Maintenance of synaptic plasticity

• Dr. Harel Z. Shouval, University of Texas Medical School in Houston, TX

Memory and learning that (hopefully) last a lifetime are believed to be embedded in synaptic efficacies. There are two problems associated with the stability of memory. 1) Instability due to protein turnover and trafficking. Synaptic efficacies are encoded by the number and conformational state of synaptic proteins, such as receptors. However, synaptic proteins, are constantly trafficked in and out of synapses and have limited lifetimes due to protein turnover. This transient nature of synaptic proteins can erase the trace left by synaptic plasticity. 2) Instability due to interference between different stored patterns. Many memories can be stored within the same neural network. This is likely to cause interference between different embedded memories. When synapses have a limited dynamical range this tends to destabilize memory (Fusi, et al. 2005). Here we will only discuss the first of these two problems, and how it can be overcome.

Synaptic plasticity is expressed by changes to the synaptic transmission machinery. Such changes include receptor phosphorylation, change in receptor number and presynaptic changes in probability of release. Changes in conformational states of receptors through phosphorylation, are counteracted by dephosphorylation, and by protein turn-over and trafficking, and therefore are likely to have a limited lifetime; much shorter than the lifetime of memory. Similarly changes in number of receptors are likely to be transient due to trafficking and turnover of receptor proteins. Similarly, presynaptic changes, although their mechanism is less well explored, are likely expressed in changes in proteins, changes that are also likely to be transient. However, although changes in protein function or number are transient, the memories they encode can last a lifetime.

The problem of synaptic stability due to the transient nature of its protein substrate was first noted by Francis Crick in 1984 who suggested that cooperative interactions among proteins can overcome this problem. Crick suggested a specific but quite abstract model of such cooperativity. Some form of cooperativity underlies most models that address this problem.

Experimental background

Protein turnover and trafficking

As discussed, molecular components of synaptic transmission including postsynaptic receptor, protein-kinase, protein-phosphatase, and even structural proteins such as actin have limited lifetime. Earlier experimental estimates of metabolic turnover rate of actin yielded approximately ~ 13 – 24 days (Sedman et al., 1986). Recent analysis suggested a much shorter lifetime (2-11 hours) of post-synaptic protein (Ehlers, 2003). This short lifetime of the molecular component is likely to affect the stability of synaptic efficacies.

In addition to molecular turnover, cellular proteins diffuse within the cytoplasm or on the cell membrane and some are actively transported by specific trafficking machinery. These trafficking processes also limit memory lifetimes: the molecular signature of plasticity in a synapse is erased as key proteins move away from it. As in protein turnover, different experimental methods resulted in different estimates of the time scale of receptor trafficking (Malinow and Malenka, 2002, Adesnik et al., 2005). However, even longer estimate (~ 16 hours) of receptor dwell-time (Adesnik et al., 2005) is much shorter than the memory lifetime.

L-LTP and L-LTD

Activity of pre- and post-synaptic cells can increase or decrease synaptic efficacies. The induced increase in synaptic efficacy is called long-term potentiation (LTP) while the decrease is called long-term depression (LTD). The properties of LTP and LTD have led researchers to the belief that they are the key molecular mechanisms underlying learning and memory. The time course of LTP (and LTD) and its dependency on protein synthesis will thus provide an important clue to understand the molecular mechanism underlying the stability of memory.

In fact, different types of induction protocols can produce LTP of different lifetime: a transient change that returns to baseline within 2-4 hours or a much longer lasting change. For example, LTP induced by a single theta burst produces a transient LTP (E-LTP) but four theta bursts produce a long lasting LTP (L-LTP). Similarly, different induction protocols can produce a transient LTD (E-LTD) or a long lasting LTD (L-LTD). A key finding is that the induction of L-LTP and L-LTD depends on the synthesis of new proteins during the induction phase (Abraham and Williams, 2008). It is interesting to note that inhibition of protein synthesis prevents L-LTP and L-LTD when applied at the induction phase but not during the maintenance phase. The time scales of E-LTP and E-LTD are roughly consistent with the lifetime of synaptic protein turnover and trafficking rate. The longer lifetimes (days ~ weeks or longer) and requirement of protein synthesis of L-LTP and L-LTD suggest that protein synthesis somehow counteracts protein turnover and trafficking when stable memory is established.

