Essence of Memory

Visit the link below to download the full version of this book:

https://medidownload.com/product/essence-of-memory/

Click Download Now

 Preface

The following volume stems from a meeting of the same name ‘‘The Essence of Memory’’ held in Montreal
from May 12 to 14, 2007 organized by the editors of this volume. The editors would like to acknowledge
the support of the Groupe de Recherche sur le Système Nerveux Central, Université de Montréal, for the
organization of the meeting. This meeting brought together an international group of investigators
studying memory from the molecular, cellular, physiological and behavioral levels. The goal of the meeting
was to see how these levels could talk to one another and to inform researchers in this field. In addition, for
this book, we have recruited a number of additional leading investigators to fill in gaps necessitated by the
limited time for presentations at the meeting. Several of these were from other attendees of the meeting, and
several were by invitation. By trying to discuss the ‘essence’ of memory, we mean to distill the fundamental
issues in the memory field that should transcend boundaries and hopefully eventually lead to an
understanding of memory at the molecular level that will be able to be translated into pharmacological and
behavioral treatments of memory disorders.
The volume begins with cellular and molecular approaches to understanding memory. There is an
overview chapter by Dr. Sossin on molecular memory traces, discussing how our growing understanding of
the molecular and cellular basis of the memory trace opens up opportunities and challenges for cellular and
systems understanding of memory. In particular, an argument is made for multiple independent memory
traces that are formed after an experience. This is followed by a series of chapters further elucidating these
distinct molecular traces. Dr. Sacktor describes his research on PKMz, which fulfills many of the
requirements for a molecular memory trace. Another memory trace requires gene expression and local
translation. Dr. DesGroseillers describes how mRNAs are transported in neurons and Dr. Klann then
describes how the translation of these mRNAs is regulated, and how dysregulation of this pathway can
disrupt memory. Dr. Sonenberg describes his recent findings that protein translation also regulates an
important transcriptional switch that determines when long-term memory is induced. Drs. Abel and
Nguyen describe the role of another critical signal transduction pathway, production of cyclic AMP, in
multiple kinds of memory traces. Together these investigators highlight the recent advances in molecular
cognition. After identifying important molecules involved in memory formation, mice are generated either
lacking these molecules, or expressing blockers of this pathway, and then behavioral tests of memory are
used to determine the importance of the molecular pathway.
The discussion of the molecular and cellular memory trace continues with a chapter from Dr. Frey, the
discoverer of synaptic tagging. The memory trace is thought to be synapse-specific; however cellular
processes such as translation and transcription may not be this local. Dr. Frey describes the mechanism of
‘tagging’ to allow for plasticity-related proteins to be restricted to activated synapses. Dr. Wang introduces
another well-understood memory trace, the trafficking of ionotropic receptors, notably the glutamate
sensing AMPA receptors, and describes how other systems like stress impact on the ability of this plasticity
to occur. Dr. Bannerman examines memory in the absence of this pathway, suggesting interesting
dissociations between different kinds of memory based on the trace that they use. Another type of memory
trace is structural; both structural modification of synapses already present, and the generation of new
synapses. Dr. Bailey and Kandel describe the evidence that structural modifications occur during memory

ix
x

in invertebrates and then the question of whether similar changes occur in the rodent hippocampus are
discussed by Dr. Muller.
An understanding of the ‘essence’ of memory must also elucidate memory at the systems level;
understanding the anatomical constraints of the circuits underlying memory and identifying at which
synapses the molecular and cellular events occur. Dr. Turrigiano discusses her work on how intrinsic
properties are important during development to set up the system for efficient memory formation.
Dr. McBain and Dr. Lacaille both discuss the importance of plasticity in inhibitory neurons both in
controlling the memory systems, and perhaps in forming memories themselves. Dr. Lu describes his work on
how molecules such as BDNF can impact the persistent activation of neurons seen during working memory.
The volume now moves into investigations of the ‘essence’ of memory at the whole animal level, where
memory traces are examined during behavior. This problem is introduced by Dr. Castellucci in an
introductory chapter. Dr. Glanzman describes the growing importance of post-synaptic changes in
regulating the defensive reflex in Aplysia during associative conditioning; Dr. Davis describes how new
molecular techniques in the fly allow monitoring of cellular memory traces during learning and the growing
knowledge of the cellular organization of memory in the fruit fly. Dr. Suzuki describes her work on
following memory formation using recordings in the monkey limbic area as they learn a new task.
The ‘essence’ of memory must reflect an understanding of our memory, human memory. In particular, a
major justification for elucidating the molecular and cellular basis of memory traces is to help solve
problems and pathologies of memory formation in humans. Over the last decade, studies of human
memory have provided cognitive and neurobiological models of memory to explain, characterize and
organize the act of memory within a coherent framework. This effort should lead to a better understanding
of how events are typically encoded and retrieved by memory. It should also shed light on the memory
changes that accompany human development and aging, the impact of memory-related disorders such as
Alzheimer’s disease or mild cognitive impairment, and the effect of stressful biological events such as
anesthesia and surgery. This part of the volume starts with a chapter by Dr. Cowan who proposes a
theoretical view regarding the distinction between important components of memory: short-term memory,
working memory and long-term memory. He presents defining principles to dissociate those forms of
memory and discusses why working memory tasks are such good predictors of high-level performance. In
the following chapter, Dr. Rugg describes imaging studies to support a neurocognitive model of episodic
memory. The proposed model addresses the correspondence between encoding and retrieval cues and the
role of binding processes in episodic memory. Dr. Hasher describes her work on the role of distraction on
memory processes and on complex cognitive tasks. She discusses that the age-related changes in attentional
regulation can result in both costs and potential benefits on older persons’ ability to complete memory and
complex cognitive activities. Dr. Belleville introduces the notion of mild cognitive impairment, a potential
precursor state of Alzheimer’s disease. She discusses its impact on episodic memory and working memory
and proposes that both components are impaired during the earliest phase of the disease. Dr. Isingrini
presents a chapter where metamemory is conceived as a crucial component of memory performance. He
discusses findings showing that this is a component critically modified by healthy aging and by Alzheimer’s
disease. Dr. Chertkow discusses the impact of Alzheimer’s disease on memory by focusing on the notion of
semantic memory, that is memory for words and concepts. In this chapter, Dr. Chertkow proposes a
cognitive and a neurobiological model to account for the semantic memory breakdown characterizing
Alzheimer’s. Finally, Dr. Caza discusses a new area of research investigating the impact of anesthesia and
surgery on memory processes. She introduces hypotheses regarding the underlying biological models for
such an impact on memory and highlights areas for future investigation.

