|Home | About | Journals | Submit | Contact Us | Français|
The slow (<1 Hz) rhythm is an EEG hallmark of resting sleep. In thalamocortical (TC) neurons this rhythm correlates with a slow (<1 Hz) oscillation comprising recurring UP and DOWN membrane potential states. Recently, we showed that metabotropic glutamate receptor (mGluR) activation brings about an intrinsic slow oscillation in thalamocortical (TC) neurons of the cat dorsal lateral geniculate nucleus (LGN) in vitro which is identical to that observed in vivo. The aim of this study was to further assess the properties of this oscillation and compare them with those observed in TC neurons of three other thalamic nuclei in the cat (ventrobasal complex, VB; medial geniculate body, MGB; ventral lateral nucleus, VL) and two thalamic nuclei in rats and mice (LGN and VB). Slow oscillations were evident in all of these additional structures and shared several basic properties including, i) the stereotypical, rhythmic alternation between distinct UP and DOWN states with the UP state always commencing with a low-threshold Ca2+ potential (LTCP), and ii) an inverse relationship between frequency and injected current so that slow oscillations always increase in frequency with hyperpolarization, often culminating in delta (δ) activity at ~1-4 Hz. However, beyond these common properties there were important differences in expression between different nuclei. Most notably, 44% of slow oscillations in the cat LGN possessed UP states that comprised sustained tonic firing and/or high-threshold (HT) bursting. In contrast, slow oscillations in cat VB, MGB and VL TC neurons exhibited such UP states in only 16%, 11% and 10% of cases, respectively, whereas slow oscillations in the LGN and VB of rats and mice did so in <12% of cases. Thus, the slow oscillation is a common feature of TC neurons that displays clear species- and nuclei-related differences. The potential functional significance of these results is discussed.
During resting sleep the human EEG exhibits a distinctive slow (<1 Hz) rhythm (Amzica and Steriade, 1997; see also Achermann and Borbely, 1997). This slow rhythm is also evident during natural sleep in cats (Amzica and Steriade, 1998, Steriade et al., 2001, Timofeev et al., 2001), rats (Pinault et al., 2001) and mice (Lee et al., 2004), and is prominent in the EEG of cats maintained under either urethane or ketamine/xylazine anaesthesia (Steriade et al., 1993a,b,c). The presence of the slow rhythm in cats during general anaesthesia has greatly facilitated the in vivo investigation of its cellular components. Specifically, studies exploiting this fact have revealed that in both cortical and thalamic neurons, the slow rhythm corresponds to a slow (<1 Hz) membrane potential oscillation consisting of distinct UP and DOWN states (Steriade et al., 1993a,b,c). In cortical neurons, these UP and DOWN states seem to be largely sculpted by alternating periods of intense synaptic barrages and disfacilitation, respectively (Steriade et al., 1993a,b; Sanchez-Vives and McCormick, 2000). In thalamic neurons, however, UP and DOWN states are more stereotypical, do not seem to involve such a great deal of synaptic activity and therefore appear to be formed more by intrinsic events (Steriade, 1993; Steriade et al., 1993c, Contreras and Steriade, 1995, Steriade et al., 1996).
An important observation in vivo is that the slow oscillation in thalamocortical (TC) neurons is abolished following decortication (Timofeev and Steriade, 1996). This, together with the findings that the slow oscillation in the neocortex is resistant to widespread thalamic lesions (Steriade et al., 1993b) and that similar activity is present in isolated cortical slices (Sanchez-Vives and McCormick, 2000), has led to the suggestion that the slow rhythm is predominantly a cortically-generated phenomenon. However, we have recently shown that following sustained activation of the metabotropic glutamate receptor (mGluR) that is postsynaptic to corticothalamic fibres, i.e. mGluR1a (Godwin et al., 1996; von Krosigk et al., 1999, Turner and Salt, 2000), TC neurons of the isolated cat dorsal lateral geniculate nucleus (LGN) maintained in vitro can exhibit an intrinsic slow oscillation (Hughes et al., 2002b, Hughes et al., 2004) which is identical to that observed during the slow rhythm in the intact brain (Steriade et al., 1993c, Contreras and Steriade, 1995, Steriade et al., 1996; Steriade, 1993). This illustrated that, at least in the LGN, TC neurons do not require rhythmic cortical input in order to generate the slow oscillation, and suggested that TC neurons might play a more active role in shaping the slow rhythm than had previously been recognised.