Which protein(s) is synthesized during induction phase (or later phase?) and/or persistent activity of which signaling protein(s) is required for L-LTP (L-LTD)? What is the relation between protein synthesis and protein degradation during different stages of synaptic plasticity (Fonseca et al. 2006, Yi and Ehlers, 2007)? These are key questions to identify the molecular substrate(s) and mechanisms underlying the maintenance of synaptic plasticity (Otmakhov et al., 1997, Ahmed and Frey, 2005, Abraham and Williams, 2008).

Theoretical ideas

Several theoretical models have been proposed to account for the maintenance of synaptic plasticity. Below I will describe some of these theories.

The bi-stable molecular switch hypothesis

One possible mechanism that can stabilize synaptic efficacies is a bi-stable molecular switch. Suppose we have a kinase that is active when phosphorylated, and can auto-phosphorylate itself ( Figure 1a). If these kinases interact with neighboring kinase (cooperativity), the resultant protein complex can exhibit an interesting property.This kinase complex, when brought into a fully activated (phosphorylated) state by a strong synaptic stimulus, can sustain its active state via cooperative interaction. If one of the kinase subunit is dephosphorylated by phosphatase or is replaced by an unphosphorylated kinase, the neighboring active kinase in the complex will quickly turn this subunit into an active state. Thus, the protein complex can sustain its active state (memory). If, on the other hand, the incoming signal is weak, phosphatase quickly brings the protein complex back to the basal state. Thus, this protein complex has two states (“bi-stable”).

Figure 1: a. Schematic diagram of an autophosphorylation loop. Here it is assumed that phosphorylation activates the kinase, this active kinase in turn phosphorylates more of the yet unphosphorylated kinase. This kinase can be dephosphorylated by a phosphatase. b. One candidate molecule for an autophosphorylation loop is CaMKII. Twelve CaMKII molecules assemble into a holoenzyme composed of two hexameric rings of CaMKII subunits. Autophosphorylation of CaMKII takes place in a cooperative manner when two Ca2+-CaM molecules are bound to two neighboring subunits.

One candidate of such a postsynaptic kinase complex is CaMKII (Ca²⁺/CaM dependent protein kinase II). In fact, CaMKII is the key regulator of LTP induction (E-LTP) and appears to have a capacity to fulfill the requirement for bi-stable memory molecule (Lisman and Goldring, 1988). Twelve CaMKII subunits assemble into a holoenzyme structure composed of two hexameric rings of subunits (Lisman et al., 2002, Bradshaw et al. 2003). CaMKII molecules are activated by binding Ca²⁺-CaM. Autophosphorylation of CaMKII takes place in a cooperative manner when two Ca²⁺-CaM molecules are bound to two neighboring subunits in the CaMKII holoenzyme. For this reason, CaMKII requires a non-zero level of signal (Ca²⁺-CaM) to maintain its activity.

Several modeling and experimental studies were conducted to test if CaMKII can serve in fact as a bi-stable switch. Some mathematical models exhibit bi-stability (Lisman and Zhabotinsky, 2001; Miller et al., 2005), while other modeling study shows no bi-stability (Kubota and Bower, 2003). Some of the kinetic parameters of the CaMKII system are still unknown and the topology of kinetic pathways in these two modeling studies differs significantly. Further computational and theoretical works are required to determine if CaMKII autophosphorylation in fact allows bi-stability or it is robustly monostable (regardless of parameter values). On the experimental side, a systematic study was conducted by Bradshaw and co-workers to test the bistability of CaMKII-protein phosphatase system. This in vitro study, however, shows no sign of bi-stability but the CaMKII-protein phosphatase system responds sharply to Ca²⁺ signals (ultra-sensitivity) (Bradshaw et al. 2003). It is extremely important to note that switching on (phosphorylation) and off (dephosphorylation) of CaMKII must NEVER be confused with bi-stability. Although an in vitro assay differs significantly from the intra-cellular environment, the bi-stability of CaMKII in vivo has not yet been demonstrated either.

