convert the RET% into a section thickness measurement (De Groot,. 1988). The four ... tified by the presence of a dark area of postsynaptic density (PSD) and.
European Journal of Neuroscience, Vol. 17, pp. 1973±1980, 2003
ß Federation of European Neuroscience Societies
Morphological changes in hippocampal dentate gyrus synapses following spatial learning in rats are transient Mark D. Eyre,1 Gal Richter-Levin,2 Avi Avital2 and Michael G. Stewart1 1 2
Department of Biological Sciences, The Open University, Walton Hall, Milton Keynes MK7 6AA, UK Department of Psychology and the Brain and Behaviour Research Centre, University of Haifa, Haifa 31905, Israel
Keywords: water maze training, synaptic alterations
Abstract The hippocampus is believed to play a crucial role in the formation of memory for spatial tasks. In the present study quantitative electron microscopy was used to investigate morphological changes in the hippocampal dentate gyrus of 3-month-old male rats at 3, 9 and 24 h after training to ®nd a hidden platform in a Morris water maze. Average escape latency (time taken to reach the platform) in all trained groups decreased progressively with increased training but data from a probe trial (quadrant analysis test) at the end of training indicated that only animals in the 9- and 24-h groups, not the 3-h group, displayed signi®cant retention of platform location. Unbiased stereological methods were used to estimate synapse and neuronal density at each time point after training. The majority of synapses had unperforated postsynaptic densities, were localized on small dendritic spines and were classed as axo-spinous. In comparison to age-matched untrained rats, signi®cant but transient increases were observed in axo-spinous synapse density and synapse-to-neuron ratio 9 h after the start of training, but not at earlier (3 h) or later (24 h) times. These changes at 9 h post-training were accompanied by transient decreases in both mean synaptic height and area of postsynaptic density. No such changes were observed in an exercisematched control group of rats, indicating that the transient synaptic changes in the dentate gyrus are most likely to be speci®cally related to processes involved in memory formation for the spatial learning task.
Introduction Cellular mechanisms of learning and memory formation are believed to involve alterations in synaptic strength (Martin & Morris, 2002). The hippocampus may be expected to play a central role in synaptic plasticity (O'Keefe & Nadel, 1978; Morris et al., 1982), which is likely to result from alterations in synaptic morphology (Marrone & Petit, 2002). Whilst unambiguous evidence of the nature of such changes in the hippocampus following memory formation has yet to be provided, studies involving a number of learning paradigms have shown a variety of morphological changes, especially alteration in dendritic spine parameters. O'Malley et al. (1998) used electron microscopy to show a transient increase (at 6 h) in spine number, in the mid-molecular layer of the dorsal dentate of rats trained on a passive avoidance task, which reverted to control levels by 72 h. The same group (O'Malley et al., 2000) also demonstrated that at 6 h post-training in a Morris water maze there was a transient doubling of spine density, in the molecular layer of the dentate, which returned to control levels after 72 h. Moser et al. (1994) used confocal microscopy on rat brain slices to show that spatial training increased spine density on CA1 pyramidal neurons, and suggested that this indicated formation of new synapses. Previously we employed quantitative electron microscopy to examine synaptic morphometry in the hippocampus of rats spatially trained in a Morris water maze (Rusakov et al., 1997). No synaptic changes were found 5 days after a 5-day training period. However, an analysis of synaptic distribution indicated a training-associated rearrangement of synaptic active zones in CA1 with synapses showing closer cluster-
ing. We concluded that spatial memory formation may involve subtle topographical changes in local hippocampal circuitry without necessarily synapse formation de novo. However, given the time of our analyses after training, the spatial reorganization of synaptic circuitry observed was more likely to represent later phases in memory consolidation rather than the processes of synaptic activation and remodelling which occur during the earlier period of memory formation. These begin with NMDA receptor activation (Bliss & Collingridge, 1993), immediate early gene expression (Bozon et al., 2002), and a cellular cascade of events leading to protein synthesis (Bourtchouladze et al., 1998; Nader et al., 2000), followed by morphological changes in the synaptic components of neural circuits (Bailey & Kandel, 1993; Moser, 1999; Poirazi & Mel, 2001). Synaptic density increases occur in the middle molecular layer of the dentate after long-term potentiation (LTP) in vivo (Stewart et al., 2000), and Abraham et al. (2002) have recently provided evidence to show that LTP meets one of the principal criteria for long-term memory storage. Here we investigated the ultrastructural morphology of synapses in the dentate gyrus of rats 3, 9 and 24 h after the start of training in a water maze to ®nd a hidden platform. Signi®cant but transient changes in synaptic morphometry were observed only in hippocampal dentate synapses of those rats which had been trained to ®nd the hidden platform.
Materials and methods Spatial memory training
Correspondence: Professor M.G. Stewart, as above. E-mail: M.G.Stewart@open.ac.uk Received 7 January 2003, accepted 6 March 2003 doi:10.1046/j.1460-9568.2003.02624.x
All experiments were approved by the University of Haifa and Open University animal ethical committees. Adult male Sprague-Dawley rats (3 months old, 300 g bodyweight) were housed in standard
1974 M. D. Eyre et al. conditions, four per cage with ad libitum access to food and water. Three groups were trained in a hidden platform version of the Morris water maze (Morris, 1984). Training was conducted in blocks, with four trials per block. Start location for each trial was randomised so that each quadrant was used as a start location within each block of four trials. A black-coloured pool 1.7 m in diameter, 50 cm high, and with water depth 30 cm, was used and the water was maintained at 23 1 8C. The platform was always located in the centre of quadrant 1 (north-east), equidistant from the edge and the centre of the pool and the two quadrant boundaries, 3 cm below the surface of the water. Allocentric visual cues consisted of high-contrast images on the walls of the experimental room. Escape latency was measured using a stopwatch for each trial, and animals were allowed to remain on the platform for 15 s, with an intertrial interval of 15 s. Animals that did not locate the platform within 60 s were guided to the platform and allowed to remain there for 15 s. For each trained group, a probe trial was given 3 h after the end of training, always starting from the centre of the perimeter of quadrant 4. The three training protocols are shown in Fig. 1: (i) a 3-h group (one block of four trials, followed by a probe trial 3 h after the start of training); (ii) a 9-h group (three blocks of four trials, each 3 h apart, followed by a probe trial 3 h after the last training trial, i.e. 9 h after the start of training); (iii) a 24-h group (®ve blocks of four trials, each 3 h apart except for 12 h between the fourth and ®fth blocks of trials, followed by a probe trial 3 h after the ®nal training trial, i.e. 24 h after the start of training). Two control groups were used, one naõÈve group with no water maze experience, and a second group allowed to `swim only', without any platform to locate. This group was introduced to the maze for the same amount of time as the average latency for each trial of all animals in the 9-h trained group, and thus experienced three blocks of four swim-only trials, each 3 h apart. These swim-only animals were not given a probe trial. Immediately following the probe trial (or 3 h after the last swimonly trial) animals were anaesthetized intraperitoneally with a mixture of urethane and chloral hydrate (40 g urethane 5 g chloral hydrate in
100 mL of 0.9% NaCl solution). Animals were then perfused intracardially with 15 mL of 0.9% saline followed by 200 mL of ®xative solution (2% paraformaldehyde and 2% glutaraldehyde in 0.1 M phosphate buffer, pH 7.4) at 7 mL/min via a peristaltic pump. Both hippocampi were dissected from the brain and 1-mm-thick perpendicular slabs were cut from the centre of the rostro-caudal axis. Slabs were prepared for electron microscopy by reaction with osmium tetroxide and dehydration in graded acetone solutions followed by in®ltration and embedding in Epon resin. More animals were trained than were analysed by electron microscopy because in a few cases the ultrastructural preservation was judged as nonoptimal. Hence the number in the 3-, 9- and 24-h groups (Figs 4±6) were two less than the number of animals actually trained (Fig. 1). Electron microscopy Ultrathin sections (70 nm thick) of silver-grey interference colour were cut from an area of the upper blade of the hippocampal dentate gyrus molecular layer using a Leica UCT ultramicrotome. Sections encompassed a region that stretched 600 mm from the blade tip toward the hilus region in one dimension, and included the dentate gyrus cell layer, the molecular layer of the granule cell dendrites and the hippocampal ®ssure in the other dimension. Not less than four sequential serial ultrathin sections were mounted onto individual copper slot grids, contrast-stained with lead and uranium salts and visualized in a JEOL 1010 transmission electron microscope. The electron scattering method was used to calculate the relative electron thickness (RET%) of each section. A standard curve which related the electron scattering method to the small-fold technique was used to convert the RET% into a section thickness measurement (De Groot, 1988). The four serial sections on a copper slot were nominally labelled A, B, C and D. The dentate gyrus middle molecular layer was identi®ed by using landmark structures in section A, and a series of digital images was taken systematically at regular intervals along the middle of the molecular layer, parallel to the upper blade, using a Gatan CCD digital camera and Digital Micrograph 3.3.1 acquisition
Fig. 1. Protocols used to train the groups of animals used in the study, indicating the temporal spacing of the training sessions. Three-hour group, n 9; 9-h group, n 10; 24-h group, n 8. Also shown is the protocol for the swim-only group, n 6. ß 2003 Federation of European Neuroscience Societies, European Journal of Neuroscience, 17, 1973±1980
Spatial learning and synaptic changes in rat hippocampus software (http://www.gatan.com). At least 12 locations were imaged per hippocampus on each of the four serial sections, the areas on sections B, C and D corresponding to those areas imaged on section A. The image scale was calibrated and a smaller area of known dimensions (8092 5783 nm) was selected from each image. These selected areas were stored as separate images and formed a data ®le of 12 sets of four aligned serial counting frames for each hippocampus. Each set of counting frames was analysed for synaptic contacts. Presumed excitatory type 1 asymmetric synaptic contacts were identi®ed by the presence of a dark area of postsynaptic density (PSD) and the presence of vesicles in the corresponding presynaptic structure; type 2 symmetric synapses, which do not contain a PSD and have ovoid-shape vesicles, are considered to be inhibitory. Synapses were distinguished also as being either axodendritic or axospinous (Fig. 2A shows two typical axospinous synapses and Fig. 2B an axodendritic synapse; all three indicated are type 1 with asymmetric PSDs). In addition, a synapse was regarded as being perforated if, in at least one of the four sections (A, B, C or D), the PSD was separated into two or more distinct areas by intervening membrane that did not contain PSD. The disector technique (Sterio, 1984) was applied to all 12 pairs of B and C counting frames in each data ®le. To avoid sampling errors associated with objects lying completely within the two planes of the disector, 70-nm sections were considered optimal. The originally acquired images were used to verify synapses which intersected the inclusion and exclusion boundaries, and two additional guard sections (A and D) were used to aid the identi®cation of synapses in the z-axis.
1975
The total number of `speci®c' synapses, i.e. those present in the lookup but not the reference frame, was then used to calculate the synapse numerical density (NV). Optimally, the counting of 100 objects is required for accurate, yet ef®cient, disector calculations (Sterio, 1984). The sampling method used in this study resulted in an average of 161 synapses being counted from each hippocampus of each animal, with only three data sets containing