In this study, we further investigated the properties of the slow oscillation in LGN TC neurons and tested whether a simple ‘reactivation’ of modulatory cortical input in vitro is also able to induce equivalent activity in TC neurons from three additional thalamic relay nuclei in the cat (ventrobasal complex, VB, i.e. somatosensory thalamus; medial geniculate body, MGB, i.e. auditory thalamus; ventral lateral nucleus, VL, i.e. motor thalamus) and from two different sensory thalamic nuclei of rats and mice (LGN and VB). We find that mGluR1a-activation can bring about a stereotypical slow oscillation in TC neurons of all these additional structures with basic properties that are shared by all cells. Specifically, in each nuclei examined the slow oscillation, i) always comprised the rhythmic alternation between discrete UP and DOWN membrane potential states with the UP state being initiated by a low-threshold Ca2+ potential (LTCP), and ii) always increased in frequency with increasing hyperpolarization, often eventually leading to δ oscillations at ~1-4 Hz (McCormick and Pape, 1990, Leresche et al., 1991). However, outside these basic common components the slow oscillation was expressed with important differences between different nuclei. In particular, the cat LGN was found to be exceptional in that a large proportion (44%) of slow oscillations in this structure exhibit ‘active’ UP states comprising sustained action potential output whereas in all other cat thalamic nuclei and in the LGN and VB of both rats and mice this proportion was considerably lower. We suggest that this may reflect a rich visual nature of sleep in higher mammals and might also provide important clues for understanding the functional role of the slow oscillation in TC neurons. Some of these results have been presented in abstract form (Blethyn et al., 2002).
Procedures involving experimental animals were carried out in accordance with local ethical committee guidelines and the U.K. Animals (Scientific Procedure) Act, 1986. All efforts were made to minimize the suffering and number of animals used in each experiment.
Young adult cats (1-1.5 kg) were deeply anaesthetized with a mixture of O2 and NO2 (2:1) and 2.5% halothane, a wide craniotomy performed and the brain removed. Sagittal slices (450-500 μm) of the thalamus containing either the LGN, VB, MGB or VL were prepared and maintained as described previously (Hughes et al., 2002b, Hughes et al., 2004). Male Wistar rats (150-200g) and adult C57BL/6 mice were deeply anaesthestized (1·5 % halothane) and killed by decapitation. In both cases, coronal slices (400-450 μm) containing either the LGN or VB were prepared and maintained as in previous studies (Hughes et al., 1999). For recording, slices were perfused with a warmed (35±1 °C), continuously oxygenated (95% O2, 5% CO2), artificial cerebrospinal fluid (ACSF) containing (mM): NaCl (134); KCl (2); KH2PO4 (1.25); MgSO4 (1); CaCl2 (2); NaHCO3 (16); glucose (10). All drugs were dissolved directly in ACSF. The following drugs were obtained from Tocris-Cookson (UK): DL-2-amino-5-phosphonovaleric acid (DL-AP5), (+)-2-methyl-4-carboxyphenylglycine (LY367385), P-(3-aminopropyl)-P-diethoxymethyl-phosphinic acid (CGP 56999A), 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX), 1,2,3,4-tetrahydro-6-nitro-2,3-dioxobenzo[f] quinoxaline-7-sulfonamide (NBQX),. 6-imino-3-(4-methoxyphenyl)-1(6H)-pyridazinebutanoic acid (SR95531), (+/−)-1-aminocyclopentane-trans-1,3-dicarboxylic acid (trans-ACPD), tetrodotoxin (TTX). Bicuculline methiodide (BMI) was obtained from Sigma (UK).
Extracellular recordings were performed using glass pipettes filled with 0.5 M NaCl (resistance: 1-5 MΩ) connected to a Neurolog 104 differential amplifier (Digitimer Ltd., Welwyn Garden City, UK) and bandpass filtered at 0.1-20 kHz. Intracellular recordings, using the current clamp technique, were performed with standard-wall glass microelectrodes filled with 1M potassium acetate (resistance: 80-120 MΩ), and in some cases 2% biocytin or neurobiotin, and connected to an Axoclamp-2A amplifier (Axon Instruments, Foster City, USA) operating in bridge mode. Impaled cells were identified as TC neurons using established electrophysiological and morphological criteria (Williams et al., 1996, Hughes et al., 2002a). In slices where neurons had been filled with biocytin, visualization of the dye was performed as described previously (Hughes et al., 2002a). All neurons recorded in the VB (cat, rat and mouse) were from the ventral posterior lateral nucleus (VPL) whereas all cells assessed in the MGB (cat only) were from the dorsal subdivision. A small proportion (<20%) of the cat LGN TC neurons included for analysis in this study were also included in a previous paper (Hughes et al., 2002b). Membrane potential and injected current records were stored on a Biologic DAT recorder (IntraCel, Royston, UK) and later analyzed using Clampfit (Axon Instruments). The apparent input resistance (RN) was estimated from voltage responses evoked at −60 mV by small (20-50 pA) hyperpolarizing current steps. Statistical significance was assessed using Student’s t-test. All quantitative results in the text and figures are expressed as mean±SEM.