Another potential problem of CaMKII hypothesis is, while the importance of CaMKII for the induction of LTP is well-established (Lisman et al., 2002), its role in the maintenance of L-LTP or long-term memory is still unclear. Several experimental studies (Malinow et al., 1989, Otmakhov et al., 1997) show that blocking CaMKII activity during the maintenance phase did not influence L-LTP. In addition, the putative target of CaMKII, AMPA receptor, seems only transiently phosphorylated after protocols that induce long-term memory (Whitlock et al., 2006).

Protein synthesis and bi-stability of a translational switch

Figure 2: A bi-stable translational switch. A schematic diagram of a positive feedback loop between translation and a translation factor X. Such positive feedback might result in bi-stability.

A detailed model for the protein synthesis dependent induction of L-LTP has been proposed by Smolen et al. (2006). However, this model does not propose a mechanism as to how L-LTP can persist beyond the characteristic time scales of the different molecules within the network.

An intriguing possibility is that a biochemical network involving protein synthesis may serve as a bi-stable switch and contribute to synaptic stability (Blitzer et al., 2005). In fact, the molecular machinery needed for protein translation resides in dendrites (even in spines) and an increasing number of experimental studies support the notion that dendritic protein synthesis is required for many forms of long-term plasticity (Sutton and Schuman, 2006). This localized protein translation could then account for the synapse specificity of plasticity. Figure 2 shows a (hypothetical) positive feedback loop involving translation factor(s) and a kinase protein X. In this diagram, the translation factor(s) regulates the translation of a specific kinase X, which in turn up-regulates the activity of the same translation factor(s). A biochemical network with such a positive feedback loop could exhibit bi-stability.

A potential problem for this hypothesis is that inhibitors of translation inhibit L-LTP (and L-LTD) only when applied before or shortly after stimulation. A widely accepted interpretation of this finding is that protein synthesis is important only during the induction (or consolidation) phase of plasticity. However, a bi-stable feedback loop of a translation factor and a protein kinase (if it exists) may respond differentially to the inhibitors during induction phase and during maintenance phase.

The cluster model

Figure 3: The cluster theory of synaptic stability. a. Receptors in the PSD tend to cluster. When a receptor is removed due to turn over or trafficking, it is rapidly replaced by another receptor. b. A simulation of a 7x7 cluster of receptors shown at times t=0, 50 and 100 (one time unit corresponds to the average dwell-time of a receptor). While receptors are rapidly removed and replaced, the cluster remains stable. c. The number of receptors as a function of time in two independent simulations.

The cluster theory of synaptic stability (Shouval, 2005) offers an alternative account for the synaptic stability. This theory is based on several assumptions: (1) Synaptic efficacy is proportional to the number of postsynaptic receptors, for example AMPA receptors. (2) Receptors in the postsynaptic density are clustered. (3) The insertion rate of a receptor in the vicinity of other receptors in the cluster is much higher than for an isolated receptor. (4) The rate of receptor removal from the cluster is independent of interactions with other receptors in the cluster. Assumptions 1-3 are essential assumptions of this model, while assumption 4 could be altered while preserving the main features of the model.

Simulations of such networks ( Figure 3b, c) show that although individual receptors within a cluster are rapidly removed or replaced, the cluster itself remains stable for an extended period of time. The exact instantaneous number of receptors in a cluster fluctuates; however, the mean number of receptors remains stable for a time period much longer than the dwell-time of a single receptor. These receptor clusters are not in a stable steady state; instead they are in a metastable state and will eventually decay.

In fact, the life-time of clusters depends linearly on the dwell-time of receptors and increases steeply with the initial size of the cluster. For example, if we assume a receptor dwell-time of 20 minutes then the median life-time for a 9x9 cluster of receptors can be more than one year. As in other theoretical models, the experimental test of this hypothesis is necessary.

