Physiol Rev 93: 681–766, 2013

doi:10.1152/physrev.00032.2012

ABOUT SLEEP’S ROLE IN MEMORY

Björn Rasch and Jan Born

Division of Biopsychology, Neuroscience Center Zurich and Zurich Center for Integrative Human Physiology,
University of Zurich, Zurich, Switzerland; and Institute of Medical Psychology and Behavioral Neurobiology,
and Center for Integrative Neuroscience, University of Tuebingen, Tuebingen, Germany

L
Rasch B, Born J. About Sleep’s Role in Memory. Physiol Rev 93: 681–766, 2013;
doi:10.1152/physrev.00032.2012.—Over more than a century of research has
established the fact that sleep benefits the retention of memory. In this review we aim
to comprehensively cover the field of “sleep and memory” research by providing a
historical perspective on concepts and a discussion of more recent key findings.
Whereas initial theories posed a passive role for sleep enhancing memories by protecting them
from interfering stimuli, current theories highlight an active role for sleep in which memories
undergo a process of system consolidation during sleep. Whereas older research concentrated on
the role of rapid-eye-movement (REM) sleep, recent work has revealed the importance of slow-wave
sleep (SWS) for memory consolidation and also enlightened some of the underlying electrophysio-
logical, neurochemical, and genetic mechanisms, as well as developmental aspects in these
processes. Specifically, newer findings characterize sleep as a brain state optimizing memory
consolidation, in opposition to the waking brain being optimized for encoding of memories. Consol-
idation originates from reactivation of recently encoded neuronal memory representations, which
occur during SWS and transform respective representations for integration into long-term mem-
ory. Ensuing REM sleep may stabilize transformed memories. While elaborated with respect to
hippocampus-dependent memories, the concept of an active redistribution of memory represen-
tations from networks serving as temporary store into long-term stores might hold also for
non-hippocampus-dependent memory, and even for nonneuronal, i.e., immunological memories,
giving rise to the idea that the offline consolidation of memory during sleep represents a principle
of long-term memory formation established in quite different physiological systems.

I. INTRODUCTION 681 A. Sleep

II. OVERVIEW OF APPROACHES AND... 684
III. MEMORY REACTIVATIONS DURING... 695 Sleep is defined as a natural and reversible state of reduced
IV. SLEEP-SPECIFIC ELECTRICAL... 702 responsiveness to external stimuli and relative inactivity,
V. NEUROCHEMICAL SIGNALING... 713 accompanied by a loss of consciousness. Sleep occurs in
VI. GENETIC APPROACHES TO... 719 regular intervals and is homeostatically regulated, i.e., a loss
VII. DEVELOPMENTAL ASPECTS OF... 726 or delay of sleep results in subsequently prolonged sleep
VIII. SLEEP-DEPENDENT MEMORY... 733 (113). Sleep deprivation and sleep disruptions cause severe
IX. CONCLUDING REMARKS 736 cognitive and emotional problems (142, 634, 1243), and
animals deprived of sleep for several weeks show tempera-
I. INTRODUCTION ture and weight dysregulation and ultimately die of infec-
tions and tissue lesions (973). Sleep probably occurs in all
The capability to form memory is critical to the strategic vertebrates, including birds, fishes, and reptiles, and
adaptation of an organism to changing environmental de- sleeplike states are similarly observed in invertebrates like
mands. Observations indicating that sleep benefits memory flies, bees, and cockroaches (209).
date back to the beginning of experimental memory re-
search, and since then have been fitted with quite different Sleep in mammals consists of two core sleep stages: slow-
concepts. This review targets this field of “sleep and mem- wave sleep (SWS) and rapid-eye-movement (REM) sleep,
ory” research, which has experienced a unique renaissance which alternate in a cyclic manner (FIGURE 1A). In human
during the last three decades. Although we have aimed at nocturnal sleep, SWS is predominant during the early part
comprehensively covering the field, we might have missed and decreases in intensity and duration across the sleep
out or overlooked some aspects, owing to the vast progress period, whereas REM sleep becomes more intense and exten-
achieved in the last years. Before we begin, we will briefly sive towards the end of the sleep period. SWS is hallmarked by
introduce the core concepts of sleep and memory, respec- slow high-amplitude EEG oscillations (slow wave activity,
tively. SWA), whereas REM sleep (also termed paradoxical sleep) is

0031-9333/13 Copyright © 2013 the American Physiological Society 681

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 BJÖRN RASCH AND JAN BORN

Sleep stages
A
W
REM sleep
REM

N1

N2

N3
Slow wave sleep

Early sleep Late sleep

23.00 0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 h

B
Field potential oscillations

Slow oscillation Spindle Sharp wave-ripple PGO wave Theta activity

up-state

down-state

C
Neuromodulators

Acetylcholine Acetylcholine

Noradrenaline/ Noradrenaline/
Serotonin Serotonin

Cortisol Cortisol

FIGURE 1. Typical human sleep profile and sleep-related signals. A: sleep is characterized by the cyclic
occurrence of rapid-eye-movement (REM) sleep and non-REM sleep. Non-REM sleep includes slow-wave sleep
(SWS) corresponding to N3, and lighter sleep stages N1 and N2 (591). According to an earlier classification
system by Rechtschaffen and Kales (974), SWS was divided into stage 3 and stage 4 sleep. The first part of
the night (early sleep) is dominated by SWS, whereas REM sleep prevails during the second half (late sleep).
B: the most prominent electrical field potential oscillations during SWS are the neocortical slow oscillations
(!0.8 Hz), thalamocortical spindles (waxing and waning activity between 10 –15 Hz), and the hippocampal
sharp wave-ripples (SW-R), i.e., fast depolarizing waves that are generated in CA3 and are superimposed by
high-frequency (100 –300 Hz) ripple oscillation. REM sleep, in animals, is characterized by ponto-geniculo-
occipital (PGO) waves, which are associated with intense bursts of synchronized activity propagating from the
pontine brain stem mainly to the lateral geniculate nucleus and visual cortex, and by hippocampal theta (4 – 8
Hz) activity. In humans, PGO and theta activity are less readily identified. C: sleep is accompanied by a dramatic
change in activity levels of different neurotransmitters and neuromodulators. Compared with waking, cholin-
ergic activity reaches a minimum during SWS, whereas levels during REM sleep are similar or even higher than
those during waking. A similar pattern is observed for the stress hormone cortisol. Aminergic activity is high
during waking, intermediate during SWS, and minimal during REM sleep. [Modified from Diekelmann and Born
(293).]

682 Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 SLEEP AND MEMORY

characterized by wakelike fast and low-amplitude oscillatory integrate newly encoded memory into a long-term store.
brain activity. In addition, REM sleep is characterized by pha- Encoding and consolidation might be mutually exclusive
sic REMs and by muscle atonia. Almost 50% of sleep in adult processes inasmuch they draw on overlapping neuronal re-
humans is marked by a lighter form of non-REM sleep (stage sources. Thus sleep as a state of greatly reduced external
“N2”) that is characterized by the occurrence of distinct (wax- information processing represents an optimal time window
ing and waning) sleep spindles (FIGURE 1B) and K-complexes for consolidating memories.
in the EEG, but minor SWA. Sleep stage N2 is not discrimi-
nated from SWS in rodents. The so-called consolidation account of memory processing
was first proposed by Müller and Pilzecker (840) who,
Is sleep essential? From an evolutionary perspective, re- based on studies of retroactive interference between learn-
duced responsiveness to potentially threatening stimuli dur- ing lists of syllables, concluded: “After all this, there is no
ing sleep represents a significant danger to survival. The fact alternative but to assume that after reading a list of syllables
that almost all animals sleep strongly argues in favor of an certain physiological processes, which serve to strengthen
adaptive role of sleep in increasing the overall fitness of an the associations induced during reading of that list, con-
organism, although its exact functions are still a matter of tinue with decreasing intensity for a period of time.” (p. 196
debate (407, 1077). Sleep has been proposed as serving an in Ref. 840, cited based on Ref. 709). The consolidation
energy-saving function (82, 1311), the restoration of energy hypothesis is now widely accepted based on numerous stud-
resources and the repairing of cell tissue (875), thermoreg- ies showing that psychological, pharmacological, and elec-
ulation (973), metabolic regulation (651, 1229), and adap- trophysiological manipulations, such as interference learn-
tive immune functions (695). However, these functions ing, the administration of norepinephrine and protein syn-
could be likewise achieved in a state of quiet wakefulness thesis inhibitors or electroconvulsive shocks, can effectively
and would not explain the loss of consciousness and re- impair or enhance memory, when administered after encod-
sponsiveness to external threats during sleep. These prom- ing (e.g., Refs. 803, 1332). Importantly, these manipula-
inent features of sleep strongly speak for the notion that tions are time dependent and have strongest effects when
sleep is mainly “for the brain” (553, 625). Here, different applied immediately after learning (for reviews, see Refs.
functions have been proposed, ranging from detoxication 194, 805). The consolidation possibly involves multiple
of the brain from free radicals (594, 978), glycogen replace- waves of stabilizing processes, which exhibit different time
ment (1041) to an involvement of sleep in memory and courses and depend on different underlying processes of
synaptic plasticity (293, 1204). In this review we discuss neuronal plasticity. Recent evidence suggests that memory
this latter function, i.e., the critical role sleep serves in the traces are not consolidated once but, upon their reactiva-
formation of memory. tion by a reminder or active retrieval, undergo a period of
reconsolidation to persist for the long term (844).

B. Memory At the neuronal level, memory formation is thought to be

based on the change in the strength of synaptic connections
1. Memory processes in the network representing the memory. Encoding induces
synaptic long-term potentiation (LTP) or long-term depres-
To form and retrieve memories is a fundamental ability of sion (LTD) as major forms of learning-induced synaptic
any living organism, enabling it to adapt its behavior to the plasticity (222, 527, 616, 629, 1209). Activity reverberat-
demands of an ever-changing environment, and allowing it ing in the neuronal representation following encoding is
to appropriately select and improve the behaviors of a given thought to promote two kinds of consolidation processes,
repertoire. Memory functions comprise three major sub- termed “synaptic consolidation” and “systems consolida-
processes, i.e., encoding, consolidation, and retrieval. Dur- tion” (330). Synaptic consolidation leads to the remodeling
ing encoding, the perception of a stimulus results in the of the synapses and spines of the neurons contributing to a
formation of a new memory trace, which is initially highly memory representation, eventually producing enduring
susceptible to disturbing influences and decay, i.e., forget- changes in the efficacy of the participating synapses (e.g.,
ting. During consolidation, the labile memory trace is grad- Refs. 555, 616, 977). System consolidation builds on syn-
ually stabilized possibly involving multiple waves of short aptic consolidation and refers to processes in which rever-
and long-term consolidation processes (803), which serve berating activity in newly encoded representations stimu-
to strengthen and integrate the memory into preexisting late a redistribution of the neuronal representations to other
knowledge networks. During retrieval, the stored memory neuronal circuitries for long-term storage (418).
is accessed and recalled. This review discusses sleep’s criti-
cal role in the consolidation of memory. We assume that 2. Memory systems
whereas the waking brain is optimized for the acute pro-
cessing of external stimuli that involves the encoding of new In neuropsychology, declarative and nondeclarative mem-
information and memory retrieval, the sleeping brain pro- ory systems are distinguished depending on the critical in-
vides optimal conditions for consolidation processes that volvement of medial temporal lobe regions, particularly of

Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org 683

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 BJÖRN RASCH AND JAN BORN

the hippocampus, in the acquisition of memory (1134). De- these problems. It assumes that memories are initially en-
clarative memory encompasses 1) episodic memories for coded into a fast learning store (i.e., the hippocampus in the
events that are embedded in a spatiotemporal context (in- declarative memory system) and then gradually transferred
cluding autobiographical memories) and 2) semantic mem- to a slow learning store for long-term storage (i.e., the neo-
ories for facts that are stored independently of contextual cortex). The fast learning store ensures quick and efficient
knowledge (1218). Declarative memories can be encoded encoding of memories, even in one attempt (one-trial learn-
intentionally or unintentionally, but are typically explicitly ing). Yet, these representations are unstable and vulnerable
(i.e., with awareness) accessible by active recall attempts. to (retroactive) interference by newly encoded information.
Episodic memories are learned very quickly, i.e., in one Over time, the information is gradually integrated in the
trial, but are also subject to fast forgetting (1332). Semantic slowly learning long-term store without overwriting older,
memories can be regarded as a result of the repeated encod- more remote memories. It is assumed that by the repeated
ing or activation of overlapping episodic memories (1327). reactivation of the new memories during off-line periods
Integrity of hippocampal circuitry is a prerequisite for re- like sleep, the slowly learning long-term store is trained and
taining an episode as well as spatial and temporal context the new memories are gradually strengthened and adapted
information in memory for more than 15 min (231, 418). to preexisting long-term memories. The transformation
new memory representations undergo in this system consol-
In contrast to declarative memories, nondeclarative memo- idation process comprises also the extraction of invariants
ries can be acquired without involvement of medial tempo- and the development of prototypes and schemas, as the core
ral lobe structures (1134). Nondeclarative memory encom- of the newly learned information is reactivated more fre-
passes quite different memory systems that rely on different quently than divergent details (734, 800, 1239). For the
areas of the brain. It includes procedural memories for mo- declarative memory system, the two-stage model has re-
tor skills (motor areas, striatum, cerebellum) and percep- ceived strong support from lesion studies, indicating that
tual skills (sensory cortices), certain forms of conditioning lesions of the hippocampus abolish the ability to acquire
and implicit learning (priming), etc. Nondeclarative mem- new declarative memory and simultaneously produce a
ories can be implicitly (i.e., without awareness) acquired temporally graded retrograde amnesia where older memo-
and recalled, and learning is slow, usually requiring multi- ries remain intact (231, 418). The time interval for a mem-
ple training trials. It is of note that experimentally disentan- ory to reach a state of hippocampus-independent retrieval
gling nondeclarative from declarative memory processing is can vary from one day to several months or years, depend-
often complicated by the fact that these memory systems ing on the acquired information and the schemas preexist-
interact during acquisition of new knowledge in the healthy ing in long-term memory (1210, 1308). The standard two-
brain. Thus acquisition of skills like language learning and stage model of memory has been also successfully applied to
finger sequence tapping, especially at the initial stages, in- nondeclarative kinds of memory, like procedural memory
corporates declarative in addition to procedural compo- (668), suggesting that the offline reactivation of recent
nents (910). memories and their redistribution from a fast encoding tem-
porary to a slowly learning permanent store could be a
3. The standard two-stage memory system general feature of long-term memory formation.

Why does the consolidation of memory have to take place II. OVERVIEW OF APPROACHES
during sleep? The hypotheses that sleep serves memory con- AND CONCEPTS
solidation is conceptually rooted in the standard two-stage
memory system which is currently the most influential In this section, we review evidence from behavioral studies
model of human memory, and has been developed as a in support of the notion that sleep benefits memory consol-
solution to several key problems arising from simple asso- idation. Key experiments for the different theoretical ac-
ciative network models of memory (175, 780, 800). The counts and concepts will be described in more or less chro-
foremost of these problems is that although simple associ- nological order, thus complementing previous reviews of
ation networks are in fact able to store information very studies on sleep and memory (23, 49, 99, 116, 121, 125,
rapidly, as is the case in the declarative memory system , the 199, 200, 226, 246, 251, 293, 297, 319, 348, 377, 379,
uptake of new conflicting information has a strong ten- 411, 475, 524, 537, 538, 734, 765, 782, 807, 883, 894,
dency to erase the older memories, thus inducing so-called 910, 915, 928, 957, 958, 967, 984, 988, 1001, 1025, 1092,
“catastrophic interference” (1005). The critical question is 1094, 1096, 1098, 1149, 1151, 1152, 1156, 1202, 1242,
how the neuronal network can learn new patterns without 1279 –1281, 1283, 1289 –1291, 1305, 1321).
simultaneously forgetting older memories, an issue that has
also been referred to as the “stability-plasticity dilemma” A. Sleep Acts by Passively Protecting
(e.g., see Ref. 3). In addition, unstructured recurrent net- Memory From Retroactive Interference
works have been demonstrated to face essential capacity
constraints (723). The two-stage memory formation mech- In 1885, Ebbinghaus, the father of experimental memory
anism first proposed by Marr (780) offers a solution to research, published a series of studies, on himself, about the

684 Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 SLEEP AND MEMORY

forgetting of lists of nonsense-word pairs that established duced the amount of forgetting. Because the time retention
the well-known “forgetting curve” indicating that forget- interval was identical for the sleep and awake conditions,
ting occurs rapidly in the first hours after learning and levels the authors concluded that “! results of our study as a
out after several days (339). He noticed already in this work whole indicate that forgetting is not so much a matter of the
that forgetting is reduced when sleep occurred in the reten- decay of old impressions and associations as it is a matter of
tion interval, a phenomenon similarly observed in fol- interference, inhibition, or obliteration of the old by the
low-up studies examining the forgetting curve (reviewed in new” (p. 612 in Ref. 603). Because sleep represents a time in
Ref. 1242). Others reported that depriving a participant of which new encoding of external and, perhaps, also internal
a night of sleep impaired his ability to remember (890). information is strongly reduced, the reduction of interfer-
Rosa Heine (528, 1046) was the first to show in a more ence by sleep appears to be crucial. However, the findings
systematic study (in 6 subjects) that learning in the evening by Jenkins and Dallenbach also pose a challenge to the
before sleep resulted in less forgetting 24 h later than learn- interference theory, because learning of highly similar ma-
ing before a daytime retention interval of wakefulness. This terial did not occur during the waking periods in these stud-
work provided the first clues as to the importance of sleep ies. Interference is considered to depend on the similarity
for memory. between learning and interference materials with stronger
interference for highly similar tasks (see Ref. 632 for a re-
Memory research in the first half of the 20th century was view). Regardless of this issue, the findings were interpreted
preoccupied with the cause of forgetting. Two concepts as evidence that any waking mental activity increases for-
were proposed, i.e., the “decay” account, assuming that getting by a kind of nonspecific interference (1332).
memory traces decay over time resulting in time-dependent
forgetting (1188), and the “interference” account, assum- Many studies subsequently confirmed the positive effect of
ing that forgetting results from learning of new information sleep on memory (63, 79, 80, 248, 341–343, 350, 408, 478,
which (retroactively) interferes and overwrites the old 479, 592, 745, 853, 855, 1068, 1131, 1192, 1241), exam-
memory traces (806). In a classic study, Jenkins and Dal- ining also longer retention intervals of from 24 h up to 6
lenbach (603) compared (in two participants, which were days (79, 80, 474, 592, 990). The underlying concept was
repeatedly examined every day and night over a period of
that sleep acts as a “temporary shelter” that simply post-
almost 2 mo) the retention of nonsense syllables across 1-,
pones the effect of interference and, thereby, passively
2-, 4-, and 8-h retention periods that were filled either with
maintains the memory traces (p. 717 in Ref. 348). However,
sleep or wakefulness (FIGURE 2). Sleep after learning re-
the pure hypothesis that simply the amount of interference
between learning and recall determines the degree of forget-
10 ting is critically challenged by the fact that effects of retro-
Participant Mc active interference are time dependent and much stronger
Participant H when occurring immediately after learning than at a later
8
time, speaking in favor of a time-dependent process of con-
Correctly recalled syllables

solidation after encoding that strengthens the original mem-

6 Sleep ory trace, rendering it less susceptible to interference with
** time (840).

4 ** A time dependency of the effects of sleep on memory for-

***
*** mation is indicated by studies showing stronger effects for
2
Wake sleep occurring shortly after learning than for sleep at a later
time (80, 343, 431, 899, 1172). For example, sleep occur-
ring within 3 h after learning vocabulary was more benefi-
0
0 2 4 6 8
cial than sleep delayed by more than 10 h (431). Further-
Hours more, recall of word pairs after 24 h was better when sleep
FIGURE 2. Effects of sleep and wake intervals of different length occurred immediately after learning than after a day of
after learning on memory for senseless syllables. Sleep after learn- wakefulness (899). Importantly, because the time between
ing leads to superior recall of syllables after the 1-, 2-, 4-, and 8-h learning and retrieval as well as the time spent sleeping was
retention interval, compared with wake intervals of the same length.
identical for the immediate versus delayed sleep conditions
Two subjects (H. and Mc.) participated in this classic study by Jen-
kins and Dallenbach (603). For each data point, each participant of these studies, the findings cannot be explained by inter-
completed 6 – 8 trials, with the different retention intervals per- ference reduction per se, but stress the importance of the
formed in random order. The study took !2 mo during which the timing of reduced interference with reference to the learning
participants lived in the laboratory and were tested almost every day period. That sleep after learning actually benefits the con-
and night. Data are based on Table 3 in Reference 603, as the
original figure contains an erroneous exchange of data points at the
solidation of memories and strengthens memory traces
4-h wake retention interval. Values are means " SE. **P ! 0.01; against future interference was compellingly demonstrated
***P ! 0.001. by Ellenbogen and co-workers (346, 347). In two studies,

Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org 685

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 BJÖRN RASCH AND JAN BORN

they revealed that the enhancing effect on word recall of ance and complex mazes, which encompass a change in the
sleep compared with wakefulness was strongly enhanced animals regular repertoire (69, 100, 312, 516, 525, 539,
when the subjects had learned an interference list shortly 644, 710, 713, 714, 787, 900, 903–906, 992, 1021, 1072,
before final recall testing. Further studies confirmed that a 1111, 1113, 1238, 1352, 1353). In contrast, long-term
90-min sleep period as well as 60-min naps, both containing memory for simpler tasks, like one-way active avoidance
mainly SWS, likewise protect memory against future inter- and simple mazes, were less consistently affected (15, 249,
ference (18, 290, 1069). 386, 390, 495, 558, 611, 644, 821, 872, 902, 907–909,
1072, 1091, 1334).
In fact, recent versions of the interference account on sleep-
associated memory consolidation have integrated this issue, REM sleep increases were mostly observed in the first hours
assuming that sleep provides a time of reduced interference after learning, partly reflecting the fact that recording was
on consolidation processes, which per se are considered to limited to these hours. With prolonged recording sessions, el-
be time dependent (809, 1332). Thus any treatment that evated periods of REM sleep occurred up to 4 – 6 days after
reduces interfering influences on consolidation should be learning, sometimes following a cyclic pattern (1111,
more effective the shorter it is applied after learning. Still, 1116). Here, REM sleep increases were typically most
these interference accounts assume a passive “opportunis- prominent during specific time periods and dependent on
tic” role of sleep in memory consolidation occurring regard- the task. In the Morris water maze task, it started more than
less of whether the brain is asleep or awake (809). Yet, with 2 h after learning and persisted for 22 h (1099). In several
the assumption that sleep generally reduces interference other avoidance tasks, REM sleep increases were less per-
from encoding of external events, this theorizing is chal- sistent, emerged later (i.e., 9 –12 h post-learning), and some-
lenged by a great body of studies indicating a dependence of times reemerged 17–20 h post-learning (1094). In appeti-
consolidation on the composition of sleep, with differential tive learning tasks, REM sleep increases started after 4 h
outcomes for sleep rich of REM sleep or SWS (63, 343, 408, and persisted for 12 h (1104). The increases in REM sleep
930, 931, 1048, 1340). Thus explaining the improving ef- during the specific time periods predicted later memory re-
fect of sleep on memory retention solely on the basis of call and reliably separated between learners and nonlearn-
reduced interference appears to be untenable unless sleep
ers (1095, 1110). Learning in these studies induced distinct
stages are thought to differ in their degree of interference,
and prolonged waves of REM increases possibly involved in
e.g., owing to associated dreaming (408).
memory formation. Based on these findings, Smith pro-
posed the concept of “paradoxical sleep windows” (PSW)
B. REM Sleep and Memory Consolidation mediating memory formation (1092, 1097). Indeed, selec-
tive deprivation of REM sleep during, but not outside of,
The hypothesis has been around for some time that REM identified PSWs impaired memory (719, 1101, 1105, 1106,
sleep contributes to memory consolidation, stimulated in 1113, 1115). Inhibition of protein synthesis by anisomycin
particular by the wakelike EEG activity during this sleep also impaired memory only when intraperitoneally injected
stage together with frequent reports of vivid dreams after during a PSW 9 h after learning a shuttle avoidance task
awakening from REM sleep. Very consistent evidence for a (1108). Similar results were obtained when blocking mus-
role of REM sleep for memory was provided by studies in carinic cholinergic receptors during a PSW by scopolamine
animals (for comprehensive reviews, see, e.g., Refs. 101, (720, 1108), pointing to a crucial involvement of protein
388, 807, 901, 936, 943, 1092, 1095, 1098, 1260). With synthesis as well as cholinergic activation in PSW-associ-
the use of a variety of tasks including classic, aversive, and ated memory processes. Interestingly, blocking of NMDA
appetitive conditioning procedures, a large number of stud- receptors was most effective in impairing memory consoli-
ies consistently revealed increases in REM sleep after learn- dation when administered after a PSW, suggesting that
ing in rats, mice, and cats (100, 284, 389, 540, 708, 711– PSW-associated memory processing induces subsequent
713, 747, 944, 1102, 1104, 1110, 1111). Rats living in NMDA-dependent plasticity (1094, 1099).
enriched compared with impoverished environments like-
wise exhibited enhanced REM sleep, although increases in Quite a number of studies in this context have been criti-
non-REM sleep were also observed (278, 493, 494, 669, cized as they employed the “flower-pot” method to deprive
823, 1086, 1168). Increasing REM pharmacologically by the animal from REM sleep. In this procedure, the rat rests
administration of carbachol into the pontine reticular for- on a small platform (i.e., the flower pot) surrounded by
mation and of corticotrophin-like intermediate lobe peptide water and, owing to complete muscle atonia, falls into the
(CLIP) as well as a REM sleep rebound after prior REM water whenever REM sleep starts (119, 387, 569). Awak-
deprivation, all improved memory for a Y-maze discrimi- enings induced in this way are highly stressful and may per
nation task when applied after learning of the task (1316). se impair later memory performance (953). However, im-
Deprivation of REM sleep (mostly without simultaneous pairing effects of REM sleep deprivation on memory reten-
sleep recording) appeared to primarily impair memory for- tion have been also demonstrated with less stressful proce-
mation on complex tasks, like two-way shuttle box avoid- dures like gentle head lifting (263), mild touching (596), or

686 Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 SLEEP AND MEMORY

after pharmacological REM suppression (720, 721, 1098, sleep, whereas the improvement in the mirror tracing task
1108). Ponto-geniculo-occipital (PGO) waves which occur correlated with the number of REMs in the fourth REM
associated with REMs in rats and cats have been proposed sleep period. Good learners with higher IQ showed greater
as a mechanism promoting plastic processes underlying increases in REM sleep (1114). In another study, REM sleep
memory formation during REM sleep (259, 260) (see sect. in the last quarter of an 8-h period of sleep, together with
IVF). the time in non-REM sleep in the first quarter, was highly
predictive for learning success in a visual texture discrimi-
The focus on REM sleep as the sleep stage that supports nation task (1157).
memory has also been criticized by several studies revealing
concomitant or even selective increases in non-REM sleep Compared with declarative learning paradigms, tasks with
after the animal’s exposure to enriched environments or a strong procedural memory component appeared to be
other learning procedures (389, 493, 495, 510, 529, 540, also more sensitive to the detrimental effects of REM sleep
645, 1092, 1104, 1168). Non-REM sleep was even pro- deprivation. Karni et al. (622) were the first to show that
posed as a factor that could explain REM sleep-depriva- sleep after training a visual texture discrimination task sub-
tion-induced memory deficits (992), and increases in non- stantially reduced discrimination thresholds on the task,
REM sleep after fear conditioning correlated with the indicating a critical importance of sleep for gains in skills
learned fear response on the next day (62). In addition, in that occur offline after training has finished. The overnight
some cases learning decreased subsequent REM sleep (795, reduction in discrimination thresholds was prevented by
1029 –1031, 1100), and this could be accompanied by a selective REM sleep deprivation, whereas awakening from
concurrent increase in non-REM sleep (754). non-REM sleep had no effect. Comparing the effects of
REM sleep and non-REM sleep deprivation on tasks with
Compared with the findings in rats, evidence for a role of strong procedural components (verbal word fragmentation,
REM sleep in memory processing in humans is surprisingly priming, Tower of Hanoi, Corsi block tapping) and declar-
inconsistent. Most studies failed to find effects of selective ative memory tasks (verbal word recognition, visual-spatial
REM sleep deprivation on the retention of declarative mem- learning), Smith and colleagues (230, 1093, 1094) found
ories when simple verbal materials (word lists, word pairs, that total REM sleep deprivation or deprivation of the last
etc.) were used (88, 185, 192, 343, 344, 729, 918, 1094, two REM episodes of postlearning sleep selectively im-
1191). Only with more complex declarative materials paired performance on the procedural memory tasks at the
(meaningless sentences, stories, etc.) did REM sleep depri- retest session 1 wk later. A similar impairment specific to
vation impair declarative memory in some studies (350, procedural memory was also revealed after alcohol-induced
1192). In narcoleptics, isolated periods of REM sleep facil- REM sleep suppression (1107).
itated memory for complex associative information, com-
pared with periods of non-REM sleep or wakefulness Based on these findings, Smith (1094, 1096) suggested that
(1054), and learning a topographical map increased subse- in humans, REM sleep is involved in the processing of pro-
quent REM sleep in healthy participants (365). Further- cedural memory, whereas REM sleep plays no role in the
more, changes in REM sleep patterns were found during formation of declarative memories, particularly with re-
intensive study periods (e.g., student exams, see Refs. 1103 spect to simple learning tasks. However, REM sleep depri-
and 811, but see Ref. 559). More consistent evidence for an vation experiments in humans, like in animals, have been
involvement of REM sleep was obtained for tasks with a criticized due to possible confounding effects of stress on
strong procedural memory component, like learning a for- memory formation (119). In addition, the idea that REM
eign language or Morse code (e.g., Refs. 488, 656, 657, sleep contributes to memory formation has been questioned
758, 1249). Increases in REM sleep were observed, for ex- based on two more fundamental concerns (387, 1076,
ample, during training of unfamiliar patterns in motor co- 1250, 1253): one was that the large differences in time spent
ordination, like trampolining (149, 150) and adaptation to in REM sleep between species, e.g., ferrets spend more than
distorted vision by a set of lenses (e.g., Refs. 20, 232, 658, 6 h per day in REM sleep, whereas humans only 2 h, do not
659, 1356, but see also Refs. 19 and 1358). REM sleep translate into any obvious and systematic differences in ca-
increased also in infants who learned a head turning re- pabilities to form memories. However, comparisons be-
sponse, in contrast to infants who did not learn the response tween species in learning capabilities are per se rather diffi-
(891). In a motor finger sequence tapping task, the amount cult and inconclusive. The second more critical concern is
of REM sleep after learning predicted sleep-dependent im- that even the complete absence of any REM sleep, e.g.,
provement in this task (383). Although less clear than in during treatment with antidepressants, does not lead to any
rats, some evidence in humans has also been provided for a obvious impairment of memory formation (e.g., Refs. 962
REM sleep window of task specific memory processing and 1252, but see Refs. 138 and 1310 for discrepant results
(1103, 1114): memory for the procedural Tower of Hanoi in rodents), suggesting that at least the overt EEG charac-
task was revealed to be most strongly correlated with in- teristics of REM sleep are not necessary for successful mem-
creases in REMs in the second REM episode of postlearning ory consolidation. On the other hand, the increase in sero-

Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org 687

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 BJÖRN RASCH AND JAN BORN

tonin and catecholamine levels induced by antidepressant increased subsequent REM sleep (730). Furthermore, prim-
intake might compensate for the memory-impairing effects ing before naps filled with REM sleep had a much stronger
of REM-sleep suppression, as these neurotransmitters have impact on later creative answers compared with naps with-
been implicated in memory consolidation processes (815, out REM or quiet resting, indicating that REM sleep im-
842, 1011). To summarize, although REM sleep may ben- proves creative problem solving instead of reducing creativ-
efit procedural memory consolidation, this effect appears to ity (163). In volunteers asked to voluntarily suppress “un-
be linked to specific conditions and to underlying, REM- wanted memories,” sleep and particularly REM sleep
sleep associated biological and molecular mechanisms that appeared to counteract this suppression as reflected by an
are so far unknown. improved retrieval for previously suppressed items, rather
than to enhance forgetting of these unwanted memories
(381). Finally, repetitive nightmares, which are highly prev-
C. Sleep and the Erasure of Information:
alent in patients suffering from posttraumatic stress disor-
Accounts of Emotional Memory ders (PTSD), do not lead to forgetting of the traumatic
event, but are rather associated with increased severity of
The idea that sleep might be involved in the erasure or the disorder and considered a risk factor in the development
filtering of information has been put forward by several of PTSD (75, 1240, 1331). Overall, these findings speak
authors (e.g., Refs. 210, 236, 362, 369, 854). In particular against the view posed by Crick and Mitchison (236).
in 1983, Crick and Mitchison (236) proposed, based on a
neurocomputational model of associative learning, that Of note, in other computational models the mechanisms
dreaming during REM sleep helps to forget “parasitic limiting the strength of synaptic weights in the neural net-
modes” of activity, thus ensuring an efficient mode of op-
work have been linked to non-REM rather than REM sleep
eration of the brain during waking. Such parasitic modes of
(210). A recent version of this idea is the “synaptic homeo-
activity particularly occurred after stimulation overload
stasis hypothesis” (1203, 1204) which will be discussed in
and included “fantasy” (i.e., the net produces far-fetched
section IVB and assumes that a global downscaling in the
and bizarre associations), “obsession” (i.e., iterates the
strength of synaptic connections takes place during SWS to
same response, irrespective of input) and “hallucination”-
prevent saturation and to reduce place and energy demands,
like responses (i.e., responds to inappropriate input sig-
thereby preparing the network for the encoding of new
nals). As a solution to this problem, the authors proposed a
information during succeeding wakefulness (565, 566).
“reverse learning” mechanism during REM sleep-dreaming
that dampens synaptic weights to reduce the probability of
If the emotional tone of a memory is considered an “un-
these parasitic activity modes and thereby also enhances the
efficacy and storage capacity of the network. Thus, accord- wanted activation,” then the “sleep to forget sleep to re-
ing to this account, dreaming reduces unwanted and bizarre member” (SFSR) hypothesis recently proposed by Walker
forms of representations in memory, which enhances new and van der Helm (1282, 1284) bears some similarities with
learning the next day as well as retrieval of memories ac- the ideas about REM sleep by Crick and Mitchison. The
quired before sleep (237). In simulation studies, repeated SFSR hypothesis assumes that REM sleep after an (aversive)
unlearning procedures indeed improved the learning capa- emotional experience strengthens the content of the respec-
bility of the network and retrieval of recently learned pat- tive representations in memory, but simultaneously reduces
terns, but concurrently weakened more remote memories the emotional tone associated with this memory, i.e., re-
(564, 1237). duces the emotional response when the memory is retrieved.
The process is not restricted to one night after encoding, but
Although computational scientists agree in that a mecha- would continue during multiple nights. In depressed pa-
nism is necessary that limits the strength of synaptic weights tients showing enhanced REM sleep, according to the SFSR
in artificial neural networks, empirical evidence for the pro- hypothesis, this enhancement would bias strengthening of
posed function of REM sleep, specifically with regard to the memories towards increased storage of negative contents,
removal of “unwanted modes” of activation, is so far lack- while suppression of REM sleep through antidepressants
ing. Several authors have reported an influence of REM counteracts this bias. The impairment of mood associated
sleep on emotional reactivity or mood (490, 491, 1243), but with this REM-related process might indicate that the
a specific influence on “obsessive” or “hallucination”-like attenuation of emotional tone by REM sleep is not func-
behaviors has not yet been tested. In contrast, after awak- tional in these patients. Similarly, in traumatized pa-
ening from REM sleep, the brain appears to remain in a tients, increased nightmare frequency would point to a
“hyperassociative” mode, in which weak semantic primes failure to attenuate memory-associated emotions during
produced distinctly stronger priming effects than during REM sleep.
consolidated wakefulness or after awakenings from non-
REM sleep (1154). Similarly, REM sleep awakenings also Consistent with the theory, several studies have shown
resulted in a 32% increase in a complex anagram solving that emotional memories are particularly strengthened
task (1288), and conversely, the need for creative thinking across sleep (560, 574, 732, 861, 895, 898, 1274, 1276,

688 Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 SLEEP AND MEMORY

1278, 1294), in particular when containing high amounts D. The Dual Process Hypothesis
of REM sleep (179, 478, 479, 1274). The enhancing
effect of postencoding sleep on emotional memories was The dual processes hypothesis assumes that different sleep
detectable even after several years (1276). Duration and stages serve the consolidation of different types of memories
latency of REM sleep significantly correlated with the (428, 765, 967, 1096). Specifically, it has been assumed that
later recall emotional memories (861). In the context of declarative memory profits from SWS, whereas the consol-
fear learning, sleep promoted the generalization of ex- idation of nondeclarative memory is supported by REM
tinction learning (878), increased intersession habitua- sleep. The hypothesis received support mainly from studies
tion to emotional stimuli (879), and in both humans and in humans, particularly from those employing the “night-
rodents, the deprivation of REM sleep impaired fear ex- half paradigm.” This paradigm, originally developed by
tinction (421, 1084, 1133). Further support for the SFSR Ekstrand and co-workers (341, 408, 1340), basically com-
hypothesis was provided by functional magnetic reso- pares retention performance across retention intervals that
nance imaging studies indicating that the strengthening cover either the early or late half of nocturnal sleep.
of negative emotional memories by sleep is accompanied Whereas in the early sleep condition, participants learn (to
by a reduced amygdala activation, i.e., a diminished emo- criterion) the memory tasks in the evening and then sleep
tional response during retrieval (1147, 1231). for 3– 4 h before a later recall test, in the late sleep condi-
tion, participants first sleep for !3 h (to satisfy the need of
However, there are also contradicting findings. For exam- SWS) and then are subjected to the learning phase, followed
ple, two studies have shown that memory recall of negative by the late night retention sleep. Due to the circadian
pictures is less impaired by sleep deprivation after encoding rhythm, early nocturnal sleep contains most of SWS,
compared with neutral pictures, suggesting a reduced de- whereas late nocturnal sleep is dominated by REM sleep.
Time in stage 1 or 2 sleep usually does not differ between
pendency of negative memories on sleep-dependent consol-
early and late sleep retention conditions. The approach thus
idation processes (46, 1147). With respect to emotional
allows for comparing the effects of sleep rich in SWS versus
reactivity, one study showed that REM sleep-rich sleep
REM sleep, elegantly avoiding possible confounding effects
amplified subjectively experienced aversion to previously
resulting from stressful repeated awakenings accompanying
viewed emotional pictures (1273). Conversely, selective
standard procedures of selective sleep deprivation. To con-
REM sleep deprivation after picture viewing reduced
trol possible confounds of the circadian rhythm, the effects
arousal ratings to negative pictures presented again on the
of early and late sleep retention periods are typically com-
next morning (704). Others found that emotional reactivity
pared additionally with the effects of wake retention peri-
decreased across wakefulness, but was preserved during
ods that cover the same early and late phases of the night.
sleep, with the preserving effect on emotional reactivity be-
These two wake control conditions are also necessary to
ing specifically linked to REM sleep (60, 481). In addition, ensure that depth of encoding during the learning phase is
amygdala responses to correctly recognized emotional comparable between the conditions, because prior sleep is
objects (with reference to neutral objects) increased known to influence encoding capabilities (328, 342, 486,
rather than decreased after sleep, compared with an as- 515, 548, 1190, 1232, 1345), and learning is preceded by
sessment after a wake retention interval (896). The in- prior sleep only in the late sleep condition, but not in the
creased amygdala response was accompanied by a stronger early sleep condition.
connectivity among limbic regions after sleep in this study.
Sleep after fear conditioning in humans increases the con- Yarush et al. (1340) were the first to report a beneficial
ditioned response and the associated amygdala activity, effect of SWS-rich early sleep on declarative memory (word
with this enhancement being positively associated with pairs), compared with retention performance across a REM
postlearning REM sleep (Menz, Rihm, Born, Kalisch, Pape, sleep-rich late sleep or across corresponding wake intervals,
Marshall, and Büchel, unpublished observation). Sleep like- and these findings were replicated in a later study of the
wise facilitated the generalization of implicit fear responses same group (408). A similar benefit for declarative (paired-
(678). In animals, REM sleep has been consistently associ- associates) memories was revealed when controlling for cir-
ated with a strengthening of conditioned fear memories (see cadian influences by placing SWS-rich and REM sleep-rich
sect. IIB; for reviews, see, e.g., Refs. 538, 807, 1092, 1094). sleep periods at the same circadian phase, between 3 and 7
Furthermore, a recent study points towards an involvement a.m. (63). While these studies quite compellingly showed
of adrenergic activity during SWS instead of REM sleep for that declarative memory for neutral materials is enhanced
the consolidation of emotional information (482). In con- by SWS, emotional declarative memories appear to addi-
clusion, it is still an open question whether the consolida- tionally benefit from REM-rich late sleep (1273, 1274).
tion of emotional memories actually differs in quality from Building on these early studies, Plihal and Born (930, 931)
that of neutral declarative memories, or whether it is the not only demonstrated a benefit for declarative memories
same consolidation process that is simply enhanced or ac- (word pairs, spatial information) from early SWS-rich
celerated by the emotional arousal that is attached to the sleep, but also demonstrated that late REM sleep-rich
representation at encoding. retention sleep selectively improved procedural and im-

Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org 689

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 BJÖRN RASCH AND JAN BORN

plicit memories (mirror tracing skill, word-stem prim- stage 2 duration and fast spindle density also predicted
ing), compared with corresponding wake-retention inter- sleep-dependent improvements in a finger tapping task (59,
vals. Later studies replicated beneficial effects of late 961, 1285, 1286). Based on these findings, Smith and co-
REM sleep-rich sleep on implicit memories (e.g., faces, workers (846, 1114) proposed that especially simple motor
masked stimuli) (1248, 1277), and altogether these find- tasks require stage 2 sleep, whereas complex motor tasks
ings fit well with the notion of an involvement of REM may require REM sleep. However, sleep stage 2 spindle
sleep in procedural and implicit memory processes as activity and spindle counts were found to similarly correlate
revealed by standard REM sleep deprivation procedures with the overnight retention of verbal and visuospatial
(described in sect. IIB). memories, suggesting that spindles and stage 2 sleep are also
involved in declarative memory formation (213, 214,
Improving the effects of early SWS-rich sleep on declarative 1016).
memory appeared to be overall more consistent with free
(or cued) recall measures than with recognition measures of
E. The Sequential Hypothesis
memory (265, 326, 667, 966; see Ref. 297, for a review),
possibly reflecting that unlike recollection, familiarity based
recognition taps to a greater extent implicit components of The sequential hypothesis stresses the importance of the
these memories (1343, 1344). Two studies found enhanced cyclic succession of SWS (or non-REM sleep) and REM
recollection (of words) after early, SWS-rich sleep, com- sleep for memory formation, with the sleep stages serving
pared with a late REM-rich sleep or wake intervals, while complementary functions in this process. The sequential
familiarity-based recognition measures remained unaf- hypothesis originally assumed that in a first processing step
during SWS, nonadaptive memories were weakened and
fected (265, 326). In a third study measuring recognition
adaptive responses were strengthened, whereas during the
right after free recall, this effect was not replicated, al-
second processing step during REM sleep, the adaptive
though SWS-rich sleep still enhanced free recall of the de-
memories would be integrated and stored in preexisting
clarative materials (966).
knowledge networks (23, 460, 461). A series of studies in
rats provided evidence for the hypothesis (24, 25, 27, 28),
Overall, studies using the night-half paradigm have pro-
for example, reporting high positive correlations between
vided substantial evidence for the dual processes hypothe-
the number of SWS periods followed by REM sleep with
sis, such as hippocampus-dependent declarative memories
memory performance on a two-way active avoidance task
preferentially profiting from SWS, whereas nondeclarative
(698). In contrast, the number of SWS periods followed by
aspects of memory, such as procedural, implicit, and
wakening correlated negatively with performance, provid-
emotional, additionally profiting from REM sleep (121, ing indirect support for a weakening of memories during
428). The hypothesis has been challenged by findings SWS, if not followed by REM sleep. In addition, it was
showing that procedural tasks like visuomotor adapta- argued that SWS led to a global depotentiation of synaptic
tion and visual texture discrimination also benefit from connections due to the slow EEG frequency as well as the
SWS (10, 433, 581, 582). However, training such skills absence of important transcriptional factors. Then, during
does not proceed entirely detached from declarative subsequent REM sleep, the high-frequency EEG and hip-
memory mechanisms, especially at an initial stage of pocampal theta activity support the strengthening of syn-
training (e.g., Refs. 198, 533, 1013, 1014, 1042). Such aptic connections (461). Further elaboration of the hypoth-
declarative components might have mediated the strong esis integrated an additional stage of transitional sleep char-
benefits for skills from SWS in those studies (910). acterized by a sudden mixing of theta and alpha waves with
the previous delta waves (23, 929). SWS-REM sleep se-
A major weakness of the night-half paradigm is that it ig- quences comprising a short period of transitional sleep ac-
nores possible contributions of stage 2 sleep to memory. curately predicted whether rats learned the active avoidance
Although the amounts of stage 2 sleep were comparable in tasks, whereas in the absence of interposed transitional
the early and late sleep conditions of the studies reported sleep, rats did not reach the learning criterion (26, 760,
above, sleep in this sleep stage may have substantially dif- 1254).
fered between the two phases, for example, with regard to
spindle density (440, 441, 550), heart rate, or levels of neu- The sequential hypothesis has also received some support
romodulators like catecholamines and cortisol (95, 124, from studies in humans. The overnight improvement on a
131, 313, 554, 960). There is consistent evidence for an visual texture discrimination task was best predicted by the
involvement of stage 2 sleep and sleep spindles in motor time in SWS in the first quarter and the time in REM sleep in
learning (139, 396, 822, 862, 921), and memory for a sim- the last quarter of the night (1157). In nap studies, discrim-
ple motor task is impaired after selective stage 2 sleep de- ination thresholds in the same task improved only after a
privation (1112, 1117). Training on motor tasks increased longer 90-min nap containing both non-REM and REM
the time spent in stage 2 and the density of fast (but not sleep, but not after a shorter 60-min nap solely containing
slow) spindles (399, 837, 920, 921, 1173, 1174), and sleep non-REM sleep (808). Similarly, studies adopting the night-

690 Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 SLEEP AND MEMORY

half paradigm revealed the greatest improvement in texture hypothesis integrates aspects of both the dual-process
discrimination thresholds after a whole night of sleep con- view and the sequential hypothesis. Central to the “active
tained both early SWS-rich and late REM-rich sleep, system consolidation” hypothesis is the assumption that
whereas early SWS-rich sleep per se had only intermediate memory consolidation during sleep originates from the
effects, and REM-rich late sleep was completely ineffective repeated reactivation of newly encoded memory repre-
(433). Experimentally fragmenting sleep such that the cyclic sentations. These reactivations occur during SWS and
structure was disturbed strongly impaired overnight-reten- mediate the redistribution of the temporarily stored rep-
tion of words, whereas the same degree of fragmentation resentations to long-term storage sites where they be-
did not impair word recall when the sleep cycles were pre- come integrated into preexisting long-term memories
served (378). In the elderly, sleep cycle organization pre- (FIGURE 3A). The slow oscillations during SWS drive the
dicted their capacity for overnight-retention of words repeated reactivation of hippocampal memory represen-
(796). It has been also suggested that differences in memory tations during sharp wave-ripples (SW-Rs; FIGURE 1B) in
retention between early and late sleep in the night-half par- the hippocampus together with thalamo-cortical spin-
adigm actually reflect differences in the cyclic organization dles, which are involved in inducing enduring plastic
of sleep rather than in the amount of SWS and REM sleep changes in cortical areas (FIGURE 3B). Thus reactivation
(379). Indeed, it appears that many findings support the and integration of temporarily stored memories into
sequential hypothesis, although this hypothesis has rarely
long-term stores accompany a qualitative reorganization
been subjected to direct testing.
(transformation) of the memory representation (system
consolidation) that needs to be stabilized in a synap-
F. The Active System Consolidation tic consolidation process assumed to take place during
Hypothesis succeeding periods of REM sleep. In claiming that mem-
ory consolidation during sleep is an active process, this
The hypothesis that sleep supports the formation of long- hypothesis contrasts with accounts that sleep only pas-
term memory in an active system consolidation process sively or opportunistically supports consolidation pro-
has been elaborated in several previous reviews (293, cesses mainly by providing a time of reduced interference
345, 734, 828, 957, 958, 984, 988, 1284, 1305). The (809, 1332).

A B
Long-Term Store
Neocortex Neocorticol
Slow Oscillations

Thalamo-Corticol
Spindles

Temporary Store
Hippocampus
Hippocampal
Ripples
Reactivation

Spindle-Ripple Event

FIGURE 3. A model of active system consolidation during sleep. A: during SWS, memories newly encoded
into a temporary store (i.e., the hippocampus in the declarative memory system) are repeatedly reactivated,
which drives their gradual redistribution to the long-term store (i.e., the neocortex). B: system consolidation
during SWS relies on a dialogue between neocortex and hippocampus under top-down control by the neocor-
tical slow oscillations (red). The depolarizing up phases of the slow oscillations drive the repeated reactivation
of hippocampal memory representations together with sharp wave-ripples (green) and thalamo-cortical spin-
dles (blue). This synchronous drive allows for the formation of spindle-ripple events where sharp wave-ripples
and associated reactivated memory information becomes nested into succeeding troughs of a spindle (shown
at larger scale). In the black-and-white version of the figure, red, green, and blue correspond to dark, middle,
and light gray, respectively. [Modified from Born and Wilhelm (125).]

Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org 691

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 BJÖRN RASCH AND JAN BORN

Compared with the previously discussed hypotheses, the tionally, sleep enhanced sequence-finger tapping perfor-
“active system consolidation” account is more concerned mance when learning occurred by observation (1234) or
with identifying and integrating the neural mechanisms me- motor imagery (274), a further hint towards qualitative
diating the beneficial effect of sleep on memory consolida- changes in the skill representations induced by sleep.
tion, going far beyond the simple differentiation of sleep
stages. In this regard, electrophysiological, neurochemical, In connection with declarative memory processes, evidence
and genetic conditions are considered that will be discussed for an active consolidation process during sleep that leads to
in detail in the next sections. In fact, at a purely behavioral a qualitative reorganization of memory representations has
level, numerous studies have demonstrated that sleep after been provided by studies showing that sleep preferentially
learning benefits declarative (17, 18, 60, 63, 79, 290, 326, supports memory for the “gist” in the learned material,
327, 343, 348, 350, 376, 408, 427, 431, 478, 479, 547, thereby supporting processes of abstraction, inference, and
560, 571, 574, 592, 603, 682, 732, 735, 861, 898, 930, insight. Thus, in adults and children, sleep promoted the
932, 959, 1069, 1172, 1192, 1217, 1274, 1276, 1278, integration of newly learned spoken words into existing
1282, 1302) as well as nondeclarative kinds of procedural knowledge networks as measured by a lexical competition
memory (135, 147, 164, 311, 322, 374, 383, 435, 447, 563, task (333, 334, 530; but see Ref. 735), with this effect being
602, 631, 664, 679, 733, 930, 995, 1003, 1286, 1287, associated with increased spindle activity (1175). Sleep like-
1330). However, these behavioral findings have but shed wise promoted grammar-related abstraction processes in
little light on the putative processes of memory reactivation language learning tasks in infants (469, 589). When partic-
and reorganization mediating the consolidation process ipants learned a hierarchy of pairwise presented elements
during sleep. (e.g., A # B and D # E), sleep after learning improved the
ability to infer the correct relation between the most distant
1. Reorganization of procedural and declarative elements (A # E) (345). Comparable results were obtained
memories during sleep after a nap using relations between different faces paired
with the same object, whereby the amount of SWS corre-
With regard to procedural memory, findings that sleep after
lated with the accuracy of relational memory (705, 706).
training of perceptual and motor skills like visual texture
Sleep also increased the production of false memories in the
discrimination and finger sequence tapping can produce
Deese-Roediger-McDermott (DRM) paradigm (1007), in
significant improvement (i.e., a gain in skill at a later retest-
which in the learning phase participants listened to lists of
ing) have pointed towards an active reprocessing of skill
semantically related words (e.g., nurse, patient, hospital,
representations occurring during sleep that sharpens the
sick, medicine, etc.), whereas the semantic “topic” of the list
respective representations (47, 57, 147, 274, 275, 291, 322,
383, 384, 433, 582, 622, 631, 664, 665, 679, 762, 770, is not presented (the “critical lure,” e.g., doctor). Compared
808, 810, 863, 930, 1003, 1153, 1157, 1286, 1287, 1290, with wake retention conditions, after sleep subjects showed
1318). However, sleep-dependent gains may not equally a greater tendency to “falsely” recall the critical lure, in line
occur for all types of skills (e.g., Refs. 221, 1002, 1015). with the notion that sleep promotes processes of abstraction
Moreover, recent studies have shown that such gains in skill and the extraction of the gist information from the list (294,
that are typically measured with reference to the perfor- 897). However, the effect of sleep in the DRM paradigm
mance level at the end of training, can occur also within a was less consistent when recall of the critical lure was tested
few hours after training in the absence of sleep, and may with a recognition procedure rather than by free recall (255,
thus partly reflect a recovery process that is independent of 296, 373; see Ref. 1159 for an overview).
sleep (364, 573, 1002–1004). Furthermore, some studies
reported that no sleep-dependent gains in procedural motor Using the Number Reduction Task (1189), Wagner et al.
tasks occur when circadian and homeostatic influences (1275) showed that sleep facilitates the gain of insight, i.e.,
(e.g., time of day, time since sleep) are controlled (164, 991, explicit knowledge of a hidden structure that was embed-
995). Also, a mere gain, for example, of speed in finger ded in strings of digits which the subjects had to process
sequence tapping could be explained solely on the basis of before sleep. Subsequent studies of the same group specified
synaptic consolidation processes that strengthen connec- that the gain of insight depended on the occurrence of spin-
tions formed during training without reorganizing the dle activity during early SWS-rich nocturnal sleep (1346 –
memory representation. However, convincing behavioral 1350). Very similar results were obtained with an approach
clues for a reorganization of skill representations during combining the classical Serial Reaction Time Task (SRTT)
sleep come from investigations of sequence-finger tapping with a so-called “generation task.” On the SRTT, the sub-
skills that show that sleep favors the emergence of an effec- ject is trained to press as fast and as accurately as possible
tor independent representation, i.e., sleep benefited pressing different keys corresponding to the changing positions of a
the sequence of target keys independent of whether the cue. Unknown to the subject, the changes in the cue position
sequence was tapped with the right or left hand, whereas follow a repeating sequence. Typically, during training sub-
the sequenced tapping movements per se appeared to ben- jects acquire an implicit knowledge of this sequence as in-
efit also from a wake retention interval (218, 1330). Addi- dicated by faster responses to cue positions that follow the

692 Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 SLEEP AND MEMORY

sequence compared with responses to random positions. which stayed awake on the night after learning. Some of the
Training does not lead to an immediate formation of ex- effects of sleep on brain activation during recall of emo-
plicit sequence knowledge, as the subjects typically cannot tional memories were still observed 6 mo later (1148). Fi-
deliberately reproduce the SRTT sequence, when explicitly nally, sleep-dependent changes in brain activity indicative
asked to do so right after training (in a generation task). for a reorganization of memory representations have also
Yet, significant explicit sequence knowledge developed been reported for procedural tasks like sequences finger
when SRTT training was followed by nocturnal sleep (324, tapping and visual texture discrimination (384, 1292,
382). Generation task performance in the wake control 1293).
group in these studies remained at chance level. The sleep-
induced extraction of explicit sequence knowledge from an 2. Selectivity of memory consolidation during sleep
implicitly trained SRTT was particularly pronounced in
children, where this extraction process was correlated with As another key feature, the “active system consolidation”
enhanced EEG slow oscillation activity during posttraining concept implies that memory consolidation during sleep is
sleep (1321). Similarly, in adults performing on a statistical selective. It is rather unlikely that off-line consolidation
learning task, sleep promoted the abstraction of probabilis- strengthens recently acquired memory traces and their syn-
tic regularities in tone sequences, with this effect being as- aptic connections globally, because such global and unse-
sociated with increased SWS during postlearning sleep lective strengthening would inevitably produce a system
(337). Complementary evidence for a neuronal reorganiza- overflow. In support of this notion, a growing body of ex-
tion of memory representations by sleep was provided by periments indicates that sleep does not equally benefit all
studies imaging brain activation during learning and re- memories, although the mechanisms determining whether
or not a certain memory gains access to offline consolida-
trieval. In experiments by Takashima et al. (1170), partici-
tion during sleep are currently not well understood.
pants napped for 90 min after studying pictures. Recogni-
tion of the pictures was tested 1, 30, and 90 days later.
Several factors have been identified. First, sleep-dependent
Activation of the hippocampus gradually decreased over
gains in skills are more robust under explicit learning con-
time, whereas cortical activation in ventromedial prefrontal
ditions (i.e., the subjects are aware about the skill to be
areas increased. The duration of SWS during the nap pre-
acquired) (383, 664, 1285) compared with implicit learning
dicted the reduction in hippocampal activity and was also
conditions (385, 1003, 1122, 1128). As learning of declar-
associated with better recognition performance on day 1. A
ative memories is explicit (and often intentional), these re-
second study of this group confirmed that one night of sleep
sults suggest that explicit encoding favors access to memory
decreases hippocampal activation and increases activation
consolidation during sleep. Second, the initial memory
in neocortex areas (1169). In another study (427), the sleep- strength might affect consolidation during sleep, although
induced improvement in word pair memories, at a retrieval the available data are not consistent. Stickgold (1151) pro-
test 2 days later, was accompanied by an increased func- posed that sleep mainly benefits memories encoded at an
tional connectivity between the hippocampus and the me- intermediate memory strength, and that the effect of sleep
dial prefrontal cortex, although contrary to Takashima et on memories with varying initial encoding levels follows an
al.’s findings, hippocampal activation per se was enhanced U-shaped curve (1320). Contrasting this view, sleep after
at recall in the sleep group. At a follow-up test 6 mo later, learning preferentially strengthened memory for word stim-
the sleep group exhibited increased cortical activation com- uli weakened by interference learning (325, 343) or retrieval-
pared with the subjects who stayed awake on the night after induced forgetting (1) or for very difficult motor move-
word pair learning. Sleep following learning spatial memo- ments (679). On the other hand, other studies revealed
ries of a virtual town produced increased activation in the greater benefits from sleep for strongly encoded memories
striatum at a recall test 3 days later, when compared with a (523, 1172) or only in well-performing subjects (1038,
group deprived of sleep after learning (873). Activation in 1215). Moreover, it is well known that the encoding of
the right striatum positively predicted way finding in the emotional events results in stronger memories. Some studies
virtual town only in the sleep group, and also functional indeed demonstrated that sleep preferentially consolidates
connectivity between the striatum and hippocampus was emotional over neutral memories (574, 602, 861, 896, 898,
modulated by sleep after learning. Similarly, the sleep-en- 1276); however, others failed to reveal such effects (60,
hancing effect of statistical learning (337) was accompanied 168, 732, 1147, 1148).
by a shift in brain activation from the medial temporal lobe
to the striatum (336). A sleep-induced reorganization of A third factor that was consistently found to favor the sleep-
neuronal representations was also revealed for emotional dependent strengthening of a memory is the relevance of an
memories (pictures) after 1 night (896) or 3 days after learn- encoded memory for an individual’s future plans (218, 298,
ing (1147). In the later study, the reorganization led to 380, 1056, 1236, 1319). In a sequence-finger tapping task,
increased hippocampal and cortical activation as well as sleep preferentially improved the goal-related aspects (i.e.,
increased connectivity between the medial frontal cortical the target keys) rather than the movement-related aspects
and hippocampal areas, compared with a control group, per se (i.e., the tapping executed with specifically the left or

Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org 693

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 BJÖRN RASCH AND JAN BORN

right hand) (218, 1330). When subjects were promised an in experiments where subjects were instructed about the
extra monetary award after learning (and before sleep) for future relevance of the learned material after the learning
performing well on one of two equally trained sequences of phase (380, 1319) remains to be investigated. However,
a sequence-finger tapping task the next day, the sleep-de- theta-related tagging during encoding might represent a
pendent gain on this sequence was greater than for the other mechanism likewise mediating the preferential consolida-
sequence that was not associated with reward (380). Impor- tion of emotional contents and reward-associated behav-
tantly, before the actual retest the next day, the participants iors during sleep, as the network activated by theta also
were informed that both sequences were equally rewarded spans brain regions implicated in the processing of emo-
to ensure that motivation to perform well was equal for tional stimuli and reward, such as the amygdala and ventral
both sequences. Likewise, sleep-associated benefits for de- tegmental area, in addition to the prefrontal-hippocampal
clarative memories (e.g., visual-spatial and verbal paired axis (423, 728, 887).
associates) were significantly greater in subjects who were
informed before sleep that they would need the materials at This view of theta-related tagging of memories integrates a
a later recall test, than in subjects who were not informed in growing number of rodent studies indicating that sleep fa-
this way (1319), and only in the informed subjects later vors the consolidation of memories that essentially depend
recall performance correlated with SWA during postlearn- on hippocampal functions. In rats and mice, sleep specifi-
ing sleep. A similar selective effect of sleep was reported cally supported consolidation of contextual fear condition-
after learning two sets of picture-location association when ing which is well known to involve the hippocampal func-
only one set was labeled as relevant for later recall testing tion, but did not alter consolidation of cued fear condition-
(1236). Importantly, in all of these experiments, the expec- ing that does not require the hippocampal function (e.g.,
tancy about the later relevance of the memories was induced Refs. 476, 165, but see Ref. 676). In a spatial maze task,
after the encoding phase, eliminating possible differences in mice that were sleep deprived after learning shifted from a
memory strength related to relevance at encoding. Also, hippocampus-dependent to a striatum-dependent response
when participants were instructed to remember some and to strategy (e.g., Refs. 498 and 500; for related results, see
forget other items during learning, sleep preferentially Refs. 94, 1106, 1115). Likewise, in studies of novel object
recognition, novel place object recognition, and temporal
strengthened the to-be-remembered items (330, 381, 968).
order tasks, retention of the place and time of an event was
However, sleep also appeared to counteracted instructed,
found to require sleep after the learning phase, whereas the
i.e., “directed” forgetting effects in these studies (2, 330,
retention of the event per se (item recognition) did not re-
381). Similar to the preferential influence on relevant mem-
quire sleep (e.g., Refs. 92 and 593, but see Ref. 882). Hints
ories, sleep after encoding benefited the memory to perform
that consolidation of memories during sleep depend on hip-
an intended action at a designated time (298, 1056), sug-
pocampal activation during prior learning have also been
gesting an enhancing effect of sleep also on prospective
provided by human studies using functional magnetic reso-
memory for future plans.
nance imaging (fMRI) (16, 733, 968). In addition, some
human studies revealed that, compared with item memory,
Collectively these findings indicate that consolidation pro-
context memory, a core function of the hippocampus, is
cesses acting during sleep are driven by motivational factors
particularly sensitive to the beneficial effects of sleep (602,
and specifically strengthen those memories, which are rele- 732, 1128, 1230). Interestingly, a recent human study indi-
vant for our goals and future behavior. However, the mech- cated that working memory capacity, a function that in a
anisms underlying this selection process are unclear. Pre- healthy brain is most closely linked to a cooperative activa-
frontal cortex executive functions mediate the processing of tion of the prefrontal cortex and hippocampus, is strongly
anticipatory aspects of behavior and, in collaboration with correlated with sleep-dependent benefits for hippocampus-
the hippocampus, these prefrontal regions also regulate the dependent declarative memories (word pair associates)
implementation of anticipated memory retrieval as well as (374). In conclusion, activation of the prefrontal-hip-
the allocation of relevance and expectancies to a memory pocampal axis in the theta rhythm during the wake encod-
(219, 511, 820, 938). In rats, prefrontal cell assemblies that ing process might be the main factor that predisposes a
fired during learning when EEG theta coherence between memory for the system consolidation process that takes
the prefrontal cortex and hippocampus was increased place during subsequent sleep (64, 782).
showed a distinctly increased probability to be reactivated
during subsequent SWS (73). Thus theta coherence in the 3. A model of active system consolidation
prefrontal-hippocampal circuitry during the encoding of during sleep
explicit memories might be a critical factor that tags these
memories for later consolidation during sleep, with the pre- The findings discussed above have been used to essentially
frontal-hippocampal circuitry integrating also emotional refine the model of active system consolidation, mainly with
and reward-predicting aspects of the encoded events (74, respect to memories that are explicitly encoded, via activa-
423, 782). Whether theta coherence likewise mediates tag- tion of prefrontal-hippocampal circuitry, in the hippocam-
ging that occurred after actual encoding of the memory, i.e., pus-dependent declarative memory system. Basically, the

694 Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 SLEEP AND MEMORY

model assumes that during wakefulness the various aspects uting the newly encoded memories to long-term storage
of an experienced episode are encoded into cortical net- sites. These reactivations occur repeatedly and offline (dur-
works with the different parts of the new memory represen- ing sleep) to enable the gradual integration of the gist of the
tation bound together by areas in the medial temporal lobe, new representations into preexisting long-term memory
especially the hippocampus. During sleep, reactivation of networks, without overwriting these older memories. There
the episodic memory originating from hippocampal net- is now ample evidence that reactivations of memories occur
works results in the activation of the different memory parts during sleep.
also at the cortical level, thereby successively strengthening
cortico-cortical connections and transforming the tempo-
rary representations into long-term memories (158). Be- A. Animal Models
cause resources for the strengthening of synaptic connec-
tions are less available during SWS, reactivations during 1. Reactivation of hippocampal place cells
SWS might only tag the involved cortical synapses for later during sleep
strengthening during subsequent REM sleep, in accordance
with the sequential hypothesis (989).
Most reactivation studies in animals examined single unit
or multiunit activity from hippocampal place cells in rats.
Processes of abstraction, insight, and integration promoted
Place cells code for the position of the animal in space
by sleep in this model are thought to be a consequence of
relative to certain landmarks, i.e., they fire at substantial
reactivation-induced reorganization of memory representa-
rates whenever the animal enters a certain spatial field (152,
tion. Thus the repeated reactivation of episodic memory
868). Running along a particular path leads to a specific
representations during sleep may be capable of identifying
sequence of place cell firing, which reoccurs when the ani-
and extracting invariant features in the learning material
mal runs along the same path at different times, indicating
simply because commonalities between acquired memories
that the pattern of firing in place cell assemblies is linked to
overall are more frequently reactivated during sleep. In-
the memory of the spatial environment. Pavlides and Win-
deed, such extraction processes might facilitate the creation
son (893) were the first to show that hippocampal place
of prototypes and the development of cognitive schemes,
i.e., memories less dependent on specific contexts in which cells exhibit signs of memory reactivation during sleep.
they were learned (166, 734), and also less sensitive to in- They recorded spiking activity from pairs of hippocampal
terference (291, 295). Efficacy of this process is probably cells with nonoverlapping place fields. When the rats were
enhanced as the reactivation of new representations also not allowed to enter the place field of one of the two cells,
spreads to closely associated older memory representations, firing rates of the exposed cell increased relative to the un-
whereby reactivations simultaneously prime the gradual in- exposed cell, and this pattern reoccurred during sleep (SWS
tegration of new memories into networks of preexisting old and REM sleep) after the wake experience.
memories. Moreover, reactivation-induced reorganization
of memory representations during SWS might enhance the In a classic study, Wilson and McNaughton (1325) re-
accessibility of memories such that implicitly acquired reg- corded multiple pairs of hippocampal CA1 place cells,
ularities are strengthened and, after sleep, become accessi- which owing to their overlapping place fields exhibited cor-
ble to explicit assessment. For this sleep-induced gain of related activity while the rat was running along a track to
explicit memory, the strengthening of ventromedial pre- obtain a food reward. During subsequent SWS, the corre-
frontal cortical connections might be particularly impor- lation pattern of these place cells was strikingly similar to
tant (256, 1321). the pattern observed during the awake performance. Im-
portantly, the place cell assemblies did not exhibit such
firing patterns during sleep before running along the track.
III. MEMORY REACTIVATIONS DURING Reactivation of firing patterns during sleep after the perfor-
SLEEP mance on spatial tasks was likewise obtained in the dentate
gyrus (1065). Based on regression analyses, the activity pat-
The assumption that memories are reactivated during the tern observed during awake task performance was esti-
consolidation phase is an integral part of standard consoli- mated to explain 10 –30% of the variance in spiking activity
dation theory as well as of the “active system consolida- in the respective place cells during subsequent sleep (673).
tion” view of the memory function of sleep (418, 780, 800) Importantly, further studies showed that cells contributing
(see sect. IIF). Reactivations of memory representations are to reactivations are not only coactivated, but a temporal
thought to transform new memories that are still labile and order of spiking in the assembly is also preserved (1089). In
prone to decay into stable memories that are preserved for fact, entire spike trains observed during a task performed
the long-term. In the standard two-stage memory model, during prior waking reemerged in the same temporal order
comprised of a fast encoding temporary store (hippocam- during succeeding sleep, with this “replay” running at a
pus) and a slow-learning long-term store (neocortex, see faster speed, i.e., in a time-compressed form (715, 716,
sect. IB3), reactivations are regarded as critical for distrib- 843).

Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org 695

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 BJÖRN RASCH AND JAN BORN

Results from initial studies showing reactivation of assem- dles supporting consolidation processes during sleep (see
bly firing patterns during sleep were criticized because the sect. IV). In addition, after learning of a new rule in a Y-
tested animals were highly overtrained in the spatial tasks maze task, reactivation of learning-related patterns in the
(538, 1204). Thus, when rats repeatedly perform stereotyp- medial frontal cortex during sleep in distinct bouts mostly
ical behaviors in a familiar environment, also the firing followed hippocampal SW-R events with a slight delay of
patterns during pretraining sleep periods become more sim- 40 ms (926), i.e., a temporal relationship consistent with the
ilar to postsleep periods, and this similarity is reduced when notion of hippocampal replay guiding neocortical replay.
the rat learns a new behavior (551). This discussion moti- Neuronal replay activity during SWS was also revealed in
vated several studies examining reactivations during sleep subcortical regions including the ventral striatum where it
after rats had engaged in exploratory behavior involving the was linked to the learning of place-reward associations
encoding of new information. These studies uniformly con- (914). Obtaining rewards during learning was associated
firmed that signs of assembly reactivation occur in the hip- with strong firing in specific striatal cell firing patterns,
pocampus and several other regions of the brain also during which reemerged during succeeding SWS in close associa-
sleep after exploration (870, 871, 985). Reactivations could tion with hippocampal SW-R. Concurrent replay patterns
last for more than 24 h, although in the earlier of these in the hippocampus occurred shortly before striatal reacti-
studies the data analysis was criticized on methodological vations (699, 700) and, in contrast to hippocampal replay,
grounds (1180). Neuron assemblies linked to place fields
reactivation in the striatum did not decay during the 40 min
that were longer or more frequently explored during wak-
recording period (914). These findings indicate a leading
ing showed stronger reactivation during succeeding sleep
role of hippocampal memory replay for replay in neocorti-
(870, 871). Replay of hippocampal place cell assembly fir-
cal and striatal areas. In addition, they are well in line with
ing during sleep has been observed in many further studies,
the idea that redistributions of memory representations
with the important findings summarized as follows (for de-
during sleep originates from the reactivation of recently
tailed reviews, see Refs. 865, 1163): 1) during SWS, firing
patterns observed during waking are replayed at a much acquired memories in the hippocampus, thereby spreading
faster (10 –20 times) rate, suggesting a time-compressed the memory traces to neocortical and striatal sites that may
form of memory reactivation (160, 551, 606, 715, 843, serve as long-term stores.
1090); 2) replay activity is typically most prominent within
the first 20 – 40 min of sleep or rest after learning and decays 3. Reactivation during REM sleep
later (65, 673, 951, 1065, 1090), although in a few cases
signs of hippocampal reactivation persisted for more than In a few studies, signs of neuronal reactivations have been
24 h (985); and 3) in hippocampal neuron assemblies, reac- observed also during REM sleep (536, 544, 744, 893, 935,
tivations are mostly observed in conjunction with SW-R 936). Interestingly, with the rat growing more familiar with
(673, 849, 871), which are prominent oscillatory phenom- certain parts of a track, respective reactivations of hip-
ena of the hippocampal EEG and occur in an irregular fash- pocampal firing patterns showed a shift in the phase of the
ion mainly during SWS but also during quiet wakefulness EEG theta cycle such that their occurrence was more likely
(see sect. IVD). during troughs of the theta cycle rather than during peaks as
typically observed during wakefulness. As spike induced
2. Reactivation in nonhippocampal areas
plasticity is known to depend on the theta phase, these data
suggest that REM sleep may help erasing episodic memory
If memory reactivations during sleep promote the redistri-
information once it becomes familiar (110, 935). No reac-
bution of representations from temporary hippocampal to
long-term stores, as it is assumed by the active system con- tivation was observed when rats were allowed to visit a
solidation view, neuronal reactivations should also occur in novel portion of the track. Louie and Wilson (744) re-
other structures than the hippocampus. In fact, in rats, con- vealed reactivation patterns during REM sleep that
jointly with reactivations in hippocampal assemblies, reac- showed even greater similarities with patterns that had
tivations have been observed in the parietal (951) and visual emerged during REM sleep preceding the performance
cortex during sleep after performance on spatial tasks, on the spatial task (circular track), which was possibly
whereby replay in the visual cortex tended to slightly follow related to the fact that the rats were highly familiar with
replay in the hippocampus (606). Similar to hippocampal the task which they had performed over several days.
replay, reactivation of spatiotemporal patterns during sleep Others did not find any hints at experience-dependent
in the medial prefrontal cortex was compressed by a factor neuronal pattern reactivation during REM sleep (e.g.,
of 6 –7 (361). Furthermore, medial prefrontal reactivation Ref. 132, 673). Indeed, most reactivation studies in rats
correlated with the density of down-to-upstate transition did not extend their analyses to periods of REM sleep.
during slow waves in simultaneously recorded cells, K-com- Thus, although the presence of consistent reactivations of
plexes, and low-voltage spindles in local field potentials assembly firing patterns during REM sleep is also sug-
(608), supporting the notion of a functional association gested by modeling approaches (521), empirical evidence
between memory reactivations, slow-waves, and sleep spin- overall remains ambiguous in this regard.

696 Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 SLEEP AND MEMORY

4. Reactivation and behavioral relevance did not affect learning, whereas cueing during SWS and also
cueing of old, remote memories during REM sleep acquired
So far, surprisingly few studies have aimed to examine the 25 days before reduced fear memories (517, 518), indicat-
functional significance of firing pattern reactivations during ing that cueing of fear memories during sleep strengthen
SWS for memory consolidation. In this regard, first clues these memories only when cueing is conducted during REM
were provided by studies of old rats in which impaired sleep and as long as these memories are fresh. The effects
hippocampal reactivation patterns during rest were associ- could be connected to cue-specific arousal responses medi-
ated with reduced capabilities for forming spatial memory ated by the mesencephalic reticular formation (MRF) as
in the Morris water maze task (448, 449). In rats learning later studies of this group showed that MRF stimulation
goal locations in a spatial task, reactivation of goal-related during post training REM sleep, but not during waking or
firing patterns during SW-Rs after learning was predictive SWS, also enhanced memory though in a different task
for later memory recall (335). Importantly, blockade of (6-unit spatial discrimination maze) (535). Recordings of
NMDA receptors critical for spatial learning eliminated the single cells in the hippocampus confirmed that re-exposure
tendency to reactivate goal-related firing patterns during to the cue during REM sleep activated firing patterns similar
SW-Rs, and impaired later memory recall. However, it re- to those observed during training of fear conditioning
mained unclear whether reactivations occurred during sleep (756). Cue-induced firing patterns during SWS were not
or waking, as sleep was not recorded in this study. investigated in these studies. Increased neural responses
were also observed after cue reexposure during REM sleep
Direct evidence for a contribution of SW-R-associated re- in the medial geniculate body of the auditory thalamus and,
activations of assembly firing patterns for memory was pro- for aversive cues, in the lateral amygdala (542, 543, 755).
vided by Nakashiba et al. (849), who employed a genetic Taken together, these studies by Hennevin’s group indicate
approach to transiently block hippocampal CA3 output, that learning-induced plasticity can be reexpressed by cue-
which strongly reduced the occurrence of SW-R events. The ing during REM sleep, whereas expression of conditioned
blockade resulted in a marked reduction of hippocampal
responses was not consistently observed during SWS (541).
pattern reactivations during sleep and also reduced subse-
Indeed, inasmuch as all of these cueing studies involved a
quent retrieval performance in a context fear-conditioning
highly emotional, mostly aversive learning component, the
task. The acquisition of spatial memory was likewise im-
REM specificity of the observed effects fits well with the
paired when hippocampal SW-Rs in resting rats were sup-
notion that this sleep stage is critically involved in the pro-
pressed by electrical stimulation of hippocampal afferents
cessing of emotional memories (1273, 1294) (see sect. IIC).
(340, 458). In the latter study, most (84%) of the single pulse
Most recently, Bender and Wilson (75) specifically exam-
stimulations of the ventral hippocampal commissure occurred
ined whether cueing during non-REM sleep is capable of
in fact during SWS. While these studies provide solid evidence
reactivating hippocampal place cell activity implicated in
that SW-R events, which act as a carrier wave for neuronal
prior learning. Rats learned by reward to associate an au-
memory reactivations, are behaviorally relevant for memory
consolidation during sleep, direct disruption of memory reac- ditory stimulus (sound L or sound R) to either the left or
tivations and its impairing consequences for memory has not right portion of a linear track (length: 1.5 m). Reexposure
yet been demonstrated. to the sounds during subsequent non-REM sleep biased
reactivation of hippocampal place cells such that sound L
5. Reactivating memories during sleep by cueing preferentially activated place cells with a left-side place
field, and sound R preferentially activated place cells with a
Cueing procedures represent an important tool in examin- right-sided place field. The effect was stronger in the early
ing the functional significance of memory reactivation dur- compared with the later portion of non-REM sleep and did
ing sleep. In this approach, contextual cues are associated not only pertain to firing rates of individual place cells but
with the learning materials, which are redelivered during the task-related sounds effectively reactivated the temporal
subsequent sleep to reactivate at least parts of the acquired structure of the replayed events. Interestingly, cueing did
memory representations. Hennevin and co-workers (534, not increase the overall number of reactivations, suggesting
538) were the first to extensively apply this approach to a capacity-limiting mechanism for replay activity during
rats. They trained rats on an active avoidance-conditioning sleep (626).
task where a mild electrical shock to the ear was associated
with a strong aversive foot shock. Redelivery of the mild ear In sum, the findings in rats have compellingly demonstrated
shock during the first six periods of posttraining REM sleep that newly encoded memories are spontaneously reacti-
significantly increased recall of avoidance performance af- vated during SWS, whereby reactivations in the hippocam-
ter sleep and also increased the time in REM sleep (519). pus appear to lead reactivation in other neocortical and
The delivery of the mild ear-shock was ineffective when the striatal regions. Consistent with predictions from computa-
food-shock had been associated with a tone (rather than a tional models (613, 800), there is now also evidence that
mild ear-shock) at training, excluding unspecific effects of hippocampal activity during SWS (i.e., SW-R, in conjunc-
stimulus presentation during sleep. Cueing during waking tion with accompanying neuronal reactivations of memory

Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org 697

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 BJÖRN RASCH AND JAN BORN

representations) serve the strengthening of these memories, sequent SWS, with the size of hippocampal reactivation
and that these reactivation patterns can be cued by task- predicting navigation performance at a later retest. No signs
related stimuli presented during sleep. Moreover, experi- of reactivation occurred during REM sleep. These findings
mentally enforced memory reactivations during REM sleep are convergent with results on replay of firing patterns in
were consistently revealed to strengthen emotional memo- hippocampal cell assemblies observed during SWS in rats,
ries. However, it remains unclear how exactly this strength- and suggest that a similar process of hippocampal memory
ening is achieved by spontaneous or cued reactivations. reactivation occurs also in humans. Learning of face-scene
associations induced combined reactivation (in fMRI) in
hippocampal and face/scene selective visual cortical areas
B. Human Studies that occurred particularly during sleep spindles (84), when
functional connectivity between hippocampus and neocor-
1. Signs of reactivation in PET, fMRI, and EEG tex is generally increased (36). Notably, reactivations did
recordings not only occur in synchrony with spindles, but their size
appeared to be also modulated by spindle amplitude, a pat-
Signs of memory trace reactivation during sleep in humans tern of findings which is altogether consistent with the view
have been mostly reported using imaging of brain activation that spindles mediate hippocampo-neocortical interactions
with positron emission tomography (PET) or functional during declarative memory processing. Using surface EEG,
magnetic resonance imaging (fMRI). Compared with the enhanced EEG coherence, though in different frequency
multiunit recordings applied in corresponding studies in bands, was found during the learning of word pairs and
rodents, these brain imaging techniques suffer from the dis- during subsequent SWS, where the coherence effects con-
advantage that temporal and spatial resolution is very low. centrated on the up-state of the slow oscillations (831).
Generally, during non-REM sleep, brain activation and
connectivity are distinctly reduced (up to 40%) compared Although the available studies provide initial evidence that
with wakefulness, particularly in prefrontal cortical areas, reactivation of brain areas can be identified in PET, fMRI,
the anterior cingulum, and several subcortical structures and also EEG recordings, more consistently so during SWS
like the basal ganglia (134, 764, 766, 768, 1335). However, than REM sleep, they need to be better characterized using,
distinct increases in activity occur associated with sleep for example, higher resolution imaging or multivariate pat-
spindles or slow waves (250, 1036, 1132). During REM tern classifiers to determine more accurately when and
sleep, some brain areas including temporal-occipital corti- where they occur.
cal regions, the hippocampus and amygdala, exhibit activa-
tion comparable to that during waking, whereas activity in 2. Reactivating memories during sleep by cueing
others is relatively reduced (e.g., parietal and prefrontal
cortices) (189, 253, 764, 769, 1335). With regard to mem- Cueing has also been used in humans to examine the func-
ory processes, Maquet and colleagues (767, 912) were the tion of memory reactivations during sleep. In early studies
first to identify learning-dependent brain activation during using a related approach, participants learned Morse codes
sleep using PET. During REM sleep following a 4-h training and the same codes were acoustically presented again (at a
on a SRTT under implicit conditions (i.e., with the subject low nonwaking intensity) during subsequent REM sleep
being unaware of an underlying regular sequence to be (331, 489). Reexposure to the codes during REM sleep
tapped), activation was enhanced in several areas including containing phasic REMs (i.e., acute eye movements) in-
the bilateral striatum, parietal and premotor cortices, com- creased performance the next day compared with reexpo-
pared with a control group which had performed on a ran- sure during tonic REM sleep without REMs or a no stimu-
dom sequence before sleep, indicating that the activation lation control condition. In another study, subjects ac-
effects were specifically linked to procedural memory for- quired a set of complex rules in the presence of a loud
mation. Also, SRTT training changes functional connectiv- ticking alarm clock and were then reexposed to the ticking
ity patterns during subsequent REM sleep (707). Learning a sound during phasic REM sleep. At retest after 1 wk, these
perceptual skill (texture discrimination) was revealed to re- subjects showed a distinct improvement (by 23%) in mem-
activate blood oxygen-dependent (BOLD) signal in the ory for the rules compared with control subjects who had
trained region of area V1 in the visual cortex during subse- not acquired the rules in the presence of the ticking clock
quent non-REM sleep, with the magnitude of reactivation (1109).
predicting improvement visual discrimination skills at re-
test (1351). Because spontaneous reactivations (of hippocampus-de-
pendent memory) are much more consistently observed
Clear signs of reactivation during SWS were obtained (by during SWS than REM sleep, recent studies have turned
PET) also following hippocampus-dependent spatial learn- towards the examination of cue-induced reactivations
ing on a virtual navigation task (911). Learning to navigate, during SWS. In a study of ours (959), we used an olfac-
as expected, activated hippocampal and parahippocampal tory stimulus (the scent of roses) to reactivate visuospa-
areas, and the same areas were again activated during sub- tial memories for card-pair locations known to involve

698 Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 SLEEP AND MEMORY

the hippocampus (1121). Odors were used because they of posttraining SWS, performance on just this reactivated
do not affect the sleep architecture (178) and are also melody was improved at a later retest, compared with
known for their strong potency to activate associated performance on the nonreactivated melody. Taken to-
memories (197). The participants learned the card-pair gether, in showing that experimentally induced reactiva-
locations in the presence of the odor, with the odor being tion of declarative and also procedural memories ro-
reexposed during subsequent SWS (FIGURE 4A), REM bustly enhance later recall of the memories, these studies
sleep or while the subject remained awake. At a later demonstrate a causal role of such reactivations for mem-
retest, participants recalled significantly more card loca- ory consolidation. Thus the cueing of memory during
tions after reexposure of the odor during SWS compared sleep has provided valuable insight into the function of
with the other two conditions. The participants were not sleep-associated reactivations, stimulating to exploit this
aware of the odor presentation during sleep. Further con- approach, in future research, also for specifying the se-
trols specified that the memory-enhancing effect of odor quels of reactivations for the representational reorgani-
exposure during SWS critically depended on whether or zation memories undergone during sleep-associated sys-
not the odor was present during the prior learning phase, tem consolidation.
indicating that the memory enhancement was caused by a
reactivation of odor-associated memories and not simply 3. Memory reactivations and dreaming
by unspecific effects of odor-exposure during SWS (FIG-
URE 4B). Functional magnetic imaging confirmed that
It is unclear whether neuronal signs of memory reactiva-
odor presentation during SWS activated the hippocam-
tion during sleep are in any way linked to the recall of
pus, again only when participants had received the odor
dreams after awakening from sleep. Highly vivid and
also during prior learning (FIGURE 4E; Ref. 959). Inter-
emotional dreams are typically reported after awaken-
estingly, hippocampal activation during odor-induced re-
ings from REM sleep, whereas more thoughtlike dream
activation in SWS was distinctly stronger than during
reports can be obtained after awakenings from non-REM
wakefulness, suggesting that during SWS, the hippocam-
sleep (199, 554). Subjectively, reported dreams often
pus is more sensitive to stimuli capable of reactivating
cover an extended time period involving a sequence of
memories compared with wakefulness. Subsequent ex-
events, whereas signs of memory reactivation are usually
periments (Rihm, Diekelmann, Born, and Rasch, unpub-
restricted to brief intervals in the range of several 100 ms.
lished observation) showed that the improvement in re-
However, time perception during dreams might be com-
call was significantly reduced, if the odor presented dur- pressed, as has been reported for replay of hippocampal
ing SWS differed from that during learning (FIGURE 4C), neuron assemblies during SWS. Nevertheless, although
and so were also associated increases in EEG delta (1.5– semantic features from past experience are often in-
4.5 Hz) and spindle (13–15 Hz) activity during sleep, cluded, only a very small portion of dream reports (1–
suggesting that olfactory-induced reactivations of mem- 2%) incorporate genuinely episodic memories experi-
ory representations in the hippocampus can promote os- enced during presleep waking (405, 1050). Also, experi-
cillatory activity facilitating plastic changes in thalamo- mental cueing of memories during sleep did not produce
cortical circuits (see sect. IV, A and C). Recent experi- specific dream reports of task-related themes (959, 1018;
ments show that odor-induced reactivations during sleep Rihm et al., unpublished observation). On the other
can even induce the generation of more creative solutions hand, several experiments showed that specific waking
to a problem encountered before sleep (999). behaviors can influence subsequent dream content (1152).
When participants played an emotionally engaging ski
To differentially reactivate individual memory traces computer game, 30% of dreams reported upon awaken-
during postlearning SWS, Rudoy et al. (1018), rather ings (from non-REM sleep stage 1 and 2) shortly after
than using a global context cue, paired the location of sleep onset contained elements related to the computer
different cards (showing objects and animals) with a game (1304). The same group also reported a link be-
characteristic sound (e.g., meow for a cat). During a post- tween dream mentation during non-REM sleep and
training nap, only half of the sounds were administered memory consolidation after a nap (1303), although the
again to reactivate respective place-object associations. effect was based on only four subjects who actually re-
At later retrieval, memory for the reactivated associa- ported task-related mentation.
tions was significantly better than for the nonreactivated
associations. In a subsequent study, reactivation of the Overall, there is so far no convincing evidence for a direct
characteristic sounds was revealed to be associated with link between the reactivation of newly encoded memory
increased activation in the right parahippocampal cortex representations during sleep as evidenced by the record-
(1235). Cueing during sleep was also effective for skill ing of neuronal activity and reported dreams. Although it
memories. Antony et al. (40) trained participants to play cannot be excluded that aside from consolidating mem-
two different melodies on a piano keyboard. When only ory, these reactivations occasionally trigger certain frag-
one of the melodies was presented again during a period ments of memory that then are incorporated into dreams.

Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org 699

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 BJÖRN RASCH AND JAN BORN

C. Memory Reactivations in the Wake State ration or running in a maze (191, 268, 556, 673, 870; see
Ref. 175 for a review). They also occur conjointly with
1. Animal studies SW-Rs that are observed at somewhat lower rates during
wakefulness than during SWS (159, 191, 870). Different
Reactivations of hippocampal cell firing patterns occur also from sequenced reactivations in hippocampal assemblies
during waking when the animal rests after task perfor- during SWS which always occur in a forward direction,
mance or during brief pauses of active behavior like explo- replay of sequences during waking can occur in both for-

A Encoding Sleep Reactivation Encoding

Odor during learning Odor / No Odor

Vehicle
Wake
REM
N1
N2
N3

20.00 24.00 04.00 08.00

B Odor during Odor only C Different odors during

learning and SWS during SWS learning and SWS
% % %
*** **
120
100 100 *
Recalled card locations

110

90 90 100

90

80 80
80

0 0 0
Placebo Odor Placebo Odor Placebo Incongruent Congruent

D Odor during Odor during E Memory-related odor

learning and SWS learning and waking re-exposure during SWS
% %
100 100

***
90 90
Recalled card locations

80 80

70 70 *
60 60

50 50

40 40

0 0
Placebo Odor Placebo Odor

700 Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 SLEEP AND MEMORY

ward and backward directions (243, 406). In a task requir- formance on the pictures. Using an interference A–B A–C
ing the rat to run back and forth on the same elevated linear paradigm, Kuhl et al. (674) observed distinct hippocampal
track, replay in the reverse order occurred mainly at the end activation during learning of the new A–C object associa-
of a run, whereas replay in a forward direction transpired in tions (interfering with the first learned A–B associations),
the anticipatory period before a new run (288, 492). Wake which was predictive for later remembering the first-learned
reactivations were found to be particularly precise when the A–B associations, suggesting that reactivation of old asso-
animal explores a novel environment, and the precision ciations during new learning prevented forgetting. Overall,
decreases when the animal becomes more and more familiar these findings provide first hints that neuronal signs of
with the spatial task (191). In larger experimental environ- spontaneously occurring memory reactivations in the wake
ments, both forward and reverse replay can occur over mul- state can be identified also in humans.
tiple SW-R events thereby covering longer distances (268).
In addition to current learning experience, more remote 3. Comparing cueing of memories during
learning experiences are also reactivated, indicating that wakefulness and sleep
wake reactivations do not depend on the actual perceptual
input (621). Also, new path sequences that had never been Whereas the reactivation of memories during wakefulness,
experienced can be constructed during replay, which could as it occurs for example during rehearsal, can strengthen
facilitate short-cut learning and the creation of an allocen- these memories in the long run (623, 1006), the reactivation
tric cognitive map. However, the exact role of wake reacti- transfers the representation into an transient unstable state
vations for memory formation and their behavioral rele- such that memory is in need of reconsolidation (470, 718,
vance has not been thoroughly studied so far. One study 731, 844, 1033). Thus, according to the “reconsolidation”
reported that both the number of SW-Rs during learning a concept, memories exist either in an active or inactive state.
spatial task as well as during subsequent rest were predic- Consolidation transforms active but unstable memories
tive for later memory performance (335). into passive but stable memories, and reactivation renders
these memories again susceptible to interfering influences.
2. Human studies There is consistent evidence from animal and human studies
that experimentally induced memory reactivations can in-
Signs of spontaneous memory reactivations during the voke a transient labilization of respective traces, and also
wake state were also observed in humans. FMRI recordings the underlying neural and molecular mechanisms have been
in subjects performing on a vigilance task indicated biased partly characterized (403, 404, 588, 590, 639, 1043, 1118;
brain activation depending on whether the subjects had for reviews, see e.g., Refs. 14, 844, 1208). Indeed, the de-
performed before on either a procedural serial reaction task stabilization of long-term memories after reactivation
or a spatial navigation (913). Prior performance on the might be highly adaptive because it presents the opportu-
procedural tasks produced relative enhanced activation in nity to update memories with respect to new experiences
striatal and supplementary motor areas, whereas prior nav- (330, 513, 718, 957).
igation performance enhanced activation in temporal lobe
regions including the hippocampus. On a shorter time scale, Whether sleep differentially acts on processes of consolida-
application of multivariate pattern classifiers revealed that tion and reconsolidation is presently not known (1034,
MEG responses to sensory inputs (pictures of indoor and 1155, 1156, 1285). Yet, of more immediate relevance in
outdoor scenes) were replayed during the 5-s delay in a this context is the question: Does reactivation during SWS
working memory task (422). The strength of the replay was like reactivation during waking, transiently destabilize
modulated by the MEG theta rhythm, with the amount of memories? Indeed, in line with the “sequential hypothesis,”
theta phase coordination predicting working memory per- it has been proposed that reactivation during non-REM

FIGURE 4. Odor-induced reactivations during SWS benefit memory consolidation. A: procedures: participants learned a visuospatial memory
task (card-pair locations) in the presence of an odor. During subsequent SWS, they were either reexposed to the same odor serving as a cue
to induce memory reactivations, or received an odorless vehicle. After sleep, retrieval was tested (in the absence of the odor). B: odor-induced
reactivation of memories during SWS distinctly increased memory for card-pair locations, compared with vehicle condition. In a control
experiment, retention of card-pairs remained unchanged when the odor was not administered during learning, excluding unspecific effects of
odor exposure on memory processing during sleep. [Modified from Rasch et al. (959), with permission from American Association for the
Advancement of Science.] C: only reexposure during SWS to the same odor as during learning effectively enhanced card-pair memory (congruent
odor condition), whereas an odor different from that administered during learning (incongruent condition) was not effective. (Data from Rihm
et al., unpublished observation). D: odor-induced reactivations during SWS immediately stabilized memories against interference (induced by
learning an interference card-pair task shortly after reactivations during SWS). In contrast and consistent with reconsolidation theory,
odor-induced reactivations during wakefulness destabilized memories, as indicated by an impaired card-pair recall when reactivations during
waking were followed by learning an interference card-pair task. [Data from Diekelmann et al. (295).] E: odor-induced reactivations of memories
during SWS activated the left hippocampus as revealed by functional magnetic resonance imaging (fMRI). Values are means " SE: *P ! 0.05;
**P ! 0.01; ***P ! 0.001. [Modified from Rasch and Born (957), with permission from Elsevier.]

Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org 701

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 BJÖRN RASCH AND JAN BORN

sleep destabilizes memories, which in turn are stabilized IV. SLEEP-SPECIFIC ELECTRICAL
during subsequent REM sleep (293, 1156). The transient OSCILLATIONS
destabilization upon reactivation during SWS could ease
the integration of the newly acquired memory into preex- Sleep and sleep stages are characterized by specific field
isting, neocortical knowledge networks (957). However, potential rhythms of brain activity. Neocortical slow oscil-
recent experiments appeared to refute these hypotheses. In lations and SWA, thalamo-cortical spindles and hippocam-
this study by Diekelmann et al. (295), memories of card- pal SW-R have been associated with processes of memory
pair locations were reactivated either during postlearning consolidation during SWS and might support the reactiva-
SWS or wakefulness using a contextual odor cue. To probe tion and redistribution of memory representations during
stability of the reactivated memory trace, after reactivation this sleep stage. Theta rhythms and PGO waves have been
(and in the sleep group after being awakened) the partici- proposed to support REM sleep-dependent processes of
pants learned an interference task using the same card pairs consolidation and might support enduring synaptic plastic
but locations that differed from the originally learned task. changes during this sleep stage.
As expected from reconsolidation studies (e.g., Refs. 844,
845, 1043), reactivation during wakefulness destabilized
A. Slow Oscillations and SWA
memories rendering them susceptible to interference learn-
ing, as indicated by impaired memory recall for the origi-
1. Generation, propagation, and
nally learned card-pair locations after interference learning
homeostatic regulation
(FIGURE 4D). In sharp contrast, reactivation during SWS
had an immediate enhancing effect on the originally learned
During human slow wave sleep (SWS), the EEG shows pre-
card-pair locations, although the participants also in this
dominant slow wave activity (SWA), which is defined by the
condition learned the interference locations right after reac-
0.5- to 4.0-Hz frequency band and includes the $1-Hz slow
tivation. Note, in this condition memories were reactivated
oscillations with a peak frequency of 0.8 Hz (7, 830).
during the first period of SWS not followed by any REM
“Delta” activity refers to the 1- to 4-Hz band of SWA. Slow
sleep, arguing against the sequential occurrence of REM
oscillations comprise alterations between periods of neuro-
sleep as another prerequisite for the stabilizing effects of
nal membrane depolarization accompanied by sustained
reactivations during SWS. Additional fMRI recordings
firing (“up-states”) and periods of membrane hyperpolar-
showed that whereas memory reactivations during SWS
ization associated with neuronal silence (“down-state”).
mainly resulted in activation of the hippocampus and pos-
Steriade’s group was the first to demonstrate and to provide
terior cortical brain areas, reactivating memories during
an in-depth analysis of slow oscillations on the level of
wakefulness primarily induced activation of the lateral pre- intracellular and local field potential recordings in anesthe-
frontal cortex (295). tized cats (228, 1137, 1138, 1140, 1143–1145). Later stud-
ies confirmed that also during natural SWS, cortical neu-
In showing that the consequences of memory reactivation rons are indeed depolarized and fire during the depth-neg-
depend on the brain state with opposing effects induced ative (surface-positive) field potential of the slow oscillation
during waking and SWS, these data refute “opportunistic” half-wave and are hyperpolarized and silent during the
theories of sleep-associated memory consolidation (809) as- depth-positive (surface-negative) half-wave, whereas neu-
suming that reactivation-induced consolidation processes rons are depolarized and tonically fire during waking and
do not basically differ between sleep and wakefulness, apart REM sleep (552, 1143, 1146, 1199). Virtually every corti-
from the fact that sleep protects the processes from external cal neuron, both excitatory as well as inhibitory neuron,
interference (see sect. IIA). Yet, the mechanisms mediating engages in the slow oscillation with the firing patterns
reactivation-induced stabilization and destabilization in the showing high synchrony across cellular populations (33,
respective brain states are unclear. One factor could be the 188, 1139, 1146, 1261, 1268). The widespread synchroni-
cholinergic tone which is high during waking but at a min- zation of cortical and thalamo-cortical networks during
imum during SWS. Cholinergic activity might thus act as a non-REM sleep is considered the major function of the slow
switch that shifts information flow from the prefronto-hip- oscillation, providing a global time frame whereby the net-
pocampal direction prevailing during wakefulness, into the work is clocked and reset by the hyperpolarizing phase
opposite direction during SWS, i.e., from hippocampal to and neuronal processing is limited to the subsequent de-
neocortical networks (520, 957) (see sect. VB2). Related to polarizing up-phases (224, 282, 748, 828, 832, 1138,
this, (prefrontal) capacities of explicit encoding and re- 1143). Indeed, phase-locked electrical and transmagnetic
trieval monitoring available during waking might be critical stimulation has revealed the depolarizing up-phase as a
for whether reactivations actually produce a destabilization period of distinctly enhanced neuronal network excit-
of memory traces (839, 1308). In combination, these exper- ability (83, 792, 1200). In the scalp EEG, the negative
iments provide emergent evidence for a key role reactiva- peak of the slow oscillation coincides with the beginning
tion plays in all phases of memory formation that essentially of the down-to-up state transition (281, 993, 1261),
depends on the brain state. whereas the depolarizing phase of sustained firing corre-

702 Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 SLEEP AND MEMORY

lates with the positive EEG deflection (33, 998, 1268). The slow oscillation is generated in cortical networks and
K-complexes during non-REM sleep stage 2 appear to can occur in isolated cortical slices (239, 1028, 1198). Slow
represent isolated slow waves (181). oscillations originating from thalamo-cortical neurons
(586) vanish when the thalamus is isolated from cortical
The cellular mechanisms of slow oscillation generation are inputs (1200), indicating the slow oscillation is a primary
not fully understood. Whereas the up-state reflects some cortical phenomenon (993). Nevertheless, the slow oscilla-
balance between excitatory and inhibitory neuronal activity tion of the intact brain likely reflects an interaction between
(505, 1017, 1074), the hyperpolarizing down-state repre- cortical and thalamic networks (239, 1196). High-density
sents a period of disfacilitation that does not contain active EEG recordings in humans as well as depth recordings in
inhibition (229, 1146). Thus a central question is how up- epileptic patients and cats indicate that the slow oscillation
states are initiated when all neurons are hyperpolarized and behaves like a travelling wave, which originates most fre-
silent during the down-state? As a mechanism, hyperpolar- quently in the frontal regions and propagates towards pos-
ization-activated depolarizing currents (Ih) that depolarize terior regions, although other origins and directions of
propagation occur (188, 748, 791, 841, 859, 1261, 1314).
a subset of layer 5 cortical neurons have been discussed
Interhemispheric connections contribute to the propagation
(1028). However, the Ih is not strong in cortical neurons
(825). EEG-based source modeling located the main origins
and may increase firing only in conditions of increased ex-
of the travelling waves in the cingulate gyrus and the left
tracellular K% concentration, e.g., during seizures (1194).
insula (841, 993). These estimates roughly correspond with
Hence, the more likely explanation is that up-states are results from PET and fMRI studies likewise indicating fron-
triggered by the occasional summation of miniature EPSPs tal as well as midline structures as major sources of slow
as a residual synaptic activity resulting from stimulus pro- waves, i.e., the bilateral medial and inferior frontal cortices,
cessing during prior wakefulness, and formed mainly by precuneus and the posterior cingulate cortex (250, 252,
activation of T-type Ca2% currents in combination with a 557). Interestingly, the largest slow waves (#140 "V) were
persistent Na% sodium inward current (1193, 1198). Am- associated with activation in the parahippocampal gyrus,
plitude and frequency of miniature EPSPs are low under cerebellum, and brain stem, whereas smaller waves were
baseline conditions. However, large neuronal constella- more related to activation changes in frontal areas (788).
tions, particularly after intense encoding during prior wak-
ing, might provide a sufficient number of synapses enabling Traditionally, SWA is regarded as a marker of the homeo-
the generation of repetitive up-states. Moreover, extracel- statically regulated sleep pressure, which increases after
lular Ca2% concentrations are increased during the down- prolonged sleep deprivation and decreases from early to late
state, which enhances synaptic efficacy, such that a single sleep (111, 113, 994, 1269). It has been argued that the
spike generated in this condition may effectively excite the decrease in SWA across sleep reflects differences in the ho-
whole network (238, 790), in particular when such spikes meostatic regulation between $1 Hz slow oscillations and
originate from neurons in deep cortical layers with larger 1– 4 Hz delta oscillations (6, 171, 1207). However, as dis-
numbers of synaptic contacts to other neurons (188). Once tinct qualitative differences between both frequency bands
initiated, the up-state is likely amplified by activities of in- have not been confirmed, the decrease in SWA across sleep
trinsic currents such as persistent Na% and high-threshold appears to be most parsimoniously explained by a decrease
Ca2% currents. Contributions of glia cells to the regulation in the incidence of high-amplitude slow waves (1267). In-
of Ca2% concentrations and excitability in neighboring neu- dependent of the decrease in amplitude, the slope of the
rons are likely (30, 31, 506, 672). Synchronization of activ- slow waves also decreases from early to late sleep, possibly
ity during depolarizing states might be partially achieved reflecting a decrease in the synchrony and speed of recruit-
via corticocortical glutamatergic synaptic connections, im- ment of neurons at the down-to-up state transition (87,
360, 994, 1267).
plicating contributions of NMDA and AMPA receptor ac-
tivation to establish long-range synchrony (32, 360, 586).
2. The relation between SWA and memory benefits
Induction of the hyperpolarizing slow oscillation down-
phase has been mainly linked to synaptic depression, acti- There is now convergent evidence that SWA and the slow
vation of Ca2%-dependent and of Na%-dependent K% cur- oscillations represent a central mechanism conveying the ben-
rents, and the inactivation of persistent Na% currents gen- eficial effect of SWS on memory consolidation, in particular in
erally disfacilitating neuronal excitability (e.g., 68, 96, 391, the declarative memory system (see sect. IID). In animals, the
392, 437, 1053). Recent results suggest that active inhibi- encoding and learning of information during waking pro-
tion is also involved (190, 1027), which may be mediated by duced consistent increases in SWA in respective cortical net-
a particular set of cortical interneurons preferentially firing works during subsequent SWS (624, 1263). Thus hemispheric
towards the end of the up-state (950). During waking and differences in SWA during sleep in rats were related to the
REM sleep, the expression of down-states might be sup- preferential use of the left or right paw during the day (1265).
pressed mainly due to enhanced cholinergic activity in these When rats had to rely on their whiskers during the activity
states. period in darkness, subsequent SWA was higher in the so-

Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org 703

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 BJÖRN RASCH AND JAN BORN

mato-sensory cortex compared with a situation in which ad- In humans, intense learning of declarative memories (word
ditional light was available during the active period (1341). In pairs) enhanced amplitudes of the slow oscillation up-states
birds, sensory deprivation of one eye while watching a docu- as well as coherence in the SWA frequency band during
mentary about birds increased SWA and the slope of slow succeeding SWS (829, 831). Also, slopes of the down-to-up
waves only in the hyperpallium (a primary visual region) con- state of the slow oscillations were steeper after learning.
nected to the stimulated eye (726). Regarding procedural skills, training on a visuomotor ad-

W
REM
1
2
3
4
Declarative, Declarative,
non-declarative, Stimulation non-declarative,
control tests control tests
Learning Recall

B Frontal Parietal Consolidation

100 100

Sham

10 10
Stimulation
6
**
5
µV2

µV2

Recalled words
4
1 1
3

2
0.1 0.1 1

0
0 2 4 6 8 10 12 14 16 18 0 2 4 6 8 10 12 14 16 18 Sham SO-tDCS
Hz Hz
C Slow oscillations Slow spindles Consolidation

Fz Fz
-**- µV21 -**-
µV2 80

0 0 6
**
C3 Cz C4 C3 Cz C4
µV2 -**- -**- -**- µV2 5
Recalled words

1
80
4
0 0
P3 Pz P4 P3 Pz P4 3
µV2 -**- -**- -**- µV2
1
80 2
0 0 1

0
Sham theta-tDCS

D Frontal Parietal Encoding

VLMT
16
10 Sham * *
10 10 SO-tDCS
14
µV2

Number of words

0.1 12
µV2

µV2

1 1
0 1 2 3
Hz 10

0.1 0.1 8

0
0 4 8 12 16 20 24 0 4 8 12 16 20 24 R1 R2 R3 R4 R5 R6

704 Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 SLEEP AND MEMORY

aptation task increased the amplitude of slow waves during nal non-REM sleep, increased endogenous SWA, and pro-
subsequent SWS (581, 582). The increase was locally re- duced a significant improvement in the overnight retention of
stricted to the motor areas mainly involved in prior training word pair memories (785). Effects on word pair memories
and correlated with the overnight improvement on the task. were even more consistent with tDCS oscillating at 0.75 Hz,
Increases in local SWA were even observed when learning i.e., a frequency mimicking the endogenous slow oscillations
took place in the morning, suggesting that the training- (FIGURE 5B, Ref. 783). This type of stimulation applied during
induced changes in SWA do not depend on the time be- early non-REM sleep specifically enhanced $1 Hz slow oscil-
tween training and subsequent sleep (751). Conversely, arm lations and additionally increased frontal slow spindle activity
immobilization during daytime resulted in reduced SWA (10 –12 Hz). tDCS at the same frequency (0.75 Hz) during late
over the contralateral motor cortex (581). Correspond- REM-rich sleep was ineffective in enhancing both endogenous
ingly, reducing SWA during sleep by the presentation of slow oscillations and overnight retention of word pairs. Fur-
tones was revealed to suppress the sleep-dependent im- ther controls ensured that the effects depended on the fre-
provement in visuomotor adaptation skills (691, 692) and quency of the oscillating stimulation and on the brain state
also in texture discrimination skills (10). (FIGURE 5C): tDCS at the 5 Hz theta frequency during early
non-REM sleep had an immediate suppressing rather than
Several studies examined the effects of repetitive transcra- enhancing effect on endogenous slow oscillation and frontal
nial magnetic stimulation (rTMS) as a tool to directly in- slow spindle activity, and impaired overnight retention of
duce synaptic potentiation, on subsequent SWA. Applica- word pairs (784). Applying the 0.75 Hz slow oscillatory stim-
tion of 5 Hz rTMS over the motor cortex induced an im- ulation during wakefulness induced a widespread increase in
mediate potentiation of the TMS-induced cortical response theta (4 – 8 Hz), rather than slow oscillation activity, and this
which was followed by a marked (40%) increase in SWA in increase in theta activity was associated with a significant im-
the same cortical region during subsequent sleep (580). provement in the encoding of declarative memories, rather
Paired associative TMS stimulation (PAS) before sleep pro- than affecting retention of these memories (FIGURE 5D, Ref.
duced increases or decreases in subsequent SWS, depending 643). These results agree with findings that reveal increased
on whether the protocol successfully induced long-term po- cortical excitability following oscillatory electrical stimulation
tentiation-like increases or long-term depression-like de- in the wake state (83, 484) and provided initial hints at a
creases in cortical excitability as measured by motor evoked causal role of EEG theta activity for the encoding of new
potentials (583). In another study, similar effects of PAS memories in humans (646, 647, 687). In fact, combination
stimulation on SWA were associated with local changes in changes observed after tDCS at different frequencies seem to
slow spindle activity (85). Changes in frontal SWA after indicate that the cortical networks oscillating at the theta fre-
PAS appeared to extend even into succeeding REM sleep quency during encoding of hippocampus-dependent memo-
(443). Overall, these findings support the view that SWA (as ries are functionally linked to the networks that oscillate at the
well as the amplitude and down-to-up state slope of the slow oscillation frequency during subsequent SWS to consol-
slow oscillation during SWS) reflect the intensity and idate these memories.
amount of encoding during prior wakefulness, with some
studies also indicating an association of these measures with
later retrieval. B. The Synaptic Homeostasis Hypothesis

Direct evidence for a causal role of slow oscillations on sleep- 1. The concept
dependent memory consolidation is provided by studies ex-
perimentally inducing slow oscillations by transcranial direct An influential concept proposed by Tononi and Cirelli
current stimulation (tDCS; FIGURE 5A). In humans, tDCS that (1203, 1204) assumes that the slow waves of SWS serve
oscillated at a very low frequency (0.003 Hz; 30-s on/30-s off) primarily to globally down-scale the strength of synapses
and was applied to the prefrontal cortex during early noctur- that were potentiated in the course of encoding of informa-

FIGURE 5. Probing the functional relevance of slow oscillatory activity for memory processes by transcranial direct current stimulation (tDCS).
A: procedures: participants learned declarative and nondeclarative tasks before sleep and recall was tested in the next morning. During early
postlearning non-REM sleep, tDCS oscillating at different frequencies was applied via electrodes attached bilaterally over the prefrontal cortex
and to the mastoids. In a sham control condition, no current was applied. B and C: effects of tDCS depend on frequency of the oscillating
stimulation. B: tDCS during non-REM sleep oscillating at the 0.75 Hz slow oscillation frequency (SO-tDCS) increased endogenous slow oscillation
activity (0.5–1 Hz) at all recording sites and slow frontal spindle activity (8 –12 Hz), and these effects were associated with an enhanced retention
(consolidation) of declarative memory (for word pairs) across sleep, compared with the sham condition. [Data from Marshall et al. (783).] C: in
contrast, tDCS oscillating at 5 Hz (theta-tDCS) decreased slow oscillation activity at all recording sites and slow frontal spindle activity, and these
effects were associated with an impaired retention of declarative memory across sleep. [Data from Marshall et al. (784).] D: effects of SO-tDCS
depend on brain state: when applied during waking (rather than during non-REM sleep), tDCS induced a widespread increase in 4 – 8 Hz theta
and 16 –14 Hz beta activity, rather than slow oscillation activity, and these increases were associated with an enhanced encoding of declarative
memory (for words), particularly in later learning trials (R5, R6), whereas consolidation across the wake retention interval remained unaffected
(not shown). Values are means " SE: *P ! 0.05; **P ! 0.01. [Data from Kirov et al. (643).]

Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org 705

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 BJÖRN RASCH AND JAN BORN

tion during prior waking. According to this concept, linear firmed in related animal and human studies (994, 1269).
downscaling across synapses enhances memory indirectly Fourth, the synaptic homeostasis hypothesis assumes that
as this process nullifies the strength of connections that were memories are enhanced by sleep as a by-product of synaptic
only weakly potentiated during wakefulness leading to the downscaling. During learning, correct (signal) as well as
improvement of the signal-to-noise ratio for more strongly erroneous information (noise) is encoded, although the lat-
encoded memory representations. Synaptic downscaling ter at a weaker strength. The proportional downscaling of
can be considered a nonspecific process complementing ac- synapses during SWS reduces the weights of weaker syn-
tive system consolidation during sleep (293). apses below a threshold, making them completely ineffec-
tive, whereby the signal-to-noise ratio and the subsequent
The synaptic homeostasis hypothesis originated from Bor- recall of the memory representation is enhanced.
bely’s “two process” model of sleep (5, 8, 111, 112), as-
sumes that, apart from a circadian process C, sleep and The concept, and specifically the association of synaptic
specifically SWS is regulated by a homeostatic process S, potentiation during wakefulness and subsequent SWA, was
which rises during waking and declines during sleep. The confirmed by several studies measuring markers of synaptic
most important marker of process S is SWA during non- plasticity like plasticity-related genes [brain-derived neotro-
REM sleep, which reliably increases as a function of time phic factor (BDNF), activity-regulated cytoskeleton-associ-
awake and shows an exponential decrease in the course of ated protein (arc), Homer, nerve-growth factor-induced
subsequent non-REM sleep. The synaptic homeostasis hy- gene A (NGFI-A)] (584). In rats, learning a reaching task
pothesis basically links process S to processes of synaptic increased the protein expression of two activity-dependent
plasticity (1203, 1204). It relies on four key assumptions. genes c-fos and arc in the motor cortex and increased sub-
First, it is assumed that wakefulness in general is a state of sequent SWA in the same brain region (510). Suppressing
information intake and encoding that induces processes of expression of LTP-related genes in rats by chronic lesion of
LTP in cortical networks resulting in a net increase in syn- the noradrenergic system produced a strong reduction in
aptic weights. Second, it is assumed that SWA is a direct the homeostatic SWA response (see, e.g., Refs. 205 and 457,
marker of the amount of synaptic potentiation during prior for similar results in flies). Levels of postsynaptic glutama-
wakefulness such that “the higher the amount of synaptic tergic AMPA receptors containing the glutamate receptor
potentiation in cortical circuits during wakefulness [is], the (GluR1)-1 subunit reliably indicate synaptic plasticity and
higher [is] the increase in slow wave activity during subse- were revealed to be high during wakefulness and distinctly
quent sleep” (p. 144 in Ref. 1203). Third, it is assumed that lower (by 40%) during sleep (1264). Correspondingly, re-
SWA is implicated in the downscaling of synaptic weights. sults from a recent study suggest that discharge patterns of
As synapses are predominantly potentiated during wakeful- pyramidal neurons during SWS promote the removal of
ness, such process would inevitably lead to increasing en- synaptic Ca2%-permeable AMPA receptors in the somato-
ergy and space demands for permanently increased synaptic sensory cortex of juvenile rats (701). Procedures assumed to
weights, ultimately saturating the network thus impairing induce local net increases in synaptic potentiation (e.g., the
further encoding of information. SWA and in particular the cortical application of BDNF and KCl) induced cortical
$1 Hz slow oscillations are assumed to proportionally spreading depression (81) and increased SWA during sub-
downscale the potentiated synapses by a long-term depres- sequent sleep, whereas BDNF receptor inhibition decreased
sion-like mechanism, as neuronal firing at frequencies $1 SWA (366, 367). Importantly, in demonstrating that the
Hz is known to preferentially induce long-term depression encoding of information during waking enhances subse-
(630, 789). Moreover, down-states of neuronal silence fol- quent SWA in specific brain regions, these studies provide
lowing up-states of enhanced firing increase the probability compelling evidence that SWA is regulated locally in addi-
that presynaptic input is not followed by any postsynaptic tion to its global regulation by brain stem and diencephalic
output, i.e., a mechanism leading to further depotentiation structures (672, 859, 975, 1267).
of the network. Also, the neuromodulatory milieu during
SWS is characterized by low levels of acetylcholine, norepi- Supplementary evidence for sleep-associated synaptic
nephrine, and serotonin and therefore favors processes of downscaling was obtained in two recent studies measuring
depotentiation. The downscaling process, and in parallel synaptic growth. Using staining techniques in fruit flies,
the propensity for SWA, is self-limiting as synapses are Bushey et al. (156) showed that synapse size or number
gradually depotentiated, reaching a constant homeostatic increased after a few hours of wakefulness and decreased
level at the end of sleep. A computational model confirmed only when flies were allowed to sleep. Increased synaptic
that a decrease in synaptic strength can fully account for the growth due to an enriched wake experience (12 h spent
gradual decrease of SWA from the beginning to the end of together with 100 other flies) led to an increase in sleep with
the sleep period (360). Specifically, the model predicted a the time spent asleep after enriched experience being nega-
decrease in the incidence of high-amplitude slow waves, a tively correlated with spine density, suggesting that sleep
decrease in slope, as well as an increase in the number of renormalized synapses after potentiation during wakeful-
multipeak waves from early to late sleep, which was con- ness. In juvenile mice in vivo, two-photon microscopy in the

706 Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 SLEEP AND MEMORY

sensorimotor cortex revealed a net decrease in spine growth by learning - is determined solely by Hebbian LTP (or any
across periods of sleep, although there was also significant other single form of synaptic plasticity)” (p. 4 in Ref. 410).
spine growth at the same time (772). In adult mice, no Also, molecular markers of synaptic potentiation like arc,
similar changes were observed. BDNF or Calmodulin-dependent-kinase (CaMK) IV are in-
volved in both the potentiation and depression of synaptic
Further support for a net increase in synaptic weights across strength or other forms of non-Hebbian scaling (1026,
the wake period and a decrease across sleep derives from 1067), questioning that the relative increase and decrease of
measures of cortical excitability and actual firing rates these markers across sleep and wakefulness, respectively, is
which should be increased with net increases in synaptic actually related to one specific form of synaptic plasticity.
strength. Indeed, in rats, slope and amplitude of the electri- Furthermore, there is evidence that protein synthesis in-
cally evoked cortical potential as indicators of synaptic volved in LTP stabilization is enhanced during sleep, with
strength progressively increased during sustained wakeful- the rate of synthesis linked particularly to SWS (848, 956,
ness and decreased after sleep, and these changes were cor- 1245) (see also effects of sleep on ocular dominance plas-
related with SWA (1264). In humans, cortical excitability as ticity discussed in section VIIA3). Recently, Chauvette and
measured by a combined TMS/EEG study was increased colleagues (187) showed in vivo that somatorsensory
after sustained wakefulness and decreased after sleep (71),
evoked potentials in cats were enhanced after a short period
although another study reported a decrease in TMS-in-
of SWS (about 5–10 min) compared with the previous wake
duced cortical excitability after 40 h of sleep deprivation
period. Further experiments in vitro confirmed that the
(445). Firing rates of cortical neurons were high after
SWS-related enhancement in synaptic strength is a calcium-
periods of sustained wakefulness and decreased during
dependent postsynaptic mechanism which requires (i) the
sleep, with the decrease correlating with SWA during
hyperpolarizing down-state of the slow oscillation and (ii)
sleep (1268). Miniature excitatory postsynaptic currents
(mEPSC) are considered residual activity resulting from the coactivation of AMPA and NMDA receptors, suggest-
prior potentiating of synapses in the course of informa- ing that SWS is linked to synaptic potentiation rather than
tion encoding. Frequency and amplitude of mEPSCs were down-scaling (118). Related to that, although stimulation
increased in the frontal cortex slices of mice and rats after at the !1 Hz slow oscillation frequency typically induces
prolonged wakefulness compared with slices obtained LTD in vitro, in vivo such stimulation can remain ineffec-
after sleep (740). tive or even induce LTP (e.g., Refs. 497, 919). In fact, the
original concept of synaptic scaling as a self-tuning process
There is also evidence that, together with increased net syn- of neuron networks assumes that periods of reduced synap-
aptic potentiation, wakefulness leads to a gradual increase tic activity, like sleep, produce net up-scaling rather than
in energy demands, e.g., glucose uptake, of the brain. In- downscaling of synapses (409, 1219, 1220).
deed, cerebral metabolism measured by 2-deoxyglucose up-
take increased after periods of wakefulness and decreased The memory-benefit from sleep, according to the synaptic
after sleep in mice (1265). However, decreases in cerebral homeostasis theory, reflects an increased signal-to-noise ra-
metabolic rates after prolonged waking have also been ob- tio for the memorized representation that occurs as a by-
served in rats (363). In humans, cerebral glucose uptake product of proportional synaptic downscaling, shifting
measured by PET did not appear to be decreased when weakly potentiated synapses below a threshold thus nulli-
measured 2– 4 h after sleeping (157). Moreover, decreases fying their weight. So far, there is no experimental data
rather than increases in metabolic rates were observed after supporting the existence of such a threshold, nor for the
24 h of sleep deprivation (1186, 1336), although it cannot assumption that synaptic scaling in the cortical network is
be excluded that distinctly extended periods of wakefulness proportional (293). Compared with older memories, new
trigger separate compensatory processes characterized by memories appear to be generally more labile, i.e., linked to
the occurrence of local slow waves in the waking state and weak synaptic weights and thus would be at a greater risk of
by reduced metabolic demands (1204, 1267). being erased during sleep. This prediction contradicts ex-
perimental findings pointing towards the preferential con-
2. Critical issues solidation of weaker over stronger representations during
sleep (325, 343, 344, 375, 679). In classic interference par-
Although the synaptic homeostasis hypothesis integrates a adigms, learning a second list of words after a first list
wide variety of findings, especially on SWA, concerns have weakens memory for the first list learned due to retroactive
been raised about both the concept of downscaling in gen- interference. Contrary to predictions from the synaptic ho-
eral, and specifically about how the theory explains sleep- meostasis theory, the benefit from sleep for the first
dependent memory benefits. In light of clear evidence that (weaker) list of words learned was significantly stronger
learning during wakefulness relies both on LTP and LTD- than for the second list learned (325, 343). There is indeed
like mechanisms (e.g., Refs. 222, 629) “ . . . it is thus most a lack of any behavioral data indicating sleep-associated
improbable that sleep need - to the extent this is determined forgetting of irrelevant memories (293).

Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org 707

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 BJÖRN RASCH AND JAN BORN

Signs of reactivations of neuronal memory representations hippocampus-dependent declarative tasks in the absence of
observed during SWS, in the synaptic homeostasis view, changes in motor skill learning, were revealed after enhanc-
reflect ongoing activity in hypermetabolic traces induced by ing sleep SWA by transcranial direct current stimulation
intense learning in these networks, without any relation to (oscillating at the 0.75 Hz slow oscillation frequency) (39).
memory consolidation (869). This contrasts with findings However, no signs of improved encoding were revealed
indicating that an external triggering of memory reactiva- after a drug-induced enhancement of SWA by sodium oxy-
tions during sleep enhances memory performance the next bate (1297). After 36 h of total sleep deprivation, encoding
day (295, 959, 1018) and that spontaneous memory reac- of emotional pictures was strongly impaired, with this effect
tivations during sleep correlate with later memory perfor- sparing negative emotional pictures (1291). Interestingly,
mance (335, 448, 449). Consequently, findings of a sleep- amygdala responses during the viewing of aversive pictures
dependent reorganization of neuronal memory representa- increased after 35 h of sleep deprivation (1345). Collec-
tions also represent a challenge to the synaptic homeostasis tively, these initial data corroborate the idea that SWA dur-
theory inasmuch as this implies that some parts of a repre- ing sleep refreshes capacities for the encoding of informa-
sentation undergo up-scaling during sleep while others are tion possibly by a synaptic down-scaling-like mechanism. It
down-scaled (see sect. III). As any synaptic downscaling per remains unclear why the effect apparently predominates in
se cannot explain the memory consolidating effects of sleep, hippocampal networks that encode information in declara-
reactivation-based consolidation has been proposed as a tive tasks, although the hippocampus itself does not gener-
core mechanism driving active system consolidation during ate slow oscillations (597).
SWS, including basically local processes of up-scaling and
synaptic strengthening. However, these processes might be
complemented and even enhanced by an unspecific process C. Spindles
of synaptic downscaling that acts on a global scale to main-
tain overall synaptic homeostasis (293). 1. Spindle generation, fast and slow spindles

3. SWA enhances subsequent encoding Spindle activity refers to regular EEG oscillatory activity
which occurs in a frequency range between !10 and 15 Hz
Importantly in the synaptic homeostasis theory, sleep-asso- and expresses in human non-REM sleep stage 2 as discrete,
ciated synaptic down-scaling primarily serves to renormal- waxing and waning spindles lasting 0.5–3 s (440). Spindles
ize synaptic weights in networks that, in the course of the occurring during stage 2 sleep low-voltage EEG can be tem-
encoding of information during prior wakefulness, were porally locked to a vertex sharp wave or a K-complex.
potentiated to close to saturation. Therefore, rather than Spindles are also present during SWS and superimposed on
support the consolidation of memory, synaptic down-scaling delta activity, and then form less clearly discrete spindles
during SWS is expected to ease the encoding of new informa- (38, 53, 303, 426, 440, 442, 785, 1083, 1221, 1355). Al-
tion during subsequent wakefulness. Indeed, there is conver- though in the beginning of SWS spindle activity can reach
gent evidence for such action in humans and rats, i.e., an im- levels similar to those in stage 2 sleep, on average spindle
pairment of learning when they were deprived of sleep (195, activity in SWS is lower than during stage 2 sleep. There is
328, 500, 634, 759, 1291, 1345). Fittingly, in rats, sleep a reciprocal relationship between spindles and SWA such
deprivation also hampered the induction of LTP in the hip- that while SWA progressively decreases across nocturnal
pocampal regions (170, 778, 1161, 1179), the expression of sleep in humans, sleep spindles and power in the 12–15 Hz
plasticity-related genes (496) as well as excitability of the (“sigma”) band tend to increase (11, 442). Sleep depriva-
hippocampal neurons (801, 802). Selective deprivation of tion typically reduces spindle activity during subsequent
REM sleep appears to be sufficient for producing an impair- recovery sleep, together with an increase in SWA (114,
ment in LTP induction (269, 270, 596, 636, 743, 970, 300 –302), and similar reciprocal changes, i.e., decreases in
1010). In humans, the capacity to encode episodic memo- SWA conjoint with increases in spindle activity, are ob-
ries deteriorates across the day and is restored after a period served after administration of GABAA receptor agonistic
of sleep (759). Prior sleep deprivation impairs verbal learn- drugs (see sect. VA2).
ing ability and reduces temporal lobe activation during
learning (328, 329). Interestingly, the ability to encode pic- Spindle activity originates in the thalamus from mutual in-
tures after a nap was already substantially diminished when teractions between GABAergic neurons of the nucleus re-
SWA during napping was selectively attenuated by contin- ticularis (670) which function as pacemaker, and glutama-
gent mild acoustic stimulation leaving the gross sleep archi- tergic thalamo-cortical projections which mediate the syn-
tecture intact (1232, 1233). The impairing effect on picture chronized and widespread propagation of spindles to
encoding was accompanied by reduced hippocampus acti- cortical regions (227, 440, 1139, 1195). The isolated retic-
vation during learning. Surprisingly, no change in activa- ular thalamic nucleus is indeed able to generate spindles,
tion in relevant neocortical areas was found, and suppress- whereas the thalamus isolated from the reticular thalamus
ing SWA also did not impair the learning of a motor skill is not (1141, 1142). Thalamic generation of spindles is
task. Complementary effects, i.e., enhanced encoding on linked to activation of T-type calcium channels (45, 67,

708 Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 SLEEP AND MEMORY

1162) and calcium-dependent small-conductance channels nomena and are restricted to specific regions and circuits,
(1326). Within neocortical networks, spindle activity is regardless of whether or not they are synchronized in phase
probably associated with a massive calcium influx into py- with central thalamic spindle generation.
ramidal cells (1061). Repeated spindle-associated spike dis-
charges efficiently triggered LTP in neocortical synapses in 2. Relationship between spindles and memory
in vitro models (1012, 1197). In vivo, synchronous spindle
activity occurred more reliably in neocortical synaptic net- Many studies in humans and animals have indicated a ro-
works that had been previously potentiated through tetanic bust association between spindle activity and memory pro-
stimulation (1315). Correspondingly in humans, the local cessing during sleep. Effects appear to be particularly con-
expression of spindle activity during non-REM sleep in- sistent for the classic fast spindles. However, not many stud-
creased or decreased depending on whether LTP-like or ies differentiated both types of sleep spindles. Intense
LTD-like plasticity had been induced through transcranial learning of declarative memories (word pairs or virtual
magnetic stimulation (TMS) before sleep (85). These obser- maze) increases the number of spindles (12–15 Hz) during
vations suggest that spindles occur preferentially in poten- subsequent sleep, particularly in the early part of the night
tiated synaptic networks and may contribute to maintain- (432, 812). In another similar study, only participants who
ing this potentiation. exhibited enhanced spindle activity (11.5–16 Hz) after
word pair learning (in comparison with sleep after a non-
Studies in humans have consistently revealed the presence learning control session) also showed a significant over-
of two kinds of spindles: fast spindles (!13–15 Hz) show a night improvement of verbal memory (1037). Encoding of a
more widespread distribution concentrating over the cen- difficult list of abstract words produced an increase in the
tral and parietal cortex, whereas slow spindles (!10 –12 power and density of spindles (11.27–13.75 Hz) compared
with sleep after encoding a list of easier, concrete words
Hz) show a more focused topography over the frontal cor-
(1044). In rats, robust increases in sleep spindles (12–15
tex and are more pronounced during SWS than stage 2 sleep
Hz) were observed after learning odor-reward associations
(34, 440, 826, 1184). The two types of spindles differ in
(354) and after avoidance training (397). Furthermore, in-
many aspects, including their circadian and homeostatic
dicators of sleep spindle expression consistently correlated
regulation, pharmacological reactivity, as well as their age-
with the amount of overnight retention of declarative mem-
related changes (440). With the use of EEG and MEG,
ories (86, 213, 214, 235, 446, 562, 822, 1016, 1024, 1037,
neocortical sources of the classic fast spindles have been
1044, 1057). Interestingly, spindle activity (11–15 Hz) also
located in the precuneus, and for slow spindles in the pre-
correlated with signs of overnight lexical integration of
frontal cortex (Brodman areas 9 and 10) (34, 763). Com-
newly learned information, suggesting that spindles con-
bined EEG/fMRI recordings revealed that both spindles are
tribute to the integration of new memories into existing
associated with increased activity in the thalamus, anterior neocortical knowledge networks (1175). Moreover, spin-
cingulate, and insula cortices (1036). However, slow spin- dles (13–15 Hz) during a postlearning nap predicted sleep-
dles were associated with increased activation in the supe- dependent memory improvement for contextual aspects of
rior frontal gyrus, whereas fast spindles recruited medial episodic memories known to most closely depend on the
frontal, midcingulate, sensorimotor, and supplementary hippocampal function (1230). Using combined EEG/fMRI
motor cortical areas. Importantly, fast spindles were also recordings, Bergmann et al. (84) observed conjoint reacti-
associated with increased activation in the hippocampus, vations in relevant neocortical and hippocampal regions
suggesting a particular relationship between classic fast that occurred in temporal synchrony with spindle events
spindles and hippocampus-dependent memory processes (12–14 Hz) during non-REM sleep after learning of face-
during sleep. It has been suspected that slow spindle activity scene associations. The strength in reactivations covaried
reflects predominant coupling among cortical networks, with spindle amplitude. Together these findings provide
whereas fast spindles may be more closely related to first hints that spindles are implicated in the hippocampo-
thalamocortical coupling (317). Indeed, optogenetic stimu- neocortical exchange of memory information mediating ac-
lation of reticular thalamic neurons can induce spindles in tive system consolidation during sleep. Interestingly, spin-
the neocortex, that are not accompanied by thalamic spin- dle activity following exposure to a novel spatio-tactile ex-
dles (507), suggesting an active role of the neocortex in the perience in rats predicted immediate early gene activity (Arc
expression of spindle oscillations. Despite such evidence, it expression) in the somatosensory cortex during later REM
remains a matter of debate whether fast and slow spindles sleep, suggesting that neocortical parts of representations
actually reflect different neural processes or are simply the become tagged for later synaptic strengthening during spin-
modulation of a single spindle generator (440, 788). The dles (989).
picture is complicated by recent MEG data as well as intra-
cranial recordings from neurological patients that identified Spindles also appear to be involved in the consolidation of
multiple local generators for neocortical spindles (38, 52, skills, especially of simple motor (e.g., pursuit rotor task,
276, 859). Importantly, this research revealed that in the simple sequential finger tapping) and visuomotor skills
neocortex, single spindles typically express as a local phe- (1114). Increases of spindles (13–15 Hz) and stage 2 sleep

Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org 709

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 BJÖRN RASCH AND JAN BORN

are particularly observed following training of tasks like In rats, learning of an odor-reward association task pro-
figure tracing (59, 394 –396, 399, 837, 920, 921, 1173, duced a strong and long-lasting (up to 2 h) increase in the
1174), which were also revealed to be vulnerable to stage 2 magnitude of ripples and the number of ripple events during
sleep deprivation (1112, 1117). In some studies the increase subsequent SWS (355). Similarly, after a spatial learning
in spindle activity was topographically restricted to the cor- task, increases in ripple density during postlearning sleep
tex areas most strongly involved in skill performance, i.e., in were significantly correlated with the formation of associa-
the contralateral motor cortex after motor tapping training tive spatial memories (955). In epileptic humans, the num-
(862) and in the parietal cortex after training in a visuospa- ber of rhinal ripples during a nap correlated positively with
tial skill (214). Also, correlations between different param- the consolidation of previously acquired picture memories
eters of spindle activity or stage 2 sleep and overnight im- (50). Two recent studies in rats demonstrated a causal role
provements in motor skills have been consistently reported of SW-Rs in memory consolidation (340, 458). In both
(562, 862, 1216, 1286, 1287), and these correlations were studies, emergent ripple events were selectively disrupted by
revealed for fast rather than slow spindles, in the studies electrical stimulation during the rest period after learning,
distinguishing these two spindle types (59, 962, 1174). without disturbing sleep, which distinctly impaired consol-
idation of the acquired spatial memories. In conclusion,
In spite of the quite robust correlational evidence for the there is now good evidence that SW-Rs are critically in-
involvement of spindles in memory processing, a causal role volved in the consolidation of hippocampus-dependent
of spindles in memory consolidation has not yet been dem- memories during sleep.
onstrated, as this requires the selective experimental manip-
ulation of spindle activity. Approaches to increase spindle
activity pharmacologically by administration of GABAA re- E. Slow Oscillations, Spindles,
ceptor agonistic drugs do not appear to be suitable in this and SW-R Interact
regard, as they concurrently decrease SWA (see sect. VA2).
In another approach, neurofeedback training to voluntarily The fine-tuned temporal relationship between the neocorti-
increase EEG power in the 11.6 –16 Hz sigma frequency cal slow oscillations, thalamic spindles, and hippocampal
band produced a small increase in sigma power also during SW-Rs originates from a top-down control of the slow os-
subsequent sleep (86). This increase, however, was not as- cillation on the two other events. In addition to the neural
sociated with any change in the overnight retention of de- synchronization in the neocortex, the synchronizing effects
clarative memories. of slow oscillations spread to thalamic and hippocampal as
well as other brain regions involved in off-line memory
consolidation. At the neocortical level, the slow oscillation
D. Sharp Wave-Ripples in terms of EEG rhythms strongly modulates the amplitude
of faster beta and gamma frequencies which is reduced to a
Hippocampal sharp waves are fast, depolarizing events gen- minimum during the hyperpolarizing down-phase of the
erated in CA3 that become superimposed by ripple activity slow oscillation (224, 240, 828, 829). Generation of spin-
[i.e., high-frequency local field potential oscillations (100 – dles in the thalamus and generation of SW-Rs in the hip-
300 Hz) originating in CA1] to form SW-R events (159, pocampus are distinctly suppressed during the down-phase
162, 196, 241, 459, 832, 1342). SW-Rs occur mainly dur- of the slow oscillation, and this is followed by a rebound in
ing SWS but also during nonexploratory wakefulness (e.g., spindle and SW-R activity during the succeeding depolariz-
drinking, grooming, and quiet wakefulness). Ripples arise ing up-state (66, 215, 597, 829 – 832, 925, 1088). As for
from an interaction between inhibitory interneurons and spindles, only the classic fast spindles display the strong
pyramidal cells via synaptic (glutamatergic, GABAergic) phase synchronization with the emergent depolarizing up-
connections and gap junctions (162). Hippocampal stimu- state of the slow oscillation. Slow frontal spindles, in con-
lation protocols that induce LTP concurrently facilitate the trast, tend to follow fast spindles by 200 –500 ms and thus
generation of SW-Rs in CA3, and SW-Rs during sleep can occur already in the up-to-down state transition of the slow
be initiated by neurons whose recurrent connectivity had oscillation (38, 826). As for the modulation of SW-Rs, this
been transiently potentiated during preceding wakefulness appears to be entirely driven by cortical inputs, as the hip-
(70). SW-Rs may conversely promote LTP and spike time- pocampus itself does not generate slow oscillations (597).
dependent plasticity in hippocampal circuits (97, 159, 241, However, membrane potentials with a slight delay (of !50
640, 777). Modulation of neuronal connectivity during rip- ms) follow the up- and down-states of neocortical slow
ples might be specific to local circuits, because firing during oscillations, particularly in dentate gyrus and CA1 (502,
single ripples involves only small subpopulations of pyra- 503, 1333). In parallel, neocortical up states might time-
midal cells and is highly variable across multiple succeeding lock spontaneous activity in the medial entorhinal cortex, a
ripples (241, 1342). Most importantly, SW-Rs during SWS major gateway between the neocortex and hippocampus
typically accompany the reactivation of neuron ensembles (501). There is also evidence for a slow oscillation modula-
active during the preceding wake experience (see sect. IIIA). tion of locus coeruleus burst activity with preferential locus

710 Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 SLEEP AND MEMORY

coeruleus firing during the down-to-up state transition of While thalamo-cortical sleep spindles enforce the genera-
the slow oscillation (353, 727). tion of hippocampal ripples, ripple events in turn feed back
to sustain the ongoing generation of spindle activity (828),
Importantly, prior learning appears to strengthen the top- and possibly also slow oscillation activity, although this is
down control of slow oscillations on spindles and ripples. In presently unclear (635, 1087). Such looplike coordination
humans, intense learning of vocabulary did not only in- goes along with an enhanced formation of spindle-ripple
crease the slope of the down-to-up state transition of the events whereby ripples (together with the reactivated hip-
slow oscillation during succeeding non-REM sleep, but pocampal memory information they carry) are fed exactly
concurrently enhanced fast spindle activity, with this in- into the excitatory phases of the spindle cycle. By still reach-
crease concentrating on the up-states of the slow oscillation, ing neocortical networks during the depolarizing up-phase
whereas no changes were observed in hyperpolarizing of the slow oscillation, the spindle-ripple event may thus
down-states (826, 829). Interestingly, these analyses also serve as an effective mechanism for transferring hippocam-
revealed that learning promoted the occurrence of trains of pal memory information towards neocortical long-term
several succeeding slow oscillations. In these trains, fast stores. Indeed, fast spindles do not only phase-lock hip-
spindles were not only driven by the depolarizing slow os- pocampal ripples but also neocortical gamma-band activity
cillation up-state but appeared to feed themselves back to as an indicator of coherent information processing in local
enforce the succeeding slow oscillation, as well as the like- neocortical networks (52). Spindle-gamma coupling might
lihood of associated slow frontal spindles. The enhance- be linked to facilitate synaptic plastic processes underlying
ment of such slow oscillation-spindle cycles might be a key the storage of information in neocortical circuitry (1012).
mechanism whereby fast spindles initiate the consolidation
of newly learned materials during sleep (826). Learning also In sum, there is now growing evidence for a dialogue be-
increases hippocampal SW-Rs (355, 955), and it is likely tween neocortex and hippocampus mediating the system
that learning-induced increases in slow oscillations consolidation of hippocampus-dependent memory, which
strengthen parallel to the synchronization of enhanced is orchestrated by a fine-tuned interaction between oscilla-
SW-R activity and the depolarizing up-state of the slow tory field potential activities. In this dialogue, the neocorti-
cal slow oscillation provides a top-down temporal frame
oscillation, which remains to be demonstrated.
that synchronizes the reactivation of hippocampal memo-
ries with the simultaneous occurrence of fast spindles to
SW-Rs are also temporally coupled to spindles (1075,
enable the formation of spindle-ripple events. Spindle-rip-
1088) which can only be partly explained by the common
ple events occurring during the slow oscillation up-states,
driving impact of the slow oscillation on both phenomena
conversely, provide a bottom-up mechanism for the trans-
(829). Event-correlation histograms derived from intracra-
fer of reactivated memory information from the hippocam-
nial recordings in rat and humans revealed that ripples are
pus to the neocortex where they might effectively support
associated with a rise in spindle activity that starts shortly
plastic synaptic processes underlying the storage of this in-
before ripple onset and then outlasts the ripple (215, 216,
formation.
829, 832). The rise in spindle activity was even more per-
sistent (up to 2 s) when the rats had performed on a learning
task prior to sleep (829). Moreover, fine-grained temporal F. PGO Waves and Theta-Rhythm
analyses revealed that the co-occurrence of spindles and of REM Sleep
ripples leads to the formation of so-called “spindle-ripple
events” where individual ripple events become temporally 1. PGO waves
nested into succeeding troughs of a spindle (216, 1075,
1317). As ripples accompany assemble reactivations in the PGO waves are driven by an intense burst of synchronized
hippocampus, spindle-ripple events might represent a activity that propagates from the pontine tegmentum to the
mechanism serving the sequenced transfer of reactivated lateral geniculate nucleus and visual cortex. They occur in
memory information towards neocortical sites (827, 1087). temporal association with rapid eye movements in rats and
Importantly, the formation of spindle-ripple events is re- cats. They are not readily identifiable in the human EEG,
stricted to classic fast spindles, which (like SW-Rs) are though fMRI studies revealed activations in pontine teg-
driven by the depolarizing down-to-up state transition of mentum, thalamus, primary visual cortex, putamen, and
the slow oscillation. In contrast, the slow frontal spindles limbic areas associated with rapid eye movements during
typically occur 200 –500 ms later in the slow oscillation REM sleep, which might be linked to PGO waves (824,
cycle (i.e., at the up-to-down state transition), and thus also 1312). PGO waves tend to occur phase-locked to theta
tend to follow hippocampal SW-Rs with the same delay oscillations (618). Like theta activity, PGO waves have been
(216). In contrast to fast spindles, the neocortex might be proposed as a mechanism supporting synaptic plasticity in
“functionally deafferented” from its hippocampal inputs the regions they reach (258, 264). Training on an active
during frontal slow spindles (925). Taken together, these avoidance task is followed by a robust increase in PGO
data suggest a looplike scenario during sleep after learning. wave density for 3– 4 h after training, and changes in PGO

Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org 711

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 BJÖRN RASCH AND JAN BORN

wave density were also proportional to the improvement in spatial tasks, these reactivations were specifically linked to
task performance between initial training and post-sleep ongoing theta activity, with reactivations expressing either
retest (259, 261, 1225). Moreover, PGO wave density dur- during a specific phase of the theta cycle or linked to a
ing posttraining REM sleep is correlated with increased specific amplitude modulation of ongoing theta activity
activity of plasticity-related immediate early genes and (744, 935) (see sect. IIIA3). In humans, scalp-recorded EEG
brain-derived neurotrophic factors in the dorsal hippocam- theta activity was enhanced during REM sleep following
pus. These increases were abolished after selective elimina- learning of paired associates (399) and was correlated with
tion of the PGO wave generating cells in the brain stem and consolidation of emotional memories (specifically over the
enhanced after cholinergic stimulation of these cells (261, right prefrontal cortex relative to the left) (861). However,
1225). the overnight reduction in amygdala activation in response
to emotional pictures presented before and after sleep was
2. Theta activity correlated with gamma rather than theta activity during
intervening REM sleep (1231). In patients with Alzheimer’s
Theta activity, i.e., periods of synchronized activity in the disease, theta activity during not only REM but also SWS
4 – 8 Hz frequency band, is a hallmark of tonic REM sleep was faster compared with age-matched controls, and fast
(REM sleep without actual rapid eye movements) (161) and theta activity correlated with better overnight memory for-
correlates with rapid eye movements and PGO waves (618 – mation (572).
620). Primary generating mechanisms appear to be located
within the hippocampus, i.e., in CA1 (472), although extra- While these data so far do not speak for any essential func-
hippocampal input, mainly from the septum, contributes. tion of REM sleep theta activity in memory consolidation,
In rodents, theta predominates in the hippocampus and the characteristics of theta activity as well as faster EEG
associated areas where it is likewise seen during awake ex-
frequencies during REM sleep point out an important fea-
ploratory behaviors. In humans, EEG theta activity during
ture relevant to putative memory processing during this
REM sleep is less coherent and persistent in hippocampal
sleep stage. Compared with wakefulness or SWS, EEG ac-
regions and is seen as well in neocortical areas, again also
tivity during REM sleep shows reduced coherence between
during wakefulness (174, 860, 1222, 1223). Theta activity
limbic-hippocampal and neocortical circuitry in a wide
in waking is considered a condition favoring the encoding
range of frequencies including theta and gamma (51, 174,
of new information and associated synaptic plastic pro-
293). Similarly, within hippocampal circuitry, gamma band
cesses in hippocampal networks (64, 371, 612, 1251). Burst
activity shows reduced coherence across the CA3 and CA1
stimulation of CA1 inputs can induce LTP or LTD depend-
regions during tonic REM sleep compared with activity
ing on whether it arrives at the peak or trough, respectively,
during wake exploration (834), suggesting altogether di-
of ongoing theta oscillations (561, 892). Likewise, LTP and
LTD, respectively, can result from slight differences in the minished coordinated information flow between hippocam-
phase of theta oscillations between pre- and postsynaptic pal input and output regions and between the hippocampus
neurons (514). Theta activity also modulates the amplitude and neocortex during tonic REM sleep. At the same time,
of high-frequency gamma oscillations (!40 Hz), i.e., a local information processing might be enhanced, as evi-
rhythm likewise thought to favor neuronal encoding and denced by increased theta and gamma synchrony between
spike time-dependent plastic processes (173, 242, 420), dentate and CA3 to levels even higher than during wakeful-
with the theta-phase coupling of such faster oscillations ness, as well as by the general high levels of theta and fast
differing between phasic and tonic REM periods (133). EEG frequencies in hippocampus and neocortex during
Phase-locking of gamma band activity during coherent REM sleep. It has been argued that such high levels of local
theta activity in prefrontal-hippocampal circuitry is information processing in the absence of coordinate long-
thought to underlie the successful (explicit) encoding of range communication, together with the specific neuro-
hippocampus-dependent memory information during chemical milieu during REM sleep, represent conditions
wakefulness (72, 74, 220, 423). that favor processes of synaptic consolidation (293). In fact,
a recent study showed that brain activation during interven-
Evidence for an involvement of theta activity during REM ing REM sleep periods might contribute to an overall down-
sleep in memory consolidation is overall meager. Rats scaling of neuronal firing rates observed across sleep (485).
showed increased theta activity during REM sleep after While discharges in hippocampal CA1 neurons increased
training on an avoidance task (397). However, after fear during single non-REM episodes, firing rates decreased
conditioning, mice exhibited reduced REM sleep theta from pre-REM non-REM periods to post-REM non-REM
(529). Interindividual variability in fear learning across periods, with the decrease being correlated with theta
sleep was related to bidirectional changes in theta coherence power during intervening REM sleep. Interestingly, the gen-
between the amygdala, medial frontal cortex, and the hip- eral decrease in firing rate during REM sleep was accompa-
pocampus during REM sleep (941). In the two studies re- nied by an increase in firing synchrony and a decrease in the
porting signs of firing pattern reactivation in hippocampal variability of cell firing specifically during ripple events.
neuron assemblies during REM sleep after performance on Specifically, across non-REM-REM-non-REM triplets, the

712 Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 SLEEP AND MEMORY

discharge rate of hippocampal neurons decreased between (see sect. IB). Sleep might enhance memory by directly fa-
ripple events and increased during ripple events, and this voring late LTP and LTD in synaptic networks that were
increase in synchrony was likewise correlated with theta potentiated during the preceding wake phase. Alternatively,
power during intervening REM sleep. These data suggest this form of synaptic consolidation may occur as part of a
that REM sleep theta is not only involved in an unspecific system consolidation process in which newly encoded mem-
synaptic downscaling but also in reorganizing and shaping ories are reactivated and redistributed to other networks
hippocampal memory representations (118). during sleep where they subsequently undergo synaptic
consolidation (see sect. IIF). Considering that reactivation
and redistribution of memory representations constitute a
V. NEUROCHEMICAL SIGNALING AND
central mechanism of memory consolidation during sleep,
MEMORY CONSOLIDATION
then the basic question arises whether these reactivations
DURING SLEEP
during sleep can newly induce LTP and LTD or merely serve
to support the maintenance of LTP and LTD induced dur-
Sleep and sleep stages are characterized by a specific neuro-
ing prior waking. These issues are presently far from being
chemical milieu of neurotransmitters and hormones (see
clear. Nevertheless, investigations of sleep-associated
FIGURE 1C), some of which contribute to memory consoli-
changes in the molecular signals that mediate the induction
dation by favoring processes of synaptic consolidation (i.e.,
and maintenance of LTP and LTD have provided some
synaptic LTP or synaptic LTD and depotentiation) or pro-
important clues about possible contributions of sleep to
cesses of system consolidation. Neurochemical studies of
persisting plastic synaptic changes. As discussed in section
sleep-associated plasticity have taken two distinct ap-
IVB, several studies show an increase in extracellular glu-
proaches: 1) they have examined signals known to be essen-
tamate concentrations as well as molecular markers of LTP
tial for synaptic plasticity, such as glutamatergic synaptic
across wake periods and a decrease after sleep periods, par-
transmission and the cascade of intracellular signaling me-
ticularly after SWS (204, 205, 207, 257, 457, 607, 940,
diating LTP, LTD, as well as changes in synaptic morphol-
1264), suggesting that the induction of new LTP is more
ogy thought to underlie long-term memory; or 2) they have
likely during wakefulness than sleep. While experimental
examined signals known to be essentially involved in the
induction of LTP and LTD is feasible during wakefulness
regulation of sleep.
and REM sleep, induction of long-lasting changes was less
likely (although not impossible) during SWS (130). Sleep
A. LTP/LTD During Sleep might, nonetheless, be critically involved in processes sup-
porting the long-term maintenance of synaptic changes,
Synaptic LTP and LTD are considered basic neurophysio- particularly in those involving protein synthesis. More im-
logical mechanisms underlying the formation of memory. portantly, most of these studies assessed expression of the
LTP and LTD have been mainly studied in excitatory glu- relevant signals on a global scale in large regions of the
tamatergic synapses but occur also in inhibitory GABAergic brain and cortex, and the extent of encoding of informa-
synapses. In glutamatergic synapses, LTP is induced via tion, i.e., learning during wakefulness before sleep, was not
activation of postsynaptic NMDA receptors and subse- systematically varied (989). Hence, this approach leaves
quent calcium (Ca2%) influx, which leads to activation of open the basic question of how sleep specifically affects
calcium-sensitive kinases like Ca2%/calmodulin kinase II newly formed memory traces and underlying synaptic plas-
(CaMKII) and protein kinase C (PKC) that in turn activate ticity in discrete neuronal circuits.
transcription factors and immediate early genes that can
eventually lead to an altered protein synthesis and resculp- 1. Glutamate and intracellular PKA signaling
turing of synapses (4, 983). An early and late phase of LTP
and LTD is discriminated with only late LTP (#3 h) requir- Blocking of ionotropic glutamate receptors during postle-
ing new protein synthesis. Maintenance of LTP, i.e., the arning sleep was used as a pharmacological tool to study
development of late LTP from early LTP, is assumed to whether sleep-dependent memory consolidation involves
require a “tagging” of synapses due to associative het- the reactivation of glutamatergic synapses. Consistent with
erosynaptic stimulation within a certain time window (!30 this hypothesis, in humans, blocking of NMDA or AMPA
min) following LTP induction (419). Heterosynaptic tag- receptors by infusion of ketamine or caroverine during
ging can involve noradrenergic and dopaminergic transmis- post-learning retention sleep completely abolished over-
sion, but also the activation of mineralocorticoid receptors. night gains on a visual texture discrimination task (435),
The expression of BDNF and PKM-# are further important presumably mediated by local synaptic plastic changes in
signals not only supporting induction of LTP but in partic- the visual cortex (1049, 1051). Surprisingly, ketamine or
ular for mediating the emergence of late LTP (746, 850, caroverine administration during retention sleep did not
1020). impair consolidation of hippocampus-dependent declara-
tive memories (word pairs) (unpublished observation).
It is likely that memory consolidation during sleep involves However, in these studies consolidation of word pair mem-
both synaptic and system consolidation processes (1305) ories during sleep distinctly benefited from administration

Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org 713

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 BJÖRN RASCH AND JAN BORN

of D-cycloserine, a partial agonist of the NMDA receptor to REM sleep deprivation have been observed after learning
that acts at the glycine binding site to enhance Ca2% influx. of hippocampus-dependent tasks (Morris water maze with
These data suggest that the consolidation of hippocampus- hidden platform, radial maze) (1115), and blockade of
dependent memory involves the reactivation of glutamater- NMDA receptors shortly after these REM sleep windows
gic synapses, although the exact mechanisms of glutamater- disrupted sleep-dependent consolidation of these tasks
gic reactivation might differ between memories with pri- (1094, 1099) (see sect. IIB).
mary hippocampal and primary neocortical represen-
tations. 2. GABA

In animals, sleep-dependent plastic changes were also re-

GABAergic inhibition via interneurons is embedded in al-
vealed to critically depend on glutamatergic and LTP-re- most all central nervous networks and contributes essen-
lated mechanisms (411, 1066). In the developing cortex of tially to both the induction of sleep via brain stem and
cats, ocular dominance plasticity induced by monocular hypothalamic mechanisms, and the local control of LTP
deprivation of one eye requires sleep (77, 412, 413). These and LTD at glutamatergic synapses. GABAergic agonists,
sleep-dependent changes are prevented by blocking NMDA like benzodiazepines that are also used clinically to improve
receptors or cAMP-dependent protein kinase (PKA) during sleep, were consistently revealed to impair LTP as well as
sleep after monocular experience (48) (see sect. VIIA3). LTP-related plasticity in glutamatergic synapses via activa-
Blockade of NMDA and PKA signaling was associated with tion of the ionotropic GABAA receptor (e.g., Refs. 225, 277,
reduced activation of the kinases CaMKII and ERK as well 549, 575, 633, 1008, 1270). In addition, GABAergic syn-
as phosporylation of GluR1 at Ser831, i.e., processes that apses themselves undergo plastic changes that predominate
are critical to the insertion of AMPA receptors into the during early development, but some of which can be ob-
postsynaptic membrane and the strengthening and mainte- served in the mature brain as well (683). Yet, it is presently
nance of LTP. In mice, treatment with the selective phos- unclear how these processes contribute to hippocampal
phodiesterases (PDE) 4 inhibitor rolipram immediately and memory formation.
2.5 h after single-trial contextual fear conditioning pre-
vented the impairing effect of sleep deprivation on retention In humans and rats, systemic administration of agonistic
of the freezing response to the conditioned context (1246). modulators of the GABAA receptor, including barbiturates,
These observations corroborate the concept that impaired benzodiazepines, and zolipedem, reduces sleep latency and
memory retention following sleep deprivation results from consistently increases both sleep in non-REM stage 2 and
the disruption of the cAMP-PKA-CREB pathway that sup- sleep spindles, but typically simultaneously reduces SWS
ports the formation of late LTP. Impaired signaling in this and SWA (684). This pattern diverges from the concordant
pathway originates from increases in PDE activity during increase in both spindles and SWA typically associated with
sleep deprivation, with rolipram counteracting this process memory processing during sleep following intense learning.
(477, 498 –500, 546, 1246). Based on this research, Abel Interestingly, nonspecifically enhancing availability of
and colleagues (546) suggested that cAMP-PKA dependent GABA in the synaptic cleft by administration of the GABA
consolidation in the hippocampus occurs as a “sort of mo- reuptake inhibitor tiagabin induced profound increases in
lecular replay” of the same molecular events recruited dur- SWS and SWA paralleled by marked declines in light sleep
ing prior encoding of information during wakefulness. This and REM sleep in both rats and humans, leaving spindle
notion fits well to a model proposed earlier by Ribeiro and activity unaffected (686, 793, 1299, 1301). Selective
Nicolelis (988), suggesting that memory encoding during GABAA or GABAB receptor agonists like gaboxadol, mus-
waking involves calcium-dependent genetic processes cimol, or $-hydroxybutyrate produced similarly robust in-
which trigger plasticity-related gene regulation during creases in SWS and SWA, while effects on spindle activity
subsequent sleep. Similar ideas have been formulated by differed between drugs (304, 684, 685, 702, 1228, 1297,
Benington and Frank (77) as well Datta and colleagues 1298, 1300). At the network level, GABAA agonistic drugs
(261). The molecular replay may mainly occur during REM like diazepam in slice preparations suppressed the genera-
sleep enabled through coactivation of muscarinic receptors tion of SW-Rs that are known to accompany the neuronal
in the presence of high global cholinergic and low seroto- reactivation of newly encoded memory representations dur-
nergic activity, although it may similarly occur in local cir- ing SWS. The effect was dose-dependent and not observed
cuitry during non-REM sleep (475, 546). following the GABAA-receptor modulator zolpidem (655).
Conversely, the GABAA agonist muscimol and also zolpi-
Also, cAMP-PKA-dependent memory formation for con- dem impaired sleep-dependent plastic changes in the visual
textual fear conditioning is selectively sensitive to the inhi- cortex in a developmental model of synaptic plasticity in
bition of PKA and protein synthesis when this inhibition cats [i.e., ocular dominance plasticity (ODP); see sect.
occurs within specific time windows after training, i.e., im- VIIA3 for details] (48, 411, 1059), suggesting that GABA
mediately or 4 h after training (128). Correspondingly, spe- plays a critical role for plastic processes occurring during
cific time windows of increased REM sleep and sensitivity sleep.

714 Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 SLEEP AND MEMORY

GABAergic effects on sleep-dependent consolidation of mutually inhibit each other to enable transitions into the
memory in humans have not been thoroughly investigated respective brain states depending on a switchlike mecha-
thus far, which is surprising given that GABA-agonistic sub- nism (146, 262, 1032). The wake-promoting network in-
stances like benzodiazepines are widely used for the treat- cludes mainly cholinergic neurons of peduncolopontine and
ment of sleep disorders. Interestingly, there is rather consis- laterodorsal tegemental nuclei (PPT, LDT), the noradrener-
tent evidence that benzodiazepines administered after the gic locus coeruleus (LC), and the serotonergic dorsal and
encoding of information can enhance memory for this in- median raphe nucleus, as well as histaminergic neurons in
formation in waking subjects, possibly by preventing the the hypothalamic tuberomammillary nucleus (TMN),
encoding of new interfering information after learning which have widespread projections to the lateral hypothal-
(247, 888, 1313, 1332). However, the few studies concen- amus, basal forebrain, and cerebral cortex. Activity in this
trating on memory consolidation during sleep point to- network is reinforced by orexin A and B (hypocretin 1 and
wards opposite effects. Early experiments in healthy volun-
2) producing neurons in the posterior lateral hypothalamus
teers suggested an impairing influence of the benzodiaz-
(adjacent to the TMN). The sleep-promoting network in-
epine trialzolam and the non-benzodiazepine zopiclone
cludes mainly the ventrolateral (VLPO) and median
administered after learning before sleep on the retention of
(MnPO) preoptic nuclei of the hypothalamus, which con-
words (1040, 1082), whereas one study did not find an
tain neurons releasing GABA and galanin to inhibit the
impairing effect of the hypnotics zolpidem and triazolam on
sleep-dependent memory consolidation of words and non- wake-promoting network at all levels.
words (813). Administration of the GABAB agonists so-
dium oxybate or baclofen before a nap increased SWS, The sleep-promoting network is stimulated by neurons
without affecting sleep-dependent declarative memory con- sensing astrocytic adenosine that accumulates extracellu-
solidation of word pairs and face-location associations larly as a rundown product of cellular metabolism at least in
(1255). In rats, administration of the GABAA agonists eszo- some parts of the brain (78, 506, 942, 1160). Signaling of
piclone and zolipem after learning before sleep impaired adenosine via A1 receptors, which are diffusely distributed
contextual memory tested 24 h later (577). Two more re- in the brain, may directly inhibit neurons of the wake-pro-
cent studies in humans showed similarly impairing effects of moting arousal system (e.g., Refs. 741, 867, 884). A2a re-
triazolam and zolpidem on sleep-dependent consolidation ceptors located close to the VLPO mediate a direct sleep-
of finger sequence tapping skills (835, 836). This is remark- inducing effect that is counteracted by A2a-receptor block-
able, as the impairment in motor skill consolidation was ers like caffeine (578, 1035).
paralleled by robust increases in the amount of non-REM
sleep stage 2 and spindle activity as well as REM sleep in Transitions between non-REM and REM sleep are medi-
these studies. Sleep-dependent consolidation of motor skill ated by a balanced interaction in brainstem pontine net-
memories was also impaired following administration of works between GABAergic neurons in the sublaterodorsal
the GABA reuptake inhibitor tiagabin, and this impairment region (precoeruleus) that fire during REM sleep (REM-on)
was paralleled by a substantial increase in SWS and SWA and GABAergic neurons in the periaqueductal gray matter
(370). No effects on declarative memory consolidation and the adjacent lateral pontine tegmentum (vlPAG, LPT)
(word pairs) were found in this study. The negative findings that fire during Non-REM sleep to inhibit the REM-on
after benzodiazepine or tiagabin administration during neurons. This core REM switch is in turn modulated by
postlearning sleep suggest that EEG phenomena-like spin- noradrenergic LC neurons and serotonergic neurons of the
dles and SWA do not represent mechanisms that per se are dorsal raphe nucleus that act on both sides of the switch to
sufficient to support sleep-dependent memory consolida- inhibit REM sleep and cholinergic neurons of the PPT and
tion. Yet it is not clear whether pharmacologically induced
LDT promoting REM sleep. Via the same core REM switch,
spindles and SWA are functionally equivalent to their en-
hypothalamic orexin neurons inhibit, whereas VLPO neu-
dogenous counterparts.
rons promote REM sleep (1032).