For intracellular recordings we applied 100 μM trans-ACPD in all cases and then varied the injected d.c. current appropriately to examine the properties of the slow (<1 Hz) oscillation, if one was present (Supp. Fig. 1A). The durations of the UP and DOWN states were taken as the time spent above and below −60 mV, respectively (Supp. Fig. 1A). In all cases, the frequency of the oscillation was taken as the reciprocal of the time between the start of successive UP states (Supp. Fig. 1A). The peak membrane potential of the UP state was computed after discarding the first 200 ms of this event. This was done to prevent a distortion of this value by the initial LTCP-mediated burst, which was always completed within this timescale. The ‘active’ phase of the UP state was taken as the period of firing that occurred after the initial LTCP burst. Thus, the peak firing rate for ‘active’ UP states, which was computed as the reciprocal of the smallest inter-spike interval evident during the ‘active’ phase, also did not take into account this burst. A ‘grouped’ δ sequence was defined as a series of rhythmic LTCP-mediated bursts that could emerge during the slow oscillation DOWN state (Supp. Fig. 1A). To assess the properties of the slow (<1 Hz) oscillation with extracellular single unit recordings, we first induced continuous tonic firing or high-threshold (HT) bursting by applying 100-150 μM trans-ACPD (non-specific Group I/II mGluR agonist) (Hughes et al., 2004). This was then gradually washed off so that a slow oscillation could become apparent (Supp. Fig. 1B). In every neuron, the slow oscillation exhibited its lowest frequency at the highest concentration of trans-ACPD for which it was present and then progressively increased in frequency as trans-ACPD washed out (Supp. Fig. 1B). Upon complete washout of trans-ACPD, TC neurons either became quiescent or exhibited rhythmic LTCP-mediated bursting at δ frequencies (~1-4 Hz) (McCormick and Pape, 1990, Leresche et al., 1991) (Supp. Fig. 1B). This approach allowed us to examine the full behavioral repertoire and range of frequencies of the slow oscillation in each recorded cell. The majority of both extra- and intracellular recordings were performed in the presence of CNQX (or NBQX), DL-AP5, BMI (or SR95531) and CGP 56999A to block fast ionotropic glutamate, GABAA and GABABB receptors.
When assessed with intracellular recordings in control conditions, LGN TC neurons exhibited an apparent input resistance (RN) of 187.0±10.1 MΩ and a resting membrane potential (Vm) of −65.0±1.1 mV (n=60). None of these neurons exhibited a slow oscillation in control conditions although 20 displayed a spontaneous δ oscillation (McCormick and Pape, 1990, Leresche et al., 1991) (Fig. 1A1). As described previously (Hughes et al., 2002b, 2004), application of 100 μM trans-ACPD expanded the oscillatory repertoire of these cells so that 60% (36 of 60) became able to generate a slow oscillation (Fig. 1A2). Trans-ACPD also caused a 142±61% increase in apparent input resistance (RN) (Fig. 1B) and a 13±0.1 mV depolarization (n=31) (see McCormick and von Krosigk, 1992, von Krosigk et al., 1999, Turner and Salt, 2000). Expression of the slow oscillation was related to a significantly higher value of RN (slow-oscillating neurons: 614.2±92.6 MΩ, n=34; non-oscillating neurons: 147.9±14.8 MΩ, n=19; p<0.001) (Table 1) and the appearance of a distinctly non-linear voltage response to small negative current steps (Fig. 1B) (Williams et al., 1997, Hughes et al., 1999, Hughes et al., 2002b).
