SYNAPSE 59:177–181 (2006)

Short Communication

Reversible Changes in Hippocampal

CA1 Synapses Associated With
Water Maze Training in Rats
RUBÉN MIRANDA,1* EDUARDO BLANCO,1 AZUCENA BEGEGA,1
LUIS J. SANTÍN,2 AND JORGE L. ARIAS1
1
Laboratory of Psychobiology, School of Psychology, University of Oviedo, Spain
2
Psychobiology Area, School of Psychology, University of Málaga, Spain

KEY WORDS stratum radiatum CA1; hippocampus; spatial learning; synapses;

electron microscopy
ABSTRACT Long-term memories seem to require protein synthesis to be estab-
lished. This process can be related with synaptogenesis resulting in changes in the form
or even in the number or proportion of synaptic contacts. Results from behavioral stud-
ies assessing quantitative changes associated with different learning tasks are contro-
versial. The aim of our work was to assess whether the number of CA1 hippocampal syn-
aptic contacts can be modified after training in different tasks in the Morris water maze
(MWM). We found transient changes in the synaptic density of the symmetric synapses
associated with place learning. A reduced synaptic density of the symmetric synapses in
the stratum radiatum of CA1 was found at 48 h posttraining, returning to control levels
72 h posttraining. The same effect was observed 1 h after training in a nonspatial task.
Synaptic changes found in the CA1 shortly after water maze training suggest a possible
participation of the hippocampus in the acquisition of nonspatial tasks together with a
role in the short-term consolidation of spatial memory. As no changes were found in the
total number of synapses counted, it is likely that subtle changes in synaptic efficacy
than new synapse generation may be sufficient to support the acquisition and mainte-
nance of new memories. Synapse 59:177–181, 2006. V 2005 Wiley-Liss, Inc.
C

INTRODUCTION correlated with several behavioral deficits (Jeffery,

Current models of memory consolidation (Dudai, 1997). Maintenance of LTP and memory consolidation
2004; Nader, 2003) assume that the storage of long- are related with an increase in multiple spine synap-
term memory (LTM) is associated with gene expres- ses at the hippocampus (Eyre et al., 2003; Toni et al.,
sion, new protein synthesis, and synaptic remodelling. 1999). These evidences suggest that problems in long-
This synaptic remodelling can involve the modifica- term storage could be due to specific alterations in
tion of preexisting synapses or even the generation of the synaptic physiology (i.e., LTP) as well as in the
new contacts and is proposed as a key process to anatomy of synapses. However, it is not clear that in
explain the long-term maintenance of memories (Bai- unaltered conditions, increases in synapse number
ley and Kandel, 1993). Previous results from studies are always correlated with the acquisition or consoli-
to assess the hypothesis that learning promotes mor- dation of memories. Several methodological issues of
phological synaptic alterations are quite controversial. ultrastructural studies in addition to the dynamic
Specifically, there is little agreement about synaptic
changes associated to spatial memory formation. On
the other hand, there are strong evidences showing Contract grant sponsor: MEC; Contract grant number: SEJ2004-07445/PSIC.
that a functional hippocampus is required for the *Correspondence to: Ruben Miranda, Ph.D., Laboratory of Psychobiology,
University of Oviedo, Plaza Feijoo, s/n 33003 Oviedo, Spain.
acquisition of spatial tasks in the Morris water maze E-mail: [email protected]
(MWM) (Eichenbaum et al., 1990; Morris et al., Received 23 June 2005; Accepted 18 October 2005
1982). Different alterations in hippocampal synaptic DOI 10.1002/syn.20229
transmission (e.g., long-term potentiation (LTP)) are Published online in Wiley InterScience (www.interscience.wiley.com).

C 2005
V WILEY-LISS, INC.
178 R. MIRANDA ET AL.

Fig. 1. Performance in the

water maze. A: Mean distance
covered during training. Animals
trained in SP task displayed
greater distances covered during
training compared to N-SP ani-
mals. Asymptotic performance is
reached after 2 days by SP ani-
mals and after 1 day by N-SP ani-
mals. B: Mean escape latencies
performed. N-SP animals showed
lower escape latencies during
training compared to SP animals.
C: Probe test results of the SP
group. Animals showed a signifi-
cant preference for the escape
quadrant (D) compared to the
rest of quadrants (Error bars ¼
SEM; *P < 0.05). No significant
differences in the mean percent-
age of time spent in escape quad-
rant were found between SP-48
and SP-72 groups.

