Dmitri A. Rusakov

Ultrastructural synaptic correlates of spatial learning in rat hippocampus.

Memory formation is believed to alter neural circuitry at the synaptic level. Although the hippocampus is known to play an important role in spatial learning, no experimental data exist on the synaptic correlates of this process at the ultrastructural level. Here, we have employed quantitative electron microscopy in order to compare the density, size and spatial arrangement of synapses in the dentate gyrus, and in area CA1, of spatially trained (water maze, invisible platform) versus control (visible platform) rats. No training-associated changes of hippocampal volume were found using a stereological estimaion (disector) of the volume density of dentate granule, or CA1 pyramidal cells. Nor were changes found in either density, or sizes of synapses (spinous or dendritic), in CA1 or dentate gyrus. However, analysis of synaptic spatial distribution showed a training-associated increase in the frequency of shorter distances (i.e. clustering) between synaptic active zones in CA1, but not dentate, thus indicating alterations in local neural circuitry. This finding indicates subtle changes in synaptic organization in area CA1 of the hippocampus following a learning experience, suggesting that spatial memory formation in mammalian hippocampus may involve topographical changes in local circuitry without synapse formation de novo.

Morphological changes associated with stages of memory formation in the chick following passive avoidance training

Memory formation following learning is presumed to result from modification in the efficacy of neural circuitry, either through strengthening of pre-existing synapses, or formation of new contacts. An ideal paradigm to investigate memory formation is one-trial passive avoidance training of day-old chicks, in which the birds learn to avoid pecking a bead coated with an aversive substance, methylanthranilate. Following training, a sequence of biochemical, electro-physiological, pharmacological and morphological events takes place within two loci in the forebrain, the intermediate and medial hyperstriatum ventrale (IMHV), and part of the paleostriatal complex, the lobus parolfactorius (LPO). Our data reviewed here suggest that the initial acquisition of memory involves population changes in the fine spatial organization of synaptic vesicles and active zones in synapses in the IMHV whereas longer-term changes are more prominent in the LPO and involve, primarily, a bilateral increase in the density of synapses and dendritic spines. The short-term synaptic changes are as dynamic as the molecular changes which have hitherto been considered the preserve of short-term correlates of memory formation.

Synapses in hippocampus occupy only 1–2% of cell membranes and are spaced less than half-micron apart: a quantitative ultrastructural analysis with discussion of physiological implications

Relatively little information exists regarding the spatial structure of synaptic neuropil in the brain. The present electron microscopic study employs unbiased stereological techniques and Monte Carlo simulations to characterise quantitatively the spatial organisation of synaptic circuitry in the dentate gyrus of the hippocampus, an area of particular importance in mechanisms of learning and the subject of a number of experimental neurobiological models of synaptic plasticity such as long-term potentiation. Firstly, tissue shrinkage/expansion resulting from embedding was assessed by imaging 300-microm thick hippocampal slices in the course of the entire embedding protocol, giving a value of 94.3 +/- 1.1% for distance measures and 84.3 +/- 2.8% for volumetric measures. Secondly, numeric synaptic density, Nv, was estimated using the disector. Thirdly, accumulated area of post-synaptic densities (PSDs) per tissue volume, Sv, and the overall cell membrane area per tissue volume, Sv*, were assessed using unbiased stereological rules coupled with image analysis of single sections. Finally, the mean area of individual PSDs was derived as a ratio Sv/Nv giving: 0.0394 microm2 for axo-spinous PSDs (thus representing approximately 1.3% of total cell membranes) and 0.0769 microm2 for dendritic shaft PSDs (approximately 0.25% of total cell membranes). From these data, the mean nearest neighbour distance between synapses was estimated using Monte Carlo simulations of a random 3D arrangement of synapses constrained by PSD sizes (a truncated Poisson process), giving a value of 0.48-0.51 microm. The physiological importance of the morphometric data obtained is discussed in terms of assessing (i) the role of synaptic environment in modifying synaptic efficacy and (ii) the plausibility of cross talk between synapses in relation to extrasynaptic neurotransmitter diffusion and transient depletion of extracellular Ca2+.