No TC neurons in the cat VB, MGB and VL exhibited a slow oscillation in control conditions. However, 54% (14 of 26), 56% (15 of 27) and 73% (11 of 15), respectively, did so following application of 100 μM trans-ACPD (Figs. 2A2, ,33 and and4).4). Trans-ACPD also led to a significant increase in RN (VB: 195.8±30.3 to 474.6±159.5 MΩ, n=26; MGB: 216.2±25.2 to 546.3±119.8 MΩ, n=27; VL: 157.5±27.8 to 552.8±205.5 MΩ, n=15) and an appreciable membrane potential depolarization (VB: −68.5±1.7 to −56.5±1.5 mV, n=21; MGB: −64.4±0.9 to −55.3±0.7 mV, n=20; VL: −68.3±2.3 to −55.7±1.6 mV, n=13) (Fig. 2). In accordance with results in the LGN, the slow oscillation in VB, MGB and VL TC neurons was associated with a significantly larger RN compared to non-oscillating cells (p<0.01 for each case, Table 1) and the appearance of a conspicuously non-linear voltage response to small hyperpolarizing current pulses (Fig. 2B). Also, as shown previously for the LGN (Hughes et al., 2002b, Hughes et al., 2004), the action of trans-ACPD in bringing about a slow oscillation in VB, MGB and VL TC neurons was via activation of mGluR1a since subsequent application of the mGluR1a-specific antagonist, LY367385 (300 μM), reversibly blocked the ability of these cells to generate this activity (n=5) (Figs. 2A3) (see also Turner and Salt, 2000).
In all types of cat TC neurons, the slow oscillation was extremely robust and could often remained unaltered and uninterrupted for long recording periods (up to 6 hours). No clear morphological differences were detected between oscillating and non-oscillating neurons in the LGN, VB, MGB or VL.
In all cat thalamic relay nuclei examined, the slow oscillation shared a number of basic common properties (Fig. 3). These included, i) the stereotypical alternation between UP and DOWN membrane potential states that were separated by ~20 mV, ii) the commencement of each UP state with an LTCP-mediated burst (Fig. 3A1), iii) a transition from the UP to DOWN state that was marked by a clear membrane potential inflection (Fig. 3A1), and iv) a hyperpolarization-activated mixed cation current (Ih)-dependent slow depolarization during the DOWN state (Fig. 3A1) (Hughes et al., 2002b). Also common to all types of cat TC neurons was the finding that hyperpolarization (achieved through injecting negative d.c. current) always increased the frequency of the slow oscillation until it was often eventually transformed into a continuous δ oscillation (Fig. 3; see also Figs. Figs.1,1, ,2,2, ,6,6, ,77 and and8).8). This increase in frequency overwhelmingly resulted from a shortening of the UP state (LGN: from 11.3±2.4 to 1.4±0.2 s, n=26 neurons; VB: from 14.1±5.6 to 1.5±0.4 s, n=8 neurons; MGB: from 17.1±3.7 to 2.3±0.6 s, n=7 neurons; VL: from 48.5±14.2 to 3.9±1.3 s, n=6 neurons; p<0.05 for each nucleus) with the duration of the DOWN state remaining essentially unchanged (LGN range: 1.0±0.2 to 1.4±0.2 s, n=26 neurons; VB range: 1.0±0.2 to 1.2±0.2 s, n=8 neurons; MGB range: 0.9±0.1 to 1.4±0.2 s, n=7 neurons; VL range: 1.2±0.2 to 1.3±0.3 s, n=6 neurons; p>0.1 for each nucleus) (Fig 3; ee also Fig. 6A3 and 6B3 and Table 1). The mean minimum frequencies of the slow oscillation recorded intracellularly in LGN, VB, MGB and VL TC neurons were 0.3±0.04 Hz (n=34), 0.12±0.04 Hz (n=11), 0.06±0.03 Hz (n=14) and 0.04±0.02 Hz (n=8) respectively.
Slow oscillations observed with intracellular recordings in the LGN occurred with a remarkably large variety of manifestations. Most neurons displayed an oscillation possessing an UP state that was always ‘quiescent’, i.e. that did not exhibit further firing, following the initial LTCP burst (‘quiescent’ slow oscillation) (24 of 36; 67%) (Figs. 4A, 4B, ,6A6A and and8A).8A). However, in the remainder of cells (12 of 36; 33%), the UP state could be characterised by the sustained firing of either single action potentials (i.e. tonic firing) (n=6) (Figs. 4C1 and 4D1), a mixture of single action potentials and high-threshold (HT) bursts (see Hughes et al. 2002b, 2004) (n=3) (Figs. 4C2 and and7A)7A) or HT bursts only (n=3) (Figs. 4C3) (‘active’ slow oscillation). Both the ‘quiescent’ (Fig. 4B) and ‘active’ (Fig. 4D) forms of the slow oscillation could exhibit ‘grouping’ of the δ oscillation (21%, n=5 of 24, and 33%, n=4 of 12, respectively) (Steriade et al., 1993c; Hughes et al., 2002b) whereby the DOWN state could be replaced by transient sequences of rhythmic LTCP-mediated bursts at ~1-4 Hz (‘grouped’ δ sequences) (Figs. 4B, 4D and and8A8A).
