What is the snow in a neural avalanche?

Neurons spike spontaneously in many experimental settings. Such firing patterns are often characterized by prolonged periods of silence followed by an unknown trigger of spontaneous activity that propagates throughout the slice. These were dubbed `neural avalanches' as they resemble avalanches in which accumulated snow rapidly flows down a mountain slope. Expanding this analogy -- if the network connectivity is equivalent to the mountain slope, what is the 'snow' that accumulates prior to a neural avalanche? Here, we propose that it is the accumulation of ATP in neuronal mitochondria, and that avalanches provide a respite from toxic conditions that arise during lower-than-baseline ATP consumption.

Neurons, presumably in anticipation of synaptic inputs, keep their ATP levels at a maximum. As metabolic recovery from synaptic inputs requires substantial energy resources, neurons are ATP-surplus/ADP-scarce during synaptic quiescence. With ADP availability as the rate-limiting step, ATP production stalls in the mitochondria when energy consumption is low, leading to the formation of toxic Reactive Oxygen Species (ROS) which disrupt many cellular processes. We hypothesize that neurons actively sense their metabolic state and trigger `metabolic spikes' to restore ATP production, to avoid ROS. To test this, we built a recurrent network in which neurons sense their metabolic state (based on recent inputs and outputs) and modulate an intrinsic metabolic current to control spiking when necessary. When the network goes silent, neurons initiate metabolic spikes to increase their own energy expenditure and avoid ROS poisoning. These first spikes trigger a domino effect of activity that ripples through the network and ceases when neurons have increased their ATP expenditure either through synaptic inputs or by spiking. This mitochondrially mediated homeostatic mechanism can account for many intrinsic firing patterns observed in neurons, as well as the avalanche-like activity, and it explains how networks maintain criticality without loss of stability.