Wayne S. Sossin
Jean-Claude Lacaille
Vincent F. Castellucci
Sylvie Belleville
Montreal, Canada
W.S. Sossin, J.-C. Lacaille, V.F. Castellucci & S. Belleville (Eds.)
Progress in Brain Research, Vol. 169
ISSN 0079-6123
Copyright r 2008 Elsevier B.V. All rights reserved

CHAPTER 1

Molecular memory traces

Wayne S. Sossin

Department of Neurology and Neurosurgery, McGill University, Montreal Neurological Institute, BT 110,
3801 University Street, Montreal, QC H3A 2B4, Canada

Abstract: To understand the essence of memory, one must examine the working of the brain on many
levels. It is important to find the appropriate level to study the particular aspect of memory under
investigation. In this review, I will focus on insights gained from examining memory at the molecular level.
I will illustrate these insights with specific examples from examining the molecular and cellular mechanisms
underlying long-term facilitation in the marine mollusk Aplysia and long-term potentiation, studied mainly
in rodents. In particular, I will discuss how molecular memory traces are formed and focus in detail on
what role increasing the level of proteins through protein synthesis and gene expression plays in memory
formation. I will point out three important constraints from molecular work that should impact on
cognitive modeling of the nervous system: (i) the induction of plasticity depends on the ‘state’ of the
synapse; (ii) there are multiple independent molecular traces formed after experience with different half-
lives; and (iii) the requirement for the conjunction of synaptic activation and new protein synthesis implies
that new conjunctions are required to induce long-term memory formation.

Keywords: Aplysia; long-term facilitation; long-term potentiation; protein synthesis; gene expression

Introduction languages or computer design. This is due to layers

of engineering software between access and writing
Brains are not computers to memory and the actual mechanism involved in
storing the memory. Being able to treat each
Since our brain is made up of neurons, not silicon module as a ‘black box’, without requiring knowl-
chips; and because the brain evolved, instead of edge of the underlying mechanism of action, is a
being designed, there are molecular constraints on strong engineering principle. The brain does not
brain function. For an alternative example, let us work this way, since it was created by evolution,
examine the storage of memory on a computer. not design. Thus, how memories are stored in the
The fundamental mechanism of memory storage brain is constrained by the limitations of biology
has changed multiple times (from electrostatic and could not be fundamentally changed as brains
memories, to magnetic memories, to semiconduc- became more complex.
tors, perhaps eventually to nanotubes), without Most neuroscientists believe that memories are
the need for changes in higher-level computer stored in the strengths of the synaptic connections
between neurons. Consequently, the properties of
neurons and the synapses that interconnect them
Corresponding author. Tel.: +1-514-398-1486; are pivotal for understanding how our memory
Fax: +1-514-398-8106; E-mail: [email protected] works. Nevertheless, just because molecules are

DOI: 10.1016/S0079-6123(07)00001-5 3
4

important does not mean that one can understand behavioral level, a number of important insights
complex mnemonic processes solely by under- have been gained into the molecular mechanisms
standing molecules or synapses; valuable insights of memory formation (Kandel, 2001). The major
are also needed from systems and cognitive level memory examined in this system is changes in the
analysis. The key from molecular studies is to defensive reflex of the animal, exemplified by the
extract the rules from which these systems can be withdrawal of the gill or siphon, based on
built up. For example, the most ambitious extrac- experience. At the simplest level, this reflex is
tion from neurons to cognitive systems was the mediated by a direct connection between sensory
parallel distributed processing concept (PDP), neurons that detect a touch and motor neurons
where cognitive scientists attempted to extract the that withdraw the gill or siphon. This reflex can be
‘essence’ of how neuronal systems work to generate modulated by experience in a number of ways
interesting models for learning (McClelland exemplifying many of the basic non-associative
and Rumelhart, 1985). A major rule that was and associative learning paradigms (Hawkins
extracted was that neurons connect to each other et al., 2006). Habituation is the decrease in the
with a specific strength, this connection can be reflex seen after multiple occurrences of a harmless
modified by some learning rule, and that each touch. This can be explained at the cellular level by
connection can be modified independently. Inter- a decrease in the strength of the sensory–motor
estingly, to make these models work, they also neuron connection, termed depression (Castellucci
required a new rule that the overall strength of et al., 1970). Sensitization is the increase in the
connections to a neuron had to remain approxi- reflex after experiencing a noxious stimulus. This is
mately the same so that as the strength of a non-associative memory since the noxious
connections increased, other connections had to stimulus does not have to be associated with the
decrease. This kind of homeostatic rule is also activation of the reflex. The noxious stimulus
present in neurons and our understanding of its causes the release of serotonin (5-HT) (Marinesco
importance for how the brain works is just and Carew, 2002), and 5-HT acts at multiple levels
beginning (Turrigiano and Nelson, 2000). This is of the circuit to increase the defensive reflex (Frost
an example of an engineering principle that is et al., 1988), including increasing the strength of
invariant, a system that works a certain way needs the sensory–motor neuron connection, a process
to encode it whether biological or silicon based. known as facilitation (Castellucci and Kandel,
However, as discussed below, there are probably 1976). Finally, if the animal is touched while it
additional important constraints not captured by receives the noxious stimulus, an additional
the PDP model that are required to understand increase in the reflex occurs that reflects the
how memories are encoded by the brain. associative change (Hawkins et al., 1983; Walters
and Byrne, 1983). Here the coupling of 5-HT and
Model systems firing of the sensory neuron leads to additional
increases in the connection between the sensory
Aplysia, a cellular model for studying memory at and motor neurons.
the molecular level One of the great strengths of this system is that a
behavioral memory has been strongly associated
The reductionist approach is a powerful way to with a specific change in synaptic strength between
determine molecular mechanisms of memory. This two identified neurons. Thus, one can study
approach assumes the mechanisms that underlie molecular mechanisms of memory at this specific
how memory works in simple systems will be synapse. Moreover, the sensory and motor neuron
conserved in higher systems due to conservation of can be cultured from the animal and in culture
molecular mechanisms affecting how synapses they re-grow their synaptic connection and one
work. Aplysia is a marine mollusk with a simple can study depression and facilitation in a reduced
nervous system. By a comprehensive examination system that mimics many of the mechanisms
of this system at the molecular, cellular and observed in the animal (Montarolo et al., 1986).
 5