In contrast to the cat LGN, the expression of the slow oscillation in the cat VB, MGB and VL was considerably less heterogeneous (Figs. 3B-D and and5).5). The vast majority of intracellularly recorded VB TC neurons exhibited slow oscillations that were of the ‘quiescent’ type (93%; 13 of 14) (Figs. (Figs.3B3B and and5A)5A) with only one neuron able to generate an ‘active’ slow oscillation (comprising tonic firing only, see below and Fig. 7B). For the VB, ‘grouped’ δ sequences were only apparent in 4 neurons that displayed ‘quiescent’ slow oscillations (Fig. 5A2). Intracellularly recorded slow oscillations in MGB and VL TC neurons were always of the ‘quiescent’ type (Figs. 3C, 3D, 5B, 5C, ,6B6B and and8B)8B) with δ ‘grouping’ being present in 8 (of 15) and 2 (of 11) cells, respectively (Figs. 5B2, 5C2, ,8B8B).
In LGN TC neurons displaying a ‘quiescent’ slow oscillation, the peak membrane potential reached during the UP state increased with depolarization (Fig. 6A). Interestingly, at its maximum (−48.3±0.7 mV; n=23) this value could exceed some of the stable membrane potentials reached by the same neurons when they were subjected to more steady depolarizing current (Fig. 6A1 and 6A2) and, specifically, was significantly larger than the average resting membrane potentials of those cells that displayed a stable membrane potential in the absence of steady current (i.e. rather than a slow oscillation) (p<0.01) (Fig. and Table 1; see also Fig. 3). The peak membrane potentials reached during the UP state of ‘quiescent’ slow oscillations in VB, MGB and VL neurons also increased as neurons were depolarized and at their maximums were −56.1±1.8 mV (n=11), −54.7±1.4 mV (n=14) and −53.3±1.6 mV (n=8), respectively (Fig. 6B and Table 1). Again, these values could be greater than some of the stable membrane potentials exhibited by the same neurons in response to additional steady depolarizing current (Fig. 6B1 and 6B2), were comparable with the average resting membrane potentials of those cells that displayed a steady membrane potential in the absence of d.c. current (p>0.1) (Table 1; see also Fig. 3), but were all significantly lower than the equivalent value for LGN TC neurons (p<0.01 for each case).
For ‘active’ slow oscillations, an increase in the ‘excitability’ level of the UP state also occurred with depolarization with this being reflected as a progressive increase in both the frequency and the extent of firing (Fig. 7). In fact, the peak firing frequency during the UP state commonly reached values that were comparable with or greater than those observed in neurons displaying continuous, spontaneous action potential output in the absence of any steady d.c. current (i.e. when neurons are subject to less steady hyperpolarizing d.c. current than during the slow oscillation) (LGN: maximum peak frequency of tonic firing during UP state: 11.6±3.4 Hz; spontaneous tonic firing at 0 pA d.c. current: 10.2±1.2 Hz, n=7; p=0.61. maximum inter-burst frequency of HT bursting during UP state: 6.5±1.2 Hz; spontaneous HT bursting at 0 pA d.c. current: 6.7±1.1 Hz, n=3; p=0.91) (Fig. 7A and Table 1) (VB: maximum peak frequency of tonic firing during UP state: 23.6 Hz, spontaneous tonic firing at 0 pA d.c current: 12.6 Hz) (Fig. 7B and Table 1).
In all types of cat TC neurons exhibiting ‘grouped’ δ activity, this ‘grouping’ occurred as a transition stage between a ‘pure’ slow oscillation and continuous δ (~1-4 Hz) activity (Fig. 8B). Thus, the overall increase in slow oscillation frequency that occurred as these cells were hyperpolarized was also affected by the dynamics of the ‘grouped’ δ sequences. Specifically, the number of LTCP bursts per sequence, and consequently the sequence duration (Fig. 8A2 and 8B2), increased as neurons were hyperpolarized whereas the inter-burst frequency remained constant (Fig. 8A3 and 8B3). ‘Grouped’ δ sequences were generally least extensive (i.e. of shortest duration) in the LGN (LGN: max. bursts per sequence: 5.5±1.1; mean inter-burst frequency: 1.7±0.2 Hz, n=9; VB: max. bursts per sequence: 7.2±1.0; mean inter-burst frequency: 1.6±0.1 Hz, n=4; MGB: max. bursts per sequence: 10.8±5.3; mean inter-burst frequency: 1.2±0.1 Hz, n=7; VL: max. bursts per sequence: 15.4±3.2; mean inter-burst frequency: 1.4±0.1 Hz, n=2) (Table 1).
