- Email: [email protected]
Sculpting memory during sleep: concurrent consolidation and forgetting
Available online at www.sciencedirect.com
ScienceDirect Sculpting memory during sleep: concurrent consolidation and forgetting Gordon B Feld1,2 and Jan Born2,3 There is compelling evidence that sleep actively supports the formation of long-lasting memory representations. Experimental cuing of memories proved that neural replay of representations during sleep plays a causal role for this consolidation, which has also been shown to promote neocortical synaptic plasticity and spine formation. Concurrently, sleep has been proposed to facilitate forgetting through processes of synaptic renormalisation. This view received indirect support by findings in humans of sleep enhancing TMS-evoked plasticity and capabilities for encoding new information. First direct behavioural evidence of sleep inducing forgetting has only recently emerged after encoding large amounts of stimuli in adults. We propose forgetting complements sleep-dependent consolidation and facilitates gist abstraction especially at high memory loads, when reactivation-based consolidation reaches capacity limits.
Addresses 1 Institute of Behavioural Neuroscience, Department of Experimental Psychology, Division of Psychology and Language Science, University College London, London, United Kingdom 2 Institute of Medical Psychology and Behavioral Neurobiology, University of Tu¨bingen, Tu¨bingen, Germany 3 Centre for Integrative Neuroscience, University of Tu¨bingen, Tu¨bingen, Germany Corresponding authors: Feld, Gordon B ([email protected]), Born, Jan ([email protected])
Current Opinion in Neurobiology 2017, 44:20–27 This review comes from a themed issue on Neurobiology of sleep Edited by Yang Dan and Thomas Kilduff
http://dx.doi.org/10.1016/j.conb.2017.02.012 0959-4388/ã 2017 Elsevier Ltd. All rights reserved.
merely passively stored but that an active process of consolidation occurs [1], which has received wide support over the past century [2]. Especially, the idea that memory is consolidated during the brain’s offline periods of deep slow wave sleep has risen to a champion of consolidation theory (Figures 1 and 2) [3–5]. Just recently, in the neocortex, the formation of dendritic spines was observed , during sleep after learning a motor task, as a neural substrate of memory formation, which was positively related to task performance [6]. At its outset, memory research mainly focused on forgetting [7]. Essentially, two forms of forgetting were proposed. The first argues that forgetting occurs due to interference, that is older memory traces are constantly being overwritten by newer encoding events. Consequently, a passive role of sleep protecting from interference has been claimed repeatedly since its first conceptualisation [8], even though this type of forgetting seems to play only a minor role for memories involving the hippocampal system [9]. The second claims that memory traces passively decay over time and this account, later on, was extended to include active decay processes [10]. Here also, sleep has been proposed to play a major role, renormalizing synaptic weights and balancing out potentiation occurring as a result of encoding information during wake (Figure 2) [11]. Covering mainly the period from 2013 to now, we review the latest developments regarding sleep’s role for consolidation and forgetting. Initially, it seems difficult to reconcile these two accounts. To the contrary, building on our previous reasoning [4,12] we will present a framework, derived from novel developments in the field, to explain how consolidation and forgetting work together to sculpt lasting memories from a day’s clay of episodic experiences. Ultimately, forgetting might arise as the fourth process of memory formation that enables the long-term function of the other three processes.
Strengthening memory during sleep Introduction The formation of long-term memory relies on the two distinct processes of encoding (or learning) and consolidation. Retrieval is a third process that contributes to memory formation by re-instating the stored information. Between encoding, that is the uptake of the information, and retrieval lies the period of retention, and already at the beginning of modern memory research it was proposed that during this retention period memories are not Current Opinion in Neurobiology 2017, 44:20–27
The hypothesis that during sleep memory benefits from the repeated replay of neuronal representations that were formed to encode information during prior wakefulness was induced from the initial finding that in rats hippocampal place cells that show correlated firing during wake encoding of a simple maze re-exhibit this correlated firing pattern during subsequent slow wave sleep (SWS) [13] and replay during sleep has recently been shown to predict reinstatement strength during retrieval [14]. This www.sciencedirect.com
Sleep-dependent consolidation and forgetting Feld and Born 21
Figure 1
Sleep REM NonREM1+2 NonREM3+4 (SWS) 23:00
3:00
7:00 h
Slow oscillation Sleep spindle Sharpwave/ ripple Current Opinion in Neurobiology
Sleep stages and oscillations. Sleep and memory research has greatly benefited from the identification of discrete sleep stages and associated oscillatory phenomena. Rapid eye movement (REM) sleep, which occurs mainly during the second half of the night, is characterised by wake-like desynchronized low amplitude EEG, which in the past made it the foremost candidate for sleep-dependent memory processing. To the contrary, reprocessing of memory representations has been shown to occur mostly during deep nonREM sleep (also slow wave sleep—SWS), which is most abundant during the first half of the night and is characterized by high-amplitude lowfrequency oscillations (0.75 Hz slow oscillations) in the surface EEG. NonREM sleep coordinates memory replay by locking hippocampal sharp-wave/ripples, which enwrap the replayed memory information, to the excitable troughs of the thalamo-cortical sleep spindle. The spindle itself is phase-locked to the up-state of the neocortical slow oscillation, thereby ensuring that the reactivated information reaches the neocortex during the excitable up-state of the slow oscillation. This process is supported by the specific neuromodulatory milieu of NonREM sleep that enables information flow from the hippocampus to the neocortex, including low levels of cortisol and acetylcholine.
research was extended to show that sequences of place cells that are replayed in the hippocampus are coordinated with replay of the sensory cortex [15] and of grid cells in the entorhinal cortex [16]. In fact, this replay can be evoked by targeted memory reactivation, that is by representing a cue (e.g. an odour or a sound) that was present during encoding again during SWS after encoding [17]. Correspondingly, in rats re-presenting a sound that predicted the correct choice in a spatial learning paradigm during sleep after learning led to the re-emergence of the associated firing pattern of place cells [18]. In humans, reactivating memory during sleep by re-presenting auditory cues from a Dutch–German paired associate paradigm learned before sleep enhances memory for the cued pairs [19]. In fact, this type of targeted memory reactivation can even be used to reduce notoriously persistent race and gender stereotypes [20] and to enhance the extinction of conditioned fear responses [21]. www.sciencedirect.com
During NonREM sleep an intricate interplay of oscillations, that is neocortical slow oscillations (0.5–1 Hz), sleep spindles (12–16 Hz) originating from thalamus, and hippocampal sharp-wave/ripple complexes, occurs that can be measured with extracranial and intracranial EEG and which is thought to coordinate consolidation. Using intracranial EEG electrodes, in human epileptic patients, it was shown that the neocortical slow oscillation groups sleep spindles to its excitable up-state and that sharpwave/ripples in turn are grouped to the spindle throughs and vice-versa [22,23]. Importantly, high gamma in the cortex has also been shown to be nested in spindle troughs in mice [24]. The efficacy of the slow oscillation to enhance memory consolidation was demonstrated in humans using closed-loop auditory stimulation. In this experiment, short bursts (50 ms) of pink noise were presented during SWS timed to the slow oscillation upstate, which effectively induced high-amplitude slow oscillatory activity with spindle power phase-locked to the up-states [25]. Interestingly, driving slow oscillatory events further by stimulating extended trains did not further enhance sleep spindles, indicating a self-limiting process [26]. Similarly, boosting sleep spindles by transcranial alternating current stimulation can improve motor memory consolidation [27] and electrically blocking ripples during sleep impairs memory consolidation [28]. Importantly, it seems to be the fine-tuned phaselocking of these different frequencies that makes consolidation during NonREM sleep possible [29]. In fact, enhancing slow oscillations pharmacologically by increasing GABAergic tone does not enhance memory consolidation, as simultaneously slow oscillation to sleep spindle phase-amplitude coupling is disrupted [30]. In addition to oscillatory prerequisites, NonREM sleep also offers a unique neurochemical pattern enabling consolidation, as low levels of cortisol and acetylcholine are essential to promote the hippocampal-to-neocortical dialogue [31–34]. Reward related signals also seem to play a major role, as dopaminergic activity during encoding enhances replay during subsequent sleep [35]. Moreover, replay activity during sleep has been shown in reward related areas such as the ventral striatum [36] and the ventral tegmental area [37]. Enhancing dopaminergic neurotransmission pharmacologically during sleep in humans enhances the consolidation of low rewarded items to match high rewarded items [38]. However, it is still unclear, if plastic processes assumed to be induced by neural replay activity during sleep, are equivalent to those active during wakefulness. For example, blocking AMPA or NMDA receptors during sleep does not impact consolidation in humans [39]. It might turn out that during sleep direct electrical coupling between neurons via gap junctions is essential (GB Feld et al., unpublished), which would fit well to the importance of oscillatory coupling for sleep-dependent memory consolidation. Current Opinion in Neurobiology 2017, 44:20–27
22 Neurobiology of sleep
Figure 2
Wakefulness Encoding
Sleep Active systems consolidation
Wakefulness
Synaptic homeostasis
Retrieval and new encoding
Long-term store (Neocortex)
Global synaptic renormalisation Inital store (Hippocampus)
Current Opinion in Neurobiology
Schematic of the classical functions of sleep for memory. In the hippocampus-dependent declarative memory domain, during wakefulness, information (yellow arrows) flows through the sensory cortices to the fast learning system of the hippocampus (green arrows), which quickly binds the information into a transient representation. During subsequent sleep, specifically NonREM and slow wave sleep (SWS), the memory trace is replayed repeatedly (red arrow), which leads to its transformation and integration into the long-lasting stores of the neocortex, that is active systems consolidation (blue arrows). Next to active systems consolidation of relevant memory traces, sleep also renormalizes synapses leading to the erasure of irrelevant information. While the synaptic homeostasis hypothesis underlines the importance of SWS for renormalization in the cortex, evidence suggests that in the hippocampus rapid eye movement (REM) sleep plays a major role. During subsequent wakefulness the consolidated information can be retrieved without recruiting the hippocampus (green upward arrows) and new information can be encoded into its replenished stores (yellow and green downward arrows).
