y, Yoo JW, Ernst SJ, Noebels JL (2009) Genetic enhancement of thalamocortical network activity by elevating alpha 1g-mediated low-voltage-activated calcium current induces pure absence epilepsy. J Neurosci 29(6): Cheong E, et al. (2009) Deletion of phospholipase C beta4 in thalamocortical relay nucleus leads to absence seizures. Proc Natl Acad Sci USA 106(51): Leresche N, Lambert RC, Errington AC, Crunelli V (2012) From sleep spindles of natural sleep to spike and wave discharges of typical absence seizures: Is the hypothesis still valid? Pflugers Arch 463(1): Song I, et al. (2004) Role of the alpha1g T-type calcium channel in spontaneous absence seizures in mutant mice. J Neurosci 24(22): NEUROSCIENCE Lee et al. PNAS December 10, 2013 vol. 110 no

7 Supporting Information Lee et al /pnas SI Materials and Methods Animals. Ca V 3.1 heterozygous mice (Ca V 3.1 +/ ) were maintained in two genetic backgrounds, 129/sv and C57BL/6J. All experiments used Ca V 3.1 / mice and their WT littermates in the F1 hybrid generated by mating Ca V 3.1 +/ mice from these two genetic backgrounds. Mice were maintained with free access to food and water under a 12-h light/12-h dark cycle, with the light cycle beginning at 6:00 AM. Animal care was provided and all experiments were conducted in accordance with the ethical guidelines of the Institutional Animal Care and Use Committee of the Institute of Basic Science, Yonsei University and the Korean Institute of Science and Technology. All experiments were conducted using 9- to 14-wk-old male mice. Drugs. Tribromoethanol (Avertin), urethane, and pentobarbital were purchased from Sigma-Aldrich. All drugs were administered by i.p. injection. Acute Electroencephalography Monitoring Under Anesthesia. Twelveto 14-wk-old male mice were prepared for acute monitoring of electroencephalography (EEG) signals. The implantation procedure was performed under 0.2% tribromoethanol (Avertin) anesthesia (20 ml/kg i.p.). An epidural electrode for EEG recording was implanted in the parietal lobe (anteroposterior, 1.7 mm; lateral, +1.5 mm), and a reference electrode was implanted in the occipital region of the skull. Barbiturates were administrated systemically at 20 mg/kg into the peritoneal cavity. EEG signals were bandpass-filtered (6 15 Hz) to analyze the spindle oscillations while excluding other brain rhythms. Chronic EEG/Electromyography Monitoring. Twelve- to 14-wk-old male mice were used for chronic monitoring of EEG/electromyography (EMG) signals. The surgical procedure for electrode implantation was identical to that described for acute EEG monitoring. A wireless telemetry transmitter (Data Science International) was implanted to record EEG and EMG patterns during sleep and awake states with the mice in their home cages. After a 1-wk recovery, the mice were placed in unrestraining chronic recording environments on a 12-h light/12-h dark schedule. EEG and EMG signals were amplified (F14-EET; Data Science International) and low-pass filtered at 100 Hz for EEG and highpass filtered at 0.1 Hz for EEG and digitized at a sampling rate of 250 Hz. Data were acquired continuously over 48 h using Dataquest A. R.T. 2.2 (Data Science International). After collection, EEG/ EMG records were scored semiautomatically using the SleepSign sleep scoring system (Kissei Comtec America) into 4-s epochs as awake, rapid eye movement (REM), and non-rapid eye movement (NREM) based on standard criteria for rodent sleep (1). This preliminary scoring was visually inspected and corrected as appropriate. EEG signals were bandpass-filtered (6 15 Hz) to analyze the spindle oscillations. In Vivo Single-Unit Recording in the Thalamocortical Region. Extracellular single-unit recording was performed as described previously (2) with some modifications. Mice were anesthetized with urethane (1.5 g/kg i.p.) and fixed in a stereotaxic device (David Kopf Instruments). For gross EEG recording, a screw electrode was inserted into the skull over the frontal lobe of the ipsilateral hemisphere. To insert an electrode for single-unit recording, the scalp was incised, and a hole was drilled into the skull above the region where VP/VL thalamic nuclei are located (anteroposterior, 1.6; lateral, +1.7). Parylene-coated tungsten microelectrodes ( 2 MΩ impedance) were positioned into the VP/VL thalamic regions (depth, mm) until single-unit spikes were observed. Pentobarbital sodium (10 mg/kg i.p.) was injected to induce spindle oscillations and to maintain a stable pattern of spindles on the EEG. EEG signals were amplified by 1,000-fold and high-pass filtered at 0.1 Hz and low-pass filtered at 100 Hz using a World Precision Instruments DAM80 amplifier. Signals from single-unit recordings were amplified approximately 5,000- fold and bandpass-filtered at 300 