Properties of the slow rhythm in mice under anaesthesia
Under anaesthesia, the EEG of mice exhibited a distinctive slow rhythm as described for other species (Fig. 4a) . The mean frequency of this rhythm was 0.88 ± 0.07 Hz (range, 0.4–2.1 Hz; n = 40). Interestingly, in 68% (n = 27 of 40) of our recordings, this slow rhythm was modulated by an additional slower oscillation occurring at 0.13 ± 0.02 Hz (range, 0.05–0.3 Hz; n = 27; Fig. 4a). As observed previously in ferrets , cats [63, 64] and humans , in local field potential (LFP) recordings obtained with glass electrodes at a cortical depth corresponding to layer 5 (0.6–0.8 mm), the negative component of the slow rhythm, which corresponds to neuronal UP states, was often accompanied by prominent oscillations in the γ frequency (~20–80 Hz) band (88%; n = 35 of 40 recordings; Fig. 4b). The mean peak frequency of γ oscillations was 67.5 ± 2.4 Hz (range, 35–98 Hz; n = 35 recordings).
Recordings of extracellular multiunit activity (MUA) obtained in combination with depth LFP recordings revealed that, as expected, the negative component of the LFP (i.e. UP state) was associated with neuronal firing (mean frequency, 45.6 ± 3.3 Hz; duration of epochs, 552 ± 65 ms; n = 200 epochs from 10 recordings), whilst the positive phase (i.e. DOWN state) was mostly related to neuronal silence (duration of silent epochs, 779 ± 85 ms; n = 200 epochs from 10 recordings; Fig. 4c, d). Examination of the average firing during UP states revealed that action potential output was at a maximum near the start of the UP state, decreased to a stable level from ~200 to 400 ms after UP state commencement and then gradually declined (Fig. 4c, right). The slower modulation of the slow rhythm as noted in the EEG was also apparent in MUA recordings (Fig. 4d).
Extracellular single-unit recordings of neocortical neurons during the slow rhythm reveal two main types of neuronal activity
Extracellular single-unit recordings from layer 5 of the neocortex during the slow rhythm revealed two separate groups of cells that exhibited distinct types of firing during UP states (Fig. 5). In the smaller of these two groups (n = 8 of 31; 26%), UP states were mainly characterised by seemingly sporadic action potential firing that did not show a consistent pattern across consecutive UP states (Fig. 5a, b), with spike frequency being essentially constant for the first ~200–300 ms before gradually declining (mean frequency during first 200 ms, 20.6 ± 2.2 Hz; n = 100 UP states from 5 recordings; Fig. 5a, b, bottom right plots). Interestingly, in this group of cells, UP states sometimes emerged from a baseline of low-frequency (~4–6 Hz) regular firing (Fig. 5b), as has previously been described in the intact brain of both cats  and humans [52, 70], as well as in brain slices from the ferret neocortex [12, 55]. In the other group of neurons (n = 23 of 31; 74%), UP states nearly always commenced with a high-frequency burst of spikes (spikes per burst, 3.4 ± 0.2; peak frequency, 86.4 ± 5.5 Hz; n = 200 bursts from 10 neurons) that was followed by varying degrees of additional firing that ranged from a few isolated (e.g. one to four) spikes (Fig. 5c) to further prominent firing (Fig. 5d). On average, the overall action potential output during UP states in these cells was highest at the start of UP states, diminished to a stable plateau that lasted from around 50 to 250 ms and then gradually decreased thereafter (Fig. 5c, d, bottom right plots). In these cells, low-frequency firing was never observed between UP states.
Intracellular recordings of neocortical neurons during the slow rhythm
To investigate the nature and diversity of UP states in neocortical neurons in more detail, we obtained intracellular recordings, again at a cortical depth corresponding to layer 5 (0.6–0.8 mm). In full agreement with extracellular recordings, we observed two basic types of UP states. In 11 of 27 neurons (41%), UP states consisted of sporadic action potential firing (mean frequency during first 200 ms, 25.3 ± 1.5 Hz; n = 80 UP states from 4 neurons) that did not show a consistent pattern between consecutive UP states (Fig. 6a, b). These cells responded to the injection of brief (600–800 ms) positive current pulses with regular firing (not illustrated) and were, therefore, classified as regular spiking (RS) neurons . The difference in membrane potential between UP and DOWN states in RS cells was 14.7 ± 0.4 mV (n = 10), whereas the average durations of UP and DOWN states were 390 ± 11 ms (n = 80 UP states from 4 recordings) and 575 ± 35 ms (n = 80 DOWN states from 4 recordings), respectively. Steady hyperpolarisation of these neurons to prevent action potential firing during UP states revealed that these events were composed of prominent barrages of synaptic activity (Fig. 6b, bottom). This activity exhibited a mean peak frequency of 36.3 ± 3.4 Hz (range, 22.1–61.8 Hz; n = 11 neurons; Fig. 6b, bottom right). Notably, in RS cells, average subthreshold UP states were of a similar duration and form to ‘full-blown’ UP states (i.e. those involving action potential firing), exhibiting an almost ‘step’-like profile (Fig. 6c, top and middle panels). In addition to conventional synaptic activity, a subset of RS neurons (n = 3 of 11) also exhibited a distinct type of subthreshold event that was reminiscent of so-called spikelets [6, 32, 33, 41, 43]. These events were remarkably well conserved within any given cell and had considerably faster rise and decay times (time to peak, 1.1 ± 0.1 ms; τ decay, 3.0 ± 0.1 ms; n = 30 events from 3 neurons) than those that would be expected for conventional excitatory postsynaptic potentials (EPSPs; Fig. 6d).
In the remainder of neurons (n = 16 of 27; 59%), UP states commenced with a prominent, high-frequency burst of action potentials (spikes per burst, 4.4 ± 0.2; peak frequency, 129.6 ± 13.6 Hz; n = 100 bursts from 5 cells; Fig. 7). As with extracellular recordings, ensuing activity during UP states in these neurons ranged from a few isolated action potentials (Fig. 7a) to considerable additional firing (Fig. 7b). These cells responded to the injection of brief (600–800 ms) positive current pulses with bursts of action potentials (not illustrated) and were, therefore, classified as intrinsically bursting (IB) neurons . The difference in membrane potential between UP and DOWN states in IB cells was 12.2 ± 0.4 mV (n = 100 UP states from 5 cells), whereas the average durations of UP and DOWN states were 343 ± 23 ms (n = 100 UP states from 5 neurons) and 501 ± 20 s (n = 100 DOWN states from 5 neurons), respectively. As observed with extracellular recording, action potential output was on average greatest at the beginning of UP states and then decreased at a variable rate as the UP state progressed (Fig. 8a, b). As with RS cells, steady hyperpolarisation of IB neurons revealed that UP states were also underpinned by barrages of synaptic activity (Fig. 8a, b, d), the time-course and form of which was markedly different to that observed in RS cells but which largely mirrored the action potential output generated by these cells during ‘full-blown’ UP states (Fig. 8a–c). The mean peak frequency of synaptic activity in these cells was 27.9 ± 2.3 Hz (n = 16; Fig. 8d, e).
Intracellular recordings of thalamocortical neurons during the slow rhythm
Previous intracellular recordings obtained from TC neurons in the cat during the slow (<1 Hz) rhythm have shown that UP states are invariably initiated by an LTCP-mediated burst of action potentials [13, 62] that is followed by a depolarised membrane potential but (usually) not by additional firing. An identical pattern of activity can also be observed in recordings from TC neurons in thalamic slice preparations and has been shown to depend on intrinsic mechanisms that mainly involve the interaction of a T-type Ca2+ current and a K+ leak current [18, 20, 21, 31, 72]. Because these ionic currents are a conserved property of TC neurons across several species [16, 19, 27, 36, 40, 49–51, 65], it is likely that a similar form of TC neuron UP state to that observed in cats will also be present in mice. Indeed, extracellular recordings from TC neurons in slices of the mouse thalamus strongly support this assertion . To fully investigate this issue, we obtained intracellular recordings from TC neurons of the ventral posterolateral nucleus in anaesthetised mice (n = 4). Similar to cats, UP states in these cells always commenced with an LTCP (Fig. 9a; see also Fig. 10a, b, bottom panels). Following these LTCPs, the membrane potential of TC neurons remained depolarised for a brief period but additional firing was never observed. Thus, as also largely observed in mouse thalamic slices , LTCP-mediated bursts in these neurons were the only type of action potential output that occurred during the slow rhythm. Overall, the average difference in membrane potential between UP and DOWN states in TC neurons was 6.3 ± 0.3 mV (n = 4), whereas the average durations of UP and DOWN states were 406 ± 36 ms (n = 80 UP states from 4 recordings) and 482 ± 43 ms (n = 80 DOWN states from 4 recording), respectively.
Interestingly, when TC neurons were subjected to a small amount of steady hyperpolarising current, additional LTCPs were often observed during the DOWN state in a manner reminiscent of that previously shown for cat TC neurons, recorded both in vivo [15, 62] and in vitro [31, 33, 73], where it has been referred to as a ‘grouped’ δ oscillation (Fig. 9a, lower traces; cf Figs. 2c and 3a–c, bottom panels). Further steady hyperpolarisation of TC neurons to prevent both action potential firing and LTCP generation revealed a striking pattern of subthreshold activity consisting of barrages of synaptic activity (mean peak frequency, 31.4 ± 2.3 Hz; n = 3) that were preceded by a stereotypical depolarising event that had the appearance of a ‘subthreshold LTCP’ (indicated by arrows in Fig. 9b). Interestingly, at times, these events occurred rhythmically in groups during DOWN states in a manner that was essentially identical to the grouping of ‘full-blown’ LTCPs described above (i.e. Fig. 9a, lower traces). These events appeared to be non-synaptic in nature because their amplitude did not markedly change when cells were depolarised or hyperpolarised (not illustrated). As with neocortical RS cells, 2 out of the 4 recorded TC neurons also exhibited spikelets (time to peak, 1.2 ± 0.4 ms; τ decay, 2.2 ± 0.2 ms; n = 20 events from 2 neurons; Fig. 9d).
Lastly, when TC neurons exhibiting UP/DOWN states were subjected to sufficient steady depolarising current, we observed HT bursting at 5.3 ± 0.4 Hz (n = 3). This HT bursting could occur simultaneously with UP/DOWN states (Fig. 10a, top) or could overwhelm UP/DOWN state generation to fully define the output of these cells (Fig. 10b, top). In all senses, the HT bursting observed in vivo was indistinguishable from that previously noted in vitro (i.e. Fig. 3a–c, upper panels) [30, 31, 33, 34, 41–43].