The CA3–CA1 synapse in vertebrates as a cellular be encoded only by changes in synaptic strength. A
model for associative learning neuron can also change the rate at which it fires
action potentials to the same inputs, a change in
The hippocampus is the anatomical region excitability (Daoudal and Debanne, 2003; Schulz,
required for many forms of associative learning 2006). In Aplysia, 5-HT increases the excitability
in vertebrates. Also, due to its organized fiber of the sensory neuron through activation of a
tracts, it is a wonderful model for examining kinase, protein kinase A (PKA), and phosphoryla-
cellular plasticity, as it is simple to activate tion of potassium channels, S channels, that are
presynaptic fibers onto postsynaptic neurons and open at rest (Siegelbaum et al., 1982). Closing these
record the synaptic strength in a field recording. channels causes the neurons to depolarize and thus
While long-term potentiation (LTP) was discov- less input is required for them to reach the
ered in the fibers entering the hippocampus (the threshold for firing an action potential. It also
perforant pathway) (Bliss and Lomo, 1973), it has closes channels that normally prevent the firing of
been studied most extensively in the connection multiple action potentials, a process known as
between the CA3 neurons and the CA1 neurons. anti-accommodation (Klein et al., 1986). In combi-
Unlike the non-associative facilitation induced by nation, these two molecular changes allow stimula-
5-HT in Aplysia, the increase in strength between tion of the sensory neuron (a touch) to fire more
CA3 neurons and CA1 neurons is associative, action potentials, and thus lead to an increase in
requiring both presynaptic firing in the CA3 the defensive reflex. Excitability changes in Aplysia
neuron and postsynaptic depolarization in the are both seen immediately after sensitization, or
CA1 neuron. This conjunction of inputs is 5-HT addition in culture, and can last at least 24 h
required to open the N-methyl-D-aspartic acid after training (Scholz and Byrne, 1987). The
(NMDA) receptor and calcium entry through this prolonged increase in sensitization is probably due
receptor induces LTP (Bliss and Collingridge, to the persistent activation of PKA (see below).
1993). One major weakness of this system is Excitability changes have also been detected in
the lack of knowledge of the specific function of vertebrate systems. During LTP there is an increase
the CA3–CA1 connection and how changes in the in excitability of the CA1 neurons as well as
strength of these connections is related to the increases in synaptic strength (Xu and Kang, 2005).
memories encoded in the hippocampus. Never- Moreover there are multiple examples of changes is
theless, there is abundant evidence that this ionic channels in dendrites affecting the propaga-
plasticity is important for memory and many of tion and integration of inputs and these changes
the fundamental molecular insights into memory can be local, either synapse or branch specific
have come from this system (Morris, 2003). (Frick et al., 2004; Frick and Johnston, 2005). It
will be interesting in the future to determine how
important these changes are for memory. Clearly
What types of cellular changes underlie memory genetic changes that affect excitability do affect
formation? memory (Nolan et al., 2004; Hammond et al.,
2006) but it is hard to determine if their effect is to
Excitability changes change the state of the neuron to undergo
plasticity, or is intrinsically involved in the
The major advantage of changing memories at plasticity itself. For example, a genetic change that
synapses is the ability to strengthen specific inputs reduces the ability of a dendritic region to
to a neuron based on those inputs coupling with depolarize will inhibit the ability of coincident
the firing of the output neuron. This can lead to the inputs to activate NMDA channels and thus
forming of neuronal associations underlying asso- inhibit the ability to induce plasticity. In contrast,
ciative memory as was famously outlined by Hebb NMDA channel opening leading to calcium entry,
(Cooper, 2005). However, while this review will activation of kinases and phosphorylation of an
focus on synaptic plasticity, not all memories need ion channel that leads to a long-term increase in the
6

ability of local excitatory postsynaptic potentials and additional PKA-dependent changes in the
(EPSPs) to propagate down the dendritic branch probability of release that have not yet been
would be an example where excitability changes are identified (Klein, 1994; Byrne and Kandel, 1996).
intrinsically involved in the memory trace. At later points, 5-HT can increase n both due to
the activation of silent synapses and the formation
Changes in synaptic strength of new synapses (see Chapter 10 by Bailey and
Kandel, this volume). Finally, 5-HT can also
Synaptic strength is regulated by three parameters, increase the number of alpha-amino-3-hydroxy-5-
the probability of release (p), the size of the methyl-4-isoxazolepropionic acid (AMPA) recep-
response to a single vesicular release (q) and tors as measured by the response to extracellular
number of functional synapses (n). All are glutamate (Trudeau and Castellucci, 1995; Chit-
probably changed in distinct memory traces. There wood et al., 2001), however it has not yet been
are also many different mechanisms to regulate shown that these receptors are synaptic (and thus
each of these parameters. For example, the value alter q) and the contribution of the increase in
of q is determined by: (i) the number and efficacy these receptors to sensitization is still under debate
of postsynaptic receptors; (ii) the amount of (see Chapter 17 by Glanzman, this volume).
transmitter/vesicle (which can be regulated by the
number of vesicular transporters); (iii) the efficacy
of neurotransmitter uptake; and (iv) the ionic Potentiation in CA3–CA1 synapses is mainly
concentrations that determine the driving force for determined by changes in the reception of
the ionotropic receptors. All of these can be neurotransmitter
modulated by plasticity (Levenson et al., 2002;
Malinow and Malenka, 2002; Woodin et al., 2003; In CA3–CA1 neurons early LTP is due to
Shen and Linden, 2005; Takamori, 2006). increased levels of AMPA receptor activity. Thus,
there is an increase in the reception of neuro-
transmitter or q, the size of the response to the
Facilitation in Aplysia sensory–motor neuron release of a single vesicle. This is due to increased
synapses is mainly determined by changes in the activity of AMPA receptors already present in the
probability of release synaptic membrane, the insertion of more AMPA
receptors into previously active synapses and the
In Aplysia, depression, the cellular model for insertion of AMPA receptors into silent synapses
habituation, is due to a decrease in p and (Liao et al., 1995). Changes in the probability of
facilitation (Castellucci and Kandel, 1974), the release have been inferred at later times after
cellular model for sensitization, is due to increases plasticity, due to a measured increase in the
in p (Castellucci and Kandel, 1976). The decrease probability of release (Bolshakov et al., 1997;
in p is not linked to a decrease in the frequency of Zakharenko et al., 2001) and the requirement of
spontaneous release (Eliot et al., 1994) and is presynaptic changes for prolonged potentiation
mainly due to a decrease in calcium-secretion after some stimulation paradigms (Huang et al.,
coupling (Zhao and Klein, 2002), the ability of 2005). It is not clear if later stages of LTP at
calcium to induce release of vesicles. Facilitation is CA3–CA1 synapses are due to increased numbers
coupled with an increase in spontaneous release of synaptic connections (see De Roo et al.,
(Dale et al., 1988), however this increase is Chapter 11).
mediated by 5-HT activation of protein kinase C The difference in the locus of the change in
(PKC), but inhibitors of PKC do not block short- Aplysia sensory–motor neuron synapses and
term facilitation (STF); thus this increase may not CA3–CA1 synapses does not reflect a fundamental
be related to facilitation (Ghirardi et al., 1992). difference between invertebrates and vertebrates,
Instead, facilitation is caused both by increased but is simply a function of the choice of model
calcium entry due to action potential broadening, synapses. Aplysia still retains the mechanisms
 7