Branching of active dendritic spines as a mechanism for controlling synaptic efficacy

Recent experimental findings (Yuste R. and Denk W. (1995) Nature 375, 682-684) suggest that dendritic spines possess excitable membranes. Theoretically, it was shown earlier that the shape of active spines can significantly affect somatopetal synaptic signal transfer. Studies of long-term potentiation in the hippocampus have related the increased synaptic efficacy to a number of structural modifications of spines, including an increased number of branched spines [Trommald M. et al. (1990) In Neurotoxicity of Excitatory Amino Acids, pp. 163-174. Raven Press, New York] and a strengthened capability for spines to alter their spatial positions [Hosokawa T. et al. (1995) J. Neurosci. 15, 5560-5573]. In the present simulation study, the potential physiological impact of several types of spine changes was examined in a compartmental neuron model built using the neuromodelling software NEURON [Hines M. (1993) In Neural Systems: Analysis and Modeling, pp. 127-136. Kluwer Academic, Norwell, MA]. The model included 30 complex spines, with dual component synaptic currents and mechanisms of Ca2+ uptake, diffusion, binding and extrusion within spine heads. The results show that local clustering properties of spine distributions along dendrites are unlikely to affect synaptic efficacy. However, partial fusion of active spines, which results in formation of spine branches, or subtle changes in spine branch positions, could alone significantly increase synaptic signal transfer. These data illustrate possible mechanisms whereby subtle morphological changes in dendritic spines (compatible with changes reported in the literature) may be linked to the cellular mechanisms of learning and memory.

Increased immunogold labelling of neural cell adhesion molecule isoforms in synaptic active zones of the chick striatum 5–6 hours after one-trial passive avoidance training

An area of the chick striatum, the lobus parolfactorius plays an important role in one-trial passive avoidance learning tasks. In the present study we report evidence that 5-6 h post-training, a significantly higher proportion of synaptic active zones in this area contain labelled epitopes of the neural cell adhesion molecule, with the greatest occurrence of labels at the edges of active zone profiles (in both control and trained groups). This suggests that there is a period after training when expression of the neural cell adhesion molecule in synaptic membranes almost doubles, and that events at active zone edges may play a specific role in mechanisms of synaptic adhesion. Cellular mechanisms of long-term memory formation are believed to include alterations in neural circuitry at the synaptic level. The involvement of the neural cell adhesion molecule (NCAM) in functional synaptic modifications has been demonstrated using a number of physiological models. Performance of rats in the Morris water maze, a spatial learning paradigm which requires the hippocampus, is impaired by either intraventricular injection of NCAM antibodies, or injection into the hippocampus of an enzyme which increases homophilic adhesion of the molecule, due to the removal of polysialic acid residuals from extracellular NCAM domains. In addition, intraventricular injections of anti-NCAM antibodies 6-8 h post-training were shown to impair memory for a one-trial passive avoidance task in the rat. An avoidance training model in the one-day-old chick indicates a similar time window, 5-6 h post-training during which memory for the task can be impaired by intraventricular injection of NCAM antibodies. In the hyperstriatum ventrale, a chick forebrain area involved in the passive avoidance task. subtle changes in the distribution pattern, but not density of NCAM molecules in synaptic membranes were revealed 5-6 h post-training. However, on the basis of studies of synaptic morphometry, a region of striatum, the lobus parolfactorius (LPO), appears to play a more important role in longer term memory storage for the task.

Quantification of dendritic spine populations using image analysis and a tilting disector

A series of image analysis routines, stochastic geometry methodology, and a design-based stereological procedure have been developed to quantify objectively the length, layout, and the true density of neuronal dendritic spines observed at the light (or confocal) microscope level. First, the image of a dendritic fragment of interest (in the plane of view) is scaled to a standard brightness scale, and the dendritic profile is separated from the background using a computerized thresholding algorithm that analyzes the histogram of grey levels. Secondly, the resulting binary image of the dendrite is transformed to a midline skeleton that underlies the dendritic geometry. Thirdly, skeletal branch lengths are directly computed (in pixels), thus giving objective measures of visible spine lengths and inter-spine distances along the dendritic stem. These raw data are the basis for (1) an estimation of the distribution of 3D spine lengths, and (2) a nearest neighbour analysis of the spine layout along the dendrite. A design-based stereological routine, the tilting disector, is suggested for unbiased estimation of the true (3D) density of spines along dendrites. The routine involves tilting the dendritic fragment of interest around its longitudinal axis for a known angular sector and scoring the number of spines seen in one angular position and unseen in the other position. Data from a study of neuronal dendrites in the chick forebrain are presented.

