sign in sign up
<<
Home , Spatial memory

Sleep and spatial consolidation: A role for replay in hippocampal assemblies?

Bachelor thesis by S.M.C. Bezema Supervised by prof. dr. D.G.M. Beersma Department of Chronobiology University of Groningen July, 2013 1

Abstract

In this thesis, the role of in spatial is investigated. The two stage model of memory consolidation predicts that acquired during the waking period are rapidly encoded by the , and consolidated (i.e. strengthened and integrated) throughout the sleeping period. The consolidation process is thought to involve the reactivation of recently encoded memory traces in the hippocampus during slow-wave sleep (SWS), a sleep stage in which highly synchronous activity patterns termed sharp-wave ripples (SWRs) are most prominently observed. These SWR events might provide the optimal conditions for neuronal plasticity and communication between the hippocampus and other areas to take place, thereby facilitating the consolidation process. SWR- related reactivation has been observed in place cell assemblies in the CA3 and CA1 areas of the hippocampus. Throughout SWS, the firing patterns observed during spatial navigation and exploration in place cell assemblies are sequentially reactivated in a temporally conserved manner during SWRs. This reactivation is thought to reflect the “replay” of recently encoded memories, and occurs in a compressed timescale, which puts the synaptic activity during replay within the timeframe of long- term potentiation. Thus, replay provides a possible neural mechanism for memory consolidation during sleep. However, SWR-associated replay is also observed during extended periods of inactivity during waking (i.e. “quiet wakefulness”), and it is accordingly thought that these instances of replay reflect consolidation as well. This suggests that consolidation of spatial memories is an opportunistic process that takes place whenever the brain is not primarily preoccupied with new memories. While this opportunistic account of consolidation is not directly at odds with the two-stage model of memory consolidation, it does provide an important addition to the current theoretical framework, and deepens our understanding of the consolidation process and the role for sleep in spatial memory consolidation.

(Picture on front page: EEG recording of slow-wave sleep, adapted from http://emedicine.medscape.com/article/1140322- overview#aw2aab6b5) 2

Contents

1. Introduction ...... 3 2. Core concepts ...... 4 2.1 Sleep ...... 4 2.2 Memory ...... 6 2.2.1 Memory stages and processes ...... 6 2.2.2 Memory systems ...... 6 2.2.3 The two-stage model of memory consolidation ...... 7 3. The brain’s spatial representation system ...... 8 3.1 Cognitive maps in the hippocampus ...... 8 3.2 Neural correlates of the spatial navigation system ...... 9 3.2.1 Place cells & the hippocampal map ...... 9 3.2.2 Other cells & entorhinal maps ...... 10 4. Spatial memory and sleep: offline consolidation? ...... 11 4.1 Encoding and of spatial memories ...... 12 4.2 Replay in place cell assemblies during sleep: consolidation at work? ...... 13 4.2.1 Replay during slow wave sleep ...... 15 4.2.2 Reactivation during REM sleep ...... 16 4.3 Replay during waking: opportunistic consolidation? ...... 16 5. Discussion & Conclusion ...... 17 References ...... 19

