We study how the nervous system controls behavior by analyzing the simple neural network that controls chemotaxis, a form of spatial orientation behavior, in the nematode worm Caenorhabditis elegans. C. elegans has a small nervous system (302 neurons), whose neurons and connectivity pattern have been completely characterized. Together with an entirely sequenced genome and high accessibility to genetic analysis, it provides an unique opportunity to study the neural and genetic basis of behavior. To reach this aim, we combine a range of experimental approaches, including quantitative analysis of behavior, neuron ablation, computer modeling, electrophysiology, and also incorporating a number of mutants deficient for chemotaxis.
Together these results provide new insights into the cellular and molecular mechanisms of information processing underlying animal behavior.
We use a quantitative analysis of worms' behavior as a way to determine the behavioral component for and neurons important for chemotaxis.
We track the movements of normal and mutant worms during chemotaxis at high spatial and temporal resolution, using an automated tracking system, to determine the behavioral strategies underlying spatial orientation in C. elegans. ("The fundamental role of pirouettes in C. elegans chemotaxis.")
Computations performed by the nervous system during chemotaxis are further evaluated by delivering controlled chemical stimulations to freely moving worms, and record their behaviors. Individual neurons in the chemotaxis network are killed with a laser microbeam to identify their role in chemotaxis behavior, thereby defining a neuronal circuit. ("The homeobox gene lim-6 is required for distinct chemosensory representations in C. elegans .")
In order to gain insights into the computations performed by the neural network for chemotaxis and to test function predicted by behavioral analysis for particular neurons, we also make whole-cell patch-clamp recording of identified neurons within this circuit. ("Active Currents Regulate Sensitivity and Dynamic Range in C. elegans Neurons.") ("Pressure polishing: a method for re-shaping patch pipettes during fire polishing.")
As a first step, we seek to understand how chemical information is represented by chemosensory neurons, in particular how chemosensory neurons encode the nature, time course and concentration of chemicals. To do so, we characterize the currents expressed by these neurons, how these currents are activated/modulated by chemical stimulation, and with what time-course they respond. Furthermore, the function of particular ion channels in shaping the response of the sensory neuron can be accessed by comparing activity in wild type vs. mutant animals for these channels.
In order to understand how the behavioral strategy for chemotaxis is implemented by the worm's nervous system, data generated by the experimental approaches are synthesized in theoretical models of the chemotaxis network. These networks allow us to generate experimentally testable prediction about the computational role of each neuron in the circuit. ("Computational rules for chemotaxis in the nematode C. elegans .") ("Chemotaxis control by linear recurrent networks.") ("Neural network models of chemotaxis in the nematode Caenorhabditis elegans .")
Specifically, we use simulated annealing on a parallel computer to train networks to perform the computations we derived from analyzing worms' behavior. Each neuron in the chemotaxis network is represented by an equivalent model neuron in the model neural network. Predictions from the model can then be tested, and refined, by neuronal ablation in real worms (comparing the deficits observed in real vs. model worms), or by recording from these same neurons (comparing the activity pattern observed in real vs. model worms).
This model network can be used to control orientation behavior in a robot. By taking the model away from the simplified representation of the computer simulated world and putting it in the real world, the reliability of this network can be accessed as it deals, like real worms, with environmental variability, noise in sensory input... The behavior of the robot can then be evaluated and compared to that of real worms, generating further hypothesis on possible behavioral strategies. ("Robust spatial navigation in a robot inspired by chemotaxis in C. elegans .")
In the future, we seek to complement the various experimental approaches described above with the use of genetically-targetable voltage probes. The development of these probes will allow non-invasive recording of activity pattern in the nervous system in behaving animals.