International Meeting for Autism Research: Synaptic Causes for Autistic Regression: A Neural Network Model

Synaptic Causes for Autistic Regression: A Neural Network Model

Thursday, May 12, 2011
Elizabeth Ballroom E-F and Lirenta Foyer Level 2 (Manchester Grand Hyatt)
1:00 PM
Y. S. Bonneh1, S. Romani2, Y. Adini3 and M. Tsodyks2, (1)Human Biology, University of Haifa, Haifa, Israel, (2)Neurobiology, The Weizmann Institute of Science, Rehovot, Israel, (3)Vision Research Inst., Kiron, Israel
Background: One of the most puzzling phenomena associated with the autism spectrum disorder (ASD) is the high incidence (15-40% reported) of a significant behavioral regression. In its strongest form, a 2 or 3 years of apparently normal development, terminate with a partial or full loss of speech, receptive-language, social skills, visual recognition, and more. Reversible regression on a short time scale was also reported, i.e. large fluctuations in performance within short time periods.

The neural mechanisms involved in regression are currently unknown. One important clue is the high incidence of epilepsy or epileptiform activity found in children with ASD, although the evidence for a link between epilepsies and regression is inconclusive.   

Objectives: to develop a simple neural-network model of autistic regression at the synaptic and local network level.

Methods: We rely upon an existing recurrent network model with short term synaptic depression, which implements a memory completion network. The model produces a sharply tuned response from a partial or noisy input and switches rapidly into a bursting regime depending, for instance, on the strength of the external input. We explored the regressive effect of the bursting regime as a model for autistic regression.

Results: The model showed degraded memory completion during the bursting regime. On a short term, this resulted in a partial loss of the "weaker" memories which were fully recovered when shifting out of the bursting regime. However, long term bursting is expected to produce long-term damage due to Hebbian synaptic changes that can only be recovered by retraining. The model thus demonstrates that a reversible short-term regression and a long-term regression could have the same origin. 

Conclusions: Autistic regression could be linked to the frequently observed spiking or epileptic activity via a recurrent network model that enters a bursting regime. The model demonstrates several features that can inspire further investigation. 

| More