3

1. Introduction

Sleep is a very widespread phenomenon, thought to occur in all vertebrates including mammals, birds, reptiles and fish. Moreover, sleeplike states have also been observed in invertebrates such as bees, cockroaches and flies (Cirelli & Tononi, 2008). This suggests that sleep serves some fundamental functions that were important enough to be sustained by natural selection throughout species and across phyla. Yet, to this day these functions have remained unclear (Rasch & Born, 2013). Although sleep is not fully understood, it is known that it is a crucially important behaviour (Cirelli & Tononi, 2008). Even moderate sleep deprivation has a negative impact on neurogenesis (Mueller et al., 2008), neural physiology (McDermott et al., 2003), and immune functions (Imeri & Opp, 2009), while extreme sleep deprivation can even lead to death (Rechtschaffen & Bergmann, 1995). Sleep loss also appears to impair the storage of recently acquired memories and interferes with the ability to acquire new ones (Hairston et al., 2005). Hence, it has been proposed that one of the important functions of sleep may be the consolidation of memories. An influential model of (mammalian) memory consolidation, the two-stage model of memory consolidation, suggests that the hippocampus rapidly encodes and stores memories during the active waking period. Then, during sleep, the newly encoded memory traces are “reactivated” by the hippocampus, with the purpose of transferring them to a long-term store in the neocortex, and (if appropriate) integrate them into pre- existing memory traces in different brain regions (e.g. Buzsáki, 1998; Marr, 1971; Squire & Alvarez, 1995). The reactivation of recent memories in the hippocampus is thought to occur primarily during slow-wave sleep (SWS), a sleep stage in which highly synchronous activity patterns termed sharp-wave ripples (SWRs) are most prominently observed (O’Neill et al., 2010). These SWR events might provide the optimal conditions for neuronal plasticity and communication between the hippocampus and other brain areas to take place. Although aspects of the two-stage model of memory consolidation remain unsettled and continue to be investigated, converging lines of evidence from suggests that consolidation of spatial memory is at least consistent with this explanation for SWS (O’Neill et al., 2006; Rattenborg et al., 2011). The rodent hippocampus is known to be involved in spatial navigation and memory (Derdikman & Moser, 2010), and the use of single-unit recording techniques in rodents during exploration and foraging has demonstrated that specific cells in the hippocampus named “place cells” code for specific locations in the environment (John O’Keefe & Dostrovsky, 1971). Moreover, these place cells indeed appear to reactivate during subsequent SWS, in a temporally conserved manner (Skaggs & McNaughton, 1996). This sequential reactivation is referred to as “replay”, and is thought to reflect the neural mechanism for memory consolidation during sleep (Buhry et al., 2011). There is now accumulating evidence for these hippocampal place cells in humans, other primates, and even bats (Burgess et al., 2002), which suggests that the neural mechanisms underlying spatial memory are conserved throughout different species of mammals. Obviously, evolving a reliable spatial navigation and memory system must have been of extreme relevance for survival, as successfully remembering where for example water, food, mates and nesting grounds are located is a fundamental requirement of everyday life. Therefore, the possibility that sleep may be necessary for the consolidation of spatial memories is exceptionally interesting. Additionally, spatial cognition is a field of research that is particularly well suited for cross-species research, since experimental (spatial) paradigms can easily be applied to animals and humans alike. As such, (invasive) neuroscientific discoveries in animals can be combined with non-invasive findings in humans (Burgess et al., 2002). Thus, the aim of this thesis is to investigate the role of sleep in spatial memory consolidation. To this end, we will discuss the brain’s spatial navigation system, the way spatial memories might be encoded and stored within the brain, and the reactivation and replay of place cells during sleep. Guiding questions will be whether replay during slow-wave sleep might reflect spatial memory consolidation, and if insights from spatial memory and sleep research are in line with the two-stage model of memory consolidation. 4

2. Core concepts

This thesis is dedicated to exploring the relation between sleep and the consolidation of spatial memories. However, before we can get to the brain’s spatial representation system and the relation between spatial memory and sleep (in chapters 3 and 4 respectively), a brief introduction to some core concepts of is in order. The goal of this chapter is not to give an exhaustive overview, but rather to provide the reader with a list of some basic concepts that will figure throughout this thesis.

