Information efficacy of a dynamic synapse

Information transmission in the brain is mediated mainly through chemical synapses. The arrival of an action potential at a chemical synapse triggers a sequence of events that leads to the release of a vesicle. The released neurotransmitter molecules open the ion channels on the postsynaptic site and create an excitatory or inhibitory postsynaptic potential. An incoming action potential does not always elicit a vesicular release. The evoked release probability varies widely among synapses. While the reliable synapses show a release probability close to one, the release probability of an unreliable synapse may be lower than one tenth. Additionally, a synapse sometimes releases spontaneously even in the absence of an action potential. Synaptic unreliability and spontaneous release can significantly alter the information transmission through the synapse. Synaptic plasticity also changes the strength of synaptic connections over a wide range of time scales. In short-term depression, the successive release of vesicles reduces the release probability of the synapse. The functional role of short-term depression in filtering and decorrelation of the presynaptic spike train has been shown in several studies. In this thesis, we investigate the function of short-term depression in modulating the rate of information transmission through the synapse. We model a synaptic release site by a communication channel, capturing the spike-evoked and spontaneous release of the synapse. The input of the channel is the presynaptic spike train and the output is the elicited postsynaptic potential. To incorporate short-term depression into the synapse model, the state of the channel is switched between a normal state and a used state. After each release, the synapse goes to the used state and the release probabilities of the channel reduce; the synapse recovers back to the normal state in the absence of release. Information theory is then employed to calculate the rate of information that is transferred from the presynaptic spike train to the postsynaptic potential. Synaptic release is energetically expensive and a neuron needs to compromise between its information efficacy and energy consumption. We consider the energetic cost of the release and calculate the energy-normalized information rate of the synapse. This measure is used to evaluate the rate of information transmission for a given energy budget. We show that the functional role of short-term depression in modulating the information transmission depends on the relative level of depression for spontaneous and spike-evoked releases. If the depression affects spike-evoked and spontaneous release equally, then the information rate and energy-normalized information rate of the synapse both decrease. However, if spontaneous release is depressed more than spike-evoked release, then short-term depression can enhance the information efficacy of the synapse. In the two-state model of depression, the synapse transits sharply between the used state and the normal state. To emulate the gradual depression and exponential recovery of short-term depression, the two-state model is extended to a communication channel with a memory of the release history. The content of the memory specifies the release probabilities of the channel based on the dynamics of short-term depression. We calculate the information efficacy of the synapse model and determine the regime of synaptic parameters in which short-term depression enhances/decreases the mutual information rate and energy-normalized information rate of the synapse. Our analysis shows how short-term depression governs the trade-off between the energy expenditure and information rate of the synapse.