Next to merely strengthening original memory traces during sleep, there is evidence that sleep qualitatively changes the memory trace towards a more generalized representation [40–43]. In fact, information that can be easily integrated into an existing schema, is more likely to benefit from sleep during the retention interval [44], an effect that is mediated by sleep spindles [45].
is evidence for renormalization of synaptic strengths occurring exclusively during the wake state in adolescent rats’ (postnatal day 27–32) visual system after monocular visual deprivation [50]. Although, it has been questioned whether monocular visual deprivation-induced plasticity represents a physiological model of normal learning or a specific form of plasticity that is only readily elicited during a sensitive period.
Forgetting during sleep The synaptic homeostasis hypothesis explains how the brain copes with the physiological demands of learning relying on potentiation of synaptic networks, which would lead to unsustainable demands in energy and space as well as escalating potentiation, if potentiation remained unchecked [11]. There is, indeed, a net reduction in the amount of cortical dendritic spines found in adolescent rats’ (postnatal day 23–44) cortex after sleep, while there is an increase after wake [46]. Likewise the amount of AMPA receptors on glutamatergic neurons is reduced across sleep [47]. However, this account has been contested [48], and it was shown that slow wave sleep (SWS) can enhance potentiation rather than depotentiation [49]. Also, there Current Opinion in Neurobiology 2017, 44:20–27
Essentially, an examination of specific brain areas and sleep stages might prove a more fruitful approach. The hippocampus, for example seems quite reliant on mechanisms of sleep-induced forgetting and here it seems to be rapid eye movement (REM) sleep that renormalizes synaptic activity [51], and such a process might rely on NMDA-receptor-dependent processes of active forgetting [52]. Fittingly, NMDA-receptor activation during sleep improved encoding of new declarative memory after sleep in humans (M Alizadeh-Asfestani et al., unpublished). Similarly, evoked responses to transcranial magnetic stimulation in humans suggested that cortical synaptic plasticity is restored after sleep, which co-occurred with an increase in brain derived neurotropic factor [53]. www.sciencedirect.com
Sleep-dependent consolidation and forgetting Feld and Born 23
Altogether the last years have supplied a vast amount of behavioural support for the role of sleep in consolidating memories (see Ref. [54] for a comprehensive review). However, the notion of sleep promoting forgetting is mainly supported indirectly by neurophysiological evidence and hints from enhanced new learning after sleep, whereas direct behavioural evidence is largely missing. Nonetheless, in a recent laborious experiment we found initial support for sleep-induced forgetting in a behavioural study [55]. Here, participants learned different amounts of word-pairs (40, 160 or 320) and then were either allowed to sleep or stayed awake. As expected, participants showed a strong beneficial effect of sleep on memory retention in the 160 word-pair condition. However, unexpectedly in the 320 word-pair condition the sleep-induced enhancement in memory was nullified, and the sleep group forgot as many items as the wake group (Figure 3). In our view, in a neurophysiological framework, this result is most convincingly explained, if the local processes of potentiation that consolidate memories are assumed to be accompanied by global processes of depression or depotentiation. In this scenario potentiation processes are accomplished via limited resources of neural memory replay that are overstretched at high informational loads leading to a net forgetting. Intriguingly, to our knowledge the only other behavioural study that has shown increased forgetting due to sleep in humans was performed in infants [56]. This likewise argues for forgetting occurring primarily in conditions of informational overload, as young children lack the extensive knowledge networks that allow adults to abstract chunks of information into schemas. Correspondingly, while in that study the infants showed a sleepdependent decline in item memory (i.e. words of an artificial language) they showed increased learning of grammatical rules. Building on the idea that sleep abstracts information because overlapping representations are reactivated more frequently than individual items [57,58], we propose that this kind of gist abstraction occurs at an increased rate, if large amounts of information are encoded, which increases the overlap among representations. We reason that, because available amounts of neural reactivation are limited during sleep, at high information loads, overlap between representations might not be resolved by separating representations [59,60] and therefore, the disparate details of item memory might become subjected to devastating processes of global depotentiation. Concurrently, the relatively increased local reactivation of overlapping representations is preserved as gist (Figure 4). Importantly, this proposition has the power to explain seemingly contradictory results in the field of sleep and memory. For example, although some studies find that www.sciencedirect.com
Figure 3
Experimental design Learning: 40,160 or 320 words-pairs
0 Difference in word-paris recalled at learning and retrival
Interactions between forgetting and consolidation
Immediate recall
Sleep or wake
Delayed recall
Sleep Wake
-2 -4 “Forgetting” -6 -8 -10 Extrapolation 40
160 320 Information load: list length
480
Current Opinion in Neurobiology
Forgetting during sleep induced experimentally by information overload. This first demonstration of sleep-induced forgetting in adult participants used lists of word-pairs with increasing length (40, 160 and 320 word-pairs) to successively enhance information load at encoding. Participants retrieved the word-pairs immediately after learning (immediate recall) and at retrieval (delayed recall). The sleep group (blue) was allowed to sleep for 8 hours after learning and the wake group (green) remained in the lab for eight hours of sleepdeprivation. Retrieval was tested after an additional 24 hours including 8 hours of recovery sleep for the wake group. The difference in wordpairs recalled at retrieval minus the word-pairs recalled at learning was calculated as a measure of retention performance. As predicted, there was a strong beneficial effect of sleep for the 160 word-pair condition. However, contrary to what is predicted if sleep keeps up at least a constant rate of memory enhancement (this prediction is indicated by light blue dashed line), increasing the list length to 320 word-pairs completely nullified the sleep-induced memory enhancement. This significant reduction in retention performance from the 160 to the 320 word-pair condition in the sleep group is indicative of a process of sleep-induced forgetting that can be observed at high information loads. It is tempting to extrapolate the data (dashed blue and green lines), which would predict an advantage of wake over sleep at even higher information loads. In fact, preliminary data from an experiment with a 620 word-pair list length condition point in this direction. Data adapted from Ref. [55].
rewarded items are retained better over sleep than unrewarded items [61,62], there have been contradictory reports [63,64]. According to our proposition, we expect the item enhancing effect of sleep to vanish at high loads of information, even in the presence of rewards. In fact, in a recent study, where reward would reduce the longer it took participants to arrive at a goal in a maze, they Current Opinion in Neurobiology 2017, 44:20–27
24 Neurobiology of sleep
Figure 4
Sleep
Wakefulness Encoding of large amounts of items
Active systems consolidation
snow
Increased gist retrieval & decreased item memory
Synaptic homeostasis
snow
snow
snow
hot
Wakefulness
cold frost
hot
cold frost
hot
frost
hot
frost
Long-term store (Neocortex)
Initial store (Hippocampus)
Global synaptic renormalisation
Current Opinion in Neurobiology
Proposed processing of memory during sleep in conditions of information overload. When large amounts of information are encoded during learning (yellow and green arrows) their hippocampal representations overlap (Here, information is symbolized by the items hot, snow and frost, but of course much more information is encoded). Limited capacity of local replay (red arrow) during subsequent sleep allows only the overlapping information to be processed effectively by active systems consolidation such that the concept ‘cold’ is strengthened within the neocortex (blue arrow). This gist word is shared implicitly by all items, but was not experienced explicitly during learning [70,71]. Without access to stabilizing replay the neocortical traces of the individual items remain labile. Importantly, concurrent global synaptic renormalization nevertheless erases the hippocampal item memory traces, thereby, precluding that these traces aid recall or replay at a later time point. Hence, during subsequent wake retrieval (green arrow) the gist memory ‘cold’ can easily be extracted from the strong neocortical trace, whereas the labile item memory has been forgotten. Note, depending on the amount of encoded information and pre-existing schemas, processes of gist abstraction and forgetting are highly variable and may extend over multiple sleep periods [72].
performed worse after sleep than their peers who received no reward at all [65]. Crucially, the participants of the reward group were constantly monitoring the current reward during learning and perceived the task as significantly more difficult, suggesting that the additional reward signal produced informational overload at encoding. This reasoning is supported by reports of rewards during learning before sleep favouring the generalisation of memory in complex tasks [66,67]. When systematically varying rewards and information load during encoding our model predicts that during subsequent sleep at low loads rewards benefit item memory and at high loads rewards benefit generalisation. Likewise interesting predictions can be made for the paradigm of targeted memory reactivation. Usually, cuing a newly encoded representation during post-learning sleep does not come at the cost of a deterioration of the uncued items [19]. However, at high amounts of encoded information, we expect cueing to deteriorate uncued representations. In fact, some preliminary Current Opinion in Neurobiology 2017, 44:20–27
evidence was provided when in a study participants learned an unusually long finger-tapping sequence and only half of it was cued during sleep, performance for the uncued part was as bad as in the wake condition [68]. Our prediction is also in line with the finding that targeted memory reactivation biases the reactivation of place cells without enhancing overall replay for cued and uncued memories [18]. Interestingly, targeted memory reactivation has been found to lead to the extinction and to the enhancement of fear memories in rodents and human brains, respectively [69]. Thus, understanding the boundary conditions of forgetting and consolidation in terms of perceived informational overload appears to be essential, before implementing any translational approaches.