required for insertion of AMPA receptors and this memory) or static (such as memory encoded on a
may play a role in associative sensitization at magnetic disc). In biological memory, this differ-
sensory–motor neuron synapses and probably at ence is not a dichotomy but a continuum where
other synapses that have not been examined in different types of memory storage require different
detail (Roberts and Glanzman, 2003). Moreover, amounts of energy to maintain. For example, if
there are numerous examples of LTP in vertebrates memory is encoded by changes in phosphorylation
that are mainly mediated by changes in p, notably (i.e. due to production of a persistently active
the mossy fiber synapses connecting the dentate kinase), the memory is volatile and inhibiting the
gyrus to CA3 (Zalutsky and Nicoll, 1990). production of the kinase should eliminate the
memory. In contrast, if memory is encoded by
constructing new synapses, inhibiting production
Molecular memory can be dissociated into temporal of new proteins should not immediately remove
phases the synapse, and thus, the memory is more static.
Below, we will examine the molecular memory
Memory has been roughly divided into three traces at the two model synapses under discussion
temporal domains. Working memory refers to and examine what traces underlie the different
retention of events that can be represented by the temporal domains in detail, concentrating on
continued firing of neurons that represent the short- and long-term traces. For discussions of
stored information. Short-term memory refers to working memory, see Chapter 19 by Suzuki, this
memories that are vulnerable to interference until volume; Chapter 15 by Galloway et al., this
they are consolidated; at the cellular and molecular volume.
level this is often associated with the lack of a
requirement for protein synthesis. Finally long- Molecular traces of different volatility are induced
term or consolidated memory represents a long- by protein phosphorylation
term store that is less vulnerable to interference and
at the cellular and molecular level requires protein Most short-term changes in synaptic strength are
synthesis. This classification serves some purposes, due to phosphorylation, either kinases adding
but does not fully capture the cellular changes that phosphates, or phosphatases removing them.
underlie the memory trace. I would argue instead While the phosphorylation event itself is usually
that there are multiple molecular traces that are short lasting, depending on the consequence of
induced by experience that last for various times phosphorylation the trace is graded in its volatility
and do or do not require new protein synthesis. (Fig. 1). The half-life of messengers (calcium or
These traces can be independent of each other. second messengers (cyclic AMP, diacylglycerol,
If memories are stored in the strength of etc.), that activate the kinases/phosphatases ranges
synapses, how are these changes stored. Is it in from seconds to minutes. Furthermore, the change
the state or level of a single molecule, or is it a can be easily reversed, i.e. a phosphate added by a
more complex interaction between many mole- kinase can be removed by a phosphatase. Thus in
cules? One of the key problems to overcome is the Aplysia, a single 5 min application of 5-HT leads to
long lifetime of a memory compared to the lifetime activation of PKA, phosphorylation of target
of an individual molecule. Moreover, unlike molecules including ion channels and proteins
computer memories in neurons, there are distinct involved in vesicular release and an increase in
types of biological synaptic changes that last for synaptic strength that lasts for about 20–30 min.
different time periods. While memory in biological Thus, this is a highly volatile memory trace.
systems is vastly different than memory in However, transient changes in phosphorylation
computers, there is a useful analogy. In computers, sites can also lead to a longer-lasting change by
memory storage devices are divided into volatile inducing other events that may not reverse when
(such as dynamic random access memory where the site is removed. For example, insertion and
constant energy is required to maintain the removal of AMPA receptors at CA3/CA1
8

Trace mediated by
phosphorylated protein A

Trace triggered by
phosphorylation B
event (I.e. insertion
of receptor)
Induction

Trace is
structure
built by activity of C
transiently
phosphorylated protein

Volatility

Fig. 1. Different molecular traces induced by phosphorylation have different volatility. (A) The simplest memory trace involves
activation of kinase that phosphorylates a substrate. Phosphorylation of the substrate itself leads to an increase in synaptic strength
through either increasing transmitter release (e.g. SNAP-25) or transmitter reception (e.g. CAMKII phosphorylation of AMPA
receptors). Once the kinase is no longer active (degradation of second messengers), phosphatases remove the phosphate and the trace is
erased. (B) Phosphorylation can induce a change in the protein (shift from hexagon to cross) that lasts longer than the phosphorylation
itself. For example, phosphorylation of AMPA receptors may lead to insertion of the receptor that can persist after the
phosphorylation has been removed. Some additional event then has to occur to erase the memory trace. (C) Phosphorylation can
activate a protein (i.e. a regulator of actin polymerization, cofilin) that leads to a structural change (assembly of blocks into a house-
like structure) that persists after the phosphorylation is removed and requires an active process to disassemble. The three changes are
graded in their volatility. (See Color Plate 1.1 in color plate section.)

synapses requires several phosphorylation events, involved in longer-lasting changes is that they are
but the insertion or removal of the receptor may required to induce the change, but in many cases
last longer than the change in phosphorylation. not for the maintenance of synaptic changes. An
This would represent a less volatile trace, however, exception to this is when the molecular trace itself is
insertion of AMPA receptors can still be reversed a persistently active kinase (see below).
by endocytosis of the receptors, and does not
persist in the absence of additional events (i.e. the
early form of LTP lasts only for a few hours). Thus State-dependence of cellular memory traces
this trace is less volatile than a simple phosphory-
lation event, but still is easily reversed (Fig. 1). An important aspect of the memory trace is
Morphological changes that are induced by kinase its induction; what is required to induce the
activation are probably not reversed simply by memory trace. The state of the synapse is a critical
removing the phosphate off the cytoskeletal regulator of the induction step, and thus depend-
regulators involved in inducing the change, and ing on the state of the synapse, the identical
may persist in the absence of an active process that stimulus can have different effects and induce
would lead to a reversal of the cytoskeletal change. different traces. The synapse regulates its own
Thus, morphological changes could represent a ability to be plastic, or metaplasticity (Abraham
more ‘static’ change still induced by phosphorylation and Bear, 1996), at the molecular level by altering
(Fig. 1). Thus, a common finding for most kinases the rate-limiting steps in the induction of plasticity.
 9