The essential pattern of findings obtained with intracellular recordings in different cat thalamic nuclei was also evident with extracellular single unit recordings (see Supp. Information and Supp. Table 1). In particular, the most widespread array of slow oscillations was clearly observed in the LGN (Fig. 11 and Supp. Fig. 2) whereas ‘grouped’ δ sequences were more extensive (i.e. of longer duration) in other nuclei (Supp. Fig. 3 and Supp. Table 1). However, one clear difference between intra- and extracellular recordings was that during the latter, we encountered a larger number of neurons that exhibited ‘active’ slow oscillations (Fig. 11 and Supp. Fig. 3). Indeed, whereas during intracellular recordings we did not observe any ‘active’ slow oscillations in the cat MGB or VL and only one in the VB, during extracellular recordings, 19% (8 of 43), 14% (6 of 42) and 14% (4 of 28) of VB, MGB and VL neurons, respectively, showed an ‘active’ oscillation. In addition, in the VB, one of these ‘active’ oscillations involved HT bursts, a phenomenon that was not observed with intracellular recordings (n=1 of 8) (Supp. Fig. 3A3). For the LGN, an ‘active’ slow oscillation was present in 56% (19 of 34) of extracellular recordings. Thus, when intra- and extracellular recordings were considered together, ‘active’ slow oscillations accounted for 44% 16%, 11% and 10% of cases in the LGN, VB, MGB and VL, respectively. A full description of the results obtained with extracellular recordings in different cat thalamic nuclei is given in the supplementary information and in Supplementary Table 1 and Fig. 11.
Rat LGN TC neurons examined with intracellular recordings in control exhibited an RN of 198.0±18.0 MΩ and a Vm of −64.8±0.9 mV (n=10) (Fig. 9A). None of these cells displayed a slow oscillation in control conditions. However, application of 100 μM trans-ACPD brought about a slow oscillation (with basic components that were equivalent to those observed in cat thalamic nuclei, see above) in 40% (4 of 10) of cases (Fig 9B1 and Supp. Fig. 4A). Intracellularly recorded slow oscillations in rat TC neurons were always of the ‘quiescent’ type. The presence of the slow oscillation was linked with a larger value of RN (slow-oscillating neurons: 1040.0±272.6 MΩ, n=4; non-oscillating neurons: 325.2±68.0 MΩ, n=6; p<0.01) and the advent of an overtly non-linear voltage response to the challenge of small negative current steps (Fig. 9A2). Also in direct correspondence with cat TC neurons, the frequency of the slow oscillation in the rat increased with hyperpolarization until a continuous δ oscillation became apparent (Fig 9B1 and 9B3). The mean minimum frequency of the slow oscillation recorded intracellularly in rat LGN TC neurons was 0.25±0.05 Hz (n=4). The peak membrane potential of the UP state of the slow oscillation in the rat increased as neurons were depolarized and at its maximum was −57.6±0.8 mV (n=4) (Fig. 9B2). This value was comparable with the resting membrane potential of these cells (−57.4±0.7 mV; n=4; p=0.89) but was less than the equivalent values obtained from all types of cat TC neurons (see above and Table 1).
Slow oscillations were also clearly evident in extracellular single unit recordings from the rat LGN and VB (Fig. 9C and Supp. Fig. 4B). Interestingly, as in the cat, these recordings revealed the presence of a small proportion of ‘active’ slow oscillations so that 14% (2 of 14) and 10% (1 of 10) of slow oscillations observed with extracellular recordings in the LGN and VB, respectively, were of this type (Fig. 9C1 and Supp. Fig. 4B2). Furthermore, in one of the LGN TC neurons, the ‘active’ oscillation involved HT bursts (Fig. 9C1), a phenomenon which was not evident with intracellular recordings. When intra- and extracellular recordings were considered together, ‘active’ slow oscillations accounted for 8% of cases in the rat LGN. A more detailed description of the results obtained with extracellular recordings in the rat LGN and VB is given in the Supplementary Information and Supplementary Table 1.