B. Neurotransmitters Involved As a result of this regulation of non-REM and REM sleep,

in Sleep Regulation widespread changes in activity occur also for the wake pro-
moting neuromodulators, mainly activity of acetylcholine,
Most neurotransmitters and neuromodulators involved in norepinephrine, and serotonin. Cholinergic activity reaches
sleep regulation are also involved in processes of memory a minimum during SWS, whereas during REM sleep it
and plasticity, speaking in favor of a high degree of overlap reaches levels well comparable or even higher compared
and functional interaction between these processes. The with cholinergic activity during wakefulness. Noradrener-
regulation of sleep and wakefulness is based on a balanced gic and serotonergic activity reaches a minimum during
interaction between wake-promoting networks centered in REM sleep, and is at an intermediate level during SWS. At
the upper brain stem and lateral hypothalamus and sleep- the neurohormonal level, early periods of SWS are associ-
promoting networks in the anterior hypothalamus, which ated with a strong activation of somatotropic activity,

Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org 715

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 BJÖRN RASCH AND JAN BORN

whereas late periods of REM sleep are characterized by an ies specifically targeting the role of neurochemical condi-
increased release of corticosteroids from the adrenals. tions during SWS for memory processing have so far con-
centrated on astrocytic adenosine signaling, the minimum
1. Effects of wake- versus sleep-promoting signals levels of cholinergic activity and intermediate-level, pulsa-
on LTP and LTD tile activity of noradrenergic systems during SWS, as well as
on SWS-related changes in hormonal systems.
Wake-promoting signals have been mainly found to sup-
port LTP induction and maintenance. The wake-promoting Regarding astrocytic adenosine, Halassa and co-workers
neuropeptide orexin robustly facilitated LTP induction in (393, 506) showed that genetic inhibition of gliatransmis-
the ventral tegmental area (VTA) and in the hippocampus, sion in mice attenuates the accumulation of sleep pressure
directly and via LC noradrenergic release, respectively (115, as indicated by decreased sleep time, decreased duration of
1296). Supporting effects on the induction and mainte- non-REM sleep bouts, and decreased SWA activity in re-
nance of LTP were also consistently revealed for histamine sponse to sleep deprivation. Despite decreased SWA, these
(677, 750), norepinephrine, as well as acetylcholine (140, mice, unlike wild-type mice, did not exhibit signs of im-
585, 598, 1187, 1309). Indeed, norepinephrine and acetyl- paired recognition memory on an object recognition task
choline are also important for the induction of late LTP (12, when they were sleep deprived after training. Infusion of the
1295) and were often found to synergistically act to en- adenosine A1 receptor antagonist 8-cyclopentyl-1,3- dim-
hance LTP (140, 1309) as well as LTD (642). Serotonergic ethylxanthine (CPT) suppressed sleep only in the wild-type
activity displaying sleep-dependent fluctuations similar to mice and mimicked the transgenic phenotype with regard to
those of noradrenergic activity appears to influence LTP both sleep and memory effects. Parallel effects were dem-
and LTD in a less homogeneous way, with its effects onstrated for the sleep deprivation-induced impairment of
strongly depending on the brain region, receptor subtype hippocampal late-phase LTP. Recently, Schmitt and col-
activated, and the type of stimulation used to induce LTP or leagues (1047) showed that the level of the adenosine A1
LTD (98, 576, 601, 628, 654, 675).
receptor activation increases during normal and prolonged
wakefulness in mice, and that this increase affects synaptic
The major sleep-promoting factor adenosine was also re-
transmission in the hippocampus as well as network activity
vealed to contribute to the regulation of synaptic plasticity.
in the cortex. These effects were prevented by genetic inhi-
Via activation of A1 receptors extracellular adenosine at-
bition of gliatransmission. Taken together, these findings
tenuates hippocampal LTP, whereas A2A receptor activa-
indicate that astrocytic adenosine, beyond its promoting
tion mediates an enhancing effect (233, 816, 972). A2A
effect on sleep, via activation of A1 receptors is also in-
receptor activation is also critically involved in the enhanc-
volved in mediating the impairing effects of sleep depriva-
ing effects of BDNF on hippocampal LTP (402). On the
tion on memory consolidation.
other hand, LTP as well as basal excitatory synaptic trans-
mission in orexin neurons of the lateral hypothalamus can
be diminished by activation of adenosine A1 receptors Suppression of cholinergic activity during SWS alleviates
(1339). It has been proposed that under basal firing condi- tonic inhibition of hippocampal CA3 and CA1 feedback
tions, glia-derived extracellular adenosine mainly activates neurons, thereby it enables spontaneous reactivations of the
A1 receptors leading to a diffuse inhibition of synaptic hippocampal networks and of the memory information en-
transmission, whereas with high-frequency synaptic stimu- coded in these networks, as well as the transfer of the reac-
lation A2A receptors are activated by adenosine which is tivated information to neocortical networks (FIGURE 6A;
formed locally by ecto-nucleotidases from synaptically re- Refs. 520, 522, 782). Consistent with this concept of low
leased ATP and overrides A1 receptor-mediated effects acetylcholine enabling systems consolidation of declarative
(245). LTP induced at hippocampal Schaffer collateral syn- memories, increasing cholinergic tone during a period of
apses can be reversed by low-frequency (1–2 Hz) stimula- SWS-rich sleep by administration of the cholinesterase in-
tion mimicking basal conditions as they could occur during hibitor physostigmine completely blocked the sleep-associ-
SWS, and this depotentiation appears to be essentially me- ated consolidation of word pair memories (FIGURE 6B;
diated through diffuse activation of A1 receptors (576). The Refs. 429, 963). Some subjects showed slight decreases in
depotentiation triggered in this way by direct neocortical SWA after physostigmine, but these were unrelated to the
inputs might represent a mechanism helping to reset synap- impairment in declarative memory consolidation, suggest-
tic transmission in the hippocampus, thus preparing these ing that cholinesterase inhibition primarily affected mem-
networks for further encoding of information (599). ory processing in hippocampal rather than thalamo-cortical
circuitry. Conversely, blocking cholinergic receptors in
2. Neuromodulation associated with sleep and SWS, waking subjects by simultaneous administration of nico-
and memory processing tinic and muscarinic receptor antagonists improved the
consolidation of declarative memories during wakefulness,
The neuromodulatory changes characterizing SWS are in- but concomitantly decreased the ability to encode new in-
deed the same that accompany the induction of sleep. Stud- formation (963). These results strongly support the notion

716 Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 SLEEP AND MEMORY

A
Encoding Consolidation
during wakefulness during slow wave sleep
(high ACh) (low ACh)

Information

Associations cortex Associations cortex

ACh

Entorhinal cortex Entorhinal cortex

ACh

DG CA1 DG CA1

ACh
CA3 CA3
ACh

Hippocampus Hippocampus

B
**
% % %
120 120 100

12%
Recognized numbers

*
Recalled word pairs

Recalled word pairs

110 110 90
**
100 100 80
9%
90 90 70

80 80 60

0 0 0
Sleep (3 h) Wake (10 h) Encoding test
Placebo Placebo Placebo
Physostigmine Scopolamine + Scopolamine +
Mecamylamine Mecamylamine
FIGURE 6. Influence of cholinergic activity on memory consolidation during wakefulness and sleep.
A: concept: during active waking, acetylcholine (ACh) levels are high. Information encoded by neocortical
structures flows through the entorhinal cortex and dentate gyrus (DG) into hippocampal region CA3 (connec-
tions less sensitive to modulation by ACh; thick arrows). Connections suppressed by ACh modulation (dashed
arrows) to region CA1, entorhinal cortex, and association cortex are strong enough to mediate immediate
retrieval, but do not overwhelm the feed-forward connectivity, ensuring efficient encoding. In contrast, during
SWS, ACh levels are low, and memories are reactivated in region CA3 during sharp wave-ripples (SW-Rs).
These waves of activity flow back through region CA1 to entorhinal cortex and neocortex, enabling an efficient
redistribution of memory representation (system consolidation) underlying long term memory storage.
[Adapted from Hasselmo (520), with permission from Elsevier.] B: in accordance with the model, increasing
cholinergic tone in humans by administration of the acetylcholineesterase inhibitor physostigmine during
postlearning SWS impairs consolidation of declarative memory (word pairs) during sleep, compared with
placebo. In contrast, combined blockade of muscarinic and nicotinic cholinergic receptors during a postlearn-
ing wake interval (by administration of scopolamine and mecamylamine) enhanced consolidation of declarative
memory (word pairs) during this wake interval. Simultaneously, the combined receptor blockade impaired new
encoding (of numbers). Values are means " SE: *P ! 0.05; **P ! 0.01. [Data from Gais and Born (429) and
Rasch et al. (963).]

Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org 717

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 BJÖRN RASCH AND JAN BORN

that high acetylcholine levels are critical for successful en- tem and a profound surge in the release of growth hormone
coding, whereas low acetylcholine levels facilitate consoli- releasing hormone (GHRH) and GH from the somatotropic
dation of memories, suggesting that acetylcholine might axis (122, 123, 1271). While activation of the latter system
function as a switch between brain modes of encoding and might support hippocampal memory processing through
consolidation as established during waking and SWS, re- brain-borne GHRH (430, 509), inhibition of the former
spectively (522). Manipulating solely nicotinic or musca- system affects limbic regions mainly via cortisol feedback
rinic receptor activity remained ineffective in these studies that is downregulated during SWS. Low cortisol concentra-
(864, 963). Since the cholinergic neurons mediating the re- tions during SWS-rich sleep benefit declarative memory
current inhibition in hippocampus receive strong inhibitory consolidation by preventing activation of glucocorticoid re-
GABAergic inputs, strengthening this GABAergic activity ceptors that mediate an inhibitory influence on hippocam-
should enhance hippocampal memory reactivation. This pal LTP and output from CA1 (648, 649, 666). Accord-
mechanism could explain why postlearning administration ingly, increasing glucocorticoid levels during sleep by
of benzodiazepines improves memory consolidation in postlearning administration of cortisol or dexamethasone
waking subjects but not during SWS when cholinergic ac- impaired consolidation of memory for word pairs and for
tivity is already minimal (454, 813). the temporal order in stories (932, 933, 1323). In addition,
spontaneously increased nocturnal cortisol levels around
Noradrenergic activity during SWS, arising from the LC midnight were found to be correlated with impaired re-
as the brain’s main source of norepinephrine, seems to be trieval of declarative memories in healthy subjects and pa-
particularly related to the depolarizing up-state of the tients with primary insomnia (58). However, a certain basal
slow oscillations as LC firing has been found to be en- release of cortisol is necessary during SWS to sufficiently
trained to these up-states (353). Moreover, a specific occupy mineralocorticoid receptors, which bind cortisol
sleep-related window of increased LC burst activity with a distinctly higher affinity and support the transition of
!120 min after learning of odor-reward associations has early into late LTP in hippocampus. Accordingly, lowering
been identified in rats (356). In humans, suppressing nor- cortisol levels in humans during early nocturnal SWS below
adrenergic LC output by administration of the alpha2- baseline levels by pharmacological blockade of cortisol syn-
autoreceptor agonist clonidine during a SWS-rich reten- thesis impaired sleep-dependent declarative memory con-
tion period reduced consolidation of odor memories, solidation (1272).
whereas the retention of these memories was enhanced
when availability of epinephrine during sleep was in- Melatonin is a circadian hormone that is released by the
creased (434). In a subsequent study, clonidine infused pineal gland during the night and, in diurnal animals and
during a SWS-rich period of retention sleep appeared to humans, substantially contributes to the entrainment of
impair specifically emotional memory processing (1031). sleep to the night-time phase (212). Melatonin inhibits hip-
Whereas temporal order of emotional stories was better pocampal LTP (223, 877, 1171, 1307) and has also been
remembered compared with neutral stories in the placebo found to impair acquisition of hippocampus-dependent
condition, clonidine blocked this superiority of emo- spatial memories in rats (372) and of an active avoidance
tional memory consolidation during SWS-rich sleep. To- task in diurnal zebrafish (971), although divergent findings
gether, these findings suggest that bursts of noradrenergic were obtained in humans (997). Whereas these observa-
activity during SWS are particularly important for the tions point to an impairing influence of melatonin on the
consolidation of memories that involve both a strong encoding of memory, there is preliminary evidence from
amygdala-mediated emotional component as well as a fMRI experiments in humans that melatonin might enhance
hippocampus-mediated declarative component, as keep- processes of hippocampal memory consolidation during
ing temporal order in the events of an episode represents sleep: parahippocampal activity patterns during retrieval of
a key function of the hippocampal formation (e.g., Refs. word-pair memories were found following a nap showed
285, 722, 761). Phasic burst of noradrenergic LC activity some similarity with those obtained when napping was re-
can enforce plasticity-related immediate early gene activ- placed with administration of melatonin (471).
ity (207, 208), and thereby contribute to LTP mainte-
nance in circuitries that were potentiated during prior 4. Neuro- and hormonal modulation associated with
encoding. In the case of odor and emotional memories, REM sleep, and memory
these representations may partly reside in the basolateral
amygdala as a main target of noradrenergic influences Cholinergic activity during REM sleep is high and compa-
(804). rable with that during waking, which might be particularly
relevant for the consolidation of procedural memories dur-
3. Hormonal modulation associated with sleep and ing this sleep stage. Blocking muscarinic cholinergic recep-
SWS, and memory processing tors during “REM sleep windows” (i.e., time periods after
learning in which REM sleep deprivation effectively im-
SWS is associated with the inhibition of glucocorticoid re- paired consolidation of respective materials, see sect. IIB)
lease from the hypothalamus-pituitary-adrenal (HPA) sys- by administration of scopolamine consistently impaired

718 Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 SLEEP AND MEMORY

memory in a habit learning version of the radial arm maze inhibitors, with decreases in overnight retention of declar-
(720, 721). Infusion of scopolamine into the dorsal stria- ative memories (word lists) or procedural skills (mirror
tum, a structure specifically involved in habit and skill tracing) (466). In healthy humans, administration of the
learning, was particularly effective. In humans, the com- SSRI fluvoxamine and the SNRI reboxetine during a postle-
bined blockade of nicotinic and muscarinic receptors dur- arning period of REM-rich sleep likewise did not lead to
ing REM-rich late sleep impaired off-line consolidation of a any impairment in the sleep-dependent consolidation of
motor skill (finger sequence tapping) without affecting de- mirror tracing or finger sequence tapping skills, although
clarative memories (961). Conversely, increasing the avail- after the SNRI REM sleep was almost completely sup-
ability of acetylcholine during post-training sleep by admin- pressed (962). On the contrary, sleep-dependent gains in
istration of an acetylcholinesterase inhibitor enhanced finger sequence tapping accuracy were even significantly
sleep-related benefits in a skill learning task (570). Antich- greater following the reuptake inhibitors than after pla-
olinergic treatment during wakefulness after learning re- cebo. The greater gains in accuracy after SNRI and SSRI
mained ineffective (961). Taken together, these results iden- administration were additionally correlated with increases
tify high cholinergic tone as an important factor contribut- in non-REM sleep spindle density. These findings challenge
ing to the off-line consolidation of procedural skills during a role of phenotypic REM sleep for procedural memory
REM sleep, in combination with other unknown processes. consolidation. If any substantial contribution of REM sleep
The effects are consistent with a role of acetylcholine in to procedural memory consolidation exists, increasing nor-
synaptic consolidation, e.g., by promoting activity of plas- adrenergic or serotonergic tone can apparently compensate
ticity-related immediate early genes (475, 615, 1182) and for it. Both monoamines support synaptic remodeling via
maintenance of LTP in cortico-striatal networks (742). increasing activity of plasticity related immediate early
genes (205, 208, 475).
There is some evidence that the increase in cholinergic ac-
tivity during REM sleep is accompanied by an increased It has been proposed that the inhibition of noradrenergic
dopaminergic activity originating from ventral tegmental, activity during REM sleep enhances procedural types of
rather than striatal, areas which innervate a network of memory representations in cortico-striatal circuitry by en-
abling spontaneous reactivations in cortical networks that
further regions such as the nucleus accumbens and medial
during wakefulness are under tonic inhibitory control by
prefrontal cortex (724, 757). Although it is not unlikely that
noradrenergic neurons (520). Similarly, it has been sug-
such REM-related increases in dopamine activity modulate
gested that REM sleep allows for a replay of amygdala-
the effects of memory reprocessing during sleep (e.g., Ref.
dependent emotional memories that, in the absence of nor-
73), the involvement of dopaminergic activity in sleep-asso-
adrenergic activation, decrease the emotional tone in these
ciated memory consolidation has not been thoroughly ex-
memories, whereas their information content is maintained
amined thus far. Legault et al. (721) reported that the do-
(1294) (see sect. IIC). Additionally, the decreasing effect on
pamine receptor blocker flupenthixol infused into the dor-
emotional tone could be further enhanced by glucocortico-
sal striatum 0 – 4 h after learning during “REM sleep
ids, the levels of which are distinctly increased during late-
windows” exerted the same impairing effect on habit con-
night REM sleep and which distinctly diminish emotional
solidation as observed after scopolamine. In humans, so far memory consolidation (1272). However, there is presently
only patients (with early Parkinson disease and schizophre- little empirical support for these theories. Thus, overall, the
nia) have been examined in this context with dopaminergic role of noradrenergic and serotonergic inhibition during
drugs that were partly rather unspecific (465, 776). No REM sleep for memory processing remains enigmatic.
consistent effects of the treatments on overnight retention
of procedural or declarative memories were found in these
patients. VI. GENETIC APPROACHES TO
SLEEP-DEPENDENT MEMORY
Whether the very low levels of noradrenergic and seroto- FORMATION
nergic activity during REM sleep, perhaps in a permissive
way, also contribute to memory consolidation, is currently Sleep is genetically controlled. Although environmental fac-
not clear. Selective serotonin or norepinephrine reuptake tors clearly contribute, a large part of the interindividual
inhibitors (SSRI or SNRI) that enhance availability of these differences in sleep architecture is likely due to genetic fac-
monoamines in the synaptic cleft are commonly used for tors. The search for genes involved in sleep regulation has
antidepressant therapy and, thereby, produce a substantial received a great boost in recent years with the discovery of
reduction in REM sleep. However, clinical observations quiescent states in simple organisms like the fruit fly and
revealed no clear memory impairments in patients treated worms that fulfill the criteria for sleep: a reversible and
with these drugs (22). Also, a more systematic clinical study repeatedly occurring period of reduced responsiveness and
of the effects of the SSRI citalopram and the SRNI rebox- relative inactivity which is homeostatically regulated (i.e.,
etine in moderately depressed patients failed to reveal any deprivation of sleep leads to subsequent longer sleep peri-
association of REM sleep diminution after the reuptake ods) (209). In these organisms, molecular methodology can

Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org 719

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 BJÖRN RASCH AND JAN BORN

be applied much more efficiently and at low costs mainly examined gene expression in the cerebellar cortex of rats
due to the reduced number of genes and neurons and the high which were killed after 8-h periods of sleep, wakefulness, or
reproduction rate. However, there are also important caveats, sleep deprivation. Five percent of the examined transcripts
as certain sleep stages (e.g., REM sleep) and oscillatory brain were differentially regulated between sleep versus wakeful-
activity characterizing sleep in mammals and humans cannot ness or sleep deprivation. Sleep was associated with a great
be discriminated in these simple organisms. Also, although number of upregulated gene transcripts, and this number
typically homologs of genes exist between species, they can be was indeed comparable with that during wakefulness, de-
much more differentiated in complex animals. For example, spite the behaviorally “inactive” state of sleep.
one Shaker gene in the fly has 16 homologs in rodents (202).
These limitations underscore the importance of genetic studies Importantly, one category of gene transcripts upregulated
in rats and mice and optimally in humans. during sleep is involved in synaptic plasticity. This includes
mRNA expression levels of calmodulin-dependent protein
kinase IV (CAMK4), a gene that is implicated in synaptic
A. Animal Models
depression and long-term memory consolidation (617), and
the expression of several other genes associated with depo-
Gene expression studies, mainly in flies and mice, show an
tentiation and depression of synaptic strength. In contrast,
upregulation of genes specifically during sleep, including
during wakefulness, expression of genes that are involved in
genes involved in protein synthesis and synaptic depoten-
LTP is upregulated, including genes coding for Arc, c-Fos,
tiation. In addition, several genes related to neural plasticity
NGFI-A, and BDNF (206, 939, 940). Based essentially on
and synaptic potentiation are upregulated during sleep de-
these gene expression patterns, the synaptic homeostasis
pending on prior learning and exposure to novel experi-
ences. Also, genes have been identified that exhibit joint hypothesis has been proposed assuming that wakefulness is
regulatory actions on sleep and memory. Intriguingly, associated with prevailing synaptic potentiation in cortical
short-sleeping genotypes often also show impairments in networks, whereas processes of synaptic depotentiation
learning and memory. and depression predominate during sleep to desaturate the
network (1204) (see sect. IVB). Subsequent microarray
1. Differential gene expression during sleep studies in the mouse and fruit fly largely confirmed that
and wakefulness expression of several genes is differentially regulated during
sleep and waking and that these genes are related to synap-
tic plasticity, response to cellular stress, energy and lipid
Early studies suggesting that gene expression differs be-
metabolism, and membrane trafficking (201, 610, 752,
tween sleep and wakefulness mainly examined overall
changes in mRNA and protein synthesis. First experiments 753, 771, 1357). In particular, the expression of genes in-
showed that an injected radioactive substance was incorpo- volved in macromolecular synthesis increased during sleep,
rated into newly synthesized RNA at a faster rate during supporting a role for sleep in the synthesis of proteins and
sleep than during wakefulness (462, 1257). Similarly, la- lipids (e.g., fats, sterols, vitamins) (753). Conversely, sleep
beled proteins in rats increased after 90 min of sleep com- deprivation strongly attenuates the expression of genes re-
pared with 90 min of sleep deprivation (103). In rats and lated to protein synthesis and thus downregulated transla-
monkeys, the rate of protein synthesis correlated positively tion in the mouse hippocampus, mainly mediated by key
with the amount of non-REM sleep (848, 956). Analysis of regulators of protein synthesis like mammalian target of
several hundreds of proteins by mass spectrometry (309) rapamycin (mTOR) (1247). In the developing visual cortex
revealed that protein levels in the mouse cerebral cortex in cats, expression of LTP-promoting genes (e.g., Arc,
were generally decreased after sleep deprivation compared BDNF) was likewise found to decrease across sleep in the
with sleep. These studies indicate that sleep favors protein visual cortex as in adult rodents (1060) (see sect. VIIA3).
synthesis, thus pointing towards a differential expression of Notably, however, translation of the corresponding pro-
genes during sleep and wakefulness. Indeed, blocking pro- teins at the same time was increased particularly during
tein synthesis during sleep impairs sleep-dependent consol- early sleep, suggesting that the first hours of sleep might be
idation of ocular dominance plasticity (ODP) in the visual a period of accelerated protein synthesis in these networks.
cortex of cats (1060) (see sect. VIIA3), indicating that the
increased protein synthesis during sleep has functional con- Gene expression studies identified Homer1a as a core gene
sequences for plastic processes underlying memory forma- of sleep loss, as its expression was consistently upregulated
tion. in different strains of mice after 6 h of sleep deprivation
(771). Homer1a expressing cells also overexpressed three
More recent studies have directly examined gene expression further genes in response to sleep loss, which all appear to
with molecular techniques like microarrays that indicate be implicated in intracellular calcium homeostasis, and may
changes in transcription of the genome, i.e., the up- or function to protect and recover neurons from glutamate-
downregulation for a great number of gene transcripts (for induced hyperactivity during periods of extended wakeful-
reviews, see Refs. 201, 752). Cirelli and colleagues (204) ness. Even after controlling for possible confounding influ-

720 Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 SLEEP AND MEMORY

ences of increased glucocorticoids due to the sleep depriva- synaptic plasticity-related genes in limbic regions, specifi-
tion procedure, Homer1a (as well as plasticity-related genes cally in the dorsal hippocampus.
including Arc and Fos) was still specifically associated with
sleep loss (833). In the later study, several other genetic Exposure to novel tactile stimuli as well as experimental
pathways that were previously associated with sleep ho- induction of hippocampal LTP during wakefulness pro-
meostasis (e.g., circadian clock genes, see Ref. 415) were no duced an upregulation of Arc and Egr-1 in neocortical and
longer affected by sleep deprivation in this study, indicating hippocampal regions during subsequent REM sleep (986,
that effects of stress and glucocorticoids need to be carefully 987, 989). The reinduction of both immediate early genes
controlled in gene expression studies examining sleep ho- during REM sleep after novelty exposure was only transient
meostasis. in hippocampal regions, and more pronounced and persis-
tent in the neocortex, where it was particularly strong in the
2. Experience-dependent local regulation of genes somatosensory areas most activated by the previous tactile
novelty experience (986, 989). Cortical expression of Arc
was correlated to EEG spindle activity during prior non-
Apart from the global regulation of gene expression during REM sleep, possibly favoring immediate early gene activity
sleep and wakefulness, experience-dependent upregulation during REM via Ca2%-dependent mechanisms (989). Spin-
of plasticity-related genes has also been observed during dles likely promote massive calcium influx into cortical py-
sleep. Using fluorescence in situ hybridization in hippocam- ramidal cells and through activation of CaMKII may set the
pal cells in rats, Marrone and co-workers (781) found that stage for the expression of immediate early genes during
exploration of a new environment increased the number of ensuing REM sleep (283, 1061). Both Arc as well as Egr-1
hippocampal CA1 neurons showing induction of the imme- are known to interact with CaMKII in synaptic remodeling
diate early genes Arc and Homer1a both during exploration during LTP. Ribeiro et al. (989) related the REM sleep-
as well as during rest after exploration. However, others associated increase in immediate early gene expression oc-
failed to reveal increased protein levels of c-Fos or Arc in curring preferentially in cortical over hippocampal regions
cortical motor areas in rats during sleep following explora- to waves of synaptic plasticity that hippocampus-depen-
tion or following training of skilled reaching, a task known dent memories undergo during successive non-REM-REM
to induce LTP in motor areas (510, 584). Still, expression of sleep cycles (see FIGURE 7, A AND B). Reactivation of hip-
the immediate early gene coding for BDNF increased fol- pocampal memory representations during non-REM sleep
lowing exploratory behavior during wakefulness, and the stimulate the redistribution of these memories to neocorti-
amount of exploration during waking (% recording time) cal areas where they are synaptically consolidated during
was correlated with subsequent increases in BDNF mea- ensuing REM sleep (293). Across several waves of plasticity
sured in the frontal and parietal cortex after a short 10-min associated with the non-REM-REM sleep cycles, memories
period of non-REM sleep, as well as with the amount of become stored mainly within neocortical networks while
SWA during this non-REM sleep period (584). they fade out in the hippocampus, the latter coinciding with
the fading of reinduction of immediate early gene activity
Distinct local upregulation of plasticity-related genes and during postencoding REM sleep in hippocampal areas (FIG-
proteins following novelty and learning experiences has URE 7C).
been revealed to be specifically related to REM sleep (261,
1022, 1225). Learning of a two-way active avoidance task 3. Deletion and gene knockout studies
in rats increased phosphorylation of the cAMP response
element-binding protein (CREB), expression of Arc protein In flies, several genes that regulate the amount of sleep have
as well as mRNA expression of Arc, BDNF and early been identified by mutagenesis screenings. Whereas wild-
growth response-1 (Egr-1, also known as Zif268) in the type flies sleep !8 –10 h/day, flies with a loss-of-function
dorsal hippocampus and amygdala during a 3-h postlearn- mutation in the genes Shaker (Sh) or hyperkinetic (Hk) sleep
ing sleep period. Elimination of pontine brain stem cells that only 2– 4 h/day (154, 202, 203). Shaker codes for the alpha
generate P-waves hallmarking REM sleep suppressed reten- subunit of a specific K% channel (alpha-subunit of a tetra-
tion of the avoidance response as well as learning-depen- meric potassium channel that mediates a voltage-activated
dent increases in the expression of phosphorylated CREB, fast inactivating IA current), and hyperkinetic codes for a
Arc protein, and mRNA of Arc, BDNF, and Egr-1 in dorsal regulatory beta-subunit of this channel that interacts with
hippocampus and amygdala (261). Conversely, cholinergic the alpha-subunits encoded by Shaker (202, 1052). Shaker-
stimulation of P-waves by microinjection of carbachol into modulated K% channels have homologs in vertebrates (Kv.1
respective pontine regions increased expression of these and other Shaker-like channels, Kv.2, Kv.3, Kv.4, etc.). In
proteins and mRNAs in the dorsal hippocampus, and was both flies and mammals, they play a major role in mem-
also associated with improved retention of the active avoid- brane repolarization and likely also modulate EEG oscilla-
ance response (263, 794). These findings indicate that con- tory phenomena like spindles (35, 155, 244, 320, 357, 358,
solidation of an active avoidance response during REM 717, 934, 1266). Importantly, both Shaker and hyperki-
sleep critically relies on P-wave-induced upregulation of netic mutants showed marked impairments in learning and

Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org 721

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 BJÖRN RASCH AND JAN BORN

A B
EXP
Pre-Experience Post-Experience
12

Neurons
1.0
HP

Hippocampus 1
12

Neurons
S1

1
12

Neurons
V1
Cortex
0.5
Novel stimulation Sleep cycles 1
0 120 300
Time boundaries of concatenated SWS episodes (minutes)
C
Time

New
Synaptic Old
changes
None

Memory acquisition Memory propagation Memory propagation

during waking during early sleep during late sleep
FIGURE 7. Different time courses of plasticity in the hippocampus and neocortex. A: concept: the hippocam-
pus undergoes a few plasticity waves before fading out. These plasticity waves are probably enough for
memories to remain in the hippocampus for weeks or months. In contrast, the cerebral cortex undergoes
plasticity waves for a much longer period of time, leading to many more cycles of memory reinforcement and
years-old memories. B: in single neuron recordings in rats, long-lasting firing rate increases after novel
spatio-tactile stimulation (EXP) occurred during SWS in primary somatosensory cortex (S1), but not in the
hippocampus (HP) or primary visual cortex (V1). Increased neuronal activity persisted for hours after experi-
ence offset during SWS in S1. Shown are the normalized firing rates during concatenated SWS episodes
spanning an entire representative experiment. Ticks at the bottom indicate SWS episode boundaries. C: model
of memory propagation from hippocampus to neocortex during sleep. Via thalamo-cortical inputs (not shown),
episodic and spatial memories are acquired during waking as new synaptic changes (red) distributed over
hippocampocortical networks of neurons (top panel). The recurrence of cortical plasticity during subsequent
sleep causes the stabilization and propagation of new synaptic changes in the neocortex. Conversely, the fast
decay of sleep-dependent plasticity in the hippocampus generates a net outflow of information, gradually
flushing memories to associated cortical networks over time. [Modified from Ribeiro et al. (989).]

memory, as well as a reduced life span. Similarly, reduced sleep (663, 703, 851, 1073, 1328, 1329), whereas in flies
sleep and life span have been reported for flies with a loss of only the activators Clk and cyc affect sleep phenotypes
the gene Sleepless (SSS) which codes for glycosylphosphati- (531). Per, which is regulated by CREB, plays a key role in
dylinositol-anchored protein with unknown function, with long-term memory formation in Drosophila (1023). Recent
the effects of this mutation possibly in part also mediated by studies in fruit flies by Shaw and colleagues revealed that
Shaker channels (202, 272, 653, 1306, 1337). However, its per, together with other clock genes, also regulates sleep-
influence on learning and memory has not been investigated dependent memory formation (314, 438). These studies
so far. first established that waking experience, i.e., a socially en-
riched environment where the flies were housed in groups of
There is strong evidence for an involvement of clock genes !40, compared with social isolation increased sleep time
in sleep-dependent plasticity. Clock genes regulate the cir- and sleep bout duration (315). Sleep need likewise increased
cadian rhythm and affect sleep (681). In mammals, genetic after a learning task (conditioned suppression of courtship
variations in both clock genes activating circadian rhythm, behavior), and this increase in sleep was critical for success-
like Clock (Clk) and cycle (cyc), and clock genes with re- ful recall of the conditioned behavior 48 h later (314, 438).
pressing functions on circadian rhythm, like period (per) Remarkably, flies mutant in the clock gene per failed to
and Cryptochrome (Cry), have been linked to variations in show this experience-dependent increase in sleep need and

722 Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 SLEEP AND MEMORY

in parallel showed no long-term memory after 48 h. Similar sleep and, if temperature was raised after learning, had an
results were observed in flies with mutations in the rutabaga enhancing effect on long-term memory for a conditioned
gene (involved in cAMP signaling) and blistered (involved olfactory avoidance response. While baseline sleep was
in synaptic LTP and contextual habituation). Rescue of the lowered in the transgenic flies, the heat-induced increase in
genes in clock neurons of the flies reestablished the experi- sleep was correlated with the gain in memory, altogether
ence-dependent increase in sleep as well as long-term mem- suggesting an important role for Fabps in mediating mem-
ory (314). ory consolidation during sleep (451).