The actions of 5-HT depend on the past history of activation and for the reversal of depression, 5-HT
the synapse requires PKC activation (Ghirardi et al., 1992).
This could be explained in two ways: (i) the mode
In Aplysia, adult animals respond to a shock of actions was the same, but based on the
similarly whether or not they were previously differences in the state of the synapse, 5-HT either
habituated, showing an increase in the reflex. activated PKA or PKC (Fig. 2A); (ii) the mode of
However, in juvenile animals, response to the actions was different, PKA phosphorylated pro-
shock occurred only after initial habituation teins important for increasing the strength of
(Rankin and Carew, 1988). Thus, dishabituation resting synapses, but a different phosphorylation
appeared to be distinguishable from sensitization. event was necessary for increasing the strength of
At the cellular level, both the reversal of depression depressed synapses (Fig. 2B). Indeed, this second
and facilitation are due to increases in transmitter model (Fig. 2B) is correct; studies of PKC
release, but at the molecular level the mechanisms activation showed that a specific isoform of PKC,
are distinct. For facilitation, 5-HT requires PKA the novel PKC Apl II, was responsible for the

PKA activity
Non-depressed
synapse
State affects
kinase activation

Depressed
PKC activity
synapse

B
Rate-limiting step

PKA activity
Non-depressed
synapse PKC
State affects
rate-limiting
Rate-limiting step synaptic mechanism

Depressed
PKC activity
synapse
PKA

Fig. 2. The state of the synapse can determine the requirements for increases in synaptic strength. (A) This figure describes the
possibility that the kinase activated can be determined by the state of the synapse. In the non-depressed synapse PKA is activated,
increases the readiness of the synaptic vesicles to be released (shift from red-green), and thus more transmitters are released after an
action potential (activity). At the depressed synapse, PKC is activated instead of PKA, but the change at the synapse is the same.
(B) This figure describes the possibility that the state of the synapse determines which kinase is required. At the non-depressed synapse,
the rate-limiting step is the readiness of the synaptic vesicles to be released. While both PKA and PKC are activated by 5-HT, only
PKA phosphorylations that increase the readiness of the synaptic vesicles to be released are affected. At the depressed synapse,
calcium-secretion coupling becomes the rate-limiting step and activation of PKC, but not PKA can remove this inhibition. The bars
represent calcium channels and the arrow represents the influx of calcium leading to the release of synaptic vesicles. (See Color Plate
1.2 in color plate section.)
10

response of 5-HT at depressed synapses, and this The state of the synapse is also regulated by
isoform was equally activated whether the synapses development
was depressed or not (Manseau et al., 2001; Zhao
et al., 2006; Sossin, 2007). Thus, the difference The state of the synapse is not only determined by
between the two states lay in the rate-limiting step the past history of the synapse, but also by the
for synaptic activation. In depressed synapses, developmental state. This has been examined in
since PKA could not remove the depression, it is detail at CA3–CA1 synapses in vertebrates where a
not sufficient to increase synaptic strength. In number of developmental switches are important
contrast, PKC could reverse depression, but could in determining the stimulation that is required for
not phosphorylate the protein that was rate limit- plasticity. In particular, developmental switches in
ing in non-depressed synapses (Fig. 2B). the ionotropic receptors are important. Both
NMDA receptors and AMPA receptors are
tetramers. For NMDA receptor, the tetramer
The rate-limiting step for GluR1 regulation depends consists of an NMDA receptor (NR)1, which is
on the past history of the synapse always the same and NR2 that consists of four
isoforms A–D, of which the most prominent are
There is an interesting parallel in state-dependence NR2A and NR2B. There is a developmental
of phosphorylation in the examination of AMPA switch from NR2B to NR2A (Sheng et al., 1994;
receptor regulation during LTP. The regulation of Cull-Candy et al., 2001). NR2B is thought to
glutamate receptor (GluR)1 receptors depends on induce plasticity more easily; this switching limits
three phosphorylation events, one mediated by plasticity in adult animals (Dumas, 2005). Indeed,
PKC, one mediated by PKA and one mediated by in animals where NR2B is expressed in the adult
calcium calmodulin-dependent kinase II (CAM- using transgenic technology, animals have
KII) (Mammen et al., 1997; Esteban et al., 2003; enhanced learning ability (Tang et al., 1999).
Boehm et al., 2006); however the important site is Similarly there are four major subunits of AMPA
determined by the previous history of the synapse. receptors GluR1-4, although each isoform also
The PKA site is normally phosphorylated at rest, contains a number of splice forms as well. GluR4
activation of phosphatases removes this site, and and a spliced version of GluR2 are abundant early
is thought to be important for the removal of this in development and are inserted into synapse more
receptor during long-term depression (LTD) easily than GluR1 (Zhu et al., 2000; Kolleker
(Malinow and Malenka, 2002). However, PKA et al., 2003). Thus, in GluR1 KOs, LTP exists in
activation is not normally required to insert the juveniles, but not in adults (Zamanillo et al., 1999;
receptor, because at rest this site is already Lee et al., 2003). These are a few examples, but
phosphorylated (Lee et al., 2000). However, many more exist, demonstrating that the molecu-
following LTD, there is now a requirement for lar constituents of the synapse regulate the ability
PKA to phosphorylate this site to ‘reverse’ the of that synapse to be modified, and thus not all
LTD (Lee et al., 2000). The CAMKII site is synapses are equal. This is an important concept
normally not phosphorylated, but is phosphory- that has not yet been used in computational
lated during LTP (Lee et al., 2000). While this site models extracted from the nervous system.
is not thought to be important for insertion
(Hayashi et al., 2000), it does lead to increased
efficacy of receptors already at the synapse Persistent activation of protein kinases
(Derkach et al., 1999). Again, after this stimula-
tion, activation of phosphatases now removes this One of the best examples of a molecular memory
site, causing a ‘reversal’ of LTP. Thus the kinases/ trace is the persistent activation of a protein
phosphatases that are important depend on the kinase. Since activation of the kinase was sufficient
previous history of the synapse. to increase synaptic strength in the short-term,
 11