3,000HzwithanAxonInstruments Cyberamp 402 amplifier. Both signals were digitized at a 10-kHz sampling rate (Digidata 1322A; Axon Instrument) and stored in a computer for off-line analysis. The electrode positions for single-unit recording were confirmed by postmortem histology. The number of spikes was quantified using Mini Analysis software (Synaptosoft). Bursts were identified according to the criteria of Sherman et al. (3), with a burst defined as a cluster of two or more spikes with interspike intervals of 4 ms, in which the first spike in the burst has a preceding interspike interval of at least 100 ms. The percentage of burst spikes was calculated as the ratio of the number of spikes within a burst to the number of all recorded spikes. Preparation of Brain Slices and Spindle-Like Oscillation Recordings. Thalamocortical (TC) slices were prepared as described previously (4). In brief, mice at postnatal day were anesthetized with halothane and then decapitated. Brains were removed and transferred into ice-cold slicing solution containing 234 mm sucrose, 11 mm glucose, 24 mm NaHCO 3, 2.5 mm KCl, 1.25 mm NaH 2 PO 4, 10 mm MgSO 4, and 0.5 mm CaCl 2 bubbled with 95% O 2 and 5% CO 2. Then 400-μm-thick TC slices were cut at 45 or 50 degrees with a Leica Microsystems VT1000 microtome. The intrathalamic oscillations were recorded in the brain slices with and those without cortical parts. The brain slices were incubated in artificial cerebrospinal fluid (ACSF) for 1 h before recordings were made. For the recordings, brain slices were placed in a probe of a multielectrode array system (MED64 system; Alpha MED Scientific) and perfused with ACSF at 32 C. In reduced [Mg 2+ ] experiments, the [MgCl 2 ] in the perfusion solution was reduced to 0.65 mm, whereas normal ACSF contains 2 mm MgCl 2. Intrathalamic oscillations were evoked by a 20- to 100-μV, 60- to 80-μs stimulus to the internal capsule (IC) through electrodes positioned at the border of the IC and the thalamic reticular nucleus (TRN). The stimulus interval was varied depending on the duration of an oscillation, to avoid intersweep variability presumably owing to rundown, but was in the range of one stimulus every s. Extracellular multiunit activity was recorded using the MED64 system at a sampling rate of 20 khz. The amplitude of action potentials ranged from 15 to 80 μv, and signals were bandpass-filtered (0.3 5 khz) using MED64 Mobius software. Preparation of Brain Slices and Whole-Cell Patch Clamping. Firing patterns of TC and TRN neurons were recorded in acutely isolated brain slices from 4-wk-old mice. Brains were blocked and sectioned in the coronal plane in ice-cold slicing solution (124 mm NaCl, 26 mm NaHCO 3, 1.25 mm NaH 2 PO 4, 5 mm MgCl 2, 1 mm CaCl 2, 3 mm KCl, and 10 mm glucose) and then incubated for at least 1 h in slicing solution at room temperature before use. Lee et al. 1of4

8 Firing patterns of thalamic relay neurons in brain slices were recorded under current clamp mode in ACSF solution (124 mm NaCl, 26 mm NaHCO 3,1.25mMNaH 2 PO 4,1.3mMMgSO 4, 2.4 mm CaCl 2, 3 mm KCl, and 10 mm glucose) bubbled with 95% O 2 /5% CO 2. Patch electrodes (4 6 MΩ) fabricated from standard-wall borosilicate glass (GC150F-10; Warner Instruments) were filled with an intrapipette solution containing 140 mm K gluconate, 10 mm KCl, 1 mm MgCl 2, 10 mm Hepes, 0.02 mm EGTA, 4 mm Mg-ATP, and 0.3 mm Na 2 -GTP, with ph adjusted to Signals were amplified with a Molecular Devices MultiClamp 700B amplifier and analyzed using pclamp 10.1 (Molecular Devices) and Mini Analysis (Synaptosoft) software. Tissue Processing. For histological analysis or post hoc staining to confirm the recording sites, mice were deeply anesthetized with tribromoethanol (Avertin) and transcardially perfused with 0.9% NaCl followed by 4% paraformaldehyde in PBS. After perfusion, brains were removed and postfixed in 4% paraformaldehyde in PBS at 4 C overnight, then transferred to 30% sucrose in 1 PBS and stored at 4 C overnight. Brains were then frozen in freezing media, cut into serial 30-μm-thick coronal sections on a freezing microtome, and collected into tissue storage buffer (5). Data Analysis. Spindles were quantified by counting the number and duration of spindles using MATLAB (MathWorks). We developed a semiautomatic analysis paradigm using MATLAB to detect the spindle waves preventing the contamination of other components. We used a method that is particularly suitable for detecting spindles and focused on the appropriate waveform detection. Threshold crossing and amplitude were used to determine the start and continuity of spindles. Waveforms with low total power and nonspindle harmonic, high-frequency content were removed automatically. This approach is more reliable than a purely power-based approach or visual inspection and can detect spindle events not contaminated by slow wave, delta wave, or other waveforms. Unless specified otherwise, the total duration of spindles and averaged spectrograms were obtained from the parietal cortex signal. Unless specified otherwise, the Student t test was used to identify statistically significant differences between groups. Cross-correlation was calculated for cortex EEG and spike events as source and reference, respectively (2,048-point fast Fourier transform and Hanning window with no overlap). A P value < 0.05 was considered statistically significant. 