Abstracts

Rationale: To investigate the hypothesis that mossy fiber sprouting may under some circumstances act as a compensatory mechanism to restore dentate gyrus (DG) function in the setting of cell loss, we studied the impact of these processes on the computational power of a DG model. We assumed that a key function of the DG is the separation and sparsification of entorhinal cortex (ER) input patterns (which in turn may minimize interference when partial retrieval cues are presented and may increase memory capacity), and we considered the observation that DG cell loss and sprouting can occur in human temporal lobe epilepsy.Methods: We studied a model based upon that of Santhakumar et al., J Neurophysiol. 2005. The model contained 100 ER, 500 granule (GC), 20 mossy, 10 basket, and 6 HIPP cells. Sprouting was modeled by increasing the density of GC to GC synapses. The morphology, membranes and synapses of DG cell types were represented in sufficient detail to approximate their known electrophysiological properties and firing patterns. DG cell and synaptic parameters were essentially as described in Santhakumar et al., J Neurophysiol. 2005. ER cells were modeled to fire 1:1 in response to input spike trains. Cell loss was represented by decreasing the number of DG cells by 50% (the reduced model). Network input was modeled as Poisson spike trains with Gaussian distributed average firing rates with mean 15 Hz and SD 5 Hz. A random 15% of ERs fired during each trial. The output of the model was considered to be the GC firing pattern. The capacity of each configuration of the model (full vs. reduced, 0%, 1%, 5% and 50% sprouting) to separate input patterns of spike trains was assessed by subjecting 3 instances of each configuration to 6 pairs of input patterns, each lasting 100 msec. Trial to trial separation between spike trains of individual neurons was quantified by convolving vectors of spike times with an exponential with a time constant of 5 msec and then subtracting, squaring and summing the resulting sequences (a van Rossum metric). Separation between patterns of spike trains was defined as the sum of trial to trial separations between constituent spike trains. Pattern separation capacity was assessed from the mean and standard deviation of pattern separations under each configuration. Sparseness was assessed from the average fraction of the GC population firing in each configuration.Results: Table 1 summarizes key results. Figure 1 shows representative data. In the full model, 5% sprouting increased pattern separation at the cost of reduced sparseness. Cell loss, without sprouting, reduced pattern separation. Lower levels of sprouting tended to restore pattern separation at the cost of decreased sparseness. 50% sprouting, in both the full model and in the setting of cell loss, resulted in erratic pattern separation and absent spareness. 50% sprouting in both configurations also resulted in paroxysmal activity (examples to be shown).Conclusions: These preliminary results suggest that under some circumstances mossy fiber sprouting may partially compensate for DG cell loss.