Professional Documents
Culture Documents
com/
Neuroscientist
Published by:
http://www.sagepublications.com
Additional services and information for The Neuroscientist can be found at:
Subscriptions: http://nro.sagepub.com/subscriptions
Reprints: http://www.sagepub.com/journalsReprints.nav
Permissions: http://www.sagepub.com/journalsPermissions.nav
Citations: http://nro.sagepub.com/content/13/5/492.refs.html
What is This?
A defining characteristic of the brain is its remarkable capacity to undergo activity-dependent functional and
morphological remodeling via mechanisms of plasticity that form the basis of our capacity to encode and retain
memories. Today, it is generally accepted that the neurobiological substrate of memories resides in activity-
driven modifications of synaptic strength and structural remodeling of neural networks activated during learn-
ing. Since the discovery of long-term potentiation, the role of synaptic strengthening in learning and memory
has been the subject of considerable investigation, and numerous studies have provided new insights into how
this form of plasticity can subserve memory function. At the same time, other studies have explored the con-
tribution of synaptic elimination or weakening; synaptogenesis, the growth of new synaptic connections and
synapse remodeling; and more recently, neurogenesis, the birth and growth of new neurons in the adult brain.
In this review, based on work in the hippocampus, the authors briefly outline recent advances in their under-
standing of the mechanisms and functional role of these four types of brain plasticity in the context of learn-
ing and memory. While they have long been considered as alternative mechanisms of plasticity underlying the
storage of long-term memories, recent evidence suggests that they are functionally linked, suggesting the
mechanisms underlying plasticity in the brain required for the formation and retention of memories are multi-
faceted. NEUROSCIENTIST 13(5):492–505, 2007. DOI: 10.1177/1073858407302725
KEY WORDS Memory, Synaptic plasticity, Long-term potentiation, Long-term depression, Synaptogenesis, Neurogenesis, Hippocampus
The capacity to form, retain, and use memories is a fun- that memories are encoded as dynamic spatio-temporal
damental property of the brain essential for survival in all patterns of synchronized cellular activity within wide-
organisms. For example, an aplysia will learn to withdraw spread neural networks and that this dynamic, reverberat-
its gill in response to noxious stimuli; a rodent will learn ing activity progressively results in altered patterns of
to map environments to remember where it can access connectivity among the coactivated neurons. Within this
food and avoid places in which danger may be apparent. framework, any memory representation would correspond
Humans have a rich array of memories associated with with specific sets of patterns of activity in overlapping
emotion, acquired skills and habits, facts about life, and networks. The neural code embedded within these pat-
specific episodes of experiences with personal tags. terns of activity in large part defies our understanding.
Collectively, these allow us to form and constantly elabo- Nonetheless, it has long been recognized that this dynamic
rate our own definition of the world, giving us our indi- activity, transient in nature, cannot persist long enough to
viduality. How do we form memories; how are they be the actual substrate of long-term memory. Thus, it has
encoded and stored in the brain? The brain is not a passive been postulated that there should exist a second state of
recorder of experiences as if information were merely pro- memory encoded as changes at the cellular level to store
jected onto a mental screen; it is a dynamic system that these representations. A process of stabilization or consol-
creates information. To process and store a lifetime of idation would lead to what Hebb called a “structural
memories, some form of plasticity in the brain that goes trace,” a memory trace that is maintained in some form of
beyond that known to occur during early development is a dormant state but has the capacity to return to an “active
required. Following Cajal’s original ideas and Hebb’s pre- state” to evoke recall whenever a subset of the original
scient and influential dual-trace theory, it is now believed information, or related information, is available. Although
it has been suggested that once a long-term memory had
been established it was stable and remained immune to
Grant sponsor: Centre National de la Recherche Scientifique. European any form of disruption, Lewis in 1979 suggested this was
Union, Grant RTN-CT-2003-504231 to SL. Ministère de la Recherche,
ACI-NEURO-NIC-0027 to SD. Fellowship from the Fondation pour la not the case. A so-called established, or consolidated,
