Skip to main content

Computational roles of Feed-Forward Loops in neural circuits

Final Report Summary - FFL FUNCTION (Computational roles of Feed-Forward Loops in neural circuits)

In this project, we aimed to decipher the functional roles a special circuit, the feed-forward loop (FFL), in neural networks. In this circuit, two given neurons synapse on a third neuron, and one of these two neurons also synapses on the second neuron. The reason this circuit raises particular interests in the fact that it is embedded in neural networks significantly more than randomly expected, suggesting it may carry key functional roles in information processing.

To study potential roles of this circuit, we focused on the neural network of C. elegans nematodes, for which the entire wiring diagram of its 302 neurons is available. Specifically, we developed novel approaches to analyze network connectivity and found that the neural network obeys to the Common- Neighbor-Rule (CNR) in which a pair of neurons is more likely to be connected the more common neighbors it shares. These common neighbors form homogenous structures that are primarily made of interconnected multi-FFLs. Moreover, these homogenous sets appear in defined layers of the network such that they can tell about their possible functional roles in the network.

For example, we found that specific structures, namely multi-input FFLs, appear primarily in the sensory layer of the neural network. Simulations of signal propagation reveal that these circuits can serve in this layer as signal amplification and short-term memory devices, and indeed these features are particularly crucial for the sensory layer. In addition, multi-output FFLs appear primarily in the motoneuron layer. Again, these features are vital in this layer specifically to promote synchronized activity that supports coordinated movement. Interestingly, when we redraw this complex network and reconstruct it based on these homogenous sets alone, we reveal a simple modular network architecture that is intuitive to understand.

Together, this study provides a novel framework for analyzing and extracting function based structure alone. This approach is particularly important for analyzing gigantic and more complex networks (e.g. the human brain), once they become available.

Inspired by these results we turned to experimentally test these predictions. We therefore generated a suite of transgenic animals that allow studying functional dynamics in individual neurons on the network-wide scale. In addition, we developed the methodologies and the software analyses tools to identify individual target neurons. We also generated transgenic lines expressing the light-activated channelrhodopsin in selected chemosensory neurons such that we can now infer which sets of neurons are co-activated and how information flows in the neural network. In addition, we developed a system for measuring neural activity from freely behaving animals. This system is unique as it does not require a custom-built microscope. We also developed a software that controls a motorized stage in a fast fashion to keep the animal at the center of the field of view while the camera grabs images of the entire body of the worm including its target neuron. This setup is now key in our lab for understanding how neural circuits and their activity dynamics dictates behavioral outputs.