nature of learning and memory processes have to be both SP and N-SP tasks. N-SP animals were pre-
taken into account to explain quantitative electron vented from using room landmarks to navigate by
microscopy data (Geinisman, 2000). Furthermore, re- placing black curtains around the pool. SP animals
gional differences may exist due to a differential par- were submitted to a 1-trial probe test with no plat-
ticipation of cellular assemblies in memory processes form after the last training trial. The performance
(Eyre et al., 2003; O’Malley et al., 2000; Ramirez- was measured as distance covered and escape laten-
Amaya et al., 1999, 2001; Rusakov et al., 1997). cies (Fig. 1). Behavioral analyses (repeated measures
Together, these results suggest that hippocampal ANOVAs) revealed that acquisition of the SP task
synaptic remodelling involving changes in the num- was gradual and animals mastered the task after two
ber of synapses is likely to occur after training in spa- days of training (12 trials) (distance, days effect
tial tasks. These changes appear to be transitory and F(3,9) ¼ 6.333, P ¼ 0.013; Bonferroni posttest P <
evident at short posttraining periods. However, there 0.05). On the other hand, the N-SP task was acquired
is no consensus about the specific nature and time faster, displaying short escape latencies since the first
course of the diverse synaptic changes found in the trials reaching asymptotic performance after one day
different hippocampal subfields. The main goal of this of training (6 trials) (latency, group effect F(1,11) ¼
work was to assess whether quantitative synaptic 8.947; P ¼ 0.012). The probe test confirmed (one-way
changes occur in hippocampal CA1 area after spatial ANOVA, F(3,4) ¼ 10.427; P ¼ 0.023) that rats recall
and nonspatial MWM training. We examined the the location of the hidden platform based on room
stratum radiatum of CA1 (SR-CA1) in control rats landmarks (Bonferroni posttest P < 0.05) (Fig. 1). No
and at different time intervals after training in two significant differences were found between SP-48 and
MWM tasks, using quantitative electron microscopy. SP-72 groups (t(5) ¼
1.098; n.s.), validating subse-
A total of 23 adult Wistar rats (300þ/
50 g) were quent comparisons of learning-induced synaptic
assigned randomly to the following groups: CC, caged changes.
control (n ¼ 8); SP-48, spatial hidden platform task For electron microscopy examination, all animals
sacrificed 48 h posttraining (n ¼ 4); SP-72, spatial were transcardially perfused, under deep anesthesia,
hidden platform task sacrificed 72 h posttraining (n ¼ with a washing solution (0.12 M phosphate buffer,
4); and N-SP, nonspatial visual fixedþcurtains task 0.02 Cl2Ca, and 8% sucrose) for 5 min, followed by fix-
sacrificed 1 h posttraining (n ¼ 5). Trained animals in ative (glutaraldehyde þ paraformaldehyde
(1:1); 0.1
the MWM were habituated to the maze in one day M; pH ¼ 7.3) for 15 min. After postfixation, coronal
with 2 trials and then submitted to a 4-day task with trimmed slabs (300 lm) of the left antero-dorsal hip-
6 trials per day. The escape platform (hidden or visi- pocampus (Paxinos and Watson, 1988) were osmicated
ble) was located in the same place over training in in a 1% osmium tetroxide-buffered solution for 90 min,
 CA1 SYNAPSES AND SPATIAL LEARNING 179
In a previous study by Rusakov et al. (1997), no
quantitative synaptic changes were observed in the
molecular layer of DG and in the SR-CA1 5 days after
water maze training in a 5-day protocol. Later, a
transient increase in the number of dendritic spines
in the middle molecular layer of DG was found after
avoidance conditioning and after a 1-session spatial
learning task. These changes were observed at 6 h
posttraining returning to basal levels at 72 h post-
training (O’Malley et al., 1998, 2000). Similarly, a
transient increase of axo-spinous synapses was found
in the same layer 9 h after the beginning of a spatial
MWM training (Eyre et al., 2003). Together, these
