Controlling the Ups and Downs of Synaptic Strength

See allHide authors and affiliations

Science  14 May 2004:
Vol. 304, Issue 5673, pp. 973-974
DOI: 10.1126/science.1098805

Both increases and decreases in synaptic efficiency can be triggered by activation of the N-methyl-D-aspartate (NMDA) type of glutamate receptor. How can activation of the same receptor lead to diametrically opposite changes in synaptic strength? This conundrum has long puzzled neuroscientists interested in how information is stored in the vertebrate brain. According to Liu et al. (1) on page 1021 of this issue, the answer to the puzzle appears to lie, at least in part, in the subunit composition of the NMDA receptor.

The dominant model of activity-dependent synaptic plasticity is long-term potentiation (LTP). This phenomenon, first described more than 30 years ago, occurs in the hippocampus, a cortical structure that in humans is essential for the formation of new episodic memories. In synapses displaying LTP, a brief high-frequency burst of synaptic activity leads to an immediate and enduring increase in synaptic efficiency (see the figure) (2). In contrast, more prolonged, lower frequency synaptic activity gives rise to long-term depression (LTD). In both cases, antagonists of the NMDA receptor block induction of the long-lasting change (2, 3). The channels formed by NMDA receptors are permeable to calcium, and a rise in the concentration of calcium in the postsynaptic cell is required for both LTP and LTD. In an influential paper, Lisman proposed that long trains of low-frequency synaptic stimulation, producing a modest but prolonged rise in intracellular calcium ions (Ca2+), led to LTD via the activation of phosphatases. In contrast, high-frequency trains, producing a greater rise in intracellular Ca2+, led to LTP via the activation of protein kinases (4). Experiments in which the intracellular Ca2+ concentration was manipulated supported the predictions of the Lisman model (5). The new results of Liu et al. (1) put flesh on the bones of an earlier suggestion (6) that the subunit composition of the NMDA receptor provides the gateway to LTP or LTD.

Secret liaisons.

Distinct NMDA receptor subtypes direct changes in synaptic strength induced by different patterns of neural activity. Stimulation of nerve fibers projecting to hippocampal pyramidal cells leads to the release of the neurotransmitter glutamate from presynaptic terminals. Glutamate binds to all glutamate receptor subtypes on the postsynaptic membrane, including NMDA receptors (the only type of glutamate receptor depicted). (A) High-frequency stimulation leads to the induction of LTP mediated by NMDA receptors containing the NR2A subunit. (B) In contrast, low-frequency stimulation leads to LTD mediated by NR2B-containing NMDA receptors. The composition and signaling properties of the receptor-associated protein complex (RAC) may be different for the two NMDA receptor subtypes. NR1 subunits (green) are obligatory members of NMDA receptors.


NMDA receptors are multimeric proteins containing, in hippocampal neurons, two obligatory NR1 subunits, usually paired with two NR2A or two NR2B subunits (see the figure) (7). Liu et al. used two NR2B-specific antagonists (ifenprodil and Ro25-6981) together with a recently developed NR2A antagonist (NVP-AAM077) to dissect subunit involvement in LTP and LTD at synapses on pyramidal cells in area CA1 of the hippocampus. Both forms of synaptic plasticity can be blocked by the NMDA receptor antagonist D-APV. The new and unexpected result is that LTP is mediated by NMDA receptors containing NR2A subunits, whereas LTD requires activation of receptors containing NR2B subunits (see the figure).

These results are satisfyingly convincing and explain the puzzling difficulty of inducing LTD in older animals: The NR2B subunit is maximally expressed in neonatal animals, and thereafter its levels decline. However, although this is the case for pyramidal cells in the hippocampus, the developmental gradient in granule cells of the hippocampal formation is much less steep, and qualitatively there is little difference in NR2B expression in neonatal and adult animals. One would therefore expect that LTD might be easier to induce in the dentate gyrus of adult animals than in area CA1. In fact, the reverse seems to be the case (8). As Liu et al. point out, the rule they have uncovered in area CA1 is not universally obeyed. In layer 2/3 of sensory cortex, for example, different rules seem to apply (9).

Appropriate patterns of stimulation can drive synaptic efficacy up or down at single hippocampal synapses (10). Given the results of Liu et al. (1), this bidirectionality implies that individual synapses on pyramidal cells express mixed populations of NMDA receptors, some with NR2A subunits and others with NR2B subunits.

How do the two NMDA receptor subtypes direct the two patterns of stimulation to the appropriate signal transduction pathway, leading to LTP on the one hand and LTD on the other? The decay time constant of synaptic responses generated by NR1-NR2B receptors is ∼300 ms, compared to 50 ms for NR1-NR2A receptors (7). Single stimuli, even when the cell is at or near its resting potential, generate small NMDA receptor-mediated Ca2+ fluxes. Stimulation at frequencies in the range 1 to 3 Hz that generate LTD might allow a temporal summation of responses via NR1-NR2B receptors that would not occur with the more rapidly decaying responses mediated by NR1-NR2A receptors. In this way, NR2B subunits could mediate the buildup of postsynaptic Ca2+ to the moderately increased tonic level required to induce LTD. The problem comes with the high-frequency trains that induce LTP. Both NR2A- and NR2B-containing NMDA receptors will be strongly activated by such trains. What is so special about NR2A-containing receptors such that only they control induction of LTP? One prosaic possibility is that even in young animals there are too few NR2B receptors to generate the required Ca2+ signal. Alternatively, the development of LTP may depend on specific and highly local interactions, perhaps between Ca2+ permeating through the channel during intense stimulation and Ca2+-dependent elements in the receptor-associated signaling complex (11) attached to the cytoplasmic tails of NR2A subunits (see the figure).

A question for future research concerns the role that subunit-specific synaptic plasticity plays in information processing in the intact, learning animal (12). Another type of plasticity that has recently come to prominence is induced by pairing afferent activity with action potentials that are backpropagated from the cell body into the dendritic tree (12). Here, timing rather than duration is all-important; LTP develops if the presynaptic action potential precedes the dendritic action potential by a few milliseconds, whereas LTD occurs if the timing is reversed. Does the latter protocol in some way engage NR2B receptors, and the former, NR2A receptors? Timing will tell.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
View Abstract

Navigate This Article