Concluding remarks The present review aims to reconcile sleep-associated processes of consolidation and forgetting, and highlights their interaction bearing the potential for abstracting gist from large amounts of information. While the causal relationship between behavioural and neurobiological www.sciencedirect.com
Sleep-dependent consolidation and forgetting Feld and Born 25
indicators of sleep-dependent memory consolidation has been investigated in detail during the last years, such a link is completely missing for forgetting during sleep. We argue that sleep-induced forgetting can only be readily measured behaviourally, if participants (humans or animals) are exposed to large amounts of information at encoding, which is not routinely done as it can be extremely laborious. Nevertheless, we believe using informational overload together with the established neurobiological methods to investigate sleep’s function represents a highly promising approach. Such research might identify additional boundary conditions of sleepinduced forgetting, that are systematically interlinked with informational overload, such as age. Importantly, understanding sleep-induced forgetting might also allow us to tackle clinical symptoms that are associated with pathological memories such as PTSD, drug abuse and anxiety by enhancing forgetting of memories that maintain these diseases. Ultimately, it might also open ways to treat pathological forgetting due to, for example Alzheimer’s disease.
10. Hardt O, Nader K, Nadel L: Decay happens: the role of active forgetting in memory. Trends Cogn Sci 2013, 17:111-120.
Conflict of interest statement
18. Bendor D, Wilson MA: Biasing the content of hippocampal replay during sleep. Nat Neurosci 2012, 15:1439-1444. Using targeted memory reactivation in rats implanted for place cell recording this study demonstrated that external cues can elicit replay of specific place cell sequences in the hippocampus. This provides a crucial link to the human literature on cued memory reactiavtion during sleep.
The authors declare no conflicts of interest.
Acknowledgements This work was supported by a grant from the Deutsche Forschungsgemeinschaft (DFG; SFB 654 “Plasticity and Sleep”). GBF is currently receiving a personal stipend from the DFG to conduct research at the University College London (FE 1617/1-1).
References and recommended reading Papers of particular interest, published within the period of review, have been highlighted as: of special interest of outstanding interest 1.
Mu¨ller GE, Pilzecker A: Experimentelle Beitra¨ge zur Lehre vom Geda¨chtnis. Zeitschrift fu¨r Psychologie, Erga¨nzungsband 1900, 1:1-300.
2.
McGaugh JL: Memory—a century of consolidation. Science 2000, 287:248-251.
3.
Feld GB, Diekelmann S: Sleep smart-optimizing sleep for declarative learning and memory. Front Psychol 2015, 6:622.
4.
Diekelmann S, Born J: The memory function of sleep. Nat Rev Neurosci 2010, 11:114-126.
5.
Marr D: Simple memory: a theory for archicortex. Philos Trans R Soc Lond B Biol Sci 1971, 262:23-81.
6.
Yang G, Lai CS, Cichon J, Ma L, Li W, Gan WB: Sleep promotes branch-specific formation of dendritic spines after learning. Science 2014, 344:1173-1178. In this study, the authors applied two photon imaging of cortical neurons to show that denditic spines can be formed during sleep in a performance-dependent manner. This supports an important prediction of active systems consolidation. 7.
Ebbinghaus H: U¨ber das Geda¨chtnis. Leipzig: Duncker und Humblot; 1885.
8.
Jenkins JG, Dallenbach KM: Obliviscence during sleep and waking. Am J Psychol 1924, 35:605-612.
9.
Sadeh T, Ozubko JD, Winocur G, Moscovitch M: Forgetting patterns differentiate between two forms of memory representation. Psychol Sci 2016, 27:810-820.
www.sciencedirect.com
11. Tononi G, Cirelli C: Sleep and the price of plasticity: from synaptic and cellular homeostasis to memory consolidation and integration. Neuron 2014, 81:12-34. This important update to the synaptic homeostasis hypothesis reviews the recent developments and, especially, integrates evidence of sleep’s beneficial effect on memory into the theory. 12. Born J, Feld GB: Sleep to upscale, sleep to downscale: balancing homeostasis and plasticity. Neuron 2012, 75:933935. 13. Wilson MA, McNaughton BL: Reactivation of hippocampal ensemble memories during sleep. Science 1994, 265:676-679. 14. van de Ven GM, Trouche S, McNamara CG, Allen K, Dupret D: Hippocampal offline reactivation consolidates recently formed cell assembly patterns during sharp wave-ripples. Neuron 2016, 92:968-974. 15. Ji D, Wilson MA: Coordinated memory replay in the visual cortex and hippocampus during sleep. Nat Neurosci 2007, 10:100-107. 16. Olafsdottir HF, Carpenter F, Barry C: Coordinated grid and place cell replay during rest. Nat Neurosci 2016, 19:792-794. 17. Rasch B, Buchel C, Gais S, Born J: Odor cues during slow-wave sleep prompt declarative memory consolidation. Science 2007, 315:1426-1429.
19. Schreiner T, Rasch B: Boosting vocabulary learning by verbal cueing during sleep. Cereb Cortex 2014. bhu139. 20. Hu X, Antony JW, Creery JD, Vargas IM, Bodenhausen GV, Paller KA: Cognitive neuroscience. Unlearning implicit social biases during sleep. Science 2015, 348:1013-1015. 21. Hauner KK, Howard JD, Zelano C, Gottfried JA: Stimulus-specific enhancement of fear extinction during slow-wave sleep. Nat Neurosci 2013, 16:1553-1555. 22. Staresina BP, Bergmann TO, Bonnefond M, van der Meij R, Jensen O, Deuker L, Elger CE, Axmacher N, Fell J: Hierarchical nesting of slow oscillations, spindles and ripples in the human hippocampus during sleep. Nat Neurosci 2015, 18:1679-1686. This study replicated findings of phase coupling between slow oscillations and sleep spindles as well as between sleep spindles and sharpwave/ripples in humans undergoing treatment for epilepsy. 23. Clemens Z, Molle M, Eross L, Barsi P, Halasz P, Born J: Temporal coupling of parahippocampal ripples, sleep spindles and slow oscillations in humans. Brain 2007, 130:2868-2878. 24. Averkin RG, Szemenyei V, Borde S, Tamas G: Identified cellular correlates of neocortical ripple and high-gamma oscillations during spindles of natural sleep. Neuron 2016, 92:916-928. 25. Ngo HV, Martinetz T, Born J, Molle M: Auditory closed-loop stimulation of the sleep slow oscillation enhances memory. Neuron 2013, 78:545-553. By applying closed loop stimulation, that is detecting the sleep slow oscillation and playing a short burst of pink noise during its up-phase, the authors induced further slow oscillations and associated sleep spindle activity. This stimulation enhanced sleep-dependent memory consolidation. 26. Ngo HV, Miedema A, Faude I, Martinetz T, Molle M, Born J: Driving sleep slow oscillations by auditory closed-loop stimulation—a self-limiting process. J Neurosci 2015, 35:6630-6638. 27. Lustenberger C, Boyle MR, Alagapan S, Mellin JM, Vaughn BV, Frohlich F: Feedback-controlled transcranial alternating current stimulation reveals a functional role of sleep spindles in motor memory consolidation. Curr Biol 2016, 26:2127-2136. Current Opinion in Neurobiology 2017, 44:20–27
26 Neurobiology of sleep
28. Girardeau G, Benchenane K, Wiener SI, Buzsaki G, Zugaro MB: Selective suppression of hippocampal ripples impairs spatial memory. Nat Neurosci 2009, 12:1222-1223.
47. Vyazovskiy VV, Cirelli C, Pfister-Genskow M, Faraguna U, Tononi G: Molecular and electrophysiological evidence for net synaptic potentiation in wake and depression in sleep. Nat Neurosci 2008, 11:200-208.
29. Maingret N, Girardeau G, Todorova R, Goutierre M, Zugaro M: Hippocampo-cortical coupling mediates memory consolidation during sleep. Nat Neurosci 2016, 19:959-964.