Memory rehearsal

An alternative idea is that synaptic efficacies themselves are not stable, and in the absence of neuronal activity they decay. However, network activity might be able to stabilize efficacies (Wittenberg et al., 2002, Smolen, 2007). According to this idea, the synaptic plasticity and network activity act together and stabilize synaptic weights. If a pattern is embedded in a network by changing appropriate synaptic weights, such a pattern is more likely to be spontaneously activated later, and when it is re-activated it is further potentiated counteracting the synaptic decay. Such spontaneous activation could be manifest during dreams. However, the study of Wittenberg et al. (2002) shows that although a single activity pattern can be stabilized by this mechanism but not multiple sets of patterns embedded in the same network: one dominant pattern takes over the network and eliminates the other patterns. It remains to be seen if this is a general property of this proposed mechanism or if it can be corrected with an appropriate choice of a plasticity mechanism.

Prion based memory

Yet another idea, motivated by experimental findings regarding long term facilitation in Aplysia, is that memory can be preserved by a self sustaining prion-like state (Si et al., 2003). This theory postulates a specific translation factor (cytoplasmic polyadenylation element binding protein - CPEB) implicated in the persistence of long term facilitation in Aplysia has prion-like properties. The Aplysia form CPEB (ApCPEB) contains a prion-like sequence and exists in different conformations: a native state that is inactive, and an active translation-promoting state that is self-perpetual and can change the conformation of other ApCPEB molecules from native to active state (prion-like behavior). The idea of memories being encoded by a self-sustaining prion-like state of ApCPEB is appealing and appears consistent with the experimental findings that ApCPEB is required for the formation of stable, synapse specific long-term facilitation in Aplysia. However, it is not clear how such memory can be reversed, or how to prevent spreading of ApCPEB in a prion-like state to nearby synapses. Mammalian CPEB is also important for L-LTP. Nevertheless mammalian CPEB structure is significantly different from ApCPEB; it has phosphorylation sites that are absent in Aplysia, and prion-like properties have not been demonstrated in mammalian.

References

• Abraham, W. and Williams, J. (2008). LTP maintenance and its protein synthesis-dependence. Neurobiology of Learning and Memory, 89:260-268.

• Adesnik H., Nicoll, R. and England, P. (2005). Photoinactivation of native AMPA receptors reveals their real-time trafficking. Neuron, 48:977-85.

• Ahmed, T. and Frey, J. (2005). Plasticity-specific phosphorylation of CaMKII, MAP-kinases and CREB during late-LTP in rat hippocampal slices in vitro. Neuropharmacology, 49:477-92.

• Blitzer, R., Iyengar R. and Landau I. E. (2005). Postsynaptic signalling networks: Cellular cogwheels underlying long-term platicity. Biological Psychiatry, 57:113-119.

• Bradshaw, J., Kubota, Y., Meyer, T., and Schulman, H. (2003). An ultrasensitive Ca2+/calmodulin-dependent protein kinase II-protein phosphatase 1 switch facilitates specificity in postsynaptic calcium signaling. Proc Natl Acad Sci U S A., 100:10512-7.

• Crick, F. (1984). Memory and molecular turnover. Nature, 312.

• Ehlers, M. (2003). Activity level controls postsynaptic composition and signaling via the ubiquitin-proteasome system. Nat. Neurosci., 6:231-242.

• Kubota, Y. and Bower, J. (2001). Transient versus asymptotic dynamics of CaM kinase II: possible roles of phosphatase. J Comput Neurosci., 11:263-79.

• Lisman, J. and Goldring, M. (1988). Feasibility of long-term storage of graded information by the Ca2+/calmodulin-dependent protein kinase molecules of the postsynaptic density. Proc Natl Acad Sci U S A., 85:5320-4.

• Lisman, J., Schulman, H., and Cline, H. (2002). The molecular basis of CaMKII function in synaptic and behavioural memory. Nat Rev Neurosci, 3:175-90.

• Lisman, J. E. and Zhabotinsky, A. (2001). A model of synaptic memory: a CaMKII/PP1 switch that potentiates transmission by organizing an AMPA receptor anchoring assembly. Neuron, 31:191-201.