In control conditions, intracellularly recorded TC neurons in the mouse LGN and VB exhibited an RN of 140.2±22.5 MΩ and a Vm of −60.2±1.2 mV (n=10). None of these neurons exhibited a slow oscillation in control conditions. Application of trans-ACPD (100 μM) led to a small change in Vm (3.5±0.8 mV; n=10) and increase in RN (44.2±12.1%; n=10) but did not bring about a slow oscillation in any cells (Fig. 10A). In addition, trans-ACPD application was not associated with the appearance of a non-linear charging pattern in response to small negative current steps in any mouse TC neuron (Fig. 10A2).
The slow oscillation is present in mouse TC neurons, however, because extracellular single unit recordings obtained from the mouse LGN and VB were characterised by this activity in 78% (28 of 36) and 27% (12 of 44) of cases respectively (Figs. 10B and 10C and Supp. Fig. 5). For LGN recordings, 89% (25 of 28) of slow oscillations were ‘quiescent’ whilst 11% (3 of 28; 1 involving HT bursts) were ‘active’ (Figs. 10B and 10C). The equivalent values for the mouse VB were 92% (11 of 12) and 8% (1 of 12), respectively. ‘Active’ firing in mouse TC neurons exhibited properties that were essentially equivalent to those observed in cat and rat TC neurons (Fig. 10C and Supp. Information and Supp. Table 1). We observed δ ‘grouping’ in one ‘quiescent’ slow oscillation in the mouse VB (Supp. Fig. 5A). Further details regarding the properties and prevalence of the slow oscillation in different rat and mouse thalamic nuclei are given in Table 1, Supplementary Table 1 and Fig. 11.
The main findings of this study are, i) the stereotypical slow (<1 Hz) oscillation is present in TC neurons of the cat LGN, VB, MGB and VL, and of the rat and mouse LGN and VB; ii) in all cases, the slow oscillation increases in frequency as the depolarizing drive to TC neurons is reduced (either through the injection of hyperpolarizing d.c. current or a decrease in the intensity of mGluR1a-activation); iii) this increase in frequency overwhelmingly occurs through a shortening of the UP state; iv) in 44% of cases, the UP state of the slow oscillation in LGN TC neurons is characterised by the sustained firing of either single action potentials and/or HT bursts (‘active’ slow oscillation). This occurs considerably less often for slow oscillations in other thalamic nuclei of the cat and in each nucleus examined in the rat and mouse; and v) the peak membrane potential reached (for ‘quiescent’ oscillations) or the peak rate of firing observed (for ‘active’ oscillations) during the UP state can exceed levels exhibited when neurons are depolarized above the range where slow oscillations occur, with this being most pronounced in the LGN.
Although we observed a variety of manifestations of the slow oscillation during this study a number of fundamental elements were shared by all neurons regardless of species or nuclei. First, in all neurons the slow oscillation was expressed as the rhythmic alternation between UP and DOWN membrane potential states. Second, the slow oscillation UP state always started with an LTCP-mediated burst. Third, the UP state was always terminated by a marked, stereotypical inflection point before transition to the DOWN state. Fourth, during the DOWN state, an overt slow depolarization (due to the mixed cation current, Ih; McCormick and Pape, 1990, Soltesz et al., 1991) was always evident before the reappearance of the UP state and the repetition of the oscillation cycle. Importantly, and as discussed previously (Hughes et al., 2002b), these common slow oscillation components are clearly evident during slow oscillations observed in TC neurons in vivo (Steriade, 1993; Steriade et al., 1993c, Contreras and Steriade, 1995, Steriade et al., 1996).
Perhaps one of the most striking observations in this study was that for any given level of injected d.c. current these basic elements of the slow oscillation were both extremely robust (i.e. they could remain unaltered and uninterrupted for several hours) and remarkably well conserved from cycle to cycle. We suggest that this combination of robustness and reproducibility renders the slow oscillation (<1 Hz) ideally suited as a reliable pacemaker of rhythmic cortical activity. Thus, although it is well established that the cortex can generate a slow oscillation in the absence of thalamic input (Steriade et al., 1993b; see also Sanchez-Vives and McCormick, 2000), and whilst we therefore do not claim that the thalamus is solely responsible for generating the slow (<1 Hz) rhythm in the intact brain, this study lends further support to the argument that the TC neuron slow oscillation, at the very least, plays an active role in shaping the overall slow rhythm. The precise way in which the cortex and thalamus interact in this manner, however, remains to be fully understood.