prolonged activation of the kinase can lead to a persistent activation of PKA, and this is necessary
prolonged memory trace. Kinases usually consist for the persistent activation of gene expression
of two domains, a catalytic domain that is a (Chain et al., 1999). However, the increase in
phosphotransferase, and a regulatory domain synaptic strength requires phosphorylation of
containing a pseudosubstrate sequence that inhi- proteins at the synapse. At synapses, the majority
bits the enzyme in the absence of activators. of PKA consists of the RII subunit, which is not
Persistent kinase activity is generated by the degraded (Liu et al., 2004; Kurosu et al., 2007). RII
removal of the regulatory domain, leaving a subunits are distinguished from RI subunits due to
persistently active catalytic domain. In this case, their specific binding to an A-kinase anchoring
inhibition of the kinase can temporarily erase the protein (AKAP). Disruption of this binding blocks
memory, although the memory could return if the both short and long-term facilitation (LTF)
inhibitor is reversible and the mechanism of suggesting that this binding is important for
persistent activation does not depend on continued localizing PKA close to its substrates at the
activation of the kinase. synapse (Liu et al., 2004). It is not clear if the
down-regulation of the RI subunit is necessary for
the persistent activation of PKA at the synapse
Persistent kinases are important for intermediate that serves as the memory trace. It is possible that
forms of facilitation in Aplysia catalytic subunits freed from RI serve as the active
kinases during the intermediate phase of facilita-
While one pulse of 5-HT increases synaptic tion, but this has not been directly demonstrated. It
strength for 20 min, five spaced pulses of 5-HT is also not clear what protein is synthesized that is
can increase synaptic strength for much longer required for the persistent activation of PKA at the
times, at least 72 h in cultured neurons (Casadio et synapse. There is a gene expression dependent up-
al., 1999). For the first 12 h, inhibitors of PKA can regulation of the RII subunit that is required for
block the increase in synaptic strength, suggesting LTF (Liu et al., 2004). While up-regulation of the
that at this point, the memory trace is due to RII subunit could shift the balance between RI and
persistent activation of PKA (Greenberg et al., RII, it cannot generate a persistently active kinase.
1987; Hegde et al., 1997; Chain et al., 1999). A different persistent kinase is activated when
Initially (for 1–3 h), the persistent activation of one pulse of 5-HT is coupled to sensory neuron
PKA and the increase in synaptic strength depend action potentials, or when a touch is coupled to a
on protein synthesis, but not gene expression. At shock (Sutton and Carew, 2000; Sutton et al.,
later times both persistent activation of PKA and 2004). In this case an intermediate facilitation/
the increase in synaptic strength require gene sensitization is induced that is independent of
expression (Muller and Carew, 1998). PKA con- protein synthesis. This form of intermediate-term
sists of two separate proteins, a regulatory subunit facilitation (ITF) is blocked by inhibitors of PKC,
and a catalytic subunit. In Aplysia, two regulatory and not PKA (Sutton and Carew, 2000; Sutton
subunits have been described, RI a type I and RII, et al., 2004). PKCs can be cleaved in the hinge
a type II subunit (Bergold et al., 1992; Liu et al., domain separating the regulatory and cataly-
2004). The RI subunit is degraded after 5-HT tic domain to form a constitutively active catalytic
treatment by the proteosome. This requires up- domain called protein kinase M (PKM). The
regulation of the messenger ribonucleic acid coupling of 5-HT and activity is necessary both
(mRNA) encoding ubiquitin hydrolase, whose to activate the calcium-dependent isoform of PKC,
activity increases levels of free ubiquitin by Apl I (Zhao et al., 2006), and to activate a protease
liberating ubiquitin from poly-ubiquitin chains. (probably calpain) that may cleave PKC Apl I into
This suggests levels of ubiquitin are rate limiting a PKM form (Sutton et al., 2004). The substrate
for this degradation of RI (Hegde et al., 1997). The phosphorylated by PKC Apl I may be SNAP-25,
down-regulation of the RI subunit will lead to the since activators of PKC, such as phorbol esters
12

increase synaptic strength through phosphoryla- Morphological changes at pre-existing synapses

tion of SNAP-25 (Houeland et al., 2007). The
coupling of activity and 5-HT also can lead to a Morphological changes can occur either on the
longer-term facilitation (Schacher et al., 1997). It is presynaptic or postsynaptic side of the synapse.
not clear if this later form still requires persistent On the presynaptic side, there can be more docked
activation of PKC for its maintenance. vesicles and an increased size of the active zone
where vesicles are released. On the postsynaptic
side there can be an increased spine head, or
A persistently active PKC is also important for the increased size of the postsynaptic density (PSD)
maintenance of LTP where the ionotropic receptors are clustered
(see Chapter 11 by De Roo et al., this volume).
The first kinase to be identified as a persistently While one can envision initial changes occurring
active kinase after LTP was CAMKII that becomes on one side only, there is a strong correlation
persistently activated through autophosphoryla- between presynaptic morphology and postsynaptic
tion. Initial excitement came from the fact that morphology (Harris and Stevens, 1989) sugge-
CAMKII is an oligomer of 12 subunits, and if sting that overtime a change in one side will
activated by intrasubunit phosphorylation (Miller cause a corresponding change on the other. For
and Kennedy, 1986), one could envision a kinase example, recent studies show that overexpre-
that could remain active for long periods of time in ssion of PSD-95, a postsynaptic density protein
the absence of additional stimulation (Lisman, can increase presynaptic release (Futai et al.,
1985). However, further mechanistic studies 2007).
showed that continued phosphorylation was not In Aplysia behavioral sensitization was shown to
possible in the absence of calcium–calmodulin, and increase the number of docked vesicles and the size
thus the persistent activity was time-limited of active zones (Bailey and Chen, 1989). In
(Hanson et al., 1994). Moreover, inhibitors of cultures, increases in staining for synaptic vesicle
CAMKII activity do not block ongoing LTP, proteins are seen in ‘silent’ synapses within hours
although they do block the induction of LTP (Chen after stimulation (Kim et al., 2003). Interestingly,
et al., 2001). In contrast, inhibitors of PKC activity the initial increase in synaptic vesicles occurred
can block the maintenance of LTP (Malinow et al., in the absence of protein synthesis, but required
1988) suggesting a role for PKM formation. Only protein synthesis to be retained (Kim et al., 2003)
one form of PKC appears to have a PKM form in (see Chapter 10 by Bailey and Kandel, in
the hippocampus, the atypical zeta form (Sacktor this volume, for more details). It is still not clear
et al., 1993). Moreover, specific inhibitors of this when these morphological changes contribute
PKC isoform can block the maintenance of LTP to the memory trace, but it is likely to be
(Serrano et al., 2005). This isoform is not generated important for 24 h LTF when PKA inhibitors can
by proteolysis, but by translation of an alternative no longer inhibit the facilitated EPSP (Hegde
transcript that does not contain the regulatory et al., 1997).
domain (Hernandez et al., 2003). Production of this There have been a number of studies that
kinase is induced at the translational level after LTP observe morphological changes after LTP (Yuste
(Osten et al., 1996) and one model is that the kinase and Bonhoeffer, 2001), also see Chapter 11 by De
activity of PKMz activates the protein synthesis of Roo et al., in this volume. In some paradigms,
its own transcript. This positive feedback system there is an increase in FM1-43 labeled release sites
could explain both the long persistence of the and in puncta of presynaptic proteins, similar to
memory trace encoded by PKMz (weeks) and the what was observed at Aplysia presynaptic terminal
ability to erase this memory trace with inhibitors of (Zakharenko et al., 2001). Increases in spine size
PKMz (Pastalkova et al., 2006). For a more have been correlated with the same synapses
complete description of the role of PKM in memory having undergone LTP, suggesting synapse speci-
(see Chapter 2 by Sacktor, this volume). ficity of these changes (Toni et al., 2001; Fedulov
 13