1. Radulovacki M, Virus RM, Djuricic-Nedelson M, Green RD (1984) Adenosine analogs and sleep in rats. J Pharmacol Exp Ther 228(2): Kim D, et al. (2003) Thalamic control of visceral nociception mediated by T-type Ca2 + channels. Science 302(5642): Sherman SM (1996) Dual response modes in lateral geniculate neurons: Mechanisms and functions. Vis Neurosci 13(2): Jacobsen RB, Ulrich D, Huguenard JR (2001) GABA(B) and NMDA receptors contribute to spindle-like oscillations in rat thalamus in vitro. J Neurophysiol 86(3): Kang SJ, et al. (2008) Expression of Kir2.1 channels in astrocytes under pathophysiological conditions. Mol Cells 25(1): Fig. S1. Interspike-interval plot showing distinct firing patterns in Ca V 3.1 +/+ and Ca V 3.1 / mice. Each spike of single-unit activities of TC neurons recorded in vivo from Ca V 3.1 +/+ mice (A) and Ca V 3.1 / mice (B) was plotted as a point at the interspike interval before the spike on the x-axis and the interspike interval after the spike on the y-axis on a log-log scale. (A) Interspike interval plot from a Ca V 3.1 +/+ neuron clearly showing numerous spikes with a long interval before (>100 ms) and a short interval after ( 4 ms), which can be classified as low-threshold burst firing according to the conventional categorization scheme. (B) Interspike interval plot from a Ca V 3.1 / neuron showing no spikes that can be classified as low-threshold burst firing. Lee et al. 2of4

9 Fig. S2. Cross-correlation and spike-triggered averages showing correlation between cortical spindle events and TC spikes. (A) Cross-correlation calculated for cortical EEG and spike events of TC neurons as source and reference, respectively. (B) Spike-triggered average of cortical EEG signals, presented as mean (black) ± SEM (red). Both correlogram and spike-triggered averages were calculated for the neurons shown in Fig. 1. Fig. S3. TRN neurons discharge single and burst spikes with diverse patterns during spindles from both Ca V 3.1 +/+ and Ca V 3.1 / mice under barbiturate anesthesia. (A) Sample traces showing cortical EEG signals over the ipsilateral frontal lobe (upper trace) and the in vivo extracellular single-unit recording from TRN regions (lower trace) in Ca V 3.1 +/+ mice (Left) and Ca V 3.1 / mice (Right). Both single and burst spikes from TRN neurons are seen in the Ca V 3.1 +/+ and Ca V 3.1 / mice. (B) Perievent histograms showing the number of spikes over 15 cortical spindle events (SP1 SP15) in a 200-ms bin. Numerous spikes were detected during the cortical spindle events, whereas some spikes were observed during the nonspindling periods. Lee et al. 3of4

10 Fig. S4. Firing properties of Ca V 3.1 / TRN neurons in vitro are not altered. Representative traces from whole-cell patch recordings showing the firing pattern of Ca V 3.1 +/+ (A) and Ca V 3.1 / (B) TRN neurons. Low-threshold burst firing was elicited in both Ca V 3.1 +/+ and Ca V 3.1 / TRN neurons by injecting prepulses to hyperpolarize the membrane potentials to 80 mv (Upper). Tonic firing was induced by various depolarizing currents, including 100 pa (Lower). Rhythmic oscillatory burst firing was observed in 11 of 18 Ca V 3.1 +/+ TRN neurons and in 17 of 25 Ca V 3.1 / TRN neurons. Fig. S5. EEG power spectral densities during the NREM sleep state. (A) Power spectral densities from unfiltered EEG data showing power density from 1 to 9 Hz, including delta (1 4 Hz) and spindle (7 14 Hz) waves, decreased in α1g / mice compared with α 1G +/+ mice during NREM sleep. The peak frequency of spindle oscillations during NREM was 9.1 Hz in Ca V 3.1 +/+ mice and 9.4 Hz in Ca V 3.1 / mice. (B) Power spectral densities from a bandpass-filtered (6 15 Hz) EEG revealing no difference in spindle power, especially at 7 10 Hz near the peak frequency of spindles (ns, not significant). Power spectral densities of EEG from NREM sleep states are expressed as mean ± SEM (mv 2 Hz), averaged for WT mice (n = 5) and mutant mice (n = 5) in 0.5-Hz bins. Horizontal black bars with a star in the graph of the NREM power spectrum indicate the frequencies with significant differences between two genotypes (P < 0.05, t test). Blue bars indicate the spindle frequency (7 14 Hz). Lee et al. 4of4