Emergence of Resonances in Neural Systems: The Interplay between Adaptive Threshold and Short-Term Synaptic Plasticity
Jorge F. Mejias, Joaquin J. Torres
PLoS ONE 6(3): e17255. doi:10.1371/journal.pone.0017255
In the chaos of environmental stimuli, animals must be able to discern features important for finding food, mates, and avoiding predators. Animal nervous systems do this with amazing precision and accuracy. What are the mechanisms for this?
Certain levels of noise can strengthen a weak signal in a nonlinear system. This phenomena is called stochastic resonance. This happens at moderate noise levels. With high or low noise levels the signal is not detected. Certain cells in the hippocampus, the brain stem and some cortical regions have been found to do this. It has always been assumed that stochastic resonance produces a single peak response. The authors have found a situation where a bimodal response has been found. In addition, this second response can be tuned to different frequencies.
Neurons are inherently plastic. They are plastic physically through growth and loss of dendrites, spines and synapses. Chemically, synapses are plastic through mechanisms such as short-term depression and short-term facilitation that can modify the post-synaptic response. Also, the voltage-dependent activation and inactivation of sodium and potassium channels allows for a dynamic firing threshold. This “allows a neuron to raise its firing threshold without the generation of previous APs.” The authors consider both of these mechanisms a potentially important to the way a neuron can respond to noise. To test this they created progressively more realistic models of a simple noisy network. In every one they got a bimodal response.
First they used an integrate-and-fire model which has the advantage of having an analytical solution. Next they used the more realistic FritzHugh-Nagumo neural model which has an adaptive threshold mechanism built-in. Finally, they used a purely stochastic model, which is the most realistic of all. This third model is very important because a discrete model can often modify or add features found in continuous models. Thus, this bimodality is a structural feature at all modeling levels from most simplified to most realistic.
As I’ve said before, the published models would be most useful. In addition, this must be experimentally verified. In a video abstract accompanying a companion paper, the authors mention a recent experiment that has tested this.
J. J. Torres, J. Marro and J. F. Mejias
2011 New J. Phys. 13 053014