Recherche Médicale to EB-J. memory when reactivated enters a dynamic but fragile
state, requiring further stabilization via synaptic changes
Address correspondence to: Serge Laroche, Laboratoire de Neurobiolo-
gie de l’Apprentissage, de la Mémoire et de la Communication, UMR to be available once again for recall, a process now known
8620, CNRS, University Paris-Sud, 91405 Orsay, France (e-mail: serge as reconsolidation (reviewed in Alberini 2005). The main
.laroche@u-psud.fr). point is that long-term memories are not, as was originally
proteins of several signal-transduction cascades (Fig. 2). and colleagues (2005) reported evidence for AMPA recep-
The most well characterized of the kinases implicated in tor trafficking and insertion in synapses of a large fraction
LTP are PKC, PKA, TyrK, CaMKII, PI3K, and the of amygdala neurons following fear conditioning, and
MAPK/ERK (reviewed in Soderling and Derkach 2000). showed this was required for efficient fear memory. Kinase
Kinases are critical mediators of the early events responsi- recruitment of genomic programs in neurons leading to de
ble for the maintenance or stabilization of synaptic novo synthesis of proteins is a critical mechanism for long-
strengthening, and to date many of these kinases have been term stabilization of synaptic changes. Known for quite
implicated in the consolidation of long-term memories, some time was the fact that inhibitors of transcription or
reinforcing the idea that LTP-type mechanisms are acti- antibiotics blocking protein synthesis prevent the stabiliza-
vated during learning (e.g., Selcher and others 2002). For tion and therefore maintenance of the longer-lasting
example, numerous experiments have shown that these phases of LTP (e.g., Frey and Morris 1997) and that LTP
kinases are phosphorylated during, or immediately after, itself induces the transcription of genes in potentiated neu-
learning and that disrupting the normal activity of rons. Arguing in favor of the hypothesis that LTP mecha-
CaMKII, PKA, PKC, PI3K, or MAPK/ERK pharmaco- nisms are engaged during learning was the demonstration
logically or genetically generally causes learning and that the same inhibitors applied during or immediately
memory deficits in a variety of tasks. Activation of signal after learning prevent long-term, but not short-term, mem-
transduction pathways has two major consequences, the ory in a variety of tasks and the more recent evidence from
first being modification of the properties of synaptic recep- numerous experiments showing that learning is also asso-
tors; the second being activation of specific genomic pro- ciated with rapid gene regulation in different cell types and
grams. Modification of receptors occurs by phosphorylation brain regions as a function of the type of task.
of their subunits, change in the composition of heteromeric How is this genomic response initiated? A critical step is
receptors, recruitment of extra-synaptic AMPA receptors the expression of a class of immediate early genes (IEGs)
to synaptic sites, and mobilization of the trafficking/recy- encoding inducible transcription factors, which activate a
cling machinery to increase the number of AMPA recep- host of downstream effector genes to regulate their expres-
tors by insertion in the synaptic membrane. The exact sion (Fig. 3). This genomic response generally occurs
contribution of each of these mechanisms in memory for- within minutes after neuronal activation and is mediated
mation is not fully clarified. Recently, however, Rumpel via kinase-dependent activation of constitutively expressed
transcriptional regulators such as CREB. These activate and trafficked to the dendrites, allowing for rapid and localized
induce the expression of IEGs, the early genomic response changes at specific synapses. Many genes and proteins
required for triggering the mechanisms underlying persist- have been shown to be up- and down-regulated in a finely
ent cell modification. At present, the best-characterized tuned and coordinated manner in correlation with specific
molecular cascade involved in synaptic strengthening and aspects of LTP and memory processes. Both the full
memory consolidation is the MAPK-dependent transcrip- genomic response of neurons and the synaptic protein sig-
tional regulation of the IEG Egr1 (also known as Zif268 or nature resulting from translation of preexisting and newly
Krox24). Many studies have shown that components of this transcribed mRNAs that are associated with these
cascade such as MAPK/ERK, CREB, and Egr1 itself are processes are still, however, largely unknown and currently
activated in specific brain circuits when animals form dif- the subject of intense investigation using newly available
ferent types of memories such as fear memories, olfactory large-scale screening methods.
