---

Plasticity of the circuitry that wires the brain is a fundamental process that is thought to underlie behavior, cognition, learning, and memory (1-5). The development of new synapses, the activity-dependent changes in the strength of existing synapses, and the elimination of synapses have been proposed to form the basis of this plasticity. N-methyl-D-aspartate-type glutamate receptors (NMDARs) expressed at excitatory glutamatergic synapses are crucial for triggering synapse plasticity and pathological neurotoxicity, and their level at the synapse critically regulates brain function (6). The NMDAR plays a pivotal role in the cellular mechanisms thought to underlie learning and memory by acting as a "coincidence detector" (6) to trigger changes in synapse strength that initiate the rewiring of neural networks: In response to afferent activity-induced depolarization of the postsynapse coincident with presynaptic transmitter release, calcium influx through the NMDAR triggers the active insertion or removal of α-amino-3-hydroxy-5-methylisoxazole-4-proprionic acid-type glutamate receptors (AMPARs) (7). Plasticity models that increase synaptic strength are termed long-term potentiation (LTP), whereas those that decrease synaptic strength are termed long-term depression (LTD). Thus, AMPARs are thought to be responsible for the expression of synaptic plasticity, and NMDARs are thought to be responsible for its control.

Synaptic States: A Mechanism of Dictating Synaptic Function
A key role of synaptic plasticity is to allow the synapse to operate over a large dynamic range. Two possible models could explain the behavior of synapses over this range. In the first, synapses undergo changes in efficacy by adjusting their strength along a continuum, in such a way that the properties of strengthening or weakening occur in a graded fashion with fixed underlying mechanisms. Alternatively, synapses might exist in discrete states that represent and underlie different levels of efficacy. In the continuum model, the cellular mechanisms regulating the insertion or removal of AMPARs do not vary anywhere across the whole dynamic range of synapse strength. In the state model, the underlying properties of synapses change with alterations in synaptic strength. This would result in synapses existing in different and discrete states with regard to plasticity.

My recent work has shown that synapses undergoing LTP or LTD do so by moving between different discrete electrophysiologically defined states (Fig. 1) (8). What state a given synapse is in, and what state it has recently occupied, then determine the ability of that synapse to undergo future synaptic plasticity and, moreover, what mechanisms it employs to do so. Previous studies examining LTP or LTD could not reveal synaptic states because they were recording from large populations of synapses. Population responses average out the diverse behavior and properties of individual synapses. I was able to reveal heterogenous synaptic states by employing paired whole-cell recordings to examine the plastic properties of single or small populations of synapses that exist between individual pairs of connected neurons. From these experiments, we have defined the following five states: Silent, Active, Recently Silent, Potentiated, and Depressed, with each state being defined by the ability of both AMPARs and NMDARs to undergo activity-dependent regulation.

Experimentally, synapses are almost always encountered in either the Active or Silent states in hippocampal slices, and these synapses can be driven to the other three states by synaptic activity protocols that induce LTP or LTD. Synapses in the Silent state can be awakened through exclusively postsynaptic mechanisms (9). Synapses in this state are crucial to providing the circuit with the largest possible dynamic range for increasing synaptic strength. These synapses have only NMDARs, and in this state, NMDARs cannot undergo LTD (10). Synapses can be awakened to the Recently Silent state via the insertion of AMPARs into the postsynaptic membrane. In this state, neither AMPARs nor NMDARs can be removed from the membrane. This property of synapses in the Silent and Recently Silent state has great potential importance to neural circuitry through the prevention of widespread glutamate receptor down-regulation or synapse elimination, thereby protecting any information stored by Silent or Recently Silent synapses. With time, short-term subcellular changes induce transition into the Active state, in which synapses can be readily potentiated or depressed as neural activity dictates. The plasticity of both AMPARs and NMDARs in this state makes these synapses the most adaptable in the face of incoming information. AMPARs are added to move synapses to the Potentiated state, where only metabotropic glutamate receptor (mGluR)-dependent depression is observed. Because mGluRs and NMDARs respond preferentially to different synaptic stimulation protocols, this suggests that state transitions could result in state-dependent "tuning" of synapses to listen to different input characteristics. Both AMPARs and NMDARs are endocytosed when Active synapses are driven to the Depressed state, or in fact are silenced, through LTD--representing a potential prelude to synapse elimination. Because currents carried by the NMDAR control and trigger synapse plasticity, the observed state-dependent regulation of NMDARs is of critical importance for future plasticity and remodeling of neuronal circuitry.

Synaptic States Enhance Information That Can Be Carried by a Synapse
Discrete states preserve synaptic heterogeneity, because no single activity protocol can alter all synapses in the same way, thus maintaining the dynamic range of the synaptic population. Synaptic plasticity that occurs in a state-dependent manner increases the information-carrying capacity of a synapse. In a continuum model, information is coded solely in the current strength of the synapse, whereas a state model adds the history of the synapse to the information coded, because the ability to undergo and mechanism for undergoing further plasticity is dictated by the previous plastic changes that have occurred at that synapse. For example, synapses potentiated from the Active state (Potentiated state) differ from synapses potentiated from the Silent state (Recently Silent) in that one can be depressed and one cannot. Even unsilenced synapses that have since transitioned to the Active state retain this memory, in that their synaptic depression differs in receptor dependence from Potentiated synapses. State-dependent plasticity represents a new paradigm for understanding both the mechanistic underpinnings of synaptic plasticity and perhaps the roles that plasticity could serve in higher brain functions. The finding that synapses exist in different plastic states demonstrates that the information-carrying capacity of a single synapse is greater than previously recognized and fundamentally changes our understanding of the way in which information is processed in neural circuits.

References

1. D. L. Alkon, T. J. Nelson, FASEB J. 4, 1567 (1990).
   
2. J. C. Eccles, The Physiology of Synapses (Academic Press, New York, 1964), xi, 316.
   
3. D. Hebb, The Organization of Behavior (Wiley, New York, 1949).
   
4. T. V. P. Bliss, G. L. Collingridge, Nature 361, 31 (1993).
   
5. E. R. Kandel, J. Cell. Physiol. 173, 124 (1997).
   
6. R. Dingledine, K. Borges, D. Bowie, S. F. Traynelis, Pharmacol. Rev. 51, 7 (1999).
   
7. R. Malinow, R. C. Malenka, Annu. Rev. Neurosci. 25, 103 (2002).
   
8. J. M. Montgomery, D.V. Madison, Neuron 33, 765 (2002).
   
9. J. M. Montgomery, P. Pavlidis, D. V. Madison, Neuron 29, 691 (2001).
   
10. J. M. Montgomery, J. Selcher, J. Hanson, D.V. Madison, in preparation.