48. Frank MG: Erasing synapses in sleep: is it time to be SHY? Neural Plast 2012, 2012:264378.
30. Feld GB, Wilhelm I, Ma Y, Groch S, Binkofski F, Molle M, Born J: Slow wave sleep induced by GABA agonist tiagabine fails to benefit memory consolidation. Sleep 2013, 36:1317-1326.
49. Chauvette S, Seigneur J, Timofeev I: Sleep oscillations in the thalamocortical system induce long-term neuronal plasticity. Neuron 2012, 75:1105-1113.
31. Mitra A, Snyder AZ, Hacker CD, Pahwa M, Tagliazucchi E, Laufs H, Leuthardt EC, Raichle ME: Human cortical-hippocampal dialogue in wake and slow-wave sleep. Proc Natl Acad Sci U S A 2016, 113:E6868-E6876. 32. Gais S, Born J: Low acetylcholine during slow-wave sleep is critical for declarative memory consolidation. Proc Natl Acad Sci U S A 2004, 101:2140-2144.
50. Hengen KB, Pacheco AT, McGregor JN, Van Hooser SD, Turrigiano GG: Neuronal firing rate homeostasis is inhibited by sleep and promoted by wake. Cell 2016, 165:180-191. Measuring in the visual cortex of adolescent rats undergoing monocular deprivation, this study shows that firing rate homeostasis, that is the return to a prior set-point firing rate, is differentially affected by sleep and wakefulness. Interestingly, homeostasis was only possible during wakefulness.
33. Plihal W, Born J: Memory consolidation in human sleep depends on inhibition of glucocorticoid release. NeuroReport 1999, 10:2741-2747.
51. Grosmark AD, Mizuseki K, Pastalkova E, Diba K, Buzsaki G: REM sleep reorganizes hippocampal excitability. Neuron 2012, 75:1001-1007.
34. Kelemen E, Bahrendt M, Born J, Inostroza M: Hippocampal corticosterone impairs memory consolidation during sleep but improves consolidation in the wake state. Hippocampus 2014, 24:510-515.
52. Sachser RM, Santana F, Crestani AP, Lunardi P, Pedraza LK, Quillfeldt JA, Hardt O, de Oliveira Alvares L: Forgetting of longterm memory requires activation of NMDA receptors, L-type voltage-dependent Ca2+ channels, and calcineurin. Sci Rep 2016, 6:22771.
35. McNamara CG, Tejero-Cantero A, Trouche S, Campo-Urriza N, Dupret D: Dopaminergic neurons promote hippocampal reactivation and spatial memory persistence. Nat Neurosci 2014, 17:1658-1660. In this study, the authors used an optogenetic approach to stimulate dopaminergic afferents to the hippocampus during learning. Stimulation compared to sham showed a marked increase of replay during subsequent sleep and increased retrieval performance. 36. Lansink CS, Goltstein PM, Lankelma JV, Joosten RN, McNaughton BL, Pennartz CM: Preferential reactivation of motivationally relevant information in the ventral striatum. J Neurosci 2008, 28:6372-6382. 37. Valdes JL, McNaughton BL, Fellous JM: Offline reactivation of experience-dependent neuronal firing patterns in the rat ventral tegmental area. J Neurophysiol 2015, 114:1183-1195. 38. Feld GB, Besedovsky L, Kaida K, Munte TF, Born J: Dopamine D2-like receptor activation wipes out preferential consolidation of high over low reward memories during human sleep. J Cogn Neurosci 2014, 26:2310-2320. 39. Feld GB, Lange T, Gais S, Born J: Sleep-dependent declarative memory consolidation—unaffected after blocking NMDA or AMPA receptors but enhanced by NMDA coagonist Dcycloserine. Neuropsychopharmacology 2013, 38:2688-2697. 40. Diekelmann S, Born J, Wagner U: Sleep enhances false memories depending on general memory performance. Behav Brain Res 2010, 208:425-429.
53. Kuhn M, Wolf E, Maier JG, Mainberger F, Feige B, Schmid H, Burklin J, Maywald S, Mall V, Jung NH et al.: Sleep recalibrates homeostatic and associative synaptic plasticity in the human cortex. Nat Commun 2016, 7:12455. Using transcranial magnetic stimulation, the authors show that cortical excitability is increased while LTP-like plasticity is decreased after sleepdeprivation compared to sleep. 54. Rasch B, Born J: About sleep’s role in memory. Physiol Rev 2013, 93:681-766. 55. Feld GB, Weis PP, Born J: The limited capacity of sleep dependent memory consolidation. Front Psychol 2016, 7. Participants learned word-pair lists of varying length before sleep or wake retention intervals. Only medium list-lengths profited from sleep. This indicates a capacity limit for replay and is the first evidence for sleepinduced forgetting in adult humans. 56. Gomez RL, Bootzin RR, Nadel L: Naps promote abstraction in language-learning infants. Psychol Sci 2006, 17:670-674. 57. Lewis PA, Durrant SJ: Overlapping memory replay during sleep builds cognitive schemata. Trends Cogn Sci 2011, 15:343-351. 58. Westermann J, Lange T, Textor J, Born J: System consolidation during sleep—a common principle underlying psychological and immunological memory formation. Trends Neurosci 2015, 38:585-597.
41. Friedrich M, Wilhelm I, Born J, Friederici AD: Generalization of word meanings during infant sleep. Nat Commun 2015, 6:6004.
59. Fenn KM, Nusbaum HC, Margoliash D: Consolidation during sleep of perceptual learning of spoken language. Nature 2003, 425:614-616.
42. Gomez RL, Edgin JO: Sleep as a window into early neural development: shifts in sleep-dependent learning effects across early childhood. Child Dev Perspect 2015, 9:183-189.
60. Drosopoulos S, Schulze C, Fischer S, Born J: Sleep’s function in the spontaneous recovery and consolidation of memories. J Exp Psychol Gen 2007, 136:169-183.
43. Calvillo DP, Parong JA, Peralta B, Ocampo D, Van Gundy R: Sleep increases susceptibility to the misinformation effect. Appl Cogn Psychol 2016, 30:1061-1067. 44. Durrant SJ, Cairney SA, McDermott C, Lewis PA: Schemaconformant memories are preferentially consolidated during REM sleep. Neurobiol Learn Mem 2015, 122:41-50.
61. Igloi K, Gaggioni G, Sterpenich V, Schwartz S: A nap to recap or how reward regulates hippocampal-prefrontal memory networks during daytime sleep in humans. Elife 2015, 4:e07903. A study showing that rewarded items are remembered better than unrewarded items over a 3 month period only if participants slept after encoding.
45. Hennies N, Lambon Ralph MA, Kempkes M, Cousins JN, Lewis PA: Sleep spindle density predicts the effect of prior knowledge on memory consolidation. J Neurosci 2016, 36:3799-3810.
62. Studte S, Bridger E, Mecklinger A: Sleep spindles during a nap correlate with post sleep memory performance for highly rewarded word-pairs. Brain Lang 2016. pii: S0093-934X(15) 30151-6.
46. Maret S, Faraguna U, Nelson AB, Cirelli C, Tononi G: Sleep and waking modulate spine turnover in the adolescent mouse cortex. Nat Neurosci 2011, 14:1418-1420.
63. Baran B, Daniels D, Spencer RM: Sleep-dependent consolidation of value-based learning. PLoS One 2013, 8: e75326.
Current Opinion in Neurobiology 2017, 44:20–27
www.sciencedirect.com
Sleep-dependent consolidation and forgetting Feld and Born 27
64. Tucker MA, Tang SX, Uzoh A, Morgan A, Stickgold R: To sleep, to strive, or both: how best to optimize memory. PLoS One 2011, 6:e21737.
68. Scho¨nauer M, Geisler T, Gais S: Strengthening procedural memories by reactivation in sleep. J Cogn Neurosci 2014, 26:143-153.
65. Stamm AW, Nguyen ND, Seicol BJ, Fagan A, Oh A, Drumm M, Lundt M, Stickgold R, Wamsley EJ: Negative reinforcement impairs overnight memory consolidation. Learn Mem 2014, 21:591-596.
69. Diekelmann S, Born J: Cueing fear memory during sleep—to extinguish or to enhance fear? Sleep 2015, 38:337-339.
66. Werchan DM, Gomez RL: Generalizing memories over time: sleep and reinforcement facilitate transitive inference. Neurobiol Learn Mem 2013, 100:70-76. 67. Javadi AH, Tolat A, Spiers HJ: Sleep enhances a spatiallymediated generalisation of learned values. Learn Mem 2015, 10:532-536.
www.sciencedirect.com
70. Deese J: On the prediction of occurrence of particular verbal intrusions in immediate recall. J Exp Psychol 1959, 58:17-22. 71. Roediger HL, McDermott KB: Creating false memories— remembering words not presented in lists. J Exp Psychol Learn Mem Cogn 1995, 21:803-814. 72. Kumaran D, Hassabis D, McClelland JL: What learning systems do intelligent agents need? Complementary learning systems theory updated. Trends Cogn Sci 2016, 20:512-534.