We observed a number of important nucleus- and species-specific differences in the slow oscillation. The most prominent distinction was that when extra- and intracellular recordings were considered together, 44% of cat LGN TC neurons exhibited a slow oscillation which could display ‘active’ UP states whereas the proportion of ‘active’ slow oscillations present in all other types of TC neurons was at most 16% (see Fig. 11). Furthermore, outside the LGN, an ‘active’ slow oscillation was only evident with intracellular recordings in one cat VB TC neuron. Since it is well known that in comparison to other cat thalamic nuclei the LGN contains TC neurons that show a diverse range of morphologies and soma sizes (Friedlander et al., 1979), it is a distinct possibility that differences in the expression of the slow oscillation are related to the morphological complexity and heterogeneity of cells in this structure. Unfortunately, we did not collect sufficient morphological data in this study to definitively address this issue. However, our preliminary data suggests that ‘active’ slow oscillations may be associated with a smaller soma size (see Fig. 4) which is clearly consistent with the general finding that ‘active’ oscillations are more prevalent in extracellular recordings compared to intracellular recordings (see Fig. 11).
Other possible reasons for the preponderance of ‘active’ UP states in LGN TC neurons might be related to the expression of the ionic currents that shape the slow oscillation. We have previously shown that the slow oscillation in TC neurons arises when mGluR activation sufficiently reduces the leak K+ current, ILeak, below a threshold where it can then interact with the ‘window’ component of the T-type Ca2+ current, ITwindow, to produce a form of intrinsic membrane potential bistability (Williams et al., 1997; Tóth et al., 1998; Hughes et al., 2002b). In simplified terms, this bistability dictates that the slow oscillation UP state corresponds to ITwindow being ‘switched on’ whereas the DOWN state is related to it being ’switched off’’ (see Figs. Figs.11 and and44 in Crunelli et al., 2005). The UP state is also profoundly enhanced by a Ca2+-activated non-selective current (ICAN) (for a detailed explanation of the ionic mechanisms underlying the slow oscillation see Crunelli et al., 2005 and Crunelli et al., 2006). Thus, we expect that the balance between the inward currents that contribute to the UP state (i.e. ITwindow and ICAN) and the outwardly rectifying K+ channels that are present in TC neurons (Huguenard et al., 1991, Huguenard and Prince, 1991, Turner et al., 1997) is especially crucial in determining the precise manifestation of the slow oscillation. As such, LGN neurons which exhibit ‘active’ UP states might possess either an especially large ICAN and/or IT, or express smaller outwardly rectifying K+ currents. Interestingly, it has been previously shown that TC neurons in the cat VB exhibit particularly strong outward rectification (Turner et al., 1997) compared to LGN TC neurons (Pirchio et al., 1997), which is coherent with the lower membrane potential reached during the UP state in these cells compared to those in the LGN and, therefore, the predominance of ‘quiescent’ slow oscillations in the VB.
When considered at the functional level, the large percentage of TC neurons in the cat LGN that exhibit ‘active’ slow oscillations is clearly consistent with the idea that during sleep, the visual systems of higher mammals exhibit an elevated level of activity compared to other modalities (see Symons, 1993). However, regardless of the precise manifestation of the slow oscillation, the UP state generally represents a substantially excited state since it exhibits either a peak membrane potential or firing rate not normally observed unless neurons are depolarized above the range where the slow oscillation occurs. This suggests that even ‘quiescent’ UP states in TC neurons are highly conducive to information processing (see Steriade, 2001, Steriade et al., 2001). In connection with this, we suggest that the presence of an LTCP-mediated burst at the commencement of each UP state may be a mechanism whereby the thalamus is sending a specific priming signal to the cortex that the ‘activated’ or ‘processing’ phase of the slow oscillation is about to begin (Steriade, 2001, Steriade et al., 2001). Of course, the idea that the LTCP-mediated bursts of the slow oscillation signal the upcoming transmission of high-priority information is essentially equivalent to that which has been suggested for the role of LTCP-mediated bursts in TC neurons during wakefulness (Sherman, 2001a, Sherman, 2001b). Indeed, we have recently postulated that similar cellular mechanisms may underlie both the slow oscillation and the ability to generate LTCP-mediated bursts from depolarized membrane potentials during the wake state (Crunelli et al., 2005).
We wish to thank Mr. T.M. Gould for technical assistance. This work was supported by the Wellcome Trust (grant 71436). Additional information regarding this and other published work from the Crunelli lab is available at http://www.thalamus.org.uk.