• Malinow, R. and Malenka, R. (2002). Ampa receptor trafficking and synaptic plasticity. Annu. Rev. Neurosci., 25:103126.

• Malinow, R., Schulman, H., and Tsien, R. (1989). Inhibition of postsynaptic PKC or CaMKII blocks induction but not expression of LTP. Science, 245:862-6.

• Miller, P., Zhabotinsky, A., Lisman, J., and Wang, X. (2005). The stability of a stochastic CaMKII switch: dependence on the number of enzyme molecules and protein turnover. PLoS Biol. e107.

• Otmakhov, N., Griffth, L., and Lisman, J. (1997). Postsynaptic inhibitors of calcium/calmodulin-dependent protein kinase type II block induction but not maintenance of pairing-induced long-term potentiation. J Neurosci., 17:5357-65.

• Sanhueza, M., McIntyre, C., and Lisman, J. (2007). Reversal of synaptic memory by Ca2+/calmodulin-dependent protein kinase II inhibitor. J Neurosci., 27:5190-9.

• Sedman, G., Jeffrey, P., Austin, L., and Rostas, J. (1986). The metabolic turnover of the major proteins of the postsynaptic density. Brain Res., 387:221{30.

• Shouval, H. (2005). Clusters of interacting receptors can stabilize synaptic efficacies. Proc Natl Acad Sci U S A, 102:14440-5.

• Si, K., Lindquist, S., and Kandel, E. (2003). A neuronal isoform of the aplysia CPEB has prion-like properties. Cell, 115:879-91.

• Smolen, P. (2007). A model of late long-term potentiation simulates aspects of memory maintenance. PLoS ONE.

• Smolen, P., Baxter, D., and Byrne, J. (2006). A model of the roles of essential kinases in the induction and expression of late long-term potentiation. Biophys J., 90:2760-75.

• Sutton, M. and Schuman, E. (2006). Dendritic protein synthesis, synaptic plasticity, and memory. Cell, 127:49-58.

• Whitlock, J., Heynen, A., Shuler, M., and Bear, M. (2006). Learning induces long-term potentiation in the hippocampus. Science, 313:1093-1097.

• Wittenberg, G., Sullivan, M., and j.Z. Tsien (2002). Synaptic reentry reinforcement based network model for long-term memory consolidation. Hippocampus, 12:637-47.

• Yi J.J and Ehlers M.D. (2005). Ubiquitin and protein turnover in synapse function. Neuron, 47:629-32.

Internal references

• Eugene M. Izhikevich (2006) Bursting. Scholarpedia, 1(3):1300.

• Howard Eichenbaum (2008) Memory. Scholarpedia, 3(3):1747.

• Harel Z. Shouval (2007) Models of synaptic plasticity. Scholarpedia, 2(7):1605.

• Peter Jonas and Gyorgy Buzsaki (2007) Neural inhibition. Scholarpedia, 2(9):3286.

• Rodolfo Llinas (2008) Neuron. Scholarpedia, 3(8):1490.

• Philip Holmes and Eric T. Shea-Brown (2006) Stability. Scholarpedia, 1(10):1838.

Recommended reading

• Reymanna, K.G. and Frey, J.U. (2007). The late maintenance of hippocampal LTP: Requirements, phases, ‘synaptic tagging’, ‘late-associativity’ and implications. Neuropharmacology 52:24-40.

• McGaugh, J.L. (2000). Memory - a Century of Consolidation. Science 287:248-251

• Steward, O. (2002). mRNA at Synapses, Synaptic Plasticity, and Memory Consolidation Neuron, 36:338:340

External links

• Shouval Lab
• Lisman Lab

See also

Models of synaptic plasticity, Synapse, Synaptic plasticity

Sponsored by: Eugene M. Izhikevich, Editor-in-Chief of Scholarpedia, the peer-reviewed open-access encyclopedia

Reviewed by: Anonymous

Reviewed by: Dr. Yoshi Kubota, University of Texas, Houston, TX

Accepted on: 2009-01-06 16:05:06 GMT