Current Opinion in Neurobiology 2017, 44:20–27
ScienceDirect Sculpting memory during sleep: concurrent consolidation and forgetting Gordon B Feld1,2 and Jan Born2,3 There is compelling evidence that sleep actively supports the formation of long-lasting memory representations. Experimental cuing of memories proved that neural replay of representations during sleep plays a causal role for this consolidation, which has also been shown to promote neocortical synaptic plasticity and spine formation. Concurrently, sleep has been proposed to facilitate forgetting through processes of synaptic renormalisation. This view received indirect support by findings in humans of sleep enhancing TMS-evoked plasticity and capabilities for encoding new information. First direct behavioural evidence of sleep inducing forgetting has only recently emerged after encoding large amounts of stimuli in adults. We propose forgetting complements sleep-dependent consolidation and facilitates gist abstraction especially at high memory loads, when reactivation-based consolidation reaches capacity limits.
Addresses 1 Institute of Behavioural Neuroscience, Department of Experimental Psychology, Division of Psychology and Language Science, University College London, London, United Kingdom 2 Institute of Medical Psychology and Behavioral Neurobiology, University of Tu¨bingen, Tu¨bingen, Germany 3 Centre for Integrative Neuroscience, University of Tu¨bingen, Tu¨bingen, Germany Corresponding authors: Feld, Gordon B ([email protected]), Born, Jan ([email protected])
Current Opinion in Neurobiology 2017, 44:20–27 This review comes from a themed issue on Neurobiology of sleep Edited by Yang Dan and Thomas Kilduff
http://dx.doi.org/10.1016/j.conb.2017.02.012 0959-4388/ã 2017 Elsevier Ltd. All rights reserved.
merely passively stored but that an active process of consolidation occurs [1], which has received wide support over the past century [2]. Especially, the idea that memory is consolidated during the brain’s offline periods of deep slow wave sleep has risen to a champion of consolidation theory (Figures 1 and 2) [3–5]. Just recently, in the neocortex, the formation of dendritic spines was observed , during sleep after learning a motor task, as a neural substrate of memory formation, which was positively related to task performance [6]. At its outset, memory research mainly focused on forgetting [7]. Essentially, two forms of forgetting were proposed. The first argues that forgetting occurs due to interference, that is older memory traces are constantly being overwritten by newer encoding events. Consequently, a passive role of sleep protecting from interference has been claimed repeatedly since its first conceptualisation [8], even though this type of forgetting seems to play only a minor role for memories involving the hippocampal system [9]. The second claims that memory traces passively decay over time and this account, later on, was extended to include active decay processes [10]. Here also, sleep has been proposed to play a major role, renormalizing synaptic weights and balancing out potentiation occurring as a result of encoding information during wake (Figure 2) [11]. Covering mainly the period from 2013 to now, we review the latest developments regarding sleep’s role for consolidation and forgetting. Initially, it seems difficult to reconcile these two accounts. To the contrary, building on our previous reasoning [4,12] we will present a framework, derived from novel developments in the field, to explain how consolidation and forgetting work together to sculpt lasting memories from a day’s clay of episodic experiences. Ultimately, forgetting might arise as the fourth process of memory formation that enables the long-term function of the other three processes.
Strengthening memory during sleep Introduction The formation of long-term memory relies on the two distinct processes of encoding (or learning) and consolidation. Retrieval is a third process that contributes to memory formation by re-instating the stored information. Between encoding, that is the uptake of the information, and retrieval lies the period of retention, and already at the beginning of modern memory research it was proposed that during this retention period memories are not Current Opinion in Neurobiology 2017, 44:20–27
The hypothesis that during sleep memory benefits from the repeated replay of neuronal representations that were formed to encode information during prior wakefulness was induced from the initial finding that in rats hippocampal place cells that show correlated firing during wake encoding of a simple maze re-exhibit this correlated firing pattern during subsequent slow wave sleep (SWS) [13] and replay during sleep has recently been shown to predict reinstatement strength during retrieval [14]. This www.sciencedirect.com
Sleep-dependent consolidation and forgetting Feld and Born 21
Figure 1
Sleep REM NonREM1+2 NonREM3+4 (SWS) 23:00
3:00
7:00 h
Slow oscillation Sleep spindle Sharpwave/ ripple Current Opinion in Neurobiology
Sleep stages and oscillations. Sleep and memory research has greatly benefited from the identification of discrete sleep stages and associated oscillatory phenomena. Rapid eye movement (REM) sleep, which occurs mainly during the second half of the night, is characterised by wake-like desynchronized low amplitude EEG, which in the past made it the foremost candidate for sleep-dependent memory processing. To the contrary, reprocessing of memory representations has been shown to occur mostly during deep nonREM sleep (also slow wave sleep—SWS), which is most abundant during the first half of the night and is characterized by high-amplitude lowfrequency oscillations (0.75 Hz slow oscillations) in the surface EEG. NonREM sleep coordinates memory replay by locking hippocampal sharp-wave/ripples, which enwrap the replayed memory information, to the excitable troughs of the thalamo-cortical sleep spindle. The spindle itself is phase-locked to the up-state of the neocortical slow oscillation, thereby ensuring that the reactivated information reaches the neocortex during the excitable up-state of the slow oscillation. This process is supported by the specific neuromodulatory milieu of NonREM sleep that enables information flow from the hippocampus to the neocortex, including low levels of cortisol and acetylcholine.
research was extended to show that sequences of place cells that are replayed in the hippocampus are coordinated with replay of the sensory cortex [15] and of grid cells in the entorhinal cortex [16]. In fact, this replay can be evoked by targeted memory reactivation, that is by representing a cue (e.g. an odour or a sound) that was present during encoding again during SWS after encoding [17]. Correspondingly, in rats re-presenting a sound that predicted the correct choice in a spatial learning paradigm during sleep after learning led to the re-emergence of the associated firing pattern of place cells [18]. In humans, reactivating memory during sleep by re-presenting auditory cues from a Dutch–German paired associate paradigm learned before sleep enhances memory for the cued pairs [19]. In fact, this type of targeted memory reactivation can even be used to reduce notoriously persistent race and gender stereotypes [20] and to enhance the extinction of conditioned fear responses [21]. www.sciencedirect.com
During NonREM sleep an intricate interplay of oscillations, that is neocortical slow oscillations (0.5–1 Hz), sleep spindles (12–16 Hz) originating from thalamus, and hippocampal sharp-wave/ripple complexes, occurs that can be measured with extracranial and intracranial EEG and which is thought to coordinate consolidation. Using intracranial EEG electrodes, in human epileptic patients, it was shown that the neocortical slow oscillation groups sleep spindles to its excitable up-state and that sharpwave/ripples in turn are grouped to the spindle throughs and vice-versa [22,23]. Importantly, high gamma in the cortex has also been shown to be nested in spindle troughs in mice [24]. The efficacy of the slow oscillation to enhance memory consolidation was demonstrated in humans using closed-loop auditory stimulation. In this experiment, short bursts (50 ms) of pink noise were presented during SWS timed to the slow oscillation upstate, which effectively induced high-amplitude slow oscillatory activity with spindle power phase-locked to the up-states [25]. Interestingly, driving slow oscillatory events further by stimulating extended trains did not further enhance sleep spindles, indicating a self-limiting process [26]. Similarly, boosting sleep spindles by transcranial alternating current stimulation can improve motor memory consolidation [27] and electrically blocking ripples during sleep impairs memory consolidation [28]. Importantly, it seems to be the fine-tuned phaselocking of these different frequencies that makes consolidation during NonREM sleep possible [29]. In fact, enhancing slow oscillations pharmacologically by increasing GABAergic tone does not enhance memory consolidation, as simultaneously slow oscillation to sleep spindle phase-amplitude coupling is disrupted [30]. In addition to oscillatory prerequisites, NonREM sleep also offers a unique neurochemical pattern enabling consolidation, as low levels of cortisol and acetylcholine are essential to promote the hippocampal-to-neocortical dialogue [31–34]. Reward related signals also seem to play a major role, as dopaminergic activity during encoding enhances replay during subsequent sleep [35]. Moreover, replay activity during sleep has been shown in reward related areas such as the ventral striatum [36] and the ventral tegmental area [37]. Enhancing dopaminergic neurotransmission pharmacologically during sleep in humans enhances the consolidation of low rewarded items to match high rewarded items [38]. However, it is still unclear, if plastic processes assumed to be induced by neural replay activity during sleep, are equivalent to those active during wakefulness. For example, blocking AMPA or NMDA receptors during sleep does not impact consolidation in humans [39]. It might turn out that during sleep direct electrical coupling between neurons via gap junctions is essential (GB Feld et al., unpublished), which would fit well to the importance of oscillatory coupling for sleep-dependent memory consolidation. Current Opinion in Neurobiology 2017, 44:20–27
22 Neurobiology of sleep
Figure 2
Wakefulness Encoding
Sleep Active systems consolidation
Wakefulness
Synaptic homeostasis
Retrieval and new encoding
Long-term store (Neocortex)
Global synaptic renormalisation Inital store (Hippocampus)
Current Opinion in Neurobiology
Schematic of the classical functions of sleep for memory. In the hippocampus-dependent declarative memory domain, during wakefulness, information (yellow arrows) flows through the sensory cortices to the fast learning system of the hippocampus (green arrows), which quickly binds the information into a transient representation. During subsequent sleep, specifically NonREM and slow wave sleep (SWS), the memory trace is replayed repeatedly (red arrow), which leads to its transformation and integration into the long-lasting stores of the neocortex, that is active systems consolidation (blue arrows). Next to active systems consolidation of relevant memory traces, sleep also renormalizes synapses leading to the erasure of irrelevant information. While the synaptic homeostasis hypothesis underlines the importance of SWS for renormalization in the cortex, evidence suggests that in the hippocampus rapid eye movement (REM) sleep plays a major role. During subsequent wakefulness the consolidated information can be retrieved without recruiting the hippocampus (green upward arrows) and new information can be encoded into its replenished stores (yellow and green downward arrows).
