Parallel Processing in the Leech Central Nervous System

The aim of this thesis is to shed light on the parallel processing in the leech nervous system. Two main problems have been addressed: 1) how a mechanical stimulus is coded by mechanosensory neurons at the first stage of the information processing in the central nervous system (CNS). 2) which statistical properties of the electrical activity of neurons in the segmental ganglion are relevant for infom1ation processing and in particular for motor reactions. A semi-intact preparation consisting of a single segment innervated by its own ganglion and the adjacent ganglia was used. Sensory stimulation was delivered by touching the skin with an appropriate device. Previous observations indicate that each point of the skin surface is innervated by several mechanosensory neurons (Nicholls and Baylor, 1968; Yau, 1976; Lewis and Kristan, 1998-c). In my thesis I analyzed how sensory information is coded in the electrical activity of a population of mechanosensory neurons. I developed and used an experimental set-up allowing the recording of the simultaneous activation of all mechanosensory neurons belonging to three consecutive segmental ganglia. The experiments were performed in high Mg2+ Ringer fluid in order to block all chemical synapses so as to eliminate synaptic responses and muscle contraction leading to a motion of the skin resulting in an additional sensory feedback. In these conditions it is possible isolate the responses of sensory neurons directly activated by the stimulus. The extent of the activation of mechanosensory neurons in three adjacent ganglia after a stimulation was analyzed. The results of this mapping show that all mechanosensory neurons are consistently activated also in their accessory receptive fields. The reproducibility of their response is the same as that observed in mechanosensory neurons responding in major receptive fields. T cells activation presents the same characteristics for stimulation ·of the major or minor receptive fields, indeed they respond to the transient phase of the stimulus. P cells show a different dynamics of activation if stimulated in their major or minor receptive field. In presence of a prolonged stimulation, P cells of the central ganglion fire action potentials continuously while P cells of accessory ganglia adapt. N cells are always activated in major and minor receptive fields showing approximately the same dynamics. When the skin is touched with a moderate tactile stimulus, i.e. exerting a force ofless than 20 mN, many different neurons (T and P cells) of the three ganglia fire action potentials. Thus sensory coding is initially redundant. If the stimulation is prolonged, many of these sensory neurons rapidly adapt and only P cells of the central ganglion respond. Thus sensory coding is dynamical and becomes very sharply tuned. In the second part of my PhD work, a different stage of the information flow has been investigated. The aim of the research was to study which statistical properties of the electrical activity of neurons in the leech ganglion are important for neural processing. In particular I analyzed the statistics of the evoked activity in response to a sensory input. The ganglion was considered as a black box with inputs and outputs; inputs are represented by sensory neurons and outputs by motor neurons or other neurons innervating roots and com1ectives. The responses of mechanosensory neurons to touch stimuli show trial-to-trial reproducibility, i.e. spikes in sensory neurons occur at very precise times. This does not happen for motor neurons. The experiment consisted in stimulating repeatedly a particular mechanosensory neuron by evoking action potentials with an intracellular electrode. The evoked activity in response to this input was recorded by suction electrodes. The extracellular recordings contain the superimposed activity of many neurons which can be separated by spike sorting procedure. The first and second order statistics were studied. Two main results were obtained by this analysis: 1) the response is distributed on many neurons; many neurons respond to the same stimulus and the same neurons are recruited for different stimuli. 2) the response is characterized by spatio-temporal variability, i.e. different neurons respond in different trials and neurons respond with jitter in time occurrences of spikes. The comparison between reproducibility of mechanosensory neurons evoked by touch stimulus and motor neurons evoked by mechanosensory neurons. activation shows that sensory cells are very reproducible, therefore the origin of motor neuron variability has to be found in synaptic transmission. The analysis of the second order statistics was used to check correlation among coactivated neurons. Joint entropies and joint probabilities for each pair of neurons were computed and compared with the sum of individual entropies and the product of individual probabilities, respectively. In all cases these quantities are consistently equal. This means that coactivated neurons are statistically independent. This property of the system can be functional for neural proeessmg. If the response of the network is distributed, then the pooling of the electrical activity of all individual responses is important and may be the key feature of the network. In this perspective, a distributed process consisting in a large number of statistically independent unreliable elements leads to a reproducible response, when the response is pooled over all elements. There is theoretical justification for this statement, indeed it is demonstrated that the pooling of a large number of statistical independent stochastic processes affected by high variability lead to a stochastic process with low variability. In conclusion, increasing the number of elements pooled, the response is more and more reliable.