memories, and memories for specific environments or for A recent study adds support to the idea that synaptic
objects (reviews in Sweatt 2001; Davis and Laroche 2006). strengthening plays a central role in memory by providing
Moreover, inhibition of MAPK/ERK, CREB, or inactiva- evidence that the suppression of LTP after learning can
tion of Egr1 prevents both the maintenance of synaptic erase a previously established memory (Pastalkova and
plasticity and the formation of several types of memories others 2006). The authors showed that inhibiting PKMζ,
(e.g., Atkins and others 1998; Jones and others 2001; a kinase involved in the maintenance of LTP, can reverse
Bozon and others 2003). Other IEGs such as BDNF and or depotentiate LTP, and when injected in the hippocam-
Arg3.1 encoding proteins that are not acting as transcrip- pus even several days after spatial learning, causes retro-
tion factors but are targeted to the synapse also contribute grade amnesia, abolishing the stored memory. This
to synaptic plasticity and memory formation (Lee and oth- finding provides strong support to the idea that the main-
ers 2004; Plath and others 2006). In addition to transcrip- tenance of LTP over days is a necessary condition for
tional regulation, synaptic modification also involves the maintenance of memory (see Bliss and others 2006).
protein synthesis from preexisting mRNAs, and mRNAs To date, compelling evidence suggests that strengthening
spatial learning in the water-maze and trace eye-blink conditions, learning can rescue young neurons from
conditioning, enhance the survival of 1-week-old new- death, perhaps by increasing activity-dependent integra-
born neurons in adult rats, whereas no effect was found tion of immature neurons. Moreover, a recent study ana-
after learning tasks that do not require an intact hip- lyzing learning-induced gene expression in young
pocampus. These findings suggest that learning can res- newborn neurons provided evidence that these neurons
cue young neurons from death. Several studies confirmed are recruited into the circuits involved in spatial memory
the pro-survival effect of hippocampal-dependent learn- (Kee and others 2007).
ing (e.g., Lemaire and others 2000; Leuner and others If neurogenesis plays a role in learning and memory,
2004), but others reported no change or sometimes a then disruption of neurogenesis is likely to result in
decrease in neurogenesis (van Praag and others 2000; impaired cognitive performance. Again here, an influential
Olariu and others 2005; Snyder and others 2005). Reasons study showed that experimental reduction of neurogenesis
for these contrasting results are not fully clarified, in the adult impairs trace eye-blink conditioning but not
although several variables may account for the discrepan- hippocampal-independent delay conditioning (Shors and
cies, including differences in learning paradigms and the others 2001). The deleterious effect of neurogenesis reduc-
degree of physical activity and stress induced, and the age tion or blockade has been observed in several other tasks,
and maturation of the cells at the time of learning (reviewed including contextual fear conditioning, some forms of spa-
in Bruel-Jungerman and others 2007). Overall, however, tial learning, and performance in working memory tasks
it seems reasonable for the time being to conclude that (e.g., Madsen and others 2003; Saxe and others 2006;
certain types of hippocampal-dependent learning can pro- Winocur and others 2006), with the exception of spatial
mote hippocampal neurogenesis and that this is sensitive learning in the water-maze (Shors and others 2002), where
to task difficulty and hippocampal demand. Under these neurogenesis seems to be involved in long-term storage
viding evidence that the type of synaptic change that is foresee in the context of learning and memory in the
brought about by LTP is an essential mechanism to store adult; however, synaptic weakening, a form of “functional
memory representations and leads to retrievable memo- elimination,” also appears critical, if not as a cellular
ries. True synaptic elimination as yet is more difficult to storage mechanisms per se then at least as an essential
homeostatic mechanism for fine-tuning of neural con- The evidence as broadly outlined here suggests these
nectivity. Similarly, important advances have been made four mechanisms of plasticity are not completely inde-
in strengthening the idea that synaptogenesis, morpholog- pendent. Synaptic strengthening, demonstrated by LTP,
ical and functional synapse remodeling, and neurogenesis is often associated with LTD at neighboring synapses;
in certain brain regions are key mechanisms underlying LTD modifies the capacity for LTP, and LTP modifies
structural shaping of neural connectivity in the process of the capacity for LTD. The induction of LTP progres-
memory stabilization. sively induces functional and structural reorganization of