Next to merely strengthening original memory traces during sleep, there is evidence that sleep qualitatively changes the memory trace towards a more generalized representation [40–43]. In fact, information that can be easily integrated into an existing schema, is more likely to benefit from sleep during the retention interval [44], an effect that is mediated by sleep spindles [45].
is evidence for renormalization of synaptic strengths occurring exclusively during the wake state in adolescent rats’ (postnatal day 27–32) visual system after monocular visual deprivation [50]. Although, it has been questioned whether monocular visual deprivation-induced plasticity represents a physiological model of normal learning or a specific form of plasticity that is only readily elicited during a sensitive period.
Forgetting during sleep The synaptic homeostasis hypothesis explains how the brain copes with the physiological demands of learning relying on potentiation of synaptic networks, which would lead to unsustainable demands in energy and space as well as escalating potentiation, if potentiation remained unchecked [11]. There is, indeed, a net reduction in the amount of cortical dendritic spines found in adolescent rats’ (postnatal day 23–44) cortex after sleep, while there is an increase after wake [46]. Likewise the amount of AMPA receptors on glutamatergic neurons is reduced across sleep [47]. However, this account has been contested [48], and it was shown that slow wave sleep (SWS) can enhance potentiation rather than depotentiation [49]. Also, there Current Opinion in Neurobiology 2017, 44:20–27
Essentially, an examination of specific brain areas and sleep stages might prove a more fruitful approach. The hippocampus, for example seems quite reliant on mechanisms of sleep-induced forgetting and here it seems to be rapid eye movement (REM) sleep that renormalizes synaptic activity [51], and such a process might rely on NMDA-receptor-dependent processes of active forgetting [52]. Fittingly, NMDA-receptor activation during sleep improved encoding of new declarative memory after sleep in humans (M Alizadeh-Asfestani et al., unpublished). Similarly, evoked responses to transcranial magnetic stimulation in humans suggested that cortical synaptic plasticity is restored after sleep, which co-occurred with an increase in brain derived neurotropic factor [53]. www.sciencedirect.com
Sleep-dependent consolidation and forgetting Feld and Born 23
Altogether the last years have supplied a vast amount of behavioural support for the role of sleep in consolidating memories (see Ref. [54] for a comprehensive review). However, the notion of sleep promoting forgetting is mainly supported indirectly by neurophysiological evidence and hints from enhanced new learning after sleep, whereas direct behavioural evidence is largely missing. Nonetheless, in a recent laborious experiment we found initial support for sleep-induced forgetting in a behavioural study [55]. Here, participants learned different amounts of word-pairs (40, 160 or 320) and then were either allowed to sleep or stayed awake. As expected, participants showed a strong beneficial effect of sleep on memory retention in the 160 word-pair condition. However, unexpectedly in the 320 word-pair condition the sleep-induced enhancement in memory was nullified, and the sleep group forgot as many items as the wake group (Figure 3). In our view, in a neurophysiological framework, this result is most convincingly explained, if the local processes of potentiation that consolidate memories are assumed to be accompanied by global processes of depression or depotentiation. In this scenario potentiation processes are accomplished via limited resources of neural memory replay that are overstretched at high informational loads leading to a net forgetting. Intriguingly, to our knowledge the only other behavioural study that has shown increased forgetting due to sleep in humans was performed in infants [56]. This likewise argues for forgetting occurring primarily in conditions of informational overload, as young children lack the extensive knowledge networks that allow adults to abstract chunks of information into schemas. Correspondingly, while in that study the infants showed a sleepdependent decline in item memory (i.e. words of an artificial language) they showed increased learning of grammatical rules. Building on the idea that sleep abstracts information because overlapping representations are reactivated more frequently than individual items [57,58], we propose that this kind of gist abstraction occurs at an increased rate, if large amounts of information are encoded, which increases the overlap among representations. We reason that, because available amounts of neural reactivation are limited during sleep, at high information loads, overlap between representations might not be resolved by separating representations [59,60] and therefore, the disparate details of item memory might become subjected to devastating processes of global depotentiation. Concurrently, the relatively increased local reactivation of overlapping representations is preserved as gist (Figure 4). Importantly, this proposition has the power to explain seemingly contradictory results in the field of sleep and memory. For example, although some studies find that www.sciencedirect.com
Figure 3
Experimental design Learning: 40,160 or 320 words-pairs
0 Difference in word-paris recalled at learning and retrival
Interactions between forgetting and consolidation
Immediate recall
Sleep or wake
Delayed recall
Sleep Wake
-2 -4 “Forgetting” -6 -8 -10 Extrapolation 40
160 320 Information load: list length
480
Current Opinion in Neurobiology
Forgetting during sleep induced experimentally by information overload. This first demonstration of sleep-induced forgetting in adult participants used lists of word-pairs with increasing length (40, 160 and 320 word-pairs) to successively enhance information load at encoding. Participants retrieved the word-pairs immediately after learning (immediate recall) and at retrieval (delayed recall). The sleep group (blue) was allowed to sleep for 8 hours after learning and the wake group (green) remained in the lab for eight hours of sleepdeprivation. Retrieval was tested after an additional 24 hours including 8 hours of recovery sleep for the wake group. The difference in wordpairs recalled at retrieval minus the word-pairs recalled at learning was calculated as a measure of retention performance. As predicted, there was a strong beneficial effect of sleep for the 160 word-pair condition. However, contrary to what is predicted if sleep keeps up at least a constant rate of memory enhancement (this prediction is indicated by light blue dashed line), increasing the list length to 320 word-pairs completely nullified the sleep-induced memory enhancement. This significant reduction in retention performance from the 160 to the 320 word-pair condition in the sleep group is indicative of a process of sleep-induced forgetting that can be observed at high information loads. It is tempting to extrapolate the data (dashed blue and green lines), which would predict an advantage of wake over sleep at even higher information loads. In fact, preliminary data from an experiment with a 620 word-pair list length condition point in this direction. Data adapted from Ref. [55].
rewarded items are retained better over sleep than unrewarded items [61,62], there have been contradictory reports [63,64]. According to our proposition, we expect the item enhancing effect of sleep to vanish at high loads of information, even in the presence of rewards. In fact, in a recent study, where reward would reduce the longer it took participants to arrive at a goal in a maze, they Current Opinion in Neurobiology 2017, 44:20–27
24 Neurobiology of sleep
Figure 4
Sleep
Wakefulness Encoding of large amounts of items
Active systems consolidation
snow
Increased gist retrieval & decreased item memory
Synaptic homeostasis
snow
snow
snow
hot
Wakefulness
cold frost
hot
cold frost
hot
frost
hot
frost
Long-term store (Neocortex)
Initial store (Hippocampus)
Global synaptic renormalisation
Current Opinion in Neurobiology
Proposed processing of memory during sleep in conditions of information overload. When large amounts of information are encoded during learning (yellow and green arrows) their hippocampal representations overlap (Here, information is symbolized by the items hot, snow and frost, but of course much more information is encoded). Limited capacity of local replay (red arrow) during subsequent sleep allows only the overlapping information to be processed effectively by active systems consolidation such that the concept ‘cold’ is strengthened within the neocortex (blue arrow). This gist word is shared implicitly by all items, but was not experienced explicitly during learning [70,71]. Without access to stabilizing replay the neocortical traces of the individual items remain labile. Importantly, concurrent global synaptic renormalization nevertheless erases the hippocampal item memory traces, thereby, precluding that these traces aid recall or replay at a later time point. Hence, during subsequent wake retrieval (green arrow) the gist memory ‘cold’ can easily be extracted from the strong neocortical trace, whereas the labile item memory has been forgotten. Note, depending on the amount of encoded information and pre-existing schemas, processes of gist abstraction and forgetting are highly variable and may extend over multiple sleep periods [72].
performed worse after sleep than their peers who received no reward at all [65]. Crucially, the participants of the reward group were constantly monitoring the current reward during learning and perceived the task as significantly more difficult, suggesting that the additional reward signal produced informational overload at encoding. This reasoning is supported by reports of rewards during learning before sleep favouring the generalisation of memory in complex tasks [66,67]. When systematically varying rewards and information load during encoding our model predicts that during subsequent sleep at low loads rewards benefit item memory and at high loads rewards benefit generalisation. Likewise interesting predictions can be made for the paradigm of targeted memory reactivation. Usually, cuing a newly encoded representation during post-learning sleep does not come at the cost of a deterioration of the uncued items [19]. However, at high amounts of encoded information, we expect cueing to deteriorate uncued representations. In fact, some preliminary Current Opinion in Neurobiology 2017, 44:20–27
evidence was provided when in a study participants learned an unusually long finger-tapping sequence and only half of it was cued during sleep, performance for the uncued part was as bad as in the wake condition [68]. Our prediction is also in line with the finding that targeted memory reactivation biases the reactivation of place cells without enhancing overall replay for cued and uncued memories [18]. Interestingly, targeted memory reactivation has been found to lead to the extinction and to the enhancement of fear memories in rodents and human brains, respectively [69]. Thus, understanding the boundary conditions of forgetting and consolidation in terms of perceived informational overload appears to be essential, before implementing any translational approaches.
