Brodfuehrer, P.D., Cellucci, C.J., Acera-Pozzi, R., Albano, A.M.
Departments of Biology and Physics, Bryn Mawr College, Bryn Mawr, PA. USA.

The initiation of swimming can be traced, neuron-to-neuron, from mechanosensory input to motor output. In this pathway a pair of subesophageal trigger neurons, cells Tr1, functions as transducers linking mechanosensory input to the segmental swim-generating network. However, the swim-triggering ability of cell Tr1 is highly variable. Stimulation of cell Tr1 can elicit swimming on one trial, but may not the next trial even though the strength of cell Tr1 stimulation is approximately constant for all trials. The underlying mechanism for cell Trl's response variability is not readily apparent from our current understanding of the organization of the leech swim-generating network. Instead of trying to identify individual neurons associated with this variable response, we are using both linear and nonlinear time series analysis, as well as techniques from information theory, to address three questions associated with defining the neural substrates controlling leech swimming. 1) Is there a characteristic pattern of activity in the leech ventral nerve cord required for eliciting swimming? 2) Does the initiation of triggered swimming depend on the 'state' of activity in the ventral nerve cord when the stimulation is presented? 3) Is the ventral nerve cord activity pattern dependent on the manner in which swimming is elicited (Tr1 initiated vs. spontaneous swim episodes)?

In a series of experiments we recorded extracellularly from an anterior (between M2-M3) and a posterior (between M16-M17) location along the ventral nerve cord while intracellularly stimulating cell Tr1. In trials where stimulation of cell Tr1 triggered swimming there was a substantial increase in extracelluarly recorded activity descending from the head ganglion and a simultaneous decrease in activity ascending from the tail ganglion. An almost identical change in connective activity occurred before the onset of 'spontaneous' swim episodes. On the other hand, when Tr1 stimulation did not lead to swimming, no consistent change was evident in the activity pattern at the anterior and posterior recording locations.

Further analysis of our physiological data using linear and statistical measures have been inconclusive. However, we have found nonlinear and information theory techniques which may be applicable for defining the neural signals controlling swimming. First, a nonlinear prediction method, based on work by Surgihara and May, is being used to determine whether the initiation of swimming depends on the 'state' of the activity in the ventral nerve. Second, average mutual information theory is being used to determine if measured 'communication', either descending from anterior to posterior or ascending from posterior to anterior, can predict whether Tr1 stimulation will lead to swimming.