Several other genes and transcripts have been identified that Using a similar approach, Donlea et al. (316) introduced a
not only strongly influence sleep quantity, but also are in- temperature-sensitive cation channel in neurons of Dro-
volved in major pathways regulating synaptic plasticity, sophila. Activating the channel by raising temperature
like CREB (477, 532, 916), the extracellular signal-regu- above 31°C resulted in a sleeplike state. Inducing 4 h of
lated kinase/mitogen-activated protein kinase (ERK/ sleep in this way after massed trials on a courtship condi-
MAPK), and the epidermal growth factor (EGF) (e.g., Ref. tioning procedure enhanced retention of the conditioned
401, for a review see Ref. 202). However, direct links be- behavior at a retest 48 h later, compared with flies that did
tween sleep and memory formation have not been demon- not sleep after training. The effect cannot be explained by
strated for most of these signals. An exception is the gene an activation of the temperature-gated neurons per se, be-
bunched, a regulator of the transmembrane receptor Notch cause sleep deprivation during periods of high temperature
which is involved in both sleep homeostasis and learning in did not enhance long-term memory. Of note, temperature-
flies (1062). Overexpression of the Notch ligand Delta as induced sleep also benefited new learning on the next day,
well as the introduction of a Notch gain-of-function allele fitting the idea that memory consolidation during sleep is
reduced sleep rebound and prevented impairments in new concurrently linked to processes preparing the brain for the
learning (aversive phototaxic suppression) induced by prior future encoding of new information (828).
sleep deprivation. Interestingly, the effects on sleep and
memory of Notch signaling were mainly observed in glia These studies exemplify the rapid development of new ge-
cells, pointing to a critical role of neuron-glia interactions netic tools that can be used as the “remote control” of sleep
for regulating sleep-dependent learning benefits (1338). (316) and, in this way, to study the function of sleep for
Similar effects of Notch signaling were observed in worms memory consolidation. First attempts conducted to control
(Caenorhabditis elegans) (1085). sleep optogenetically in mice have revealed that fragment-
ing postlearning sleep by activating hypothalamic hypocre-
Knockout procedures have also been used in numerous tin/orexin neurons impaired consolidation in a novel object
studies as a genetic tool to examine the impact of neu- recognition task (9, 1009). Future developments of these
rotransmitter systems on sleep regulation (reviewed in Ref. transgenic techniques will enable the reliable activation of
202; see also sect. VB) and oscillatory EEG activity during specific neuronal memory traces, thus providing promising
sleep (reviewed in Ref. 37; see also sect. IV). For example, a tools to a more fine-grained analysis of the fate of a memory
genetic region on chromosome 13 (containing several trace during sleep (739).
genes) was identified by quantitative trait loci analysis
(QTL) that explained almost 50% of the variance in the
rebound in non-REM sleep EEG delta activity after sleep B. Human Studies
deprivation in rodents (414). However, for none of these
genes have any links to sleep-dependent memory formation 1. Stability and heritability of sleep
been examined in a more systematic manner.
Human sleep and particularly oscillatory EEG activity dur-
3. Heat-induced sleep in transgenic animals ing sleep is highly heritable and very stable within individ-
uals (689). In addition, large interindividual differences in
Gerstner and co-workers (450 – 452) engineered heat-re- the sleep EEG exist, which are more than 10 times higher
sponsive transgenic flies to examine the role of fatty-acid than those, observed within subjects across multiple nights
binding proteins (Fabps) for sleep-associated memory con- of sleep. Already in 1937, Geyer (453) reported a higher
solidation. Fabps bind small lipids and act as transporters in similarity in sleep profiles for monozygotic compared with
various cells and tissues. Of the nine mammalian genes, dizygotic twins. Systematic twin studies using large samples
mRNA coding for Fabp7 is expressed in the brain, and in indicate that genetic differences account for 30 – 45% of the
rodents its expression follows a circadian rhythm. In flies, variance in subjective sleep quality and sleep disturbances
mRNA expression coding for the Fabp7 homolog dFabp (186, 526, 889) and for !50% in physiological measures of
also shows a circadian rhythm, with elevated levels during sleep stages (137, 737, 738). Highest heritability values are
the night. Inducing mRNA expression of genes coding for observed for REM density (h2 & 64 –95%) as well as EEG
Fabp7 or dFabp in temperature-sensitive transgenic flies by spectral power particularly in the alpha (8 –11 Hz) and
elevating ambient temperature from 20 to 30°C increased spindle (11–15 Hz) frequency bands (h2 & 76 –96%) (29,

Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org 723

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 BJÖRN RASCH AND JAN BORN

266, 444, 736, 1227). Similar heritability estimates (h2 & were more vulnerable to sleep deprivation in the early
40 – 60%) have been reported for sleep amount and sleep morning hours (e.g., Refs. 1256 and 483, but see Ref. 467)
organization in rodents (1165, 1166, 1226), with even and exhibited a stronger decrease in brain activity assessed
higher estimates of genetic control for oscillatory electrical by fMRI in prefrontal areas from morning to evening
activity during sleep (37, 414, 416, 417). (1244). Altogether, these findings suggest that circadian ge-
notypes contribute to differences in the allocation of cogni-
Genetic factors also contribute to sleep differences between tive resources, thus possibly also affecting memory process-
ethnicities (e.g., Ref. 1019, for a meta-analysis) and gender ing during sleep.
(e.g., Ref. 976). Women sleep longer and still have a higher
SWS percentage than men, whereas men show increased With regard to noncircadian genes, Landolt and co-workers
non-REM sleep stages 1 and 2 (424, 587, 652, 952). In a identified a genetic difference (G-#A transition at codon
study in 2,600 participants aged 37 to 92, gender explained 22) in the gene encoding the adenosine metabolizing en-
14.6% of the variance in SWS percentage and 10.9% of zyme (adenosine deaminase, ADA) that was significantly
stage 2 sleep percentage (976). associated with SWS duration and SWA, in particular in the
$2 Hz range (e.g., Refs. 54, 798, 982; but see Ref. 797).
Additionally to the strong heritability, the human sleep A-allele carriers who exhibited more SWS also showed
EEG is remarkably stable within individuals across multiple poorer performance in focused attention (as indicated by
nights (151, 1176, 1177, 1201). Stability as indicated by the d2 test), but not in several other tests of short-term
intraclass coefficients is highest for SWS (73%), but also memory and executive functions (verbal, figural memory,
significant for other parameters like the amount of REM digit span, Stroop test. etc.) (see Ref. 688 for a review).
sleep (48%), stage 2 sleep (56%), and total sleep duration
(46%). Power in the delta frequency band (0.75– 4.5 Hz) Differences in the widely researched functional Val158Met
shows particularly high stability (78 – 89%). Interindi- polymorphism in the gene coding for the catechol-O-meth-
vidual differences in delta power were on average !10 lytransferase (COMT) were associated with more global
times greater than the rebound in delta activity observed alterations in alpha peak frequency as well as in spectral
following sleep deprivation (818, 1213). The interindi- power in the 11–13 Hz band, occurring during REM, non-
vidual differences in the sleep EEG are remarkably robust REM sleep, and also during wakefulness (104, 105). Inves-
also against sleep disturbances, first night effects, prior sleep tigation of the Val66Met polymorphism, i.e., a valine to
deprivation, and the administration of sleep-promoting methionine amino acid substitution at codon 66 of the
agents (881, 1213). BDNF gene, revealed that Met-allele carriers (compared
with Val/Val homozygotes) exhibit less SWS and reduced
2. Genetics, sleep and cognitive function delta and theta activity during non-REM sleep in baseline
nights as well as in response to sleep deprivation (54).
Specific genetic markers of sleep have been identified mainly BDNF is particularly expressed in the prefrontal cortex and
in studies of different populations with disordered sleep, the hippocampus in humans (927), and the Val66Met
such as fatal familial insomnia, narcolepsy, restless legs syn- BDNF polymorphism has been repeatedly associated with a
drome, and circadian rhythm disorders (266, 299, 638, range of cognitive functions including memory (e.g., see
954, 1058, 1167). However, the relationships of these ge- Ref. 614, for a meta-analysis). Unfortunately, none of these
netic markers to sleep-dependent memory processes are en- studies included direct tests of sleep-dependent memory for-
tirely unknown so far, except one observation indicating mation.
that a single nucleotide polymorphism (SNP) in the prion
protein gene (PRNP) implicated in fatal familial insomnia in 3. Genetic studies on sleep-dependent
human and sleep regulation in mice (579) affects learning memory consolidation
performance in healthy participants (886).
Studies mainly on flies indicated that similar genes are in-
Studies in healthy humans revealed clear associations be- volved in sleep regulation and memory processing, and that
tween sleep and genes involved in regulating circadian memory and plasticity-related genes are expressed differen-
rhythm, mainly the clock gene PER3 that can occur in a tially during sleep and wakefulness. No such studies are
different number of repetitions, so-called “variable number available in humans. In addition, no genome-wide associa-
of tandem repeats” (VNTR) polymorphisms. Homozygosis tion study in humans with regard to sleep and memory
for the 5-repeat allele (PER35/5) has been repeatedly associ- parameters has been conducted thus far, although such
ated with morning preferences (morning types) compared studies would be highly informative. In fact, presently, di-
with homozygotes for the 4-repeat allele (PER34/4) and rect evidence for a genetic contribution to the sleep-depen-
heterozygotes (PER34/5) (41, 61, 349, 609, 874, 917), as dent formation of memory is entirely lacking.
well as with increased SWS percentage and SWA during
non-REM sleep and increased theta power during REM There is a strong genetic influence on sleep parameters (dis-
sleep and wakefulness (1256). In addition, PER35/5 carriers cussed above) and also on memory formation. Twin studies

724 Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 SLEEP AND MEMORY

have estimated that 50% of the variance in learning perfor- individual variance and intraindividual stability of these
mance, typically measured after short retention periods of sleep EEG parameters (discussed above). Several studies
5–30 min, is due to genetic factors (127, 885, 1164, 1262). reported correlations between SWS or SWA and the over-
Against this background, as both sleep and memory mea- night retention of declarative memory (55, 58, 464, 1319)
sures strongly differ among individuals, a putative genetic as well as procedural motor skills (581, 582). However,
determination of sleep-dependent memory consolidation because in these studies the sleep EEG was assessed after the
should basically express itself in significant correlations be- learning phase in highly homogenous subject samples,
tween the respective sleep and memory parameters on the learning-induced changes in EEG activity could not be
interindividual level. Hence, the demonstration that sleep clearly dissociated from trait-dependent variance.
and memory processes share significant portions of their
interindividual variance represents a first and most im- Historically, memory function has been much more often
portant step towards the demonstration of any genetic associated with parameters of REM sleep than non-REM
contribution to sleep-associated memory formation, al- sleep. Evolutionary increases in encephalization are posi-
though it does not prove such contribution because the tively associated with sleep time allocated to REM sleep,
associations can basically result from parallel, but inde- even when the effects of phylogenetic similarity between
pendent, influences of the putative genetic factor on the species are controlled (e.g., Ref. 725, but see Refs. 1077,
sleep and memory processes of interest. Also, it cannot be 1078). Analyses of sleep in seven inbred strains of mice
concluded from interindividual sleep-memory associa- revealed a high correlation between the relative and abso-
tions whether a putative genetic factor primarily affects lute time in REM sleep and the performance level avoidance
sleep to change memory processing or, conversely, affects conditioning or maze learning across strains (880). In hu-
memory processing to change sleep, although clues can mans, mentally retarded children generally exhibit less
be derived from temporal relationships. For example, REM sleep, longer REM sleep latencies, and less REM den-
Dionne et al. (310) showed in a longitudinal twin study sity relative to normal controls (182–184, 211, 368, 369,
that longer and more consolidated sleep at the age of 6 425, 922–924, 1045). However, correlations between
mo predicted better language development at 18 and 30 REM sleep and IQ measures in healthy children and adults
mo. As the type of sleep was highly heritable, the findings revealed mixed results (126, 153, 922, 923, 1183). In some
suggest a sleep-mediated genetic influence on later lan- studies, gifted children tended to sleep longer and had more
guage proficiency. stage 2 sleep, compared with controls (153, 1183). In addi-
tion to baseline amounts of REM sleep, the magnitude of
With regard to non-REM sleep, spindles have been consis- learning-induced increases in REM sleep parameters have
tently revealed to be associated with memory-related pa- been linked to capabilities of learning in animals (1110) and
rameters and thus represent a most promising candidate humans (1114).
mediating a genetic link between sleep and memory. In
children and adults, spindle number and density strongly The studies discussed so far examined associations between
correlate with IQ scores (106, 395, 439, 847, 1038, 1039) sleep and memory in rather small samples (12–30 partici-
as well as overnight retention of memories (213, 1057, pants). Together with the testing of multiple sleep parame-
1230). Abnormally large spindles or the absence of spindles ters (like amounts of SWS, REM sleep, spindle numbers,
were observed in mentally retarded and dyslexic children density, etc.) and with multiple measures of memory and
(93, 148, 400, 455, 456, 1070, 1071). Contrasting with the learning capabilities, this can lead to an overestimation of
findings in healthy humans, elevated baseline spindles pre- the strength of associations due to multiple comparisons.
dicted poor performance on a two-way shuttle box avoid- To overcome such limitations, in recent study, we assessed
ance task in rats (397). However, learning-dependent in- sleep (at home by portable polysomnography) together
creases in sleep spindles only occurred in rats that success- with measures of pictorial and procedural motor learning
fully learned the task, and these increases were positively and overnight memory retention in a large sample of 855
correlated with post-sleep improvements in performance. young healthy participants (Ackermann, Pappassotiropou-
Based on these findings, Fogel and Smith (398) proposed los, De Quervain, and Rasch, unpublished results). In ac-
a curvilinear relationship between sleep spindles and ca- cordance with previous studies, the number of fast spindles
pabilities of learning, with high spindle numbers reflect- (13–15 Hz) showed a positive (but rather small) correlation
ing either highly efficient or pathological memory pro- with short-term recall (tested after 10 min) of pictures.
cessing in thalamocortical systems. However, this correlation was entirely explained by gender
differences, as women remembered more pictures and ex-
Although the importance of SWA and slow oscillation for hibited also an increased number of spindles. Short-term
memory processing is well established (see sect. IVA), inter- recall of pictures was also correlated with the percentage of
individual correlations between these EEG oscillations and REM sleep (r & 0.09), with this correlation surviving con-
learning capabilities and intelligence have been less fre- trol for gender. Also, with regard to overnight retention of
quently reported. This is remarkable given the great inter- pictures, the only significant correlation was revealed for

Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org 725

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 BJÖRN RASCH AND JAN BORN

time in REM sleep: increased REM sleep was associated A. Early Development: Animal Models
with diminished overnight retention of pictures (r &
'0.13), independently of whether these pictures were emo- Filial imprinting in domestic chicks, song learning in ze-
tional or neutral. No such associations were observed for bra finches, and ocular dominance plasticity in the visual
SWS-related measures, spindles, or sleep length. Overall, cortex of cats have been studied as models of develop-
these data indicate a surprisingly weak association between mental brain plasticity and memory formation in animals
interindividual differences in sleep and memory parame- in conjunction with sleep. Recent research has indicated
ters. Specifically, they suggest that the genetic determina- that the formation of these three types of memory indeed
tion of sleep, particularly of SWS-related measures and critically depends on sleep.
spindles, does not substantially contribute to sleep-depen-
dent memory formation processes, as these parameters do 1. Filial imprinting
not appear to be associated with each other on the interin-
dividual level. It is important to note, however, that the lack Imprinting is a very strong and early form of social recog-
of association between sleep and memory parameters nition memory, often studied in the domestic chick. Under
among subjects does not at all rule out a basic role of sleep natural conditions, during imprinting the chick learns the
in memory formation. For example, one individual can dis- characteristics of its mother to selectively follow her rather
play higher SWA than others due to neuroanatomical and than any other adult. In the laboratory, the chick is repeat-
related genetic factors in the absence of any differences in edly exposed to a moving object, and so acquires a prefer-
memory among these individuals, yet increasing SWA in ence for this object over an alternative, novel stimulus, with
this same individual might still lead to a distinct improve- this preference used as measure of the imprinting memory
ment in memory consolidation during sleep. (107, 567). Imprinting over repeated training sessions leads
to a gradual increase in the number of neurons in the inter-
VII. DEVELOPMENTAL ASPECTS OF mediate and medial mesopallium (IMM, also termed inter-
SLEEP-RELATED BRAIN PLASTICITY mediate and medial hyperstriatum) that selectively respond
to the imprinting stimulus which is correlated with the
Only a handful of studies have been conducted to explore chick’s preference for the imprinting stimulus (144, 568).
memory-consolidating functions of sleep during early post- Imprinting increases also the amount of activated CaMKII,
natal periods, infancy, and adolescence in humans. The de- the size of dendritic spine postsynaptic densities, and a de-
velopmental approach might be particularly important for layed upregulation of NMDA receptors in the IMM, with-
understanding the mechanisms underlying sleep-dependent out changing the number of synapses per se (129, 567, 799).
memory consolidation, mainly for two reasons. First, com-
pared with adults, infants and children sleep longer and An involvement of sleep in imprinting was suggested by an
deeper. In parallel, during early life the brain exhibits par- early study (1120) in which a single imprinting session start-
ticularly strong plasticity shaping memory systems and un- ing 12 h after ecclosion produced a significant increase in
derlying neuronal circuit formation. During limited periods the number of episodes and the amount of REM sleep,
early during development, the brain is particularly sensitive whereas after pseudo-imprinting REM sleep decreased. In a
to certain experience that instructs neuronal circuits to rep- different study, changing the imprinting object from one
resent respective information as very persistent memories day to the other produced a bias towards increased left
determining performance throughout later life (650). How- hemispheric sleep, possibly related to increased consolida-
ever, the mechanisms underlying neuronal plasticity during tion processes in the left hemisphere (102). Unilateral sleep
such critical periods differ from that during adulthood. As a typically covers !1–2% of sleep time in chicks. Results
consequence of maturational processes, critical period plas- from a more recent study compellingly demonstrated a crit-
ticity is linked to a distinct biochemical milieu, a change in ical involvement of sleep in the consolidation of the imprint-
the subunit composition of NMDA receptors in forebrain ing stimulus (600). In this study, one group of chicks was
synapses from a predominance of NR2B subunits at birth to allowed to sleep undisturbed after imprinting training for 6
a prevalence of synaptic NR2A subunits towards the end of h before being retested and then was subjected to a 6-h
the critical period, a specific adjustment of the balance be- period of (slightly) disturbed sleep. Final testing after the
tween excitatory and inhibitory inputs to neuronal net- second sleep period revealed a stable memory for the im-
works, and eventually also to different features of spike printing stimulus together with an overall doubled number
timing-dependent plasticity (332, 352, 545, 838, 1212). of IMM neurons responding to the imprinting stimulus
The specific characteristics of plasticity during critical peri- (compared to imprinting training before sleep). In contrast,
ods prevent a straightforward generalization of the effects chicks that experienced disturbed sleep in the first 6-h inter-
of sleep on plasticity in the adult brain. Nevertheless, it is val after training and undisturbed sleep in the second 6-h
striking that studies of the developing brain, including those interval, at final testing did not exhibit any significant mem-
targeting critical period plasticity, show an influence of ory for the imprinting stimulus and showed a strong de-
sleep on the formation of memory that is even more pro- crease in the number of IMM neurons responsive to the
found than that observed in the adult brain. imprinting stimulus. In addition, chicks with undisturbed

726 Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 SLEEP AND MEMORY

sleep after imprinting displayed an increase in EEG SWA structure achieved the best final imitation of the tutored
covering the 0 – 6 Hz frequency band during this period. song at the end of the 45-day monitoring period, suggesting
These findings point to a particular importance of SWS, a functional role of the sleep-related deterioration in singing
occurring in a window !1.3–5 h after stimulus exposition, behavior for the overall learning process. Deterioration of
for forming a stable memory for the imprinting stimulus. song structure occurred also after induction of sleep during
Interestingly, tracking of individual IMM neurons revealed the daytime by the administration of melatonin excluding
a pattern suggesting that sleep shortly after imprinting confounding effects by the circadian rhythm.
training stabilizes the memory representation by supporting
the recovery of responsiveness of cells that for any reasons The neuronal mechanisms during sleep that mediate the
had ceased to respond to the imprinting stimulus in the deterioration in morning song structure are not clear. The
course of training or shortly afterwards, with this effect major auditory and song system pathways of the songbird
requiring some hours to occur (600, 1150). The mecha- brain have been described in terms of a functional hierar-
nisms of this sleep effect are unclear. Given that the IMM chy. The caudal medial nidopallium (NCM) and the caudal
corresponds to parts of the mammalian neocortex receiving mesopallium (CM) receiving inputs from primary auditory
significant input from the hippocampus (819, 979), such structures contribute to the formation of auditory represen-
inputs during sleep may drive the stabilization of IMM tations likely including those forming the sensory template
representations, although hippocampal neurons do not ap- of the tutored song (108, 463). These structures represent a
pear to selectively respond to the imprinting stimulus after major source of the inputs to the vocal control “song sys-
training (600, 858). tem” comprising chiefly the nucleus HVC (“high vocal con-
trol”) and the robust nucleus of the arcopallium (RA) rep-
2. Song learning in birds resenting the motor cortex analog of the song system (774).
Along with the deterioration of song structure, HVC neu-
Song learning in birds was one of the first models used to
rons show a decline in burst activation across sleep in juve-
systematically study the role of sleep in developmental
nile songbirds.
learning. It bears great similarity with speech learning in
human infants (321, 774). The surprising discovery that
In adult songbirds, RA premotor neurons have been iden-
sleep is essential to a bird’s acquisition of a song fostered a
tified that show replay during sleep, i.e., patterns of burst
major advance in this field (773, 774). Song learning in
activity during sleep which, in terms of their temporal se-
birds has been mainly studied in zebra finch males which
quence and spike sequence structure within each burst, are
develop their song between day 30 (after hatching), when
very similar to those observed during singing (FIGURE 8B).
they start producing unstructured sounds, and day 90,
when they exhibit a well-developed song to be used as a The replay in RA neurons is driven by input from HVC
complex social signal (321). Song learning is based on an neurons (267, 504) which, in juvenile songbirds, show a
innate predisposition to imitate vocalizations, and requires decline in burst activation across sleep, in parallel with the
exposure to a song model. Under experimental conditions, deterioration of song structure (271). Several studies by
the birds are typically subjected to a standard protocol, i.e., Margoliash’s group provided evidence that the replay activ-
they are first raised by females who do not sing and then, at ity during sleep indeed reflects a processing of memory rep-
the appropriate age, gain limited access to a tutor song by a resentations producing changes in the representation and
form of instrumental conditioning, e.g., by pecking a cer- also in subsequent song performance. Thus, in juvenile ze-
tain key (774, 1181). Song learning has been conceptualized bra finches, the first exposure to a tutor song distinctly
as a two-stage process (660, 661): first, the bird acquires a enhanced high-frequency bursting in RA neurons during
sensory model (“template”) of the tutored song which in a subsequent sleep, with corresponding changes in singing
second sensorimotor step the bird gradually learns to imi- occurring not until the following day (1063). Moreover, the
tate, whereby auditory feedback is essential for this process. sequence structure of spiking within bursts identified during
sleep was significantly correlated with that observed while
Derégnaucourt et al. (279) provided compelling evidence the bird was listening to the tutor song during the preceding
that the sensorimotor phase of song development is driven wake phase. Both the increase in burst occurrence and burst
by sleep. Monitoring the structure of the song and its sylla- structure in RA after initial exposure to the tutor song also
ble features by frequency-based automated song analyses required that the bird receive auditory feedback from its
across the whole period of song development in juvenile own singing, as these increases were drastically reduced
zebra finches, they revealed that song structure profoundly when the birds were experimentally deprived from auditory
deteriorated after nocturnal sleep compared with the eve- feedback or were surgically muted.
ning before sleep, and was regained only after intense morn-
ing singing (FIGURE 8A). The effects occurred on the back- The findings suggested a model where RA burst activity
ground of a gradual day-to-day increase in song structure during sleep and the forming of sensorimotor representa-
and were not seen in adult birds. Interestingly, the young tions of a song requires a twofold of inputs during waking,
birds showing the greatest morning deterioration in song i.e., on the one hand, auditory feedback activity from actual

Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org 727

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 BJÖRN RASCH AND JAN BORN

A 2.0 2.0
2.0

Start
Entropy variance

Entropy variance

Entropy variance
1.5 training 1.5 1.5
Day 46 Night Day 47

1.0 1.0 1.0

Day 89 Night Day 90

0.5 0.5 0.5

0.0 0.0 0.0

40 50 60 70 80 90 08:00 20:00 08:00 08:00 20:00 08:00
Age (days) Time of day Time of day

Start
Entropy variance

1.5 training

1.0

0.5

0.0
40 50 60 70 82 90
Developmental time (days)

Singing

Sleep
(ongoing)

100 ms

FIGURE 8. Sleep-dependent formation and reactivation of song memory in birds. A: the measure of Wiener
entropy variance (EV) of song structures reveals a continuous improvement in song structure in young birds
over the 45-day developmental period starting with the first exposure to the tutor song (start of training). In
spite of this overall improvement, during early development (day 46, middle panel) overnight sleep induces an
acute decrease in song performance as indicated by a stronger deviation in song structure in the morning after
sleep compared with presleep performance. This overnight decrease is not any more present in the end of the
learning period (day 89, right panel). Bottom panel indicates continuous tracking of EV values over the 45-day
period. [Modified from Derégnaucourt et al. (280), with permission from Nature Publishing Group.]
B: neuronal trace of an arcopallium (RA) neuron emitting 10 distinct bursts of 2–7 spikes/burst (”singing“).
The bursts are precisely timed to when the bird sang a song whose motif consisted of a sequence of five
syllables (see spectrograph, frequency vs. time representation; top). For each song bout, the sequence of
syllables and the structure of each spike burst (timing of spikes and numbers of spikes) were highly reliable. The
same pattern of spike-bursts reoccurred during recording during sleep. [Modified from Dave and Margoliash
(267), with permission from American Association for the Advancement of Science.]

singing during waking that serves a permissive role in struc- conditions (i.e., in the absence of actual auditory feedback),
turing night-time RA bursting and organizing the song rep- may manifest themselves in a deterioration of song struc-
resentation and on the other hand activity related to the ture in the morning after sleep (280, 463, 856, 857). Early
sensory template of the tutor song (774, 775). Given that during song development, reactivated sensory representa-
RA burst activity during sleep reflects sensory experience of tions and sensorimotor feedback as experienced during ac-
the tutored song during the wake phase, reactivated “tem- tual singing are poorly correlated, resulting in large differ-
plate” activity during sleep may drive plastic changes in the ences in singing between morning and evening performance
song representation that, as they occur under unsupervised which disappear with increased learning. In addition to

728 Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 SLEEP AND MEMORY

modifying sensorimotor representations, reactivations dur- translation of plasticity-related mRNAs (e.g., Arc and
ing sleep may also strengthen the song template (136). BDNF) was increased during sleep, overall speaking for the
Whether SWS or REM sleep is more important for the pro- notion that sleep is a period of enhanced protein synthesis in
cessing of song memories during sleep cannot be answered, these networks. ODP was paralleled by increased multiunit
as RA burst activity appeared to be equally affected in both activity during both SWS and REM sleep, and was corre-
sleep stages (1063). In addition, it remains unclear whether lated to the time spent in non-REM sleep (412). Of the
sleep actually benefits reorganization of memory traces be- GABAA agonistic substances zaleplon, eszopiclone, zolpi-
tween brain structures as assumed for episodic memories in dem, and triazolam, only zolpidem impaired sleep-depen-
humans (964, 965). Although this model as proposed by dent ODP, although all substances distinctly reduced REM
Margoliash and co-workers (774, 775) aims to explain de- sleep and increased non-REM sleep (48, 1059). The differ-
velopmental song learning, it may account also for song ential efficacy of these substances in diminishing ODP may
learning and the maintenance of learned songs in adult be related to differences in their pharmacodynamics and
songbirds. There is evidence for song replay and for system- selectivity for binding specific GABAA receptor subtypes,
atic changes in RA burst activity patterns across sleep in but is independent of their profound changes in sleep archi-
adult zebra finches as well (969). Interestingly, RA burst tecture, as determined by standard EEG sleep recordings.
activity during sleep in some of these cases appeared to be ODP was impaired following application of the atypical
more similar to that observed during singing after sleep than hypnotic trazodone, probably acting via blocking 5-HT2c
before sleep, suggesting a kind of “preplay” that occurs receptors during sleep (48). Since serotonin activity is at a
during sleep and creates new song features (323). minimum during REM sleep and at intermediate levels dur-
ing non-REM sleep, this finding suggests that processes dur-
3. Ocular dominance plasticity ing non-REM are more important for the expression of
ODP. Interestingly, there is evidence that ODP critically
Ocular diminance plasticity (ODP) has been studied mainly depends on the T-type calcium channel function (1224),
in cats as a developmental model of synaptic plasticity that which also plays an essential role in the generation of spin-
is induced by specific stimulus conditions, i.e., monocular dles and slow oscillations during non-REM sleep (282,
deprivation (77, 411). In critical periods during develop- 283).
ment (in cats between about postnatal days 28 and 40),
blocking vision in one eye causes a massive rewiring of In sum, this research indicates that sleep plays an essential
cortical circuitry in favor of the open eye. ODP is sleep- role for behaviors that are formed during critical or sensi-
dependent inasmuch as sleep following a period of monoc- tive periods in early postnatal life. The mechanisms medi-
ular deprivation resulted in an almost twofold increase in ating synaptic plasticity during critical periods in essential
synaptic remodeling in visual cortical areas, whereas after a aspects differ from those during later life (838) and differ-
similar waking interval in complete darkness, a tendency ences in the regulation of synaptic plasticity may still exist
for erasure of the effects induced by the preceding monoc- between juveniles and adults (772). Yet, this does not nec-
ular deprivation was observed (412). The beneficial effect of essarily imply that the manner in which sleep contributes to
sleep on consolidation of ODP mainly leads to a strength- critical period plasticity also basically differs from the effect
ening of the cortical responses to stimulation of the nonde- of sleep on memory formation in the adult brain. Sleep
prived eye (48). Inactivating the sleeping visual cortex by during the early postnatal period and in adults mainly dif-
administration of the Na% channel blocker lidocaine or the fers in quantity of total sleep, time in SWS and REM sleep,
GABAA agonist muscimol inhibited ODP, indicating that intensity of SWS and SWA, etc., suggesting that the impact
postsynaptic activity during sleep is required for consolidat- of sleep on memory formation during early development is
ing the plastic changes induced by a monocular experience overall stronger than during adulthood, rather than differ-
(411, 605). Sleep-dependent ODP is prevented by blocking ing in quality. This view may account especially for devel-
glutamatergic NMDA receptors or cAMP-dependent pro- opmental learning processes that are less clearly linked to an
tein kinase (PKA) during sleep after monocular deprivation early postnatal critical period but extend towards adoles-
(48). Blockade of NMDA and PKA signaling was associated cence, like song learning in birds or language learning in
with reduced activation of the kinases CaMKII and ERK as children. Indeed, the power of sleep in forming memory
well as phosphorylation of GluR1 at Ser831, i.e., processes during early life and the fact that these memories are the
that are critical to the insertion of AMPA receptors into the first to be formed may suffice to explain the unique strength
postsynaptic membrane and the strengthening and mainte- of these memories being difficult to overwrite or reverse.
nance of LTP. Most recently, Seibt et al. (1060) demon-
strated the critical dependence of ODP on protein synthesis
during sleep, as respective plastic changes in visual cortex B. Early Development: Human Studies
could be prevented by inhibition of mTOR-mediated pro-
tein synthesis through rapamycin. Rapamycin had no effect Investigations of the effects of sleep on memory formation
on plasticity induced during wakefulness. Furthermore, during early development in humans focused on declarative
phosphorylation of regulators of protein synthesis as well as and procedural memory processes similar to those exam-

Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org 729

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 BJÖRN RASCH AND JAN BORN

ined in adults, rather than on behaviors linked to sensitive tween the groups, a significant positive correlation between
periods, like social imprinting and amblyopia as a conse- slow oscillation power during non-REM sleep was only
quence of monocular deprivation. Overall, these studies observed in the healthy children, possibly reflecting im-
point towards the particular relevance of SWS for sleep- paired functionality of this rhythm in ADHD children. In
dependent memory consolidation in children. In adults, 14- to 16-yr-old adolescents, the restriction of sleep (to up
SWS has been revealed to play a causal role for the consol- to 5 h for 4 consecutive nights) did not impair either reten-
idation of hippocampus-dependent declarative memories tion of word pairs or procedural memories for mirror trac-
(783, 957), but may play a similar role for explicitly ac- ing skill (1259). However, the adolescents showed a re-
quired procedural memory (e.g., Ref. 691, see also sect. markable increase in the proportion of SWS during the
IVA). restriction period, suggesting a high capability of the ado-
lescent’s brain to compensate effects of sleep restriction by
Infants and children sleep longer than adults. However, flexibly increasing the sleep depth.
compared with adults, the proportion of REM sleep in chil-
dren increases only during the first 6 mo of life, whereas the A beneficial effect of sleep in children (6 – 8 years) was
proportion of SWS remains enhanced throughout develop- found also for the consolidation of visuospatial memories in
ment until adolescence (480, 866). The EEG during SWS a two-dimensional object-location task known to involve
also shows typical changes during development: the ampli- the hippocampal function (1318). Unexpectedly, in this
tude and slope of slow waves as well as SWA (i.e., spectral study comprising a comparison of memory performance in
power in the 0.5– 4 Hz band, including both $1 Hz slow adults, the size of the sleep effect was closely comparable
oscillation and 1– 4 Hz delta activity) increases until the between children and adults, although the amount of SWS
beginning of puberty at the age of 10 –12 years to levels during experimental nights was on average more than two-
distinctly higher than in adults, and thereafter starts to de- fold higher in the children. However, the direct comparison
crease (169, 604, 680). Several studies indicated that sleep of retention rates for declarative materials between children
disturbances and sleep loss impairs learning and school per- and adults is generally hampered by the fact that consoli-
formance in children (for reviews, see Refs. 44, 246, 286, dation can be sped up depending on the availability of as-
662, 1178). Considering the importance of SWS specifically sociative schemas in long-term memory that can integrate
for the consolidation of hippocampus-dependent memory new declarative information (1210, 1211). In the adult
in adults, it was hypothesized that the effect of sleep on the brain, more of such schemas may be available than in the
consolidation of declarative memories is even greater dur- child’s brain. Thus, irrespective of this confound, these studies
ing development (1321). overall point towards a particular relevance of the high
amounts of SWS and associated SWA for declarative memory
1. Consolidation of declarative memory consolidation during development.