Concluding remarks The present review aims to reconcile sleep-associated processes of consolidation and forgetting, and highlights their interaction bearing the potential for abstracting gist from large amounts of information. While the causal relationship between behavioural and neurobiological www.sciencedirect.com
Sleep-dependent consolidation and forgetting Feld and Born 25
indicators of sleep-dependent memory consolidation has been investigated in detail during the last years, such a link is completely missing for forgetting during sleep. We argue that sleep-induced forgetting can only be readily measured behaviourally, if participants (humans or animals) are exposed to large amounts of information at encoding, which is not routinely done as it can be extremely laborious. Nevertheless, we believe using informational overload together with the established neurobiological methods to investigate sleep’s function represents a highly promising approach. Such research might identify additional boundary conditions of sleepinduced forgetting, that are systematically interlinked with informational overload, such as age. Importantly, understanding sleep-induced forgetting might also allow us to tackle clinical symptoms that are associated with pathological memories such as PTSD, drug abuse and anxiety by enhancing forgetting of memories that maintain these diseases. Ultimately, it might also open ways to treat pathological forgetting due to, for example Alzheimer’s disease.
10. Hardt O, Nader K, Nadel L: Decay happens: the role of active forgetting in memory. Trends Cogn Sci 2013, 17:111-120.
Conflict of interest statement
18. Bendor D, Wilson MA: Biasing the content of hippocampal replay during sleep. Nat Neurosci 2012, 15:1439-1444. Using targeted memory reactivation in rats implanted for place cell recording this study demonstrated that external cues can elicit replay of specific place cell sequences in the hippocampus. This provides a crucial link to the human literature on cued memory reactiavtion during sleep.
The authors declare no conflicts of interest.
Acknowledgements This work was supported by a grant from the Deutsche Forschungsgemeinschaft (DFG; SFB 654 “Plasticity and Sleep”). GBF is currently receiving a personal stipend from the DFG to conduct research at the University College London (FE 1617/1-1).
References and recommended reading Papers of particular interest, published within the period of review, have been highlighted as: of special interest of outstanding interest 1.
Mu¨ller GE, Pilzecker A: Experimentelle Beitra¨ge zur Lehre vom Geda¨chtnis. Zeitschrift fu¨r Psychologie, Erga¨nzungsband 1900, 1:1-300.
2.
McGaugh JL: Memory—a century of consolidation. Science 2000, 287:248-251.
3.
Feld GB, Diekelmann S: Sleep smart-optimizing sleep for declarative learning and memory. Front Psychol 2015, 6:622.
4.
Diekelmann S, Born J: The memory function of sleep. Nat Rev Neurosci 2010, 11:114-126.
5.
Marr D: Simple memory: a theory for archicortex. Philos Trans R Soc Lond B Biol Sci 1971, 262:23-81.
6.
Yang G, Lai CS, Cichon J, Ma L, Li W, Gan WB: Sleep promotes branch-specific formation of dendritic spines after learning. Science 2014, 344:1173-1178. In this study, the authors applied two photon imaging of cortical neurons to show that denditic spines can be formed during sleep in a performance-dependent manner. This supports an important prediction of active systems consolidation. 7.
Ebbinghaus H: U¨ber das Geda¨chtnis. Leipzig: Duncker und Humblot; 1885.
8.
Jenkins JG, Dallenbach KM: Obliviscence during sleep and waking. Am J Psychol 1924, 35:605-612.
9.
Sadeh T, Ozubko JD, Winocur G, Moscovitch M: Forgetting patterns differentiate between two forms of memory representation. Psychol Sci 2016, 27:810-820.
www.sciencedirect.com
11. Tononi G, Cirelli C: Sleep and the price of plasticity: from synaptic and cellular homeostasis to memory consolidation and integration. Neuron 2014, 81:12-34. This important update to the synaptic homeostasis hypothesis reviews the recent developments and, especially, integrates evidence of sleep’s beneficial effect on memory into the theory. 12. Born J, Feld GB: Sleep to upscale, sleep to downscale: balancing homeostasis and plasticity. Neuron 2012, 75:933935. 13. Wilson MA, McNaughton BL: Reactivation of hippocampal ensemble memories during sleep. Science 1994, 265:676-679. 14. van de Ven GM, Trouche S, McNamara CG, Allen K, Dupret D: Hippocampal offline reactivation consolidates recently formed cell assembly patterns during sharp wave-ripples. Neuron 2016, 92:968-974. 15. Ji D, Wilson MA: Coordinated memory replay in the visual cortex and hippocampus during sleep. Nat Neurosci 2007, 10:100-107. 16. Olafsdottir HF, Carpenter F, Barry C: Coordinated grid and place cell replay during rest. Nat Neurosci 2016, 19:792-794. 17. Rasch B, Buchel C, Gais S, Born J: Odor cues during slow-wave sleep prompt declarative memory consolidation. Science 2007, 315:1426-1429.
19. Schreiner T, Rasch B: Boosting vocabulary learning by verbal cueing during sleep. Cereb Cortex 2014. bhu139. 20. Hu X, Antony JW, Creery JD, Vargas IM, Bodenhausen GV, Paller KA: Cognitive neuroscience. Unlearning implicit social biases during sleep. Science 2015, 348:1013-1015. 21. Hauner KK, Howard JD, Zelano C, Gottfried JA: Stimulus-specific enhancement of fear extinction during slow-wave sleep. Nat Neurosci 2013, 16:1553-1555. 22. Staresina BP, Bergmann TO, Bonnefond M, van der Meij R, Jensen O, Deuker L, Elger CE, Axmacher N, Fell J: Hierarchical nesting of slow oscillations, spindles and ripples in the human hippocampus during sleep. Nat Neurosci 2015, 18:1679-1686. This study replicated findings of phase coupling between slow oscillations and sleep spindles as well as between sleep spindles and sharpwave/ripples in humans undergoing treatment for epilepsy. 23. Clemens Z, Molle M, Eross L, Barsi P, Halasz P, Born J: Temporal coupling of parahippocampal ripples, sleep spindles and slow oscillations in humans. Brain 2007, 130:2868-2878. 24. Averkin RG, Szemenyei V, Borde S, Tamas G: Identified cellular correlates of neocortical ripple and high-gamma oscillations during spindles of natural sleep. Neuron 2016, 92:916-928. 25. Ngo HV, Martinetz T, Born J, Molle M: Auditory closed-loop stimulation of the sleep slow oscillation enhances memory. Neuron 2013, 78:545-553. By applying closed loop stimulation, that is detecting the sleep slow oscillation and playing a short burst of pink noise during its up-phase, the authors induced further slow oscillations and associated sleep spindle activity. This stimulation enhanced sleep-dependent memory consolidation. 26. Ngo HV, Miedema A, Faude I, Martinetz T, Molle M, Born J: Driving sleep slow oscillations by auditory closed-loop stimulation—a self-limiting process. J Neurosci 2015, 35:6630-6638. 27. Lustenberger C, Boyle MR, Alagapan S, Mellin JM, Vaughn BV, Frohlich F: Feedback-controlled transcranial alternating current stimulation reveals a functional role of sleep spindles in motor memory consolidation. Curr Biol 2016, 26:2127-2136. Current Opinion in Neurobiology 2017, 44:20–27
26 Neurobiology of sleep
28. Girardeau G, Benchenane K, Wiener SI, Buzsaki G, Zugaro MB: Selective suppression of hippocampal ripples impairs spatial memory. Nat Neurosci 2009, 12:1222-1223.
47. Vyazovskiy VV, Cirelli C, Pfister-Genskow M, Faraguna U, Tononi G: Molecular and electrophysiological evidence for net synaptic potentiation in wake and depression in sleep. Nat Neurosci 2008, 11:200-208.
29. Maingret N, Girardeau G, Todorova R, Goutierre M, Zugaro M: Hippocampo-cortical coupling mediates memory consolidation during sleep. Nat Neurosci 2016, 19:959-964.