findings suggest that a specific time-window exists
during which quantitative synaptic changes induced
by training will be evident. This posttraining period
seems to spread from 6 h to 48 h posttraining,
although some variations may emerge depending on
the type and amount of training as well as on the
brain region. For instance, synaptogenesis in the stra-
tum lucidum of CA3 was observed even 7 days after
Fig. 2. A: Quantified field in the SR of CA1. Asterisks indicate
the location of synapses. Magnification 20,0003. B: Detail of an water maze training (Ramirez-Amaya et al., 2001).
asymmetric synapse. Asterisk indicates presynaptic area. Arrow in- Previous studies have stressed the importance of
dicates the presence of postsynaptic density. C: Detail of a symmet-
ric synapse. Asterisk indicates presynaptic area.
CA1 for memory processes (Zola-Morgan et al., 1986)
and its relation with spatial learning and LTP (Tsien
et al., 1996). Other authors have shown that enriched
experience can promote increases in spines and syn-
dehydrated, and embedded in araldite. Ultrathin sec- apse density in CA1, resulting in an improvement in
tions (70 nm thick) of silver-gray interference color spatial abilities (Moser et al., 1994; Rampon et al.,
were collected on 300 mesh copper grids, counter- 2000). However, no quantitative synaptic changes
stained with uranyl acetate and lead citrate, visual- were observed 5 days after water maze training
ized in a Zeiss EM-109 transmission electron micro- (Rusakov et al., 1997). More recent evidences suggest
scope, and photographed with a 35-mm black and that learning-induced synaptic changes may occur at
white film camera. Synapses were counted in the SR- shorter posttraining periods (Eyre et al., 2003; O’Mal-
CA1 area at two different levels of the antero-dorsal ley et al., 2000).
hippocampus. For each level, systematic sampling of Hence, we examined the number of asymmetric and
two different areas of the SR-CA1 was done taking a symmetric synapses of the SR-CA1 at 48 h and 72 h
minimum of 18 photos per area. At least 36 fields after a 4-day training protocol in the SP version of
were photographed in each level, resulting in a total the MWM. To account for nonspecific synaptic changes
of 72 fields per animal. Images were taken at associated with water maze training, another group
20,0003 and, after conventional developing, negatives was trained in a nonspatial hippocampus-independent
were digitally scanned at 1200 dpi (HP photosmart task and synapses were examined 1 h posttraining.
s20). Synapses were classified as asymmetric (type I) In agreement with previous anatomical and neuro-
or symmetric (type II) (Gray, 1959). Only synapses chemical data, in control conditions almost 90% of
identified inside the measure frame were counted synapses were asymmetric (presumably excitatory)
(Fig. 2). The number of synapses per sampled area and 10% were symmetric (presumably inhibitory) in
was calculated for each type of synapse and for the the SR-CA1 (Megias et al., 2001; Vizi and Kiss, 1998).
total. Sampling area was calculated as the sum of One-way ANOVAs showed non significant changes in
measure frames overlaid in the total of quantified the density of asymmetric synapses (F(3,17) ¼ 0.381;
images. n.s.) and in the total number of synapses counted
Overall, our results reflect a transient downregula- (F(3,17) ¼ 0.552; n.s.) after training. However, a
tion of the SR-CA1 symmetric synapses 48 h after reduction in the density of symmetric synapses were
completing a spatial task (SP) and 1 h after training found at 48 h posttraining, but not at 72 h posttrain-
in a nonspatial task (N-SP) in the MWM. On the ing (F(3,17) ¼ 5.703; P ¼ 0.007, Bonferroni posttest
other hand, no differences were found in asymmetric P < 0.05). On the other hand, a significant reduction
synapses and in the total number of synapses counted in the percentage (F(3,17) ¼ 4.522; P ¼ 0.017) and
between groups. density (F(3,17) ¼ 5.703; P ¼ 0.007) of symmetric syn-
180 R. MIRANDA ET AL.

