The cortex takes part in perception, action, and object representation. Large areas of the mammalian cortex contain neurons that respond to physical objects. For example, single cells in the fusiform face area in humans respond to specific faces.
Though population coding is a rule of thumb in object representation, it depends on single cells that acquire a specific receptive field during learning. The rules that govern synaptic plasticity on a single cell level eventually give rise to any neuronal network representation.
During the neural proliferation period, axonal sprouting results in myriad non-specific connections between neurons. Over time, synaptic elimination occurs, and responses to specific patterns start to emerge. We are interested in revealing the activation rules that shape those patterns.
The basic rule for synaptic plasticity is said to be Hebbian learning. Many theoretical studies showed that Hebbian learning can serve as the basic rule for synaptic plasticity. Surprisingly it was realized exclusively in pairs of neurons and not in multiple projections to single-cell as exists in the cortex (and brain-wide).
The precise mechanisms underlying pattern detection in the cortex remain elusive. Furthermore, the acquisition of a new receptive field under Hebbian rules by controlled input was not yet demonstrated.
We hypothesize that a neuron will learn to detect repeated patterns impinging on its dendrites using simple Hebbian rules. To test the hypothesis, we will use two-layer uni-directional neuronal culture, with which we will precisely control the activity patterns of the input layer using optogenetics and read the output layer activity using calcium imaging. We predict that specific neurons will “learn” to respond to a particular repeated pattern and be insensitive to other random activation patterns.
The system for all-optical interrogation will write-in using a pattern illuminator that will drive the neurons expressing fChrimson in any desired soma-targeted pattern. The activity read-out will be made using sCMOS camera by collecting the GCaMP8m signals.