et al., 2007). The increase in spine size is associated an increase in the translational capacity of the
with increases in the size of the PSD and increases synapse that is initiated at the same time as the
in the number of AMPA receptors, although the synaptic growth (Casadio et al., 1999). New
increase in spine size can be dissociated from synaptic connections are seen both in cultures
increases in AMPA receptors (Kopec et al., 2006). neurons and in the animal after sensitization
Changes in spine size are due to regulation of the training. These connections are new connections
actin cytoskeleton. Indeed LTP induces changes in between the sensory and motor neuron and there is
the actin cytoskeleton (Colicos et al., 2001) and no evidence that new connections between neurons
inhibitor of actin reorganization can block LTP that were not previously synaptically connected
(Kim and Lisman, 1999) suggesting a requirement occurs.
for these changes. It should be noted that actin Training in enriched environments leads to rode-
reorganization is required for recruitment of nts with a significantly higher number of synapses
AMPA receptors (Korkotian and Segal, 2007) than caged animals (Volkmar and Greenough,
and formation of PKMz (Kelly et al., 2007), 1972); however no specific increase in the density
and thus whether the requirement of actin of synapses after a specific learning event has been
dynamics for establishing the memory trace is due demonstrated. At CA3–CA1 synapses there is little
to its requirement for a change in spine size or due evidence for new synaptic connections forming
to other downstream actions is still an open soon after LTP (Sorra and Harris, 1998), however
question. LTP induces new filipodia (Engert and Bonhoeffer,
1999), and these may turn into synapse at later
New synaptic connections times (Nagerl et al., 2007). In other areas of the
brain, long-term imaging studies in living mice do
The formation of new synaptic connections observe new synapse formation, largely between
requires coordination between the pre- and post- already connected neurons; however stabilization of
synaptic sides. Synapse formation require a these new synapses is rare (Holtmaat et al., 2005).
number of steps: (i) new growth or sprouting from Clearly new techniques may be required to visualize
a neurite, or often from a pre-existing synapse; (ii) if new synapses are formed after learning in the
this growth has to connect to its partner and this rodent model.
connection has to initiate biochemical cascades
leading to synapse formation; (iii) the nascent
synapse has to become functional (i.e. release and Protein-synthesis/gene-expression-dependent
receive transmitter; (iv) the synapse needs to be memory
stabilized. This process has been studied exten-
sively at the sensory–motor neuron synapse in Studies using various methodologies at both the
Aplysia (see Chapter 10 by Bailey and Kandel, this cellular and behavioral level suggest that memories
volume). New growth can be seen after a single that persist require protein synthesis and gene
pulse of 5-HT; thus the requirement for multiple expression. These do not only involve using
pulses is not to induce the synaptic growth (Udo inhibitors of protein synthesis and gene expression,
et al., 2005). It requires activation of PI-3 kinase but also biochemical and genetic manipulations
(Udo et al., 2005) and often initiates from that specifically regulate translation and transcrip-
presynaptic varicosities that already contain pre- tion factors. Most convincingly, stimulation that
existing synapses (Hatada et al., 2000). While the would normally lead to a short-term change leads
induction of growth does not require protein to long-term changes when manipulations are
synthesis (Udo et al., 2005), the stabilization of conducted such that the stimulation now activates
the growth and the formation of synapses does gene expression. In Aplysia, activation of the
require protein synthesis and eventually gene cyclic-AMP response element binding protein
expression (Grabham et al., 2005; Udo et al., (CREB) is required for gene-expression-dependent
2005). Stabilization of the new synapses requires memory (Bartsch et al., 1998). After adding of an
14