formed when neurons have less excitatory input, are

maintained by optimal activation, and are lost when
activation is too high (Harris, 1999).
The reduction in the proportion of symmetric syn-
apses found in our study could reflect a subtle reor-
ganization involving neurochemical changes of the
hippocampal synaptic transmission. Discrepancies
with studies demonstrating increases in the total num-
ber of synapses after spatial learning (Eyre et al., 2003;
O’Malley et al., 2000), may be explained by differences
in the amount or schedule of training. Although spatial
tasks can be acquired under both massed and spaced
training schedules, it is well known that LTM is stron-
ger after spaced training (Commins et al., 2003; Spreng
et al., 2002) and performance can be improved due to
overtraining correlated with CA3 mossy fiber synapto-
Fig. 3. Group comparison in the synapse density values for
asymmetric and symmetric synapses. SP-48 and N-SP groups pre- genesis (Ramirez-Amaya et al., 1999). O’Malley et al.
sented a significant decrease in symmetric synapses compared to (2000) found a significant increase in spine density in
CC animals (Error bars ¼ SEM; *P < 0.05). the mid-molecular layer of DG after a 5-trial protocol
at 6 h posttraining, and Eyre et al. (2003) used a 12-
trial protocol carried out in 6 h, finding an increase in
apses was also found 1 h after training in the nonspa- axo-spinous synapse density in DG 3 h later, whereas
tial task compared with controls (Bonferroni posttests no changes were found 3 h after a 4-trial task or 24 h
P < 0.05) (Fig. 3). This change was concomitant to a after the beginning of a 20-trial protocol fulfilled in
significant increase in the percentage of asymmetric 21 h. As opposed to the results obtained in CA3 (Ram-
synapses in the same group compared to CC (F(3,17) ¼ irez-Amaya et al., 1999, 2001), these evidences suggest
3.71; P ¼ 0.032, Dunnett posttest P < 0.05). These that, at least in DG, synaptic changes occur only in
synaptic changes found shortly after nonspatial train- short posttraining periods after massed training with
ing could be due to the enriching experience (Moser few trials. Our findings, together with those of Rusakov
et al., 1994; Rampon et al., 2000; Sandi et al., 2003) et al. (1997), indicate that similar to DG, CA1 major
in parallel to a general involvement of the hippocam- synaptic alterations are likely to be found after massed
pus even in the processing of nonspatial information training, since no changes in the total number of syn-
(Whishaw, 1998). Given that the posttraining period apses exist after training with protocols taking 4 or
in which synapses were observed was different for 5 days, even when ultrastructural examination is done
the spatial and nonspatial tasks, it is difficult to a few hours after training. Probably, subtle synaptic
establish direct comparisons in terms of spatial vs. reorganization modifications could be due to the
nonspatial memory structural correlates. Basically, extended practice rather than to memory consolidation
the present design allows us to study synaptic changes and, the difficulties in observing quantitative altera-
associated with spatial memory consolidation (SP-48 tions associated with gradual learning, can be derived
and SP-72 groups) and whether similar changes can from the possibility that multiple waves of remodelling
occur in the hippocampus shortly after water maze occur when training protocols comprise more than one
training (N-SP), related with the procedural dimen- session (Dudai, 2004; Eyre et al., 2003). Nevertheless,
sion of the task and not with hippocampus-dependent the functional role of quantitative synaptic changes
memory consolidation. after training with massed protocols that are not
Our results showing no changes in the total num- expected to promote a robust LTM remains unex-
ber of synapses counted are in agreement with earlier plained.
studies in different animal models and learning tasks In summary, transient synaptic changes were
and on spatial learning in MWM (Rusakov et al., observed in CA1 after water maze training involving
1997), and support the notion that memories follow- a downregulation of the proportion of symmetric syn-
ing new learning may not necessarily involve a net apses, although the total number of synapses counted
synaptogenesis (Geinisman, 2000). Other authors have remains stable. This synaptic remodelling was ob-
pointed out that greater neural remodelling involving served in the spatial and nonspatial versions of the
new synapse formation appears to occur when animals MWM shortly after training, suggesting a possible
require great effort to learn a task (Sandi et al., 2003, participation of the hippocampus in the acquisition of
2004). Moreover, studies on LTP suggest that neurons nonspatial tasks with a time-limited role in spatial
may homeostatically regulate input through spine memory consolidation. The absence of changes in the
number. This hypothesis predicts that more spines are total number of synapses reflects that, under spaced