2.1 Sleep Sleep can be defined as a natural and reversible mental and physiological state of reduced consciousness and responsiveness in which volitional sensorimotor activity is deferred and metabolism is altered (Dudai, 2012; Steriade & McCarley, 2005). Sleep takes place in recurring intervals and is homeostatically regulated (Beersma, 1998; Borbély & Achermann, 1999). In mammals, sleep consists of cyclic phases that are defined by their characteristic brain activity, tonic muscle activity and coordinated eye movements. Two core sleep stages can be discerned for mammals: rapid eye movement (REM) sleep, and non-REM (NREM) sleep. In human nocturnal sleep, NREM and REM normally alternate in a cyclic ultradian pattern of roughly 90 minutes. NREM in humans is further divided into substages (N1-3) that reflect the depth of sleep. The deepest sleep occurs in NREM stage N3 (previously stages 3 and 4), and is alternatively referred to as slow-wave sleep (SWS) because of the prevalence of slow high-amplitude electroencephalographic (EEG) oscillations (Dudai, 2012). SWS is predominant in the first half of the sleep period and decreases in both duration and intensity as the night progresses, while REM sleep is more predominant towards the end of the sleep period (fig. 1A) (Rasch & Born, 2013). As the descent into NREM starts, brain activity begins to slow down. Throughout stage 2 NREM, phasic electrical events are present, such as K-complexes (large sharp waves in the EEG), and sleep spindles (short synchronized waxing and waning waveform oscillations in the EEG) (fig. 1B). Stage 2 NREM marks around 50% of adult human sleep, but is not discriminated from SWS (stage 3) in all mammals, such as rodents. SWS is characterized by slow-wave activity (SWA), which is activity in the EEG within the delta range (0.5-4 Hz) and below the delta range (<1 Hz). SWA is thought to be a reflection of underlying cortical synchrony (Walker, 2005). During SWS, both spindles and transient sharp-wave ripples (SWR) occur in the hippocampus (Diekelmann & Born, 2010; Dudai, 2004; Rasch & Born, 2013; Walker, 2005). With the onset of REM sleep, EEG oscillations desynchronize again, as wakelike high frequency and low-amplitude oscillatory brain activity in the gamma range (30-80 Hz) starts to take place. In addition, muscle tone drops significantly compared to wake and NREM sleep, and the typical phasic rapid horizontal eye movements occur (fig. 1B) (Walker, 2005). In cats and rats, ponto-geniculo- occipital (PGO) waves and theta oscillations (4-8 Hz) are hallmarks of REM sleep, but these are not readily distinguished in humans (Rasch & Born, 2013). In addition to the occurrence of the specific patterns of electrical field potential oscillations in NREM and REM sleep, there are also dramatic changes in the neurochemical balance of the brain during the sleep cycle. For example, activity of aminergic and cholinergic neurons are significantly reduced during NREM compared to waking. During REM sleep, activity of aminergic neurons drops even more, reaching a minimum, while levels of cholinergic activity are similar to those in the awake state (fig. 1C) (Rasch & Born, 2013; Walker, 2005) 5

Figure 1 – Overview of a typical human sleep profile (A), sleep-related field potential oscillations (B) and changes in activity levels of neurotransmitters and neuromodulators (C). (A): A schematic representation of the cyclic occurrence of rapid eye movement (REM) sleep and non-REM (NREM) sleep. NREM sleep comprises the deepest sleep stage, slow-wave sleep (SWS) or N3, and the lighter stages N1 and N2. The beginning of sleep is dominated by SWS, whereas REM sleep is more pervasive in the second half of the sleeping period. (B): Particular oscillations in the electrical field potential. SWS is characterized by slow oscillations (~ 0.8 Hz), spindles (10-15 Hz), and hippocampal sharp wave-ripples (SWR, 100-300 Hz). In animals, ponto-geniculo- occipital (PGO) waves, and hippocampal theta activity (4–8 Hz) can be observed, but these are less readily identified in humans. (C): Throughout the sleeping period, dramatic changes can be observed in the neurochemical balance of the brain. Cholinergic activity is greatly reduced in SWS compared to waking, while levels during REM are comparable to those during waking. Aminergic activity levels are attenuated during SWS compared to waking, and reach a minimum during REM sleep. (Adapted from Rasch & Born, 2013)

