Astrocyte regulation of sleep circuits: experimental and modeling perspectives

Abstract

Integrated within neural circuits, astrocytes have recently been shown to modulate brain rhythms thought to mediate sleep function. Experimental evidence suggests that local impact of astrocytes on single synapses translates into global modulation of neuronal networks and behavior. We discuss these findings in the context of current conceptual models of sleep generation and function, each of which have historically focused on neural mechanisms. We highlight the implications and the challenges introduced by these results from a conceptual and computational perspective. We further provide modeling directions on how these data might extend our knowledge of astrocytic properties and sleep function. Given our evolving understanding of how local cellular activities during sleep lead to functional outcomes for the brain, further mechanistic and theoretical understanding of astrocytic contribution to these dynamics will undoubtedly be of great basic and translational benefit.

Keywords: ATP; adenosine; astrocytes; computational models; glia; neuronal networks; sleep; slow oscillations.

Figure 1

Astrocytes are characterized by complex morphology. (A,B) Single-plane fluorescence images of an astrocyte expressing soluble GFP (A) or a membrane-bound YFP (B). In A the cell body and the principal cellular processes are clearly visible. When the fluorophore is bound to the membrane (B), small distal processes are more easily identified because of the increased surface-to-volume ratio of these thin cellular compartments. Scale bar 20 μm. Courtesy of A. M. De Stasi and T. Fellin.

Figure 2

Astrocytes release chemical transmitters to modulate information transfer at the synapse. (A–C) Schematic drawings showing the effect of astrocytic neuromodulation on synaptic function. Astrocytes release ATP which, after rapid extracellular degradation to adenosine, activates adenosine A₁ receptors located at pre-synaptic sites and leads to a decrease in the release of neurotransmitter (B). Astrocytes can also release D-serine which potentiates the current flowing through post-synaptic NMDA receptors thus leading to increased post-synaptic responses (C).

Figure 3

Astrocytes modulate slow oscillations. (A) Representative patch-clamp recordings from WT (top) and dnSNARE (bottom) anesthetized animals in vivo, showing reduced slow oscillations in transgenic mice with impaired gliotransmission (dnSNARE). (B,C) Cumulative distribution showing shorter UP (B) and longer DOWN- (C) state durations in dnSNARE mice (gray line) compared to WT controls (black line). Modifed from Fellin et al. (2009).

Figure 4

Weakening of synaptic transmission by astrocyte-derived ATP and adenosine could promote bursting at the onset of the UP state. (A) A stereotypical synaptic input alternating a phase of intense presynaptic firing to a relative quiescent phase, reminiscent of UP and DOWN states, respectively, is fed into a model of cortical neuron. (B) In control conditions, for low extracellular levels of ATP/adenosine (Adn), synaptic release probability is high, and the average synaptic conductance (red traces) is shaped by short-term depression while neuron fires (black trace) at sustained rate during the UP state. (C) For increased levels of ATP/Adn, the firing rate dramatically decreases due to the upstream reduction of these purines of synaptic release probability, but as shown in the histogram in (D), due to the modulation of the synaptic filtering characteristics by increased extracellular ATP/Adn concentration, the neuron fires a burst of actions potentials at the onset of UP states at higher frequency than in (B) in control conditions (n = 100; Bar + Error bar: Mean + STD; χ² test, p < 0.001). Synaptic release and ATP/Adn modulation of it were modeled as in panel 4A in De Pittà et al. (2011). Postsynaptic currents were computed as the product of postsynaptic conductance and membrane voltage. Each input spike contributed to a change of postsynaptic conductance proportional to the amount of synaptically-released resources by a α-function such as α (t) = g_max · exp(1 − t/τ) · t/τ (Ermentrout and Terman, 2010) with g_max = 500 mS/cm² and τ = 20 ms. The postsynaptic neuron was modeled as a regular spiking (RS) neuron according to (Destexhe, 2009).