48. Frank MG: Erasing synapses in sleep: is it time to be SHY? Neural Plast 2012, 2012:264378.
30. Feld GB, Wilhelm I, Ma Y, Groch S, Binkofski F, Molle M, Born J: Slow wave sleep induced by GABA agonist tiagabine fails to benefit memory consolidation. Sleep 2013, 36:1317-1326.
49. Chauvette S, Seigneur J, Timofeev I: Sleep oscillations in the thalamocortical system induce long-term neuronal plasticity. Neuron 2012, 75:1105-1113.
31. Mitra A, Snyder AZ, Hacker CD, Pahwa M, Tagliazucchi E, Laufs H, Leuthardt EC, Raichle ME: Human cortical-hippocampal dialogue in wake and slow-wave sleep. Proc Natl Acad Sci U S A 2016, 113:E6868-E6876. 32. Gais S, Born J: Low acetylcholine during slow-wave sleep is critical for declarative memory consolidation. Proc Natl Acad Sci U S A 2004, 101:2140-2144.
50. Hengen KB, Pacheco AT, McGregor JN, Van Hooser SD, Turrigiano GG: Neuronal firing rate homeostasis is inhibited by sleep and promoted by wake. Cell 2016, 165:180-191. Measuring in the visual cortex of adolescent rats undergoing monocular deprivation, this study shows that firing rate homeostasis, that is the return to a prior set-point firing rate, is differentially affected by sleep and wakefulness. Interestingly, homeostasis was only possible during wakefulness.
33. Plihal W, Born J: Memory consolidation in human sleep depends on inhibition of glucocorticoid release. NeuroReport 1999, 10:2741-2747.
51. Grosmark AD, Mizuseki K, Pastalkova E, Diba K, Buzsaki G: REM sleep reorganizes hippocampal excitability. Neuron 2012, 75:1001-1007.
34. Kelemen E, Bahrendt M, Born J, Inostroza M: Hippocampal corticosterone impairs memory consolidation during sleep but improves consolidation in the wake state. Hippocampus 2014, 24:510-515.
52. Sachser RM, Santana F, Crestani AP, Lunardi P, Pedraza LK, Quillfeldt JA, Hardt O, de Oliveira Alvares L: Forgetting of longterm memory requires activation of NMDA receptors, L-type voltage-dependent Ca2+ channels, and calcineurin. Sci Rep 2016, 6:22771.
35. McNamara CG, Tejero-Cantero A, Trouche S, Campo-Urriza N, Dupret D: Dopaminergic neurons promote hippocampal reactivation and spatial memory persistence. Nat Neurosci 2014, 17:1658-1660. In this study, the authors used an optogenetic approach to stimulate dopaminergic afferents to the hippocampus during learning. Stimulation compared to sham showed a marked increase of replay during subsequent sleep and increased retrieval performance. 36. Lansink CS, Goltstein PM, Lankelma JV, Joosten RN, McNaughton BL, Pennartz CM: Preferential reactivation of motivationally relevant information in the ventral striatum. J Neurosci 2008, 28:6372-6382. 37. Valdes JL, McNaughton BL, Fellous JM: Offline reactivation of experience-dependent neuronal firing patterns in the rat ventral tegmental area. J Neurophysiol 2015, 114:1183-1195. 38. Feld GB, Besedovsky L, Kaida K, Munte TF, Born J: Dopamine D2-like receptor activation wipes out preferential consolidation of high over low reward memories during human sleep. J Cogn Neurosci 2014, 26:2310-2320. 39. Feld GB, Lange T, Gais S, Born J: Sleep-dependent declarative memory consolidation—unaffected after blocking NMDA or AMPA receptors but enhanced by NMDA coagonist Dcycloserine. Neuropsychopharmacology 2013, 38:2688-2697. 40. Diekelmann S, Born J, Wagner U: Sleep enhances false memories depending on general memory performance. Behav Brain Res 2010, 208:425-429.
53. Kuhn M, Wolf E, Maier JG, Mainberger F, Feige B, Schmid H, Burklin J, Maywald S, Mall V, Jung NH et al.: Sleep recalibrates homeostatic and associative synaptic plasticity in the human cortex. Nat Commun 2016, 7:12455. Using transcranial magnetic stimulation, the authors show that cortical excitability is increased while LTP-like plasticity is decreased after sleepdeprivation compared to sleep. 54. Rasch B, Born J: About sleep’s role in memory. Physiol Rev 2013, 93:681-766. 55. Feld GB, Weis PP, Born J: The limited capacity of sleep dependent memory consolidation. Front Psychol 2016, 7. Participants learned word-pair lists of varying length before sleep or wake retention intervals. Only medium list-lengths profited from sleep. This indicates a capacity limit for replay and is the first evidence for sleepinduced forgetting in adult humans. 56. Gomez RL, Bootzin RR, Nadel L: Naps promote abstraction in language-learning infants. Psychol Sci 2006, 17:670-674. 57. Lewis PA, Durrant SJ: Overlapping memory replay during sleep builds cognitive schemata. Trends Cogn Sci 2011, 15:343-351. 58. Westermann J, Lange T, Textor J, Born J: System consolidation during sleep—a common principle underlying psychological and immunological memory formation. Trends Neurosci 2015, 38:585-597.
41. Friedrich M, Wilhelm I, Born J, Friederici AD: Generalization of word meanings during infant sleep. Nat Commun 2015, 6:6004.
59. Fenn KM, Nusbaum HC, Margoliash D: Consolidation during sleep of perceptual learning of spoken language. Nature 2003, 425:614-616.
42. Gomez RL, Edgin JO: Sleep as a window into early neural development: shifts in sleep-dependent learning effects across early childhood. Child Dev Perspect 2015, 9:183-189.
60. Drosopoulos S, Schulze C, Fischer S, Born J: Sleep’s function in the spontaneous recovery and consolidation of memories. J Exp Psychol Gen 2007, 136:169-183.
43. Calvillo DP, Parong JA, Peralta B, Ocampo D, Van Gundy R: Sleep increases susceptibility to the misinformation effect. Appl Cogn Psychol 2016, 30:1061-1067. 44. Durrant SJ, Cairney SA, McDermott C, Lewis PA: Schemaconformant memories are preferentially consolidated during REM sleep. Neurobiol Learn Mem 2015, 122:41-50.
61. Igloi K, Gaggioni G, Sterpenich V, Schwartz S: A nap to recap or how reward regulates hippocampal-prefrontal memory networks during daytime sleep in humans. Elife 2015, 4:e07903. A study showing that rewarded items are remembered better than unrewarded items over a 3 month period only if participants slept after encoding.
45. Hennies N, Lambon Ralph MA, Kempkes M, Cousins JN, Lewis PA: Sleep spindle density predicts the effect of prior knowledge on memory consolidation. J Neurosci 2016, 36:3799-3810.
62. Studte S, Bridger E, Mecklinger A: Sleep spindles during a nap correlate with post sleep memory performance for highly rewarded word-pairs. Brain Lang 2016. pii: S0093-934X(15) 30151-6.
46. Maret S, Faraguna U, Nelson AB, Cirelli C, Tononi G: Sleep and waking modulate spine turnover in the adolescent mouse cortex. Nat Neurosci 2011, 14:1418-1420.
63. Baran B, Daniels D, Spencer RM: Sleep-dependent consolidation of value-based learning. PLoS One 2013, 8: e75326.
Current Opinion in Neurobiology 2017, 44:20–27
www.sciencedirect.com
Sleep-dependent consolidation and forgetting Feld and Born 27
64. Tucker MA, Tang SX, Uzoh A, Morgan A, Stickgold R: To sleep, to strive, or both: how best to optimize memory. PLoS One 2011, 6:e21737.
68. Scho¨nauer M, Geisler T, Gais S: Strengthening procedural memories by reactivation in sleep. J Cogn Neurosci 2014, 26:143-153.
65. Stamm AW, Nguyen ND, Seicol BJ, Fagan A, Oh A, Drumm M, Lundt M, Stickgold R, Wamsley EJ: Negative reinforcement impairs overnight memory consolidation. Learn Mem 2014, 21:591-596.
69. Diekelmann S, Born J: Cueing fear memory during sleep—to extinguish or to enhance fear? Sleep 2015, 38:337-339.
66. Werchan DM, Gomez RL: Generalizing memories over time: sleep and reinforcement facilitate transitive inference. Neurobiol Learn Mem 2013, 100:70-76. 67. Javadi AH, Tolat A, Spiers HJ: Sleep enhances a spatiallymediated generalisation of learned values. Learn Mem 2015, 10:532-536.
www.sciencedirect.com
70. Deese J: On the prediction of occurrence of particular verbal intrusions in immediate recall. J Exp Psychol 1959, 58:17-22. 71. Roediger HL, McDermott KB: Creating false memories— remembering words not presented in lists. J Exp Psychol Learn Mem Cogn 1995, 21:803-814. 72. Kumaran D, Hassabis D, McClelland JL: What learning systems do intelligent agents need? Complementary learning systems theory updated. Trends Cogn Sci 2016, 20:512-534.
Current Opinion in Neurobiology 2017, 44:20–27