6

2.2 Memory

2.2.1 Memory stages and processes The ability to form and retrieve memories is a fundamental capability for organisms, as it allows for the appropriate selection and improvement of behaviours from an established repertoire, and the contextualization of new experiences in a demanding and ever-changing environment. Classically, the memory process is thought to encompass three different stages, i.e. encoding, consolidation, and retrieval. A memory is encoded when the perception of stimuli leads to the formation of a new memory trace. Initially, this trace is very vulnerable to decay (i.e. ) or (retroactive) interference. Thus, it is thought that the newly formed memory trace is gradually stabilized and integrated into a network of pre-existing knowledge during consolidation. During retrieval, the stored memory is accessed and recalled (Rasch & Born, 2013). Müller and Pilzecker first proposed the term consolidation in 1900 (see Lechner et al., 1999). Since then, the consolidation hypothesis has received support from numerous studies showing that memory can be effectively enhanced or impaired after encoding by electrophysiological, pharmacological, and psychological manipulations (see McGaugh, 2000; Wixted, 2004). The effects of these manipulations are time dependent, in the sense that their effects are strongest when they are applied directly after (for reviews see Lechner et al., 1999; Squire & Schacter, 2002). Consolidation is thought to involve multiple waves of stabilizing processes, that work on different timescales, and which are dependent on different underlying neuronal plasticity processes. Recently, it is being suggested that memory traces are not consolidated just once, but are in actuality reconsolidated whenever they are reactivated by a cue or active retrieval (Rasch & Born, 2013; also see Dudai, 2012; Nader & Hardt, 2009). This has led some to claim that “retrieval” is actually a misleading term, since it implies that a memory is a fixed entity that can be accessed in its entirety. Especially memories of events embedded in the context of time and space (“episodic memories”, see paragraph 2.2.2) appear to be stored in a distributed fashion over different brain systems according to the qualitative nature of the information they represent (Nadel et al., 2012). Thus, it has been argued that involves a “reconstruction” of such memories, rather than accessing a fixed entity. On the neuronal level, the memory formation process is believed to depend on changes in the strength of synaptic connections in the network that represents the memory trace (Rasch & Born, 2013). When a new memory is encoded, two major kinds of learning-induced synaptic plasticity can be brought about, namely synaptic long-term depression (LTD) and long-term potentiation (LTP) (Collingridge et al. 2010; Kandel, 2001). Reverberating activity in a newly encoded network is thought to promote two types of consolidation processes, named “synaptic consolidation” and “systems consolidation” (Dudai, 2012). Synaptic consolidation reflects the cellular and synaptic level of description, and refers to the remodelling of the spines and synapses of the neurons that are part of the newly encoded memory trace. Ultimately, this produces lasting changes in the efficacy of the contributing synapses (Kandel, 2001; Rasch & Born, 2013). Systems consolidation builds on synaptic consolidation and reflects the brain systems level of description. It refers to the reorganization of the neural memory representations over distributed brain circuits for long-term storage, possibly through reactivation of previously encoded memory traces (Dudai, 2012; Rasch & Born, 2013).

2.2.2 Memory systems Memory is not a single entity, or at least it is not in humans. The word memory is a catch-all term for several different systems, that are discriminated based on the involvement of particular neuro- anatomical regions. Currently, the most widely held taxonomy is based on the distinction between declarative and non-declarative memory. For declarative memory, it is thought that the medial temporal lobe regions, and in particular the hippocampal formations, are critically involved in the acquisition of declarative types of memories. In contrast, the formation of non-declarative memories is not believed to be dependent on these regions (Walker, 2005). 7

Declarative memory is the (mostly) conscious memory for fact-based information, and is further subdivided into 1) semantic memories for non-contextual facts and 2) episodic memories for events embedded in a spatio-temporal context, including autobiographical memories (Rasch & Born, 2013). In general, declarative memories can be encoded either unintentionally or intentionally, and are usually consciously accessible by active attempts of remembering. Semantic memories are most often the result of either repeated encoding efforts, or activation of multiple corresponding episodic memories. Episodic memories are learned very quickly - in a single trial (i.e. “one-shot” encoding), but are also forgotten more rapidly (Rasch & Born, 2013). It has been shown that the integrity of the hippocampal circuitry is a necessary precondition for the ability to retain semantic and episodic memories for longer periods (usually more than 15 minutes) (Peigneux et al., 2001). As mentioned, non-declarative memories are not dependent on medial temporal lobe regions, and instead rely on different regions of the brain. The non-declarative memory system encompasses several different subsystems, such as 1) for perceptual skills (involvement of sensory cortices) and motor skills and habits (striatum, cerebellum, motor areas); 2) implicit learning/ (neocortex); 3) classical conditioning (cerebellum, amygdala); and 4) nonassociative learning (reflex pathways). Non-declarative memories are usually implicitly learned and recalled, and often require multiple training trials (Rasch & Born, 2013; Squire & Zola, 1996). Lastly, it is noteworthy that more often than not, the disentanglement of the different memory systems and their involvement in certain tasks is complicated by their interaction during encoding and recall (Rasch & Born, 2013). Many of our daily tasks (if any) do not neatly separate into either declarative or non-declarative categories, and this might be reflected in the way the brain processes such experiences and commits them to memory. The taxonomy of the different memory systems is useful and aids comprehension, but time will tell whether such sharp conceptual distinctions actually “carve nature at its joints”.

