IntroductionTo investigate possible morphological correlates to long-term potentiation (LTP), three-dimensional reconstruction of serial electron micrographs was employed. LTP was induced in the perforant path/dentate granule cell synapse in two rats. The surgically isolated contralateral side served as control, along with two untreated animals. Longitudinally sectioned and transversally sectioned dendrites were sampled from the middle fifth of the molecular layer and all visibly connected spines were identified. A mixed, unbalanced, nested variance component model was used to make a valid statistical comparison between the LTP and control groups.The spine density was higher in the experimental than in the control groups. The changes were statistically significant in both the longitudinal and transverse sample. In addition, spines with a divided stem and two heads (bifurcating spines) were seen at a higher frequency in the LTP material compared with the contralateral material. From a subset of dendrites all connected spines were reconstructed and detailed measurements of head, neck, and PSD dimensions were made. We failed to find significant differences following LTP on either of the dimensions measured. The results suggest that new spine synapses are formed following LTP, including some of the bifurcating type.1Corresponding author.The hippocampal formation is important for spatial navigation in rodents. Surgical or excitotoxic lesions of the hippocampal formation results in reduced learning behavior in a spatial learning task like the water maze (Morris et al. 1986; Morris et al. 1990). Rodents that are dependent on good spatial navigation like long-ranging male prairie voles have larger hippocampi than closely related species with smaller territories or nonroaming female voles (Jacobs et al. 1990). Furthermore, food-storing species of birds also have a larger volume of the dorso-medial cortex, a likely avian counterpart of the hippocampus (Krebs et al. 1989). Hippocampal damage in man causes large learning and memory deficits, including those coupled to spatial behavior (Squire 1992).In a search for basic mechanisms that might subserve the hippocampal participation in learning behavior, a process that strengthens certain synapses would be an interesting candidate, in particular if it had sufficient duration to fit the long duration of spatial learning. Long-term potentiation (LTP) is a cellular model that has several properties suitable for a role in learning behavior (Bliss and Collingridge 1993). First, LTP can be induced by repetitive activity of afferent impulses within the physiological range of hippocampal neurons. Second, the effect is very long-lasting, up to several weeks or even months in an awake, behaving animal (Bliss and Gardner-Medwin 1973;Barnes 1979). Most important, long-term potentiation allows associativity between synapses belonging to two afferent systems. Simultaneous activation of NMDA (N-methyl-n-aspartate)-receptor operated synaptic channels and AMPA (0t-aminohydroxy-isoxaz...

In the hippocampus, most excitatory synapses are located on dendritic spines. It has been postulated that the geometry of spines and/or postsynaptic density (PSD) influences synaptic efficiency and may contribute to the expression of plastic processes such as learning or long-term potentiation (LTP). Based on three-dimensional reconstructions of dentate granule cell dendrites from serial electron micrographs, we have measured head dimensions, neck cross-sectional areas, neck length, and PSD area and form of 115 spines of dentate granule cells in the medial perforant path termination zone. All dimensions showed a large variability, with up to 100-fold differences in values. A calculated diffusion index for transport of molecules through the reconstructed neck varied over a 100-fold range. The neck and head dimensions were moderately positively correlated, whereas the PSD area was strongly correlated with head volume. Distribution histograms and scatter plots of various spine dimensions did not reveal any systematic clustering, suggesting that there is a continuum of spine geometries rather than distinct classes for granule cell dendritic spines in the middle molecular layer. Transversely (n = 13) and longitudinally (n = 27) sectioned dendrites had mean spine densities of 2.66 and 1.01 spines/microns, respectively, uncorrected for so-called hidden spines. Bifurcating spines made up 2.1% of the total spine number in transversely and 2.3% in longitudinally sectioned dendrites. The twin spine heads never shared the same presynaptic bouton. Fenestrated or split PSDs shared the same presynaptic element in all but two cases, arguing against PSD division as an intermediate step in synapse formation.

No abstract