 CA1 SYNAPSES AND SPATIAL LEARNING 181
training, physiological changes in synaptic efficacy, tial learning in adult rats suggests the formation of new synapses.
Proc Natl Acad Sci USA 91:12673–12675.
instead of new synapse generation, may be sufficient Nader K. 2003. Memory traces unbound. Trends Neurosci 26:65–72.
to support the acquisition and maintenance of new O’Malley A, O’Connell C, Regan CM. 1998. Ultrastructural analysis
reveals avoidance conditioning to induce a transient increase in
memories. Further studies are required to clarify the hippocampal dentate spine density in the 6 hour post-training
functional role of synaptogenesis in memory processes. period of consolidation. Neuroscience 87:607–613.
O’Malley A, O’Connell C, Murphy KJ, Regan CM. 2000. Transient
spine density increases in the mid-molecular layer of hippocampal
ACKNOWLEDGMENTS dentate gyrus accompany consolidation of a spatial learning task
in the rodent. Neuroscience 99:229–232.
The authors thank D. Carlos Villa and D. Fernando Paxinos G, Watson C. 1998. The rat brain in stereotaxic coordinates.
London: Academic Press.
Jañez for their expert technical support inside the Ramı́rez-Amaya V, Escobar ML, Chao V, Bermúdez-Rattoni F. 1999.
electron microscopy facility of the University of Synaptogenesis of mossy fibers induced by spatial water maze
overtraining. Hippocampus 9:631–636.
Oviedo, and Dr. Sandra Rubio for her comments on Ramı́rez-Amaya V, Balderas I, Sandoval J, Escobar ML, Bermúdez-
the manuscript. Rattoni F. 2001. Spatial long-term memory is related to mossy
fiber synaptogenesis. J Neurosci 21:7340–7348.
Rampon C, Tang YP, Goodhouse J, Shimizu E, Kyin M, Tsien JZ.
REFERENCES 2000. Enrichment induces structural changes and recovery from
nonspatial memory deficits in CA1 NMDAR1-knockout mice.
Bailey CH, Kandel ER. 1993. Structural changes accompanying Nature Neurosci 3:238–244.
memory storage. Annu Rev Physiol 55:397–426. Rusakov DA, Davies HA, Harrison E, Diana G, Richter-Levin G,
Commins S, Cunningham L, Harvey D, Walsh D. 2003. Massed but Blis TVP, Stewart MG. 1997. Ultrastructural synaptic correlates
not spaced training impairs spatial memory. Behav Brain Res of spatial learning in rat hippocampus. Neuroscience. 80:69–77.
139:215–223. Sandi C, Davies HA, Cordero MI, Rodriguez JJ, Popov VI, Stewart
Dudai Y. 2004. The neurobiology of consolidations, or, how stable is MG. 2003. Rapid reversal of stress induced loss of synapses in CA3
the engram? Annu Rev Psychol 55:51–86. of rat hippocampus following water maze training. Eur J Neurosci
Eichenbaum H, Stewart C, Morris RGM. 1990. Hippocampal repre- 17:1–10.
sentation in place learning J Neurosci 10:3531–4. Sandi C, Cordero MI, Merino JJ, Kruyt ND, Regan CM, Murphy
Eyre MD, Richter-Levin G, Avital A, Stewart MG. 2003. Morphologi- KJ. 2004. Neurobiological and endocrine correlates of individ-
cal changes in hippocampal dentate gyrus synapses following spa- ual differences in spatial learning ability. Learn Mem 11:244–252.
tial learning in rats are transient. Eur J Neurosci 17:1973–1980. Spreng M, Rossier J, Schenk F. 2002. Spaced training facilitates
Geinisman Y. 2000. Structural synaptic modifications associated long-term retention of place navigation in adult but not in adoles-
with hippocampal LTP and behavioral learning. Cereb Cortex 10: cent rats. Behav Brain Res 128:103–108.
952–962. Toni N, Buchs PA, Nikonenko I, Bron CR, Muller D. 1999. LTP pro-
Gray EG. 1959. Axo-somatic and axo-dendritic synapses of the cere- motes formation of multiple spine synapses between a single axon
bral cortex: an electron microscopic study. J Anat 93:420–433. terminal and a dendrite. Nature 402:421–424.
Harris K. 1999. Structure, development, and plasticity of dendritic Tsien JZ, Huerta PT, Tonegawa S. 1996. The essential role of hippo-
spines. Curr Opin Neurobiol 9:343–348. campal CA1 NMDA receptor-dependent synaptic plasticity in spa-
Jeffery K. 1997. LTP and spatial learning–where to next? Hippo- tial memory. Cell 87:1317–1326.
campus 7:95–110. Vizi ES, Kiss JP. 1998. Neurochemistry and pharmacology of the
Megı́as M, Emri ZS, Freund TF, Gulyás AI. 2001. Total number and major hippocampal transmitter systems: synaptic and nonsynaptic
distribution of inhibitory and excitatory synapses on hippocampal interactions. Hippocampus 8:566–607.
CA1 pyramidal cells. Neuroscience 102:527–540. Whishaw IQ. 1998. Place learning in hippocampal rats and the path
Morris RGM, Garrud P, Rawlings JNP, O’Keefe J. 1982. Place navi- integration hypothesis. Neurosci Biobehav Rev 22:209–220.
gation impaired in rats with hippocampal lesions. Nature 297:681– Zola-Morgan SM, Squire LR, Amaral DG. 1986. Human amnesia
683. and the medial temporal region: enduring memory impairment
Moser MB, Trommald M, Andersen P. 1994. An increase in dendritic following a bilateral lesion limited to field CA1 of the hippocam-
spine density on hippocampal CA1 pyramidal cells following spa- pus. J Neurosci 6:2950–2967.