2.2.3 The two-stage model of memory consolidation So far, research on sleep and memory have been discussed separately - but where do they intersect? The connection between sleep and memory goes back quite some time (see Diekelmann & Born, 2010), and is conceptually rooted in the standard two-stage model of memory consolidation, which is still the most influential model of (human) memory (Rasch & Born, 2013). Originally proposed by Marr (1971), the two-stage memory formation mechanism was offered as a solution to a number of problems that arose from models used to study the formation of memories on the network level. Essentially, the question that researches faced was how a neuronal network could learn new patterns without forgetting older ones at the same time. In other words, how is it possible that new memories are acquired without causing “catastrophic interference” with respect to the older memories? Additionally, these early models also faced essential capacity constraints, reflecting the limited capacity of actual neuronal networks. The two-stage model of memory consolidation provided a solution; it holds that memories are first encoded into a fast-learning store, and are subsequently progressively transferred to a slow-learning store for long-term storage (Rasch & Born, 2013). In the brain, it is suggested that the hippocampus functions as the former, rapidly and efficiently encoding memories, even in a single attempt (i.e. one-shot learning). But as these memory traces in the hippocampus are unstable and susceptible to interference from newly encoded memories, they must be gradually transferred to the neocortex (i.e. the slow-learning long-term store) without overwriting older memories. Whether or not memory retrieval remains associated with or assisted by the activation of the hippocampus remains a matter of debate (O’Neill et al., 2010). Returning to the connection of sleep and memory, it is postulated that sleep provides the optimal state for systems consolidation as described by the two-stage memory model. During the awake state, the brain is bombarded with new information and experiences that must be committed to memory, but the retrieval of old memories is also needed. Accordingly, it is hypothesized that the brain in the waking state is primarily geared to rapidly encoding new memories and remembering others. Encoding and consolidation are thought to depend on overlapping neuronal resources in the 8 hippocampus, and as such, it may be that they are (at least to some extend) mutually exclusive processes. If this is the case, then a state of greatly reduced external input (i.e. sleep) might be the optimal time window for consolidation, as the new and recently activated memory traces are relatively well protected from interferences from outside stimuli. Thus, it is thought that the repeated reactivation of newly encoded memories during offline periods such as sleep trains the neocortex’s long-term store, and new memory traces are slowly strengthened and integrated in or adapted to pre- existing long-term memories (Buzsáki, 1989; Rasch & Born, 2013; Walker, 2005). The two-stage model of memory consolidation is supported by lesion studies: lesions of the hippocampus severely compromise the ability to acquire new declarative memories, while at the same time producing a temporally graded retrograde in which older memories remain intact (Corkin, 2002; Frankland & Bontempi, 2005). We will return to the relation between sleep and (systems) consolidation of memory in chapter 4, when the reactivation of place cells during off-line periods is discussed. But first, we turn to the spatial representation system in the brain.

3. The brain’s spatial representation system

The hippocampus is critically involved in declarative type memories such as episodic memories in humans. As such, damage to these structures results in profound memory problems. At the same time, there is little dispute that the hippocampal formation in non-human animals (specifically rodents) serves spatial navigation and is accordingly also involved in spatial memory, and this function seems to be preserved in humans (see Burgess et al., 2002 and articles reviewed therein; Derdikman & Moser, 2010). As we will see in paragraph 3.1, these findings are in accordance with the theory of hippocampal function. In the remainder of this chapter, the way the brain is thought to represent space will be discussed. Specifically, we will focus on place cells in the hippocampus and other neural correlates of the spatial navigation system in paragraph 3.2. However, in the paragraph that follows, we will first turn towards the theory that the hippocampus supports a cognitive map of the environment.

