bluelogo.gif (1643 bytes)
Gatsby Computational Neuroscience Unit
Alexandra House, 17 Queen Square, LONDON, WC1N 3AR, UK
Tel: +44 (0) 20 7679 1176, Fax +44 (0) 20 7679 1173,,


annual report
the greater gatsby



16-18 September 2002
By invitation only

B10 Seminar Room, Alexandra House, 17 Queen Square, London, WC1N 3AR

Input-output relationships of cortical synaptic connections during complex patterns of activation
Hugh Robinson, Department of Physiology, University of Cambridge, UK
The postsynaptic response at a cortical synaptic connection appears to depend on its activation history over a wide range of timescales, and also has significant randomness. Because the trajectory of responses is shaped by stochastic release behaviour at individual terminals, which influences subsequent transmitter release, the dynamics of responses to individual synaptic connections is richer and more difficult to describe than the ensemble average response. The number of functional contacts between pairs of cortical cells is typically very low, so that the time-correlated behaviour of individual connections may be functionally more relevant than their average behaviour. We have used a new approach to characterising the input-output map of individual synaptic connections during arbitrary, complex stimulation. The history of activation and response is captured by a delay vector (or embedding) of both response amplitudes and of intervals between presynaptic APs. Synaptic transmission is represented by the mapping of delay vectors to their subsequent response amplitude. Defining a measure of distance between different vectors, the reliability and variability of the synaptic input-output map can be quantified by the extent to which near-neighbour delay vectors lead to similar responses. We applied this to paired recordings between pyramidal neurons in young rat cortical slices, with long (> 30 minutes), non-repeating, natural-like presynaptic spike trains. Near-neighbour prediction of synaptic responses was much more accurate than a best-fit model of short-term plasticity. The timescale of activation which measurably contributed to predictability was usually several minutes, but depended on the structure of the stimulus train. With Poisson stimulation, delay vectors were much more highly clustered than expected by chance, reflecting a wide spread in the transmission reliability of patterns within the stimulus. Reliability also distinguished classes of response with different amplitude distributions: reliably-stimulated responses had tightly-clustered amplitude distributions, while unreliably-stimulated responses did not. We are using this method to classify the dynamics of synaptic connections, and to examine the effects of long-term plasticity on the synaptic input-output relationship during natural, complex stimulation patterns.
Joint work with Ingo Kleppe
Supported by the BBSRC and the Boehringer Ingelheim Fonds.