Timed decisions: causal mechanisms and TAGging of neuronal ensembles in the prefrontal cortex – TimeTag

As a key property of the nervous system, timing is essential in decision sequences, and is impaired in many neuropsychiatric conditions. Yet timing is poorly understood and is a difficult capability to control in artificial intelligence. Causally linking brain activity and decision timing could therefore lead to great advances, but remains difficult.
Different from automatic motor timing at the millisecond scale, generating a decision at the right moment, at the second scale, engages thalamo-striato-cortical circuits. Just before the decision, neurons in the prefrontal cortex increase their electrical activity, and inhibiting the prefrontal cortex globally alters the timing of the action. This raises two fundamental questions: 1) what activity in the prefrontal cortex controls the timing of the decision, and 2) by which neural mechanisms is this activity generated?
It has been theorized that prefrontal activity would gradually increase until the right timing was reached. Our preliminary data and some of the literature indicate that the increase in activity at the right timing emerges more abruptly. This suggests that "silent" mechanisms that do not rely on the amplification of discharge activity but on the state (open, closed, inactivated) of receptors and ion channels would trigger spiking at the right time. These competing hypotheses remain speculative because 1) correlating neuronal activities with time is not sufficient to demonstrate causality in the control of timing and 2) the state of synapses and ion channels is not visible within extracellular recordings during a behavior.
To overcome these difficulties, our Timetag project proposes to combine recordings from the prefrontal cortex in mice with specific perturbations of the neuronal ensembles activated during precisely timed actions. Our data will help select the mathematical model (incorporating silent and non-silent mechanisms) that best explains the generation of activity at a specific timing, providing theoretical predictions that we will test experimentally.
To selectively manipulate the activity of neuronal ensembles involved in the control of timing, we will use transgenic mice allowing the expression of a fluorescent tag only in neurons activated during a session where the mice press levers according to a learned timing. We can thus selectively express proteins allowing to manipulate neuronal activity in the neuronal ensembles activated during lever presses at the learned timing. In our preliminary results, pharmacogenetically inhibiting the tagged neuron sets alters the timing of lever presses.
We will characterize (neural types, layers) the tagged neuronal ensemble. We will compare the effect of our selective inhibition in different nodes of the thalamo-striato-cortical circuit. We will record the activity of prefrontal neurons (in particular those previously tagged) while mice press levers according to the learned timing. We will characterize ex-vivo the properties (synapses, channels) of the prefrontal neurons of interest (before and after learning, tagged or not). These properties will be incorporated into our neural network model, in order to determine which mechanisms can generate the neuronal activity observed in the behavioral task. The most plausible model will provide predictions of the type of noise that can disrupt behavioral timing. We will test these predictions by perturbing tagged neuron ensembles with light.
Our TimeTag project, by testing mechanisms underlying electrical activity during a behavior and by specifically manipulating the neuronal ensemble correlated with timing, will provide answers to this simple and essential question: how does a group of neurons generate the right action at the right time?