antibody to a repressor of the transcription factor at the end. A major difference between these
CREB, making CREB easier to activate, a single models is that the parallel model predicts that
pulse of 5-HT was sufficient to give LTF (Bartsch short-term memories are not required for long-
et al., 1995). In vertebrates, the CREB repressor is term memories. Thus while normally both short-
translationally regulated downstream of eukaryo- and long-term memory traces are induced by
tic initiation factor 2 (eIF2)-alpha phosphorylation the same stimulus, by looking downstream of
(Costa-Mattioli et al., 2005). Decreasing eIF2- the stimulus one should be able to discover
alpha phosphorylation decreased levels of the manipulations to dissociate the two memory
CREB repressor and allowed lowering of not only stores.
the stimulation required for gene-expression- In Aplysia, the evidence is strong for the
dependent late LTP (L-LTP), but also the stimula- parallel mechanism of memory formation. First,
tions required to generate a long-term memory in a one can generate long-term increases in synaptic
number of different paradigms (Costa-Mattioli strength in the absence of short-term changes as
et al., 2007), see Chapter 5 by Costa-Mattioli and activation of gene expression in the sensory
Sonenberg, this volume. The fact that gene- neurons is sufficient to cause increases in synaptic
expression is the rate-limiting step for the forma- strength that last for 24 h in the absence of STF
tion of long-term memory has many implications (Casadio et al., 1999). However, since longer-
for how memories are stored and raises a number lasting changes do require synaptic activation
of important questions that I will address in turn. (Casadio et al., 1999), does this suggest a serial
First, is protein synthesis and gene expression pathway? To one extent it does, the longest
required to ‘consolidate’ or maintain changes in lasting memory in Aplysia is probably stored in
synaptic strength, or do they represent a parallel new synaptic connections between the sensory
mechanism to make longer-lasting memories. neuron and the motor neuron. This requires
Second, is the role of protein synthesis and/or initial activation at the synapse to induce synaptic
gene expression to make new ‘memory’ molecules growth, which in a protein-synthesis and gene-
or to increase the levels of proteins that already expression-dependent manner is eventually con-
exist, but where the number of these molecules is solidated into a new synaptic connection (see
rate limiting for synaptic strength. Finally, we will above). However, even at 24 h, this new synaptic
discuss the challenge that a requirement for gene connection is not required for either synaptic
expression means for synapse specificity. facilitation or memory. Thus, in cultures, while
new synapses are observed at 24 h, there is no
Serial vs. parallel mechanisms for protein-synthesis- difference in the level of facilitation even when
dependent memory the formation of new synapses is blocked by not
applying 5-HT to the synapse (Sun and Schacher,
A major issue in the molecular mechanism of 1998). Moreover, behavioral sensitization leads to
memory is whether long-term memory is simply a normal 24 h behavioral memory in the absence of
‘consolidation’ of short-term memory or whether new synapse formation (Wainwright et al., 2004).
long-term memory represents a parallel molecular Thus, there is another mechanism, involving
pathway. This can be made clear by an analogy increases in synaptic strength at previously
where long-term memory is represented by a brick existing synapses, that is responsible for main-
house. The serial model has postulates that the taining memory at earlier times, while in parallel
frame of the house is the short-term memory new synapses are being built. It is important to
that is then consolidated (with plaster and brick) note, that not only is gene expression required for
to solidify the structure into a long-term memory. the stabilization and consolidation of new
In contrast, memory could work similarly to the synapses, but also for the long-term changes at
houses of the three little pigs; houses of different pre-existing synapses. So in this case, the parallel
stability (straw, stick and brick) are built in pathways both involve gene-expression-dependent
parallel, but only the most stable house is present steps.
 15

In CA3–CA1 synapses there are a number of trace in the presence of an additional latent trace
indications that parallel processes occur during (a tag), but the tag is independent of the initial
LTP, although the evidence is not as strong as in memory trace. We will use the theoretical produc-
Aplysia. LTP can be induced as either early LTP tion of GluR1 in CA1 neurons to illustrate this
(E-LTP) that is protein-synthesis-independent, or concept (Fig. 3), although the same logic will be
L-LTP that is blocked by inhibitors of protein true for almost any memory trace. In the first
synthesis and gene expression (Nguyen et al., possibility, protein-synthesis-independent insertion
1994). Insertion of AMPA receptors is critical for of GluR1 receptors into synapses leads to increases
E-LTP (Malinow and Malenka, 2002), but is in synaptic strength, but perhaps the pool of
L-LTP simply anchoring of these receptors, or is receptors to be inserted has a short half-life and
it due to a parallel pathway? In genetic models must be continually generated through protein
with loss of the GluR1 subunit of the AMPA synthesis. Blocking protein synthesis will remove
receptor, E-LTP is abolished, but later forms of this pool of GluR1 and thus block the maintenance
LTP may be spared (Hoffman et al., 2002). Indeed, of plasticity; however the stimulation that generates
in these mice short-term memories show severe the plasticity does not need to regulate GluR1
impairment, while some long-term memories synthesis (Fig. 3A). In this model, no regulation of
(within the same task) are spared (Schmitt et al., translation is required for memory. The second
2003) and see Chapter 9 by Sanderson et al., this possibility is quite similar, but in this case increases
volume. As mentioned earlier, retention of L-LTP in translation of GluR1 stimulated by plasticity are
depends on translation of PKMz (Serrano et al., required to generate the replacement pool required
2005). However, overexpression of PKMz is to maintain the memory trace (Fig. 3B). Indeed,
sufficient to increase the number of AMPA LTP stimulation is known to increase local protein
receptors independently of E-LTP (Ling et al., synthesis and genetic disruption of these pathways
2002, 2006). Thus, it is possible that this phase of can lead to deficits in L-LTP and learning (see
L-LTP does not depend on the previous insertion Chapter 4 by Banko and Klann, this volume).
of AMPA receptors during E-LTP. This would Since the production of the new protein is only
suggest a parallel pathway. A major product of required to maintain insertion in the face of
gene expression downstream of L-LTP is brain turnover, the new proteins will have no effect
derived neurotrophic factor (BDNF) (Barco et al., unless the non-translation dependent insertion
2005). However BDNF acts mainly in paradigms occurred first (Fig. 3B). Finally, the third explana-
where presynaptic changes are critical for LTP tion is that increases in translation are sufficient to
(Zakharenko et al., 2003). This could act as a third generate a change in synaptic strength at synapses
parallel pathway underlying L-LTP. that have been tagged, but the tag is a parallel
pathway to the increase in synaptic strength
What is produced by protein synthesis and/or gene (Fig. 3C). In this case a signal is set in the synapse
expression? to allow newly synthesized GluR1 to be inserted
into the synapse, but this signal is independent of
The requirement for new proteins can be inter- the protein-synthesis-independent insertion of
preted in three ways. First, protein synthesis GluR1 in E-LTP. The conjunction of the signal
inhibitors block basal translation, reducing levels and the newly synthesized protein allows for a
of critical proteins required to maintain the parallel path of insertion that lasts longer than the
memory trace. The second possibility is that initial phosphorylation-mediated pathway, but
protein synthesis inhibitors block an increase in does not depend on E-LTP.
translation induced by plasticity, and these new Depending on the synapse and the plasticity, it
proteins act to maintain the memory trace by is probable that all of these explanations may
stabilizing the initial protein-synthesis-independent be appropriate. An example where translation
trace. The third possibility is that the new proteins is required but not sufficient is metabotropic
induced by plasticity are sufficient for the memory glutamate receptor (mGLUR)-mediated LTD.