Backhaus et al. (56) reported beneficial effects of sleep on the 2. Consolidation of procedural memory
consolidation of declarative memories (word pairs) in 9- to
12-yr-old children. The children’s recall performance after In contrast to the strong beneficial effects of non-REM sleep
sleep was not only better than that after the wake-retention on the consolidation of declarative memory in children,
intervals, but also compared with recall performance tested studies examining procedural memories consistently dem-
before sleep, suggesting a genuine gain of declarative mem- onstrated that in children, unlike in adults, posttraining
ory that is produced by intervening sleep in children. Reten- sleep does not produce a robust improvement of skill. This
tion of word pairs across the sleep interval was positively finding is surprising considering that the acquisition of skills
correlated with the time spent in non-REM sleep and neg- like learning to walk and to speak, to write, and to ride a
atively with the time in REM sleep. Similarly, several other bicycle represent major challenges of childhood. Also, the
studies demonstrated profits from sleep in children between neuroanatomic structures underlying skill learning mature
7 and 14 years of age for word memories as well as for the quite early during development, i.e., within the first 3 years
integration of novel words into lexical knowledge (141, of life (180, 468). In fact, the majority of studies on sleep-
530, 945). Similarly, in 7- and 12-yr-old children, memory dependent procedural memory formation in children indi-
for novel words profited from an off-line consolidation in- cated an impairing rather than improving effect of sleep on
terval involving sleep (141). Sleep in children also preferen- skill (385, 947, 949, 1318), remarkably similar to findings
tially enhanced emotional declarative memories (recogni- in juvenile song birds where overnight sleep likewise dete-
tion of emotional pictures), whereas no effect of sleep on riorated performance on the tutored song (280). In the re-
procedural memory consolidation (mirror tracing) was ob- spective studies, the sleep-dependent overnight gains in skill
served (947). A comparison of healthy children (10 –16 were measured with reference to performance before sleep
years) with age-matched children with attention deficit/hy- rather than with reference to performance changes across a
peractivity disorder (ADHD) revealed a superior sleep-de- corresponding wake interval, which excludes that the ab-
pendent benefit for the retention of pictures in the healthy sent gain of skill across sleep in children is due to a relatively
children (948). Although time in SWS did not differ be- greater capability in children to stabilize memories during

730 Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 SLEEP AND MEMORY

wakefulness (318). Interestingly, sleep restored daytime mance. A recent study indeed confirmed a striking superi-
deficits in procedural memory (serial reaction time task) in ority of sleep in 8- to 11-yr-old children, compared with
children with attention-deficit hyperactivity disorder, while adult’s sleep, to extract explicit declarative sequence knowl-
again no overnight improvements were observed in healthy edge from a coarse motor SRTT that was trained under
children (949). implicit conditions (i.e., without being aware of the under-
lying sequence) before sleep (1322). On the generation task
One factor explaining the missing overnight gain in skill in after retention sleep requiring the subjects to deliberately
children might be their fairly low initial skill levels. In generate the eight-element sequence of cue positions under-
adults, the level of skill performance at learning has repeat- lying the SRTT trained before sleep, the children clearly
edly been demonstrated to modulate sleep-dependent mem- outperformed the adults. The children’s performance was
ory consolidation, whereby benefits from sleep appeared to also distinctly better than in control children who per-
be most robust with intermediate performance levels (16, formed the generation task after training before the sleep
297, 679, 1151). In children, skill performance is generally interval. Superior explicit knowledge after sleep correlated
much slower, less accurate, and less automated than in with the amount of SWA in both age groups, in line with
adults, particularly in the initial stage of training (318, 385, studies in adults indicating that SWS is critically involved in
1185), which gives rise to the question of whether sleep the extraction of explicit knowledge (1347).
enhances skill memories also in children if presleep perfor-
mance is improved to a level comparable with that in adults. Consistent with the notion that sleep in children particu-
Indeed, children (aged 4 – 6 years) who received high larly benefits explicit task aspects and the extraction of
amounts of training in a finger-tapping task did develop a abstract knowledge, 15-mo-old infants already showed a
significant gain in tapping skill across sleep, whereas no sleep-dependent benefit for extraction of grammatical rules
such gain was observed in low performing children who during language learning (469). The results were confirmed
had received the standard amount of training (1320). In the in a second study of this group employing an extended
adults of this study, only the group with minimum training retention interval of 24 h (589). Both studies used the in-
showed a consistent benefit from the nap. In combination, fant’s orienting response towards the auditory word-strings
the findings in children and adults suggest that across age to assess memory recall. Orienting is a well-known hip-
groups, the likelihood that sleep produces a distinct gain in pocampus-dependent function (1119), suggesting that the
skill performance is greatest when presleep performance is retrieval test tapped hippocampus-dependent declarative
at an intermediate level. Owing to their generally quite low aspects of the memory, although this issue deserves further
performance of skills, children reach levels of strength at examination.
which sleep-dependent benefits directly translate into ben-
efits in behavioral speed and accuracy only after extensive Together, the findings of a facilitating effect of sleep during
levels of training. development on the generation of hippocampus-dependent
explicit knowledge about sequences and grammatical rules
3. Children extract more explicit knowledge from are well in line with the assumption of an active system
implicitly trained tasks than adults consolidation established during sleep (293, 1321, 1327).
According to this concept, processes during SWS support-
Alternatively, the lack of sleep-dependent gains in skill per- ing the redistribution of memories from temporary hip-
formance in children might be ascribed to a competitive pocampal to extrahippocampal long-term storage sites, do
interaction between declarative and procedural memory not only strengthen these memory representations, but also
systems (385, 1318). Particularly in the initial stages of bring about qualitative transformations, manifesting them-
training, contributions of explicit (i.e., conscious) learning selves behaviorally in increased knowledge about invariant
mechanisms involving prefrontal-hippocampal circuitry repeating and relevant features in the learned materials.
can be essential for skill acquisition, with the parallel stor- Due to increased SWS in children, such transformations
ing of procedural and hippocampus-dependent declarative could be stronger during development than during adult-
aspects favoring interactions and competitions between the hood.
two memory systems (16, 43, 937, 1042), which can result
in an impaired implicit task performance (e.g., slowing of
reaction times) (1064, 1135). Importantly, this competitive C. Sleep and Memory Formation in Aged
interaction between the memory systems might extend to Rats and Humans
processes of consolidation during sleep (143, 147, 273,
1000). On the basis of the assumption that SWS predom- Sleep undergoes characteristic changes in the course of ag-
inantly supports consolidation of hippocampus-depen- ing. Older adults sleep less and awaken more frequently
dent declarative memory, it was argued that sleep in chil- during the night. Beginning already around the age of 40,
dren, because of its great amounts of SWS, preferentially there is a distinct decrease in SWS (167, 866). Non-REM
strengthens hippocampus-dependent explicit aspects in a sleep in the elderly is characterized by decreases in SWA,
skill representation, thereby hampering implicit perfor- especially over the prefrontal cortex, as well as decreases in

Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org 731

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 BJÖRN RASCH AND JAN BORN

spindle density (176, 177, 690, 786). The time in REM in the aged due to deficits in SWS, the direct comparison of
sleep, on the other hand, remains relatively unchanged in benefits from sleep between age groups might be con-
late life. However, there is a significant decrease in the den- founded by the fact that age also diminishes memory reten-
sity of phasic REMs in elderly persons (254). tion across wakefulness, which is typically used as reference
in these studies.
1. Hippocampus-dependent declarative memory
2. Procedural memory
Considering that the major age-related changes in sleep ar-
chitecture concern SWS and associated SWA, prominent With regard to procedural memory, several studies relying
impairments in the consolidation, especially of hippocam- on different versions of the SRTT revealed age-related im-
pus-dependent declarative memory, may be expected (512). pairments in the formation of motor skill memories during
Consistent with this notion, aging rats at rest after spatial sleep. Comparing older (45– 80 years) and younger adults
memory encoding showed weaker reactivations of tempo- (18 –24 years) on the SRTT under implicit and explicit
ral firing patterns in hippocampal neuron ensembles, and learning conditions, Spencer et al. (1127) found that only
this was associated with diminished spatial memory perfor- the younger but not the older subjects distinctly improved in
mance in the Morris Water maze at the final testing day performance in both task versions across overnight sleep,
(e.g., Ref. 448, see sect. IIIA4). The weakening of tempo- compared with performance changes across a daytime re-
rally sequenced reactivation patterns in aged rates did not tention period of wakefulness. Another study revealed
appear to be a consequence of age-related decreases in SWS. sleep-dependent performance gains on a deterministic
Interestingly, simple reactivation patterns in CA1 cell en- SRTT in middle-aged (35–55 years), but not in 55–70 years
sembles, not taking into account the relative temporal order old participants (1324). In contrast, using a finger sequence
of cell-pair firing, was comparable between aged and young tapping task, one study reported that elderly subjects
rats (449). (60 –79 years) maintained their performance across 24 h
including sleep, while their performance was impaired after
In humans, comparing retention of word pairs across noc- 12 h of wakefulness, suggesting a role for sleep in optimal
turnal sleep between young (18 –25 years) and middle-aged motor consolidation also in the elderly (1214). However,
adults (48 –55 years), Backhaus et al. (55) revealed a signif- the overnight improvements typically observed in young
icantly lower retention of word pairs in the group of mid- subjects did not occur in these elderly participants. Sleep
dle-aged participants. The impairment was observed specif- also did not benefit the performance of older (50 –75 years)
ically for a retention interval that covered early nocturnal subjects on a continuous motor tracking task where the
sleep, which is typically dominated by SWS. SWS in this moving target to be tracked could follow a repeated se-
interval was strikingly lower in the middle-aged than young quence or move randomly similar to the SRTT (1079 –
subjects. Also, percentages of time spent in SWS, but not in 1081). Interestingly, age-matched stroke patients in this
REM sleep, were strongly correlated with later retention study did show an improvement in tracing accuracy selec-
performance. Overall, these data strongly suggest that the tively across the sleep retention interval for both explicit
age-related deficits in declarative memory consolidation and implicit performance conditions. Finally, sleep also did
during sleep are due to a decline in SWS. In line with these not improve performance in a second-order probabilistic
findings, in aged individuals (69 – 80 years), the improve- SRTT in old subjects (60 – 80 years) (852). However, sleep-
ment in the recollection of hippocampus-dependent epi- dependent benefits are also not consistently revealed in
sodic memories (for personally experienced events) after younger subjects for this task, possibly due to the great
retention periods of nocturnal sleep, with reference to rec- complexity of the underlying sequence rules associated with
ollection after daytime wakefulness, was less pronounced a higher degree of implicitness at learning (852, 1122,
than in young (19 –29 years) subjects (21). A recent study 1123).
reported no beneficial effect of sleep on memory and no
correlation between SWS and consolidation of word pairs To what extent the age-related decreases in sleep benefits
across sleep in elderly subjects (60 – 84 years), whereas a for skill learning are related to specific alteration in sleep, in
positive correlation with SWS and a sleep-dependent mem- particular of SWS, is presently unclear as most of the rele-
ory improvement was observed in younger participants vant studies in aged subjects did not include polysomno-
(18 –22 years) (1055) A diminished capability to consoli- graphic recordings. REM sleep does not appear to be in-
date hippocampus-dependent memory (for word pairs) was volved as neither experimental REM sleep deprivation nor
not confirmed in a recent study by Wilson et al. (1324) REM sleep augmentation (i.e., the rebound after depriva-
including, besides young subjects, middle-aged (30 –55 tion) produced any significant performance change in aged
years) as well as old (55–70 years) participants. With refer- subjects on a procedural mirror tracing task (570). Never-
ence to retention rates across daytime wake intervals, sleep theless, like in young adults, high cholinergic activation
improved retention rates in all age groups to the same ex- during REM sleep seems to contribute to sleep-dependent
tent. Although overall the findings point towards a dimin- consolidation of procedural memory also in aged individu-
ished sleep-dependent consolidation of declarative memory als, as administration of the acetylcholinesterase inhibitor

732 Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 SLEEP AND MEMORY

donezipil in aged subjects, together with an increase in and retrieval (641, 695). Also, consolidation includes a cell-
REM density, significantly enhanced sleep-dependent gains cell interaction in which the antigenic information is trans-
in mirror tracing performance (570, 961). The contribu- ferred from a temporary to a long-term store, with both
tions of impaired SWS to the age-related decline in proce- stores represented by different cellular networks. Specifi-
dural memory consolidation remain to be specified in future cally, in this conceptual approach (see FIGURE 9A), encod-
studies. ing in the immune system would denote the uptake of anti-
genic information by APC. Consolidation would refer to
processes at the immunological synapse that is subsequently
VIII. SLEEP-DEPENDENT MEMORY formed between APC and T cells (338, 473) and comprise
CONSOLIDATION IN THE the transfer of the antigenic information from the APC, only
IMMUNE SYSTEM serving as temporary store, to the T cells serving as long-
term store. Recall would refer to the faster and facilitated
immune response when the antigen is re-encountered.
A. Memory Formation in the
Immune System
B. Effects of Sleep on the Formation of an
The immune system forms long-lasting memories for an Adaptive Immune Response
antigen in a multistep process. The mounting of a so-called
adaptive immune response is at the core of this process. 1. Experimental vaccination
Once antigens invade the organism at certain barriers, they
are taken up and processed by antigen presenting cells Vaccination as an experimental model of infection has been
(APC). APC carry the antigenic information to secondary used as a straightforward approach to comprehensively as-
lymphoid organs where they present fragments of the pro- sess effects of sleep on the formation of immunological
cessed antigen to naive T cells, which become only activated memory mainly in humans where this approach quite con-
if they express the specific cognate receptor that can recog- sistently revealed an enhancing effect of sleep on the adap-
nize this particular antigenic epitope. The T cell recognizing tive immune response and measures of antigenic memory.
its antigen, together with the APC, form the so-called im- Human volunteers who slept normally in the first night after
munological synapse. Activation of the T cell is linked to the a single vaccination against hepatitis A, 4 weeks after the
release of different proinflammatory cytokines by APC, vaccination displayed a twofold increase in antigen-specific
among which interleukin-12 (IL-12) is central for the sub- antibody titers, compared with participants who stayed
sequent differentiation of the activated T cell into T-helper awake on the first night after vaccination (697). Similar
1 (Th1) cells. T-cell growth, proliferation, and differentia- effects were found in recent experiments using repeated and
tion is additionally supported by the release of IL-2, combined inoculations with hepatitis A and B antigens (FIG-
whereas the release of IL-10 in this context plays an antag- URE 9B, Ref. 694). In this study, nocturnal sleep after the
onistic role favoring differentiation of Th2 cells and the inoculations doubled blood counts of antigen-specific Th
recruitment of immediate but preformed humoral immune cells, as marker of antigenic memory. During antigen reen-
responses. Th1 cells also support the differentiation of B counter, these Th cells stimulate the production of antibod-
cells and their production of antigen-specific antibodies. ies specifically directed against hepatitis A and B. Moreover,
Whereas most effector cells differentiated during acute in- sleep profoundly increased the proportion of T cells pro-
fection do not survive for a long time after the infection, ducing proinflammatory and Th1-cytokines like IL-2, inter-
there is a subset of differentiated T and B cells that main- feron-$ (IFN-$), and tumor necrosis factor-% (TNF-%). The
tains the antigenic memory for the long-term. These mem- effects persisted over 1 year. In a further study, short sleep
ory cells mediate a speedy and more effective immune re- duration after hepatitis B vaccination predicted a decreased
sponse upon reencounter of the antigen (13, 1258). antibody response to the immunization (946). Together,
these studies provide a compelling demonstration that noc-
There are apparent differences between memory formation turnal sleep after vaccination suffices to strengthen the
in the immune system and in the central nervous system. evolvement of an adaptive, i.e., memory-forming immune
Most obvious, in the immune system cells migrate to act in response to a clinically relevant extent. A strengthening role
different tissues and body compartments and, unlike neu- of sleep for immunological memory formation is also sup-
rons, they often show strong proliferation in response to ported by observations of reduced total IgG under condi-
(antigenic) stimulation. These conditions may partly ex- tions of sleep restriction (to 4 h/night) for several days after
plain why memory formation in the immune system takes vaccination against influenza (1130).
much longer (several days to weeks) than in the central
nervous system. Nevertheless, immunological memory, like In contrast to studies in humans, which examined effects of
neurobehavioral memory, is formed as part of an adaptive sleep on primary immune response developing to the first
organismic response to an environmental stressor, and encounter of an antigen, studies in animals have so far con-
shares the basic subprocesses of encoding, consolidation, centrated on secondary immune responses, i.e., the effect of

Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org 733

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 BJÖRN RASCH AND JAN BORN

A Slow wave sleep

Growth hormone Cortisol

Prolactin

Anti-
bodies

Antigen
IL-12 B cells
APC Th1
response

Th cell
Immunological Cytotoxic
synapse T cells

Uptake of the Redistribution into Facilitated response

antigen long-term store upon re-encounter

Encoding Consolidation Recall

B C
0.20
** **
CD40L+ HAV-specific

0.15 (*)
Sleep
* **
Th cells [%]

Slow wave activity [µV2]

(* )
0.10 * 800 800

600 600
Waking
0.05
* 400 400

200 r = 0.72* 200 r = 0.94**

0 1 2 4 8 9 10 12 16 17 18 20 52 0.0 0.1 0.2 0.3 0.4 0.00 0.05 0.10 0.15 0.20
HAV-specific Th-cells [%] HAV-specific Th-cells [%]
Weeks at weeks 18-20 at week 52
FIGURE 9. Sleep supports the initiation of an adaptive immune response. A: concept: the invading antigen is
taken up and processed by antigen presenting cells (APC) which present fragments of the antigen to T helper
(Th) cells, with the two kinds of cells forming an “immunological synapse.” The concomitant release of
interleukin (IL)-12 by APC induces a Th1 response that supports the function of antigen-specific cytotoxic T cells
and initiates production of antibodies by B cells. This response finally generates long-lasting immunological
memory for the antigen. Sleep, in particular slow wave sleep (SWS), and the circadian system act in concert
to generate a proinflammatory hormonal milieu with enhanced growth hormone and prolactin release and
reduced levels of the anti-inflammatory stress hormone cortisol. The hormonal changes in turn support the
early steps in the generation of an adaptive immune response in the lymph nodes. In analogy to neurobehavioral
memory formed in the central nervous system, the different phases of immunological memory might be divided
in an encoding, a consolidation, and a recall phase. In both the central nervous system and the immune system,
sleep specifically supports the consolidation stage of the respective memory types. [Modified from Besedovsky
et al. (90).] B: sleep enhances the hepatitis A virus (HAV)-specific T helper (Th) cell response to vaccination
(three shots at weeks 0, 8, and 16, vertical syringes) in two groups of human subjects who either slept (black
circle, thick line) or stayed awake (white circle, thin line) in the night following inoculations. The immune
response as indicated by the frequency of CD40L% HAV-specific Th cells (percentage of total Th cells) at weeks
18 –20 (left panel) and particularly at week 52 (right panel) is strongly predicted by the amount of slow wave
activity (averaged across the three postinoculation nights). Values are means " SE: *P ! 0.1; **P ! 0.05;
***P ! 0.01. [Data from Lange et al. (694).]

734 Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 SLEEP AND MEMORY

sleep on the recall of an antigenic memory that was already not been studied thoroughly in the context of sleep. Cells
established during a preceding primary response. Overall, centrally involved in the formation of adaptive immunity,
these studies revealed inconsistent results with sleep, com- like naive and central memory T cells, show a circadian
pared with conditions of sleep deprivation, enhancing or rhythm with the highest numbers of circulating cells at night
weakening markers of recall, or showing no effect (145, in humans. This circadian rhythm is mainly driven by cor-
980, 981, 1206). Because of their focus on secondary im- tisol which upregulates the chemokine receptor CXCR4 on
mune responses, these animal studies basically cannot be these cells and thereby facilitates their redistribution to the
compared with the studies of primary responses in humans. bone marrow during daytime (305, 693). On top of this
circadian rhythm, nocturnal sleep, compared with wakeful-
The strengthening effects of sleep on the human primary ness, induces a slight but significant reduction in circulating
immune response to infection were revealed with sleep oc- naive and central memory T cells (695), which indeed might
curring within the first 36 h after inoculation. This time reflect a redistribution of these cells to lymph nodes that is
window suggests that sleep affects the cascade of responses specific to sleep. In rodents, counts of lymphocytes as well
to the viral challenge at a rather early stage. Indeed, re- as percentages of Th cells showed a variation across the
search has identified two main targets that are affected by 24-h period that ran in parallel in blood, spleen, and lymph
sleep in this process, i.e., 1) the interactions between APC nodes, with peak counts reached during the rest period,
and T cells at the immunological synapse and 2) the migra- suggesting a quick redistribution of T cells from the vascu-
tion of these cells to secondary lymphoid tissues. lar compartment to secondary lymphatic tissues. Two stud-
ies, one in sheep (289) and the other in humans (351),
2. Sleep supports APC-T cell interactions provided more direct evidence that the efflux of lympho-
cytes from lymph nodes is reduced during sleep. In rats,
There is converging evidence that sleep promotes Th1 im-
lymphocyte counts in lymph nodes indeed increased after
mune responses as an immediate sequel of APC-T cell inter-
recovery sleep that followed a period of sleep deprivation
actions. Thus sleep in humans enhances the production of
(1354). The mechanisms underlying this sleep-induced
IL-12 by monocytes and premyolid dendritic cells (pre-
homing of T cells to lymph nodes are not clear. Homing to
mDC) which represent the most important precursors of
the lymph node via high endothelial venules relies on adhe-
mature professional APC circulating in blood (307, 696).
sion molecules and chemokine receptors (CD62L, CCR7,
Concurrently, sleep diminishes the production of anti-in-
and CD11a) expressed on the T-cell membrane. In addition
flammatory IL-10 by monocytes (696). Furthermore, sev-
to the suppression of cortisol release during early nocturnal
eral reports indicated an enhancing influence of sleep on
SWS, sleep may ease T-cell trafficking to lymph nodes via
T-cell production of IL-2 which acts to further promote
sympathetic and peptidergic innervation of lymph nodes
T-cell growth, proliferation, and differentiation (120, 595),
and influences on cytokine release (117, 234, 876).
although other studies failed to reveal such effects when
examining IL-2 production in specific T-cell subpopula-
In combination, these data support the view that sleep fa-
tions (109, 308). Sleep increased proliferation of Th cells as
vors an accumulation of T cells in lymph nodes, and there
well as the activity of natural regulatory T (nTreg) cells
might be a parallel effect on APC migration (307, 627).
(109). Concurrent upregulation of nTreg activity might re-
These migratory effects probably occur in concert with syn-
flect a counterregulatory response preventing overshooting
ergistic actions of the circadian rhythm and eventually
of Th1 responses. Moreover, early nocturnal sleep in-
might serve to enhance the rate of APC-Th cell interactions
creased the ratio in IFN-$/IL-4 production by Th cells, in-
in secondary lymphoid tissues during sleep. In contrast to
dicating a shift in the Th1/Th2 cytokine balance towards
sleep, wakefulness appears to support the recruitment of
preponderant production of Th1 cytokines (308). In pa-
cells in circulating blood that have high cytotoxic effector
tients with allergic rhinitis, wheal reactions to a skin prick
potential, like proinflammatory monocytes, terminally dif-
test and spontaneous IgE production of peripheral blood
ferentiated cytotoxic T cells, and cytotoxic natural killer
mononuclear cells were increased after a night of total sleep
(NK) cells, and serve the immediate immune defense against
deprivation, indicating a shift from Th1 to prevailing Th2
pathogens invading the organism preferentially during the
immune defense (637). Finally, there is evidence that sleep
activity period (e.g., Refs. 42, 287).
increases blood concentrations of IL-7, a cytokine fostering
differentiation of T memory cells (76). Together, these find-
4. Contribution of SWS to memory formation in the
ings speak for the notion that sleep specifically benefits
immune system
APC-T cell interactions towards the development of pre-
dominant Th1 immune responses.
Although effects of the experimental manipulation of spe-
3. Effects of sleep on the migration of T cells and cific sleep stages on the adaptive immune response have not,
APC to lymphoid tissues thus far, been systematically assessed, there is converging
though indirect evidence that the strengthening effects of
Sleep appears to affect also the migratory pattern of leuko- sleep on the formation of long-term antigenic memory orig-
cytes, although the trafficking of T cells as well as APC has inates from processes during SWS. The sleep-induced in-

Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org 735

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 BJÖRN RASCH AND JAN BORN

crease in antigen-specific Th cells after hepatitis A vaccina- tory and Th1 cytokine activity (693). Probably, other fac-
tion was strongly correlated with the amount SWS and EEG tors (orexin, leptin, autonomic innervation of lymphoid tis-
SWA on the night after vaccination (693). The increase in sues, etc.) add to this mediation of sleep effects, but these
percentage of hepatitis A virus specific antigen-specific Th are as of yet entirely unexplored. Overall, these findings
cells measured 18 –20 wk after vaccination was correlated corroborate the view that during SWS, which dominates the
with r & 0.72, and that measured 1 year later with a coef- early night, an endocrine milieu is established that supports
ficient of r & 0.94 with SWA. Other measures, including the initiation of an adaptive immune response as a basis for
REM sleep-related parameters, were not predictive of this the formation of long-lasting antigenic memory. Con-
increase. Corresponding findings were obtained in rabbits, versely, there are hints that owing to an overshooting pro-
where the amount of SWS predicted the survival rate after a inflammatory cytokine response, the immediate defense of
pathogen challenge (1205). Moreover, in humans, early antigen is impaired during sleep, particularly during SWS
nocturnal SWS-rich sleep is associated with a shift towards (508, 627, 693, 779).
Th1 cytokine activity predominating over Th2 activity
which is replaced by a shift towards the predominance of It is important to note here that sleep, and particularly SWS,
Th2 cytokine production during late sleep when SWS is has been likewise identified as a factor critically involved in
fading and REM sleep becomes prevalent (308). In fact, the formation of long-term memories in the neurobehav-
peaks in Th1 cytokine activity and, more generally, in pro- ioral domain (292). These findings indicate that memory
inflammatory cytokine activity during the rest period were formation during sleep in both the immune and the central
mainly detected during the early SWS-dominated portion of nervous system, beyond conceptual analogies, shares causal
sleep, in humans as well as in animals in different tissues mechanisms. This, indeed, conveys the idea that forming
including the brain (487, 671) and lymph nodes (359), but long-term memory is a general function of sleep, which
also in blood and circulating leukocytes (109, 120, 172, eventually serves to strategically adapt the organism to en-
306, 308, 487, 696, 749, 1158). Interestingly, some of these vironmental stressors in entirely different domains.
cytokines were found, conversely, to promote SWS, point-
ing towards a positive-feedback loop between the central
nervous and immune systems (693). IX. CONCLUDING REMARKS

5. Endocrine mediation of immunological memory For more than a century it has been known that memory
formation during sleep benefits from sleep, and research in this field has put for-
ward different explanations for this phenomenon. Here we
The strengthening effect of sleep on immunological mem- aimed to establish a review covering the progress of re-
ory formation has been ascribed to the endocrine milieu as search in this field of sleep and memory in its entireness,
it is generated specifically during SWS. In humans, early simultaneously taking into account the vastly differing ap-
nocturnal SWS is associated with an increased release of the proaches that have been adopted to clarify the mechanism
immune-supportive hormones prolactin and growth hor- mediating the memory benefit from sleep. Whereas initially
mones while release of the immunosuppressive glucocorti- it was commonly assumed that sleep improves memory in a
coids is suppressed (117). Sleep, specifically SWS, contrib- passive manner, by protecting it from being overwritten by
utes to the generation of this unique endocrine pattern in interfering external stimulus inputs, the current theorizing
concert with circadian mechanisms (91, 1125, 1126, 1129). assumes an active consolidation of memories that is specif-
Not only SWS, but also the enhanced release of GH and ically established during sleep, and basically originates from
prolactin, together with the simultaneously suppressed cor- the reactivation of newly encoded memory representations.
tisol levels observed during SWS-rich sleep after hepatitis A In parallel, the perspective on the researched memory has
vaccination, were revealed to be highly predictive for the changed. Whereas initial research largely concentrated on
number of antigen-specific Th cells detected in blood 4 wk the stability of memory, demonstrating how sleep contrib-
and 1 year after vaccination (696). In fact, blocking of min- utes to the persistence of a memory in all of its qualities,
eralocorticoid receptors during early SWS-rich sleep pro- recent research has begun to concentrate on the dynamics of
duces an enhancement in naive T-helper cell counts (89). memory formation, systematically examining the changes a
Moreover, in vivo as well as in vitro studies identified high memory representation undergoes during sleep-dependent
prolactin and low cortisol levels as factors most strongly consolidation. The active system consolidation process as-
contributing to the nocturnal upsurge of IL-12 production sumed to take place during sleep leads to a transformation
(307, 696). GH and prolactin are well known to promote and a qualitative reorganization of the memory representa-
T-cell proliferation and differentiation as well as Th1 cyto- tion, whereby the “gist” is extracted from the newly en-
kine activity (193, 217, 436, 814), and to develop adjuvant- coded memory information and integrated into the long-
like actions when given shortly after vaccination (1124, term knowledge networks.
1136). Low glucocorticoid levels during SWS-rich sleep
contribute to these effects as these hormones are highly Most recently, the focus of research in the field has broad-
potent anti-inflammatory agents suppressing proinflamma- ened, indicating that sleep benefits memory not only in the

736 Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.
 SLEEP AND MEMORY

neurobehavioral domain, but also in the formation of im- REFERENCES

munological long-term memories, stimulating the idea that
forming long-term memories represents a general function 1. Abel M, Bäuml KT. Retrieval-induced forgetting, delay, and sleep. Memory 20: 420 –
of sleep. There are first cues that sleep-dependent memory 428, 2012.

formation in the immune and central nervous system share 2. Abel M, Bäuml KT. Sleep can eliminate list-method directed forgetting. J Exp Psychol
common mechanisms, that in both domains appear to be Learn Mem Cogn 2012.
linked to SWS. Also in the immune system, sleep appears to 3. Abraham WC, Robins A. Memory retention: the synaptic stability versus plasticity
support the intercellular reorganization of memory repre- dilemma. Trends Neurosci 28: 73–78, 2005.
sentations such that during the APC-T cell interaction the 4. Abraham WC, Williams JM. LTP maintenance and its protein synthesis-dependence.
epitopic information is extracted from the antigen to be Neurobiol Learn Mem 89: 260 –268, 2008.
stored by T cells. Certain features of active system consoli-
5. Achermann P. The two-process model of sleep regulation revisited. Aviat Space
dation, like cell assembly reactivation of neurobehavioral Environ Med 75: A37– 43, 2004.
representations, can occur also during wakefulness, but
6. Achermann P, Borbély AA. Low-frequency (1 Hz) oscillations in the human sleep
with different consequences. The effective reorganization of electroencephalogram. Neuroscience 81: 213–222, 1997.
the representation requires a specific milieu of neurotrans-
mitter and endocrine activity as it is established only during 7. Achermann P, Borbély AA. Temporal evolution of coherence and power in the
human sleep electroencephalogram. J Sleep Res 7 Suppl 1: 36 – 41, 1998.
sleep, specifically during SWS. Subsequent REM sleep may
then be involved in strengthening the reactivated and reor- 8. Achermann P, Borbély AA. Mathematical models of sleep regulation. Front Biosci 8:
s683–s93, 2003.
ganized representations on a molecular and synaptic level.
Thus sleep and wakefulness appear to be associated with 9. Adamantidis A, Carter MC, Lecea de L. Optogenetic deconstruction of sleep-wake
different and mutually exclusive modes of memory process- circuitry in the brain. Front Mol Neurosci 2: 31, 2010.

ing, with sleep favoring processes of memory consolidation 10. Aeschbach D, Cutler AJ, Ronda JM. A role for non-rapid-eye-movement sleep ho-
that are incompatible with the efficient encoding and re- meostasis in perceptual learning. J Neurosci 28: 2766 –2772, 2008.
trieval of stimuli, as required while coping with environ- 11. Aeschbach Borbély. All-night dynamics of the human sleep EEG. J Sleep Res 2: 70 –
mental demands in the wake phase. 81, 1993.

12. Aguiar CL, Romcy-Pereira RN, Szawka RE, Galvis-Alonso OY, Anselmo-Franci JA,
ACKNOWLEDGMENTS Leite JP. Muscarinic acetylcholine neurotransmission enhances the late-phase of
long-term potentiation in the hippocampal-prefrontal cortex pathway of rats in vivo:
a possible involvement of monoaminergic systems. Neuroscience 153: 1309 –1319,
We thank Maren Cordi, Anja Otterbein, Ursula Rasch, Ju- 2008.
lia Rihm, Thomas Schreiner, Sarah Schoch, and Manuela
13. Ahmed R, Gray D. Immunological memory and protective immunity: understanding
Steinauer for assistance in preparing the manuscript and
their relation. Science 272: 54 – 60, 1996.
Sandra Ackermann-Wohlgemuth, Luciana Besedovsky, Su-
sanne Diekelmann, and Ines Wilhelm for helpful comments 14. Alberini CM. Mechanisms of memory stabilization: are consolidation and reconsoli-
dation similar or distinct processes? Trends Neurosci 28: 51–56, 2005.
on earlier versions of the manuscript. In particular, we
thank Drs. Ted Abel, Igor Timofeev, and Paul Shaw for very 15. Albert I, Cicala GA, Siegel J. The behavioral effects of REM sleep deprivation in rats.
Psychophysiology 6: 550 –560, 1970.
constructive comments and corrections.
16. Albouy G, Sterpenich V, Balteau E, Vandewalle G, Desseilles M, Dang-Vu T, Darsaud
Addresses for reprint requests and other correspondence: J. A, Ruby P, Luppi P, Degueldre C, Peigneux P, Luxen A, Maquet P. Both the hip-
Born, Institute of Medical Psychology and Behavioral Neu- pocampus and striatum are involved in consolidation of motor sequence memory.
Neuron 58: 261–272, 2008.
robiology, Univ. of Tuebingen, Gartenstraße 29, 72074
Tuebingen, Germany (e-mail: [email protected]); 17. Alger SE, Lau H, Fishbein W. Delayed onset of a daytime nap facilitates retention of
declarative memory. PLoS One 5: e12131, 2010.
or B. Rasch, Institute of Psychology, Div. of Biopsychol-
ogy, Univ. of Zurich, Binzmühlestrasse 14, Box 5, CH- 18. Alger SE, Lau H, Fishbein W. Slow wave sleep during a daytime nap is necessary for
8050 Zurich Switzerland (e-mail: [email protected]). protection from subsequent interference and long-term retention. Neurobiol Learn
Mem 98: 188 –196, 2012.

19. Allen RP, Wagman AM. Do sleep patterns relate to the desire for alcohol? Adv Exp
GRANTS Med Biol 59: 495–508, 1975.

This work was supported by Deutsche Forschungsgemein- 20. Allen SR, Oswald I, Lewis S, Tagney J. The effects of distorted visual input on sleep.
Psychophysiology 9: 498 –504, 1972.
schaft Grant SFB 654 Plasticity and Sleep and Swiss Na-
tional Foundation Grant PP00P1_133685. 21. Aly M, Moscovitch M. The effects of sleep on episodic memory in older and younger
adults. Memory 18: 327–334, 2010.

22. Amado-Boccara I, Gougoulis N, Poirier Littre MF, Galinowski A, Loo H. Effects of

DISCLOSURES antidepressants on cognitive functions: a review. Neurosci Biobehav Rev 19: 479 –
493, 1995.
No conflicts of interest, financial or otherwise, are declared 23. Ambrosini MV, Giuditta A. Learning and sleep: the sequential hypothesis. Sleep Med
by the authors. Rev 5: 477– 490, 2001.

Physiol Rev • VOL 93 • APRIL 2013 • www.prv.org 737

Downloaded from journals.physiology.org/journal/physrev at UNIVERSIDAD DE CHILE (200.089.068.005) on April 19, 2024.