Impaired Spatial Learning after Saturation of Long-Term Potentiation

See allHide authors and affiliations

Science  25 Sep 1998:
Vol. 281, Issue 5385, pp. 2038-2042
DOI: 10.1126/science.281.5385.2038


If information is stored as activity-driven increases in synaptic weights in the hippocampal formation, saturation of hippocampal long-term potentiation (LTP) should impair learning. Here, rats in which one hippocampus had been lesioned were implanted with a multielectrode stimulating array across and into the angular bundle afferent to the other hippocampus. Repeated cross-bundle tetanization caused cumulative potentiation. Residual synaptic plasticity was assessed by tetanizing a naı̈ve test electrode in the center of the bundle. Spatial learning was disrupted in animals with no residual LTP (<10 percent) but not in animals that were capable of further potentiation. Thus, saturation of hippocampal LTP impairs spatial learning.

An important prediction of the hypothesis that activity-dependent synaptic plasticity in the hippocampus (such as LTP) plays a critical role in certain kinds of learning (1, 2) is that physiological saturation of synaptic weights should disrupt new memory encoding. Saturation of an intrinsic pathway can be viewed as a neural state in which no further potentiation is feasible, at least for a period of time, at any site in the pathway (3). Repeated tetanization at a single site in the perforant path has been reported to block spatial learning when leading to cumulative LTP in the dentate gyrus (4), but this result has not been replicated successfully (5–7), and some studies have even found enhanced learning (8). The reasons for this failure clearly include the possibility that the hypothesis is wrong but also that tetanization at a single site in the perforant path produces saturation only at selected synaptic loci and only along part of the longitudinal axis of the hippocampal formation (9).

Saturation is most likely to be achieved by an electrode array that straddles an afferent pathway and by a stimulation protocol that consists of multiple tetanization episodes with cathodal stimulation at different cross-sectional sites. The variable success of such an arrangement must be assayed by a separate test-stimulation electrode that selectively (but randomly) samples fibers within that pathway. If LTP can still be induced by tetanization of the test electrode, saturation cannot be claimed to have occurred. Thus, to reinvestigate the relation between saturation and spatial learning, we induced LTP through a multielectrode array across the angular bundle of the perforant path fibers in rats.

To increase the sensitivity of animals to the saturation of plasticity at synapses that might be used for learning, we first decided to decrease the volume of hippocampal tissue by making unilateral ibotenic acid lesions of the hippocampus and dentate gyrus (10). Two weeks later (day 14), the specially designed array of three bipolar stimulating electrodes and one recording electrode was implanted into the nonlesioned side of the brain. Two electrodes were implanted so that they straddled the angular bundle of the perforant path at the point passed by a high proportion of cortical afferents destined for the dorsal hippocampus (Fig. 1A) (11). The vertical placement of each electrode was adjusted so that the use of either the tip or the shaft of one of these concentric electrodes as a cathode and the use of either the tip or the shaft of the other electrode as an anode resulted in high-amplitude dentate field potentials (Fig. 1B). The field potentials were recorded by means of an electrode in the hilar zone of the ipsilateral dentate gyrus. Acute mapping experiments that were conducted under urethane anesthesia in nonlesioned animals revealed that cross-bundle stimulation was able to induce >10-mV amplitude field potentials at sites extending from the septal pole and along the dorsal 60% of the longitudinal axis of the dentate gyrus. We did not record signals in the temporal part of the hippocampus, which is unable to support spatial learning with the present training protocol (12).

Figure 1

Saturation of LTP in perforant path synapses of the dentate gyrus. (A) The placement of two bipolar electrodes on each side of the medial and lateral parts of the angular bundle (black bars) and one bipolar electrode in the middle (white bar). (Left) Induction of LTP by cross-bundle tetanic stimulation (horizontal and diagonal lines), with anode and cathode sites at different sides of the bundle (a, b, c, and d). (Right) A test of LTP saturation by tetanic stimulation through the naı̈ve central electrode at the end of the experiment. (B) Representative evoked potentials shown before cross-bundle stimulation (left traces), after the final cross-bundle stimulation session (middle traces), and after the residual LTP induction by the central electrode (right traces). The top and bottom rows show traces from a HF- and a LF-stimulated rat, respectively. Arrows indicate tetanic stimulation episodes. (C) The normalized values for the EPSP slope for HF- and LF-stimulated rats (means ± SEM, high stimulation intensity, six responses per animal per session). The recording was conducted at 1.5-hour intervals, except for the last session, which occurred 7 hours after the end of the last tetanic stimulation. The cross-bundle stimulation (arrows) was delivered immediately after the third through seventh recording sessions at 0, 1.5, 3, 4.5, and 6 hours. The HF stimulation gradually increased the fEPSP slope values. Error bars indicate SEM. The dotted line indicates the mean EPSP slope during baseline recording. (D) Representative traces taken during tetanic stimulation at 0 s and at 4-s intervals during the minute after tetanization. There is an absence of afterdischarges. (E) The change in the EPSP slope at the highest stimulation intensity for 1 hour after tetanic stimulation at the central stimulation electrode at the end of the experiment. The animals were considered to have residual LTP if the slope of the fEPSP was enhanced by >10% at this pulse intensity. Error bars indicate SEM. The dotted line indicates the mean EPSP slope during baseline (2 min before tetanization).

The third stimulating electrode was positioned between the other two and aimed at the center of the perforant path (Fig. 1A). This served as a low-frequency (LF) test electrode during both baseline recording and induction of cumulative LTP. It also served as the tetanization test electrode to check whether the cumulative LTP that was induced from the other electrodes was saturated. An animal could be said to have saturated LTP if the cumulative LTP had reached an asymptote and if the later attempt to induce LTP from this separate electrode was unsuccessful. Once positioned, the electrodes were cemented in place, and the animals were allowed to recover from the acute effects of surgery for 2 weeks.

High-frequency (HF) tetanization was then conducted on a single day (day 28) with a cathode on one side of the bundle and an anode on the other side. All possible combinations of cathode (tip or shaft) and anode (tip or shaft) were used (Fig. 1A) (13). The animals were placed in dark, enclosed chambers in which, to reduce the attenuation of LTP by stress (14), they had been familiarized on the preceding 3 days. After baseline responses had been sampled, five series of cross-bundle tetanization episodes were given, starting at 0, 1.5, 3, 4.5, and 6 hours after the last baseline recording. The fifth episode was an anode and cathode arrangement that was identical to the first episode in order to check whether there would be further cumulative potentiation. Low-frequency control animals received the same stimulation sequence of cathode and anode locations, but only single pulses were given at each location. Nonstimulated (NS) controls, with electrodes implanted, were handled and placed in the recording chambers.

The stimulation resulted in cumulative LTP with waveforms showing a gradual increase in the early rising portion of the extracellular field potential (Fig. 1B, middle trace) over the course of the recording period (Fig. 1C). Little change was seen immediately after the first tetanization episode, possibly because the test electrode used for the measurement of the degree of potentiation was in the center of the angular bundle. The field potential slope at 7 hours after the last tetanization session was significantly elevated above the pre-tetanization baseline in the HF group and significantly elevated above the LF group [groups: F(1,13) = 6.7, P < 0.05 ; groups × session: F(4,52) = 3.3, P < 0.05 ] ( F is the variance ratio and P is the probability). The level of LTP after the fifth episode of tetanization was comparable to the level after the fourth episode, with the mean LTP level of fibers in the center of the perforant path being comparable to the level obtained in studies where the tetanization electrodes were placed in the center of the bundle (4–7). Thus, cross-bundle stimulation did not induce a greater magnitude of LTP than previous studies did, but the cross-bundle stimulation may have induced LTP on a higher proportion of fibers afferent to the hippocampal formation. The trend toward a slight decline in slope in the LF group may be a temperature effect (15), as these animals became less active across the recording sessions; however, no direct recordings of hippocampal temperature with implanted thermistors were made in this study. High-frequency cross-bundle tetanization did not result in seizures in traces recorded at 4-s intervals for 1 min after each tetanization in a subset of eight animals (Fig. 1D) (16).

After the last recording session, all animals were trained in an open-field water maze to find a platform that was hidden at a single location in the pool (17). All animals showed a decline in escape latency across the 10 trial blocks of training (Fig. 2A). Analysis revealed significant effects of groups [ F(2,24) = 5.4, P < 0.01 ] and groups × block [F(18,216) = 2.4, P < 0.001] that reflect the higher mean escape latency of the tetanized group toward the end of training. Probe tests, in which the pneumatic platform was kept submerged for the first 40 s of the trial before raising, showed a gradual increase in time spent in a platform zone of 35-cm radius around the center of the platform in blocks 1, 6, 8, and 11 (the final probe test) (Fig. 2B). Low-frequency and NS test animals showed the most focused searching in the correct zone, with representative swim paths shown in Fig. 2C. High-frequency test animals showed a distribution, with some animals doing quite well but with most animals swimming all over the pool with no spatial bias toward the target area. Statistical analysis revealed significant groups × quadrants [ F(6,72) = 8.5, P < 0.001 ] and groups × quadrants × probe test [ F(18,216) = 2.2, P < 0.005 ] interactions.

Figure 2

The effect of LTP saturation on performance in a water maze learning task (means ± SEM). (A) The latency to enter the platform of rats receiving HF stimulation, LF stimulation, or no stimulation at all. Error bars indicate SEM. (B) The development of spatial behavior across trial blocks in tetanized rats (with and without residual LTP at the central stimulation electrode) and LF and NS control rats. The search time in a circular area (radius of 35 cm) around the platform zone was measured during the first 40 s of four trials with the platform submerged to the bottom of the pool. The dotted line indicates the chance level. Error bars indicate SEM. (C) Records of the search pattern of a representative animal from each group during the final spatial probe test (60 s). (D) Time spent inside a circle (radius of 35 cm) around the platform position (black bar) and in corresponding, equally large zones in the three other pool quadrants (diagonally striped, horizontally striped, and white bars) during the final spatial probe test (60 s). The dotted line indicates the chance level. Error bars indicate SEM.

The reason for the distribution of the search pattern by individual animals in the HF group became apparent when we returned the animals to the recording chamber to examine the extent to which the cumulative LTP that was previously observed reflected a true saturation of synaptic plasticity (18). The critical test involved the use of the stimulating electrode located at midbundle as a new site at which to induce LTP (Fig. 1E). Up to this point, the midbundle electrode had only been used for LF test pulses. Midbundle tetanization gave LTP [defined as a >10% enhancement of the slope of the field excitatory postsynaptic potential (fEPSP)] in all LF test animals. The HF group was divided into animals that showed <10% LTP on the test pathway (the saturated subgroup; n = 7) and animals that showed >10% LTP (the nonsaturated subgroup; n = 6). Analysis of the potentiation induced on the test pathway showed a significant effect [groups F(2,15) = 10.3, P < 0.005 ] (19). The hypothesis that saturation of LTP will result in a learning deficit predicts that the saturated subgroup should have learned less about the location of the hidden platform than the nonsaturated subgroup. This prediction was upheld (Fig. 2, C and D). An analysis of variance (ANOVA) of the proportion of time spent in the target zone during the final transfer test revealed an overall difference between groups [ F(3,26) = 7.5, P < 0.001 ]. Subsequent planned orthogonal comparisons revealed that the animals with >10% residual LTP did not differ from the LF group ( F = 1.1, not significant), but these two groups performed better than the animals with <10% residual LTP ( F = 7.7, P < 0.025 ). These three groups, all of which had electrodes implanted and were stimulated, also performed more poorly than the NS controls. Thus, successful saturation of LTP did impair spatial learning in the water maze.

These results uphold a key prediction of the “LTP and learning” hypothesis and can explain previous failures to see the effects of cumulative LTP on spatial learning (5–7). First, previous studies used only single bipolar tetanization electrodes in the angular bundle of the perforant path and may have activated only a small proportion of the entorhinal afferents. Thus, some studies would succeed in seeing a behavioral effect of tetanization and others would not. Second, our use of animals with a unilateral hippocampus may have increased the sensitivity of the behavioral task to a disturbance of synaptic plasticity in the dorsal hippocampus, which is the region of hippocampal formation whose integrity is essential for this form of spatial learning (12). Third, previous studies did not check whether the cumulative LTP was, in practice, saturated. Assuming that our test electrode sampled a representative subset of fibers traveling in the angular bundle, its use constitutes an independent identification of animals that show saturated LTP from those that merely show cumulative LTP. Although it was not possible to induce further LTP by means of the test electrode in the subset of animals that failed to learn where the platform was located, we do not know the proportion of maximally potentiated synapses in these animals (3). However, the effects of saturation of LTP on subsequent learning are likely to follow a sigmoidal function where deleterious effects will be observed well before a saturation maximum is achieved (7).

The fact that impaired and nonimpaired animals in the tetanized group received identical stimulation suggests that a blockade of learning after saturation of LTP is unlikely to be caused by nonspecific side effects of the HF stimulation of large populations of fibers (20). Although such side effects remain a theoretical possibility, their deleterious effects on behavior would have had to covary with the capacity to induce residual LTP on the terminals of the perforant path. The induction of seizures could be such a factor (6), but afterdischarges were not seen with our stimulation paradigm.

The procedure of cross-bundle tetanization of the perforant path demonstrably induced LTP in the dentate gyrus, but the procedure may also have induced LTP in the terminal zone of the perforant path in area CA3 or may have affected synaptic transmission at synapses at the outer dendritic portion of area CA1, where fibers emanating from layer III entorhinal cells terminate. Some LTP may also have been induced transsynaptically (21). Thus, this tetanization procedure does not speak directly to the issue of whether a blockade of dentate LTP alone is sufficient to impair spatial learning. The possibility that dentate LTP is unimportant has recently been raised by studies of mice harboring mutations of genes that affect dentate but not CA1 LTP (22). Further analyses of this mutant have revealed, however, that some residual LTP is present when studied in freely moving mice (23).

Several current models of hippocampal function emphasize its role as a distributed associative memory system that is responsible for capturing event-related information online with an LTP-like synaptic mechanism (2, 24). In these models, the distributed nature of information representation within the hippocampus and dentate gyrus provides opportunities for pattern completion in response to partial cues. Also, these models predict that artificial saturation of synaptic weights across a substantial proportion of cortical afferents should disrupt the representational capacity of the system and hence disrupt learning. Our results support those models in indicating that saturation of LTP can disrupt one form of hippocampal-dependent learning.

The link between LTP and learning rests on three pillars: blockade, saturation, and erasure. The disruption of spatial learning associated with a blockade of hippocampal LTP is well established (25). The present findings reestablish the predicted impairment of learning after saturation of LTP. However, it remains to be shown that an erasure of LTP causes forgetting.

  • * To whom correspondence should be addressed. E-mail: edvard.moser{at}


View Abstract

Stay Connected to Science

Navigate This Article