Reduction in spine density associated with long-term potentiation in the dentate gyrus suggests a spine fusion-and-branching model of potentiation.

Approximately 2,700 dendritic spines in Golgi‐impregnated hippocampal granule cells were quantified via image analysis 24 h after the unilateral induction of long‐term potentiation in seven rats. Stereological corrections were made using a tilting disector and analytical unfolding technique. In the potentiated hemisphere the mean spine density along dendrites was reduced by ∼20%. The relative frequency of shorter, thicker spines was increased in potentiated tissue. Physiological consequences of two morphological changes leading to a reduction in spine density (retraction or fusion of spines) were examined using a compartmental model of a simplified granule cell. The model was constructed in the NEURON modeling environment and included a realistic population of 60 dendritic spines (with dual‐component synapses and active Ca2+‐dependent mechanisms). Simulations demonstrated that potentiation of postsynaptic responses was compatible with fusion (with branching) of a proportion of spines with their neighbors but was not compatible with retraction of spines. This result held over wide variations of model parameters as long as dendritic membranes were assumed to be excitable. Hippocampus 1997;7:489–500.

Re-structuring of synapses 24 hours after induction of long-term potentiation in the dentate gyrus of the rat hippocampus in vivo.

In male rats, long-term potentiation was induced unilaterally in the dentate gyrus, either by high frequency (200Hz) or theta rhythm stimulation. Structural synaptic changes were examined 24h after induction using quantitative electron microscopy. A disector technique was employed in order to estimate the density of synapses (using 70-80-nm sections) and of granule cell nuclei (using 2-microm sections) in the middle, and inner molecular layer in both hemispheres. Synaptic height and total lateral areas of synaptic active zones per unit tissue volume were assessed via assumption-free stereological techniques coupled with image analysis. The results obtained indicated that both synaptic density and number (corrected per neuron) of axo-spinous, but not axo-dendritic, synapses were approximately 40% higher in the middle, but not inner molecular layer of the potentiated hemisphere compared to the contralateral (control hemisphere). No significant inter-hemispheric difference was found in the volume densities of lateral areas of active zones. These data suggest that 24h after long-term potentiation induction, active zones of existing axo-spinous synapses either split forming separate contacts, or decrease in size while new synapses are formed.

Lateral patterns of the neural cell adhesion molecule on the surface of hippocampal cells developing in vitro

In monolayer cultures of hippocampal neurons from newborn rats, an immunocytochemical quantitative study was carried out to investigate age-dependent arrangement of the neural cell adhesion molecules in different parts of cell membranes. On the fifth and 12th day in vitro, neural cell adhesion molecules were labelled with specific antibodies and protein A conjugated to colloidal gold particles. Samples of randomly selected electron micrographs that displayed labelled membrane fragments of cell bodies, growth cones, and axons were numerically analysed for the five- and 12-day in vitro neurons. Neural cell adhesion molecules surface topography was quantitatively described and compared, using a statistical stereological approach. The mean surface density of labelled neural cell adhesion molecules was found to be approximately 2.5 times higher in growth cone membranes relative to somatic and axonal membranes in five-day in vitro neurons. By the 12th day in vitro, this density decreases in somatic membranes (approximately 18%) and increases in axonal membranes (approximately 60%). Representative spectra of lateral intervals between labels as well as images that show typical topography of label on membrane surfaces were simulated. The results revealed regular patterns of neural cell adhesion molecules on the somatic surface and allowed consideration of neural cell adhesion molecules arrangement in a view of membrane adhesion properties. Participation of cytoskeleton in neural cell adhesion molecules rearrangement is discussed.

Spatial re-arrangement of the vesicle apparatus in forebrain synapses of chicks 30 min after passive avoidance training

A quantitative ultrastructural study of synapses was carried out in the forebrain IMHV (intermediate and medial hyperstriatum ventrale) of 1-day-old chicks 30 min after training to avoid pecking at a bead coated with methyl anthranilate. In 10 birds (5 control and 5 trained), the length, curvature and the number of the synaptic active zone profiles were measured, and the active zone profile length was observed to increase in trained chicks. The spatial arrangement of synaptic vesicles with respect to the active zone was examined using a statistical stereological approach. This showed that vesicles are not located uniformly but accumulate in two spatial pools which appear to rearrange following training, with a greater number of vesicles near the active zone. These data may reflect subtle changes in the functional efficacy of synapses in the IMHV in the initial phases of memory formation.