3.1 Cognitive maps in the hippocampus Nearly 70 years ago, Tolman suggested that animals might form internal cognitive maps of their spatial environment (Derdikman & Moser, 2010). Some time after that, O’Keefe and Dostrovsky (1971) discovered “place cells” – cells in the rat hippocampus that fire whenever the animal is at a specific location in its environment. O’Keefe and Nadel (1978) subsequently expanded these observations into a large-scale theory, arguing that the hippocampus was the locus of the Tolmanian cognitive map. Essentially, they proposed that the specific firing pattern of a place cell represents the animal’s location within an overarching internal map of the surroundings, with event and object information linked to particular locations in the map. This theory for the most part has held up to scrutiny, although important additions have been made (but see Eichenbaum, 1999 for a critique). Recent studies indicate that space is likely represented in several brain systems rather than just the hippocampal formation, with each of these systems hosting different types of representations of space, and involving functionally specific cell types (Derdikman & Moser, 2010). As such, it is now suggested that space might be represented in these structures by a number of rapidly interacting maps that are generated in conjunction, by specific cell types such as “grid cells” and the aforementioned place cells in the hippocampus. Briefly, before examining the neural basis of spatial cognition and memory in greater detail, it should be clarified that there are two different ways to relate to space. Firstly, in egocentric representation, the body is used as the frame of reference in order to relate to the environment (e.g. “the museum is located to my right”). These self-referenced first person representations of space are used, for example, in episodic memories. In contrast, allocentric representations are space-based and 9 are thus guided by landmarks and cues in the immediate environment (i.e. “the museum is located near the bell tower”) (Buzsáki, 2006). These concepts will figure in our discussion of spatial navigation in the brain.

3.2 Neural correlates of the spatial navigation system

3.2.1 Place cells & the hippocampal map Place cells are thought to be pyramidal neurons in the CA3 and CA1 areas of the hippocampus (Moser et al., 2008). Place cells exhibit a high firing rate whenever the head of the rat or mouse is in a certain location within its (enclosed) environment, referred to as the cell’s “place field”. Each place cell establishes its own specific place field, so that the entire environment is represented in the activity of the local cell population within the hippocampus (fig. 2) (O’Keefe, 1976; Wilson & McNaughton, 1994). When the animal is placed in a new environment, the place field of each place cell will be established within minutes, and the resulting representation of space usually remains stable for days. When transferred to another environment, a new map is formed (again in minutes) in a process termed “remapping” (see for instance Colegin et al., 2008). When the experimental environment is somehow altered (e.g. elongated etc.), remapping also takes place (Derdikman & Moser, 2010). The same place cell can participate in the representation of different environments, but the relationship between the firing fields differs from one setting to the next (John O’Keefe & Conway, 1978; Pavlides & Winson, 1989). There is no evident topography amongst the firing fields of place cells, but the brain is (probably) able to determine the position of the animal by reading out the sequential activity of a local population of place cells (Redish et al., 2001). In experiments where the activity of a large number of cells is recorded, the rat’s movement can be reconstructed by the experimenters with remarkable accuracy (Wilson & McNaughton, 1993). According to Derdikman and Moser (2010), this, taken together with the remapping evidence, indicates that the population of place cells forms a spatial map of the environment. Further evidence for the existence of spatial maps comes from the discovery of replay, i.e. the reactivation of place cells in their preserved temporal order during offline periods such as sleep. We will return to this in chapter 4. The firing of place cells seems to depend primarily on visual distal cues in the environment (Knierim et al., 1995), including boundaries (Burgess, 2008). However, it is that place cell firing persists in the dark, suggesting that other sensory modalities (e.g. , etc.) contribute as well (Derdikman & Moser, 2010; Knierim et al., 1995). At any rate, place cells are thought to provide the animal with a continuously updating, dynamic, allocentric (i.e. world based) map of the environment, and the animal’s position in that space (Moser et al., 2008). There is now accumulating evidence for hippocampal place cells (and enthorinal grid cells, which we will discuss next) in a number of different mammalian species besides rodents, including primates (Rolls, 1999), humans (Ekstrom et al., 2003), and bats (Ulanovsky & Moss, 2007). This suggests that the hippocampus plays a key role in spatial representation and memory in a broad(er) evolutionary context. The cortical map theory described in the previous paragraph is currently very influential, but it must be noted that there are other theories concerning hippocampal function (see Bird & Burgess, 2008), and that there is some debate as to whether the (human) hippocampal formation is truly dedicated to cortical maps (Eichenbaum et al., 1999; Mackintosh, 2002). Obviously, the human hippocampal formation has evolved to process more than just space, and spatial memory and the relation between episodic and spatial memory remains unclear. It is worth mentioning at this point that place cells have been observed to convey more than spatial information alone. For example, they also respond to the temporal sequence of events, and conjunctions between olfactory and textural information, suggesting that place cells are able to link location to a raft of salient experiences in the environment. This capacity could be valuable for encoding in the hippocampus, but future research is needed to establish the relation between episodic and spatial memory (Derdikman & Moser, 2010). 10

Figure 2 – Schematic illustration of hippocampal place cell activation during exploration. The top panel shows the location of the place fields of four hippocampal place cells (A-D) in a rectangular environment. The animal first visited the location corresponding to the place field of place cell D, followed by C, B, and A, respectively. Note that the place fields of cells A and B overlap. The bottom panels on the left shows raster plot for the same four place cells during a sequence of exploration. The order of firing of the four place cells matches the animals movement through the environment. The bottom panel on the left indicates the firing probability of each place cell. ( Adapted from O’Neill et al., 2010)

3.2.2 Other cells & entorhinal maps The brain contains a second class of spatially modulated neurons, located in the medial (MEC) (Hafting et al., 2005), although they also appear in the presubiculum and the parasubiculum (Boccara et al., 2010). These cells exhibit sharply tuned spatial firing, similar to place cells, but differ in the sense that each cell has multiple firing fields. The many fields of each cell collectively form a hexagonal grid that tessellates the entire space available to the animal (Hafting et al., 2005), hence their name “grid cells” (fig. 3A & B). According to Moser et al. (2008), grid cells can differ from each other in their grid spacing (distance between grid fields), grid orientation (tilt relative to an external reference axis), and grid phase (xy displacement relative to an external reference point). Much like place cells, grid cells can be used to reconstruct the position of the animal in its environment (Fyhn et al., 2004), thus possibly functioning as a map. Additionally, grid cells also appear to be sensitive to extrinsic cues, like place cells (John O’Keefe & Conway, 1978). Grid cells are believed to constitute a source of input to place cells, and might pass on metric information to allow place cells to accurately position their place fields in space (Jeffery & Burgess, 2006). The aforementioned parahippocampal regions that contain grid cells also accommodate two other types of cells thought to be relevant for spatial mapping, namely “head-direction cells” and “border cells”. Head-direction cells only fire when the animal’s head is facing a specific direction. There are different head-direction cells for different allocentric orientations, and all of these directions are equally represented in the cell population (Derdikman & Moser, 2010). Head-direction cells may influence grid field orientation (Moser et al., 2008). Border cells fire when the animal is in the vicinity of a boundary or border in the proximal environment (Solstad et al., 2008). Taken together, Derdikman and Moser (2010) suggest, these different cell types could be part of a metric navigation system that is able to map distances through grid cells, directions via head-direction cells, and vicinity to boundaries with border cells. 11

Figure 3 - Two plots of activity in a spatial environment. (A): Example of an entorhinal grid cell’s firing pattern. The animal’s course through a rectangular environment is plotted (indicated by a black line), while the firing location of one grid cell is indicated by a red dot. Firing occurs at different locations throughout the environment, and is arranged in a regular hexagonal grid across the environment. (B): Example of three neighbouring grid cells that were simultaneously recorded. The firing location of each cell is shown in either red, blue or green. (Adapted from Burgess, 2008)

Additionally, it has been shown that animals are also able to navigate (simple) environments in an egocentric manner. Apparently, animals are able to keep track of their changing position by monitoring their own movements in a process referred to as “” or “dead reckoning”. The landmark-based system and the self-referenced system are not mutually exclusive, and are likely to interact. As of yet however, it remains a matter of speculation as to how the different spatial map- like systems exactly interact, and a review of the possibilities is beyond the scope of this thesis (see for instance Burgess, 2008; Derdikman & Moser, 2010; Moser et al., 2008; also see Buzsáki, 2006). At any rate, research has made it clear that spatial representation in the brain is an incredibly intricate and highly dynamical affair. It likely involves the interaction and integration of different allocentric and egocentric mechanisms.

4. Spatial memory and sleep: offline consolidation?

We have seen that place cells, together with grid cells, head direction cells, and border cells seem to form an allocentric map of the environment as the animal moves through it. The system is highly dynamic, as remapping seems to take place whenever the animal enters into different surroundings. This chapter will be devoted to the link between this spatial representation system and sleep-related consolidation of spatial memories. We will see in paragraph 4.2 that the activation patterns of place cells that are observed during waking reoccur during subsequent sleep in a conserved temporal order, in what appears to be neural underpinnings of the consolidation process at work. However, there is also evidence pointing towards a role for consolidation during periods of reduced input, called quiet wakefulness. This possibly undermines the assumed privileged role for sleep in memory consolidation, as we will see in paragraph 4.3. But before we can explore the link between sleep and spatial memory 12 consolidation, we must first discuss how spatial memories seem to be encoded and stored within the hippocampus. This will be the topic of the next paragraph.

4.1 Encoding and storage of spatial memories

There are several computational models of spatial memory formation in the hippocampus (see Burgess & O’Keefe, 2011 for an overview). One dominant class of computer models seems to suggest that online encoding of spatial (and episodic) memory traces is facilitated by the firing of spatially and temporally proximate place cell units through the promotion of plasticity between sequentially activated cell units (Rolls, 2010; Treves & Rolls, 1994). Others focus less on interactions between network cells and emphasize mechanisms within single place cells (Burgess & O’Keefe, 2011). At any rate, it is currently thought that over time, and possibly in multiple consolidation waves, individual memory traces are integrated into an abstract and cohesive representation of the spatial environment. Together with grid cells, head direction cells, and border cells in the entorhinal cortex, spatial maps in the hippocampus could be used for a wide range of navigational functions (Derdikman & Moser, 2010; Sadowski et al., 2011). Interestingly, periods during which the animal is engaged in spatial exploration of an environment, the (rodent) hippocampus displays robust modulation of network activity within the theta frequency range. As an animal moves through a place field, the firing of place cells will change with respect to the phase of the theta cycle (fig. 4). This “theta phase precession” is accordingly believed to provide a phase code for location that is independent from the firing rate code of place cells (O’Keefe & Recce, 1993) We have seen in paragraphs 3.2 that the latter also codes for location, but may additionally be modulated by other (external) factors such as odours, objects or actions that occur at the location of the place field (Huxter et al., 2003). Sadowski et al. (2011) write that due to the degree of overlap between place fields, the sequential firing pattern of multiple place cells can be compressed into a single theta cycle for any given location, while their relative firing order is maintained. In line with this is the observation that theta power increases as a function of running speed. It thus appears that theta modulation couples place cell firing across time and space. This in turn is thought to allow for synaptic plasticity to take place during periods of exploration, providing the animal with an opportunity to learn about its surroundings (Sadowski et al., 2011). Theta phase precession has also been reported for grid cells in the entorhinal cortex (Düzel et al., 2010). As for the storage of spatial memories in the (rodent) hippocampus, the retention of the spatial and temporal organisation of the firing patterns during waking would provide the basis of a memory trace. The CA3 region is a possible storage for such traces, because it possesses highly modifiable synapses and intense interconnectivity, which would allow jointly firing cells to form associations through adjustments in synaptic weights (O’Neill et al., 2010). Theoretical work has pointed towards a possible mechanism for storing memories in the hippocampus, namely “attractor dynamics”. According to Jeffery (2011), the attractor dynamics hypothesis was based on previous ideas that the CA3 network operates like an auto-associative memory store. Attractor networks form a special kind of (artificial) auto-associative memory, and their defining characteristic is their tendency to move towards and settle into a stable pattern of activity. This process is referred to as attractor dynamics, and it is thought to provide a possible mechanism for the storