Neuronal metabolism governs cortical network response state

Abstract

The level of arousal in mammals is correlated with metabolic state and specific patterns of cortical neuronal responsivity. In particular, rhythmic transitions between periods of high activity (up phases) and low activity (down phases) vary between wakefulness and deep sleep/anesthesia. Current opinion about changes in cortical response state between sleep and wakefulness is split between neuronal network-mediated mechanisms and neuronal metabolism-related mechanisms. Here, we demonstrate that slow oscillations in network state are a consequence of interactions between both mechanisms. Specifically, recurrent networks of excitatory neurons, whose membrane potential is partly governed by ATP-modulated potassium (K(ATP)) channels, mediate response-state oscillations via the interaction between excitatory network activity involving slow, kainate receptor-mediated events and the resulting activation of ATP-dependent homeostatic mechanisms. These findings suggest that K(ATP) channels function as an interface between neuronal metabolic state and network responsivity in mammalian cortex.

Fig. 1.

The profile of SWO in the entorhinal cortex. (A) The example of a DC field potential recording from LIII revealed slow oscillations consisting of alternating periods of activity and quiescence. An array of four extracellular electrodes were placed across the layers of the cortex revealed the anatomical location of the slow rhythm as the superficial entorhinal cortex. Note, no slow rhythm was seen in the deep layer V. (Scale bars, 0.2 mV, 20 s.) (B) The characteristics of the slow rhythm depended on glucose concentration. Switching from 10 mM glucose to 2.5 mM caused a significant prolongation of the down phase (*, P < 0.05, n = 5). (C) Membrane potential bistability was specific to certain cell types during the slow rhythm. The DC field potential, recorded concurrently with the LIII pyramidal cell data, is shown for reference. (i) LIII pyramidal cells demonstrated the most robust membrane potential bistability. The example trace shows four consecutive transitions from down to up phase. Note the increase in background EPSP activity before transition to up phase (asterisks) also seen in model (Fig. 4E). The histogram shows the distribution of membrane potentials over a 120-s epoch. (Inset) Neurolucida reconstruction of the recorded neuron. (Scale bars, 50 μm.) Bistability was also evident in LII basket cells (ii) but not LII stellate cells (iii). (Scale bars, 20 mV, 5 s.)

Fig. 2.

Kainate receptor activation was critical for generation of the slow oscillation. (A) Concurrent recordings of the LIII field potential and a LIII pyramidal neuron in the absence (Upper) and presence (Lower) of the specific AMPA receptor blocker SYM2206 (25 μM). There was little change in the slow rhythm in the presence of the drug. Control membrane potential was bistable between −72 ± 2 and −62 ± 3 mV, with sym bistability was between −70 ± 3 and −62 ± 4 mV. (B) In contrast, application of the GluR5 receptor antagonist UBP302 (20 μM) abolished the rhythmic response-state transitions. At the asterisk, a depolarizing current step demonstrates the continued ability to evoke neuronal responses. Control membrane potential was bistable between −70 ± 1 and −60 ± 1 mV, with UBP302 there was no bistability, membrane potential was monostable around −62 ± 3 mV. [Scale bars, 0.5 mV (field), 20 mV (neuron), 5 s.] (C) Spontaneous excitatory events in LIII pyramids changed dramatically with GluR5 blockade. Data show averaged EPSPs in the presence (gray) and absence (black) of UBP302 (20 μM). (Scale bars, 0.5 mV, 250 ms.)

Fig. 3.

Rhythmic response-state transitions depend on K_ATP channel activity. (A) The LIII extracellular recordings and corresponding spectrograms show the effects of blockade (tolbutamide, 0.5 mM) or activation (diazoxide, 0.3 mM) of K_ATP channels on the transition from up to down phases of the oscillation. Each drug was tested on a different set of entrorhinal cortex slices. Note the wide spectral content of population activity within the up phase and the prolongation of the up and down phases with K_ATP blockade and activation, respectively. [Scale bars, 0.5 mV (field), 1 s.] Longer epochs of data, with drug wash-out examples, are illustrated in Fig. 6, which is published as supporting information on the PNAS web site. (B) Intracellular recordings from LIII pyramidal neurones taken concurrently with the data shown in A. Action potentials are truncated to display the predominant membrane potential during the disrupted slow oscillation. Each recording was taken 10–20 min after impalement in each of the three experimental conditions. The red line indicates the resting membrane potential (−75 mV). Shown below are the corresponding membrane potential histograms demonstrating disrupted bistability when K_ATP channels are exogenously activated or blocked. [Scale bars, 20 mV, 1 s.] (C) Distribution of Kir6.2 immunoreactive cells in layer III (Left). (Center and Right) Examples of layer III pyramidal cells positive for Kir6.2. Arrowheads point to labeled dendrites. (Scale bar, 30 μm in Left and 20 μm in Center and Right.) (D) Membrane potential triggered averages of 10 consecutive periods of slow oscillation (ic) in a LIII pyramid in control conditions (black) and in a pyramid recorded with an electrode filled with 50 mM MgATP (red) to block K_ATP channels only in the recorded cell (n = 5 cells). Note the absence of a post-up-phase hyperpolarization with MgATP and a slowed transition to peak up phase. (Scale bars, 5 mV, 1 s.) Corresponding averages (n = 10 periods) of field data (LIII, ec) show no change in population activity when K_ATP channels are blocked in a single neuron. (Scale bars, 0.2 mV, 1 s.)

Fig. 4.

Computational model predicts network bistability generated by interactions between recurrent excitatory network and K_ATP channel activity alone. (A) Raster plot of spiking in a network of 100 neurons showing slow rhythmic transitions between up and down phases synchronously throughout the network. Each neuron was coupled into an excitatory network by EPSPs with decay time constant of 100 ms, and each neuron had gK_ATP. Each dot corresponds to one action potential. (B) Concurrent plot of [Na⁺]_i showing gradual elevation in intracellular sodium ion concentration during the up phase of the rhythm. (C) As [Na⁺]_i increases [ATP]_i falls as the neurons attempt to restore electrochemical equilibrium via Na⁺/K⁺ATPase activity. Note changes in [Na⁺]_i and [ATP]_i are in antiphase. (D) The dynamics of [ATP]_i lead to a concurrent modulation in gK_ATP, which peaks on transition from the up to down phase. This, coupled with the resulting modulation of recurrent network activity (shown here as the total excitatory synaptic input current to one neuron, ΣI_syn E), leads to a temporal pattern of membrane potential change associated with the slow rhythm in entorhinal cortical LIII pyramids. Note the model also reproduces the increase in background EPSP generation before transition to up phase also seen in experiment (asterisks, see Fig. 1Ci). Further examples of the modle’s behavior with manipulation of gK_ATP are illustrated in Fig. 5.