Manfred's Column, September/October 2005: Learning, LTP and the hippocampus


September/October 2005

Learning, LTP and the hippocampus

From an evolutionary perspective, the hippocampus is an old part of the brain, which lies deep in the temporal lobe and has, for a long time, been associated with learning and memory. Two studies shed a new light on this structure. In one, it has been proven that the hippocampus is also involved in unconscious learning, and in the other, the relationship between the modification of synaptical strengths through LTP (long-term potentiation) and the hippocampus and learning, respectively, has been more closely linked.

LTP takes place in special glutamate receptors, the so-called NMDA receptors. These receptors have a special characteristic: glutamate can only activate the NMDA receptors when the membrane in which the receptors are located is strongly depolarised. Thus NMDA receptors register if two connected neurons are simultaneously active. Only when the postsynaptic neuron is depolarised, i.e. is active, and an impulse from the presynaptic neuron arrives at the same time, are the conditions for the activation of NMDA receptors, and thus LTP, given, and learning takes place (i.e. a strengthening of the transmission at the given synapse).

There have been neurobiologists who doubted this relationship. For example, Rose (1992) considered LTP as an artefact or at most an artificial model of learning, but not its mechanism [1]. The work of Tan and colleagues (1999) however, weakens this point of view [2]. Their investigations leave hardly any doubt about the relationship between LTP, NMDA receptors and learning.

NMDA receptors are subdivided into an obligatory NR1 subunit and various NR2 subunits. There are different types of NR2 subunits and these are termed NR2A, NR2B, NR2C and NR2D. The functions of an NMDA receptor depend on which NR2 subunits it is composed of. Thus it is known that the NMDA receptors of younger animals, for example, mainly contain the NR2B subunit, besides the obligatory NR1 subunit. These NR2B subunits arrange for receptor activation, which is nearly twice as long. Therefore, the changes that cause LTP and, consequently, learning, are more pronounced. These findings fit with the observation that younger organisms – whether they are birds, monkeys or humans – learn faster than older ones, which predominantly have a different NR2 subunit in their NMDA receptors.

Tang et al. used transgenic mice, which had been modified so that the NR2B subunit of the NMDA receptor – the one found in juveniles – was more highly expressed in the adult animals.. Tang et al. propounded the theory that, if these animals learned more effectively, then LTP and learning are closely interconnected.

This is exactly what they did find. First it was ensured that the genetically modified mice did not differ from the unaltered animals (the wild-type) in respect to their anatomy, microanatomy and receptor function of AMPA receptors (these mediate the normal signal transduction at glutamatergic synapses), but only in respect to their juvenile receptor type. In this way, Tang et al. could, using electrophysiological measurements, show that LTP works better in the genetically modified mice than in the wild-type mice.

It is especially meaningful that various behavioural tasks showed the genetically modified mice exhibiting superior ability in learning and memory compared to the wild-type mice. Tang et al. used an object exploration task in order to investigate recognition memory. During the training session, the mice were confronted with two objects, which they were allowed to explore for 5 minutes. In the memory testing session, one of the familiar objects used during training was replaced by a new object, which the mouse could also explore freely for 5 minutes; the point being that a mouse that remembers the familiar object will spend more time exploring the new object. Thus the time of exploration is a measurement of familiarity with (of having learned) the presented objects. When the memory test was run one hour after the training session, both the transgenic mice and the wild-type mice showed a similar preference for the new object. This indicates that both types of mice were equally able to retain the memory of the old object for one hour. However, when the memory tests were conducted either one day or three days later, the transgenic mice exhibited a much stronger preference for the new object than did the wild-type mice, indicating that the transgenic mice had a better long-term memory. Likewise, transgenic mice showed their superiority in further tests, which investigated hippocampally mediated learning. These findings clearly showed the relationship between the hippocampus, LTP and learning.

Traditionally in memory research, two fundamentally different forms of memory are distinguished: conscious, declarative memory, i.e. the memory for facts and events, and unconscious, non-declarative memory, such as the memory for skills [3]. A series of studies argues for the hippocampus being the neural substrate of declarative memory. An investigation by Chen and Phelps [4] challenges this notion by showing that the hippocampus is also involved in unconscious learning. The authors compared unconscious learning in healthy humans with unconscious learning in amnesic patients, who had hippocampal damage. They used a test that required the subjects to search visual displays for a rotated T target presented among 11 rotated L distractors. Upon detection of the target T, participants had to press one of two labelled response buttons in order to indicate whether the target T was located on the right or on the left side. The Ts and Ls appeared fully randomised across 120 trials. In another 120 trials, a previously demonstrated display was presented again. However, the participants were unaware of this, since trials with a repeated configuration and trials with an unknown, randomised configuration were presented in random order. By using these two sets of trials it could be shown that healthy subjects were able to find the T faster in the repeated trials than in the trials with an unknown, randomised configuration. This means that they learned without realising it. This unconscious learning was, however, absent in the amnesic patients. This investigation, therefore, challenges the widely held notion regarding the functioning of the hippocampus. It is clear that the hippocampus plays a role in the unconscious learning of complex visuo-spatial tasks.


[1] Rose S., (1992), The Making of Memory: From Molecules to Mind, Anchor Books, New York, pp. 238-40.
[2] Tang Y.P., et al. (1999), "Genetic Enhancement of Learning and Memory in Mice", Nature, Vol. 401, No. 6748, pp. 63-69.
[3] Schacter D.L. and E. Tulving (eds.) (1994), Memory Systems, MIT Press, Cambridge, MA.
[4] Chun M.M. and E.A. Phelps (1999), "Memory Deficits for Implicit Contextual Information in Amnesic Subjects with Hippocampal Damage", Nature Neuroscience, Vol. 2, No. 9, pp. 844-47.

(Translation of Spitzer, Manfred (2000),  “Lernen, LTP und Hippocampus”, in Geist, Gehirn & Nervenheilkunde: Grenzgänge zwischen Neurobiologie, Psychopathologie und Gesellschaft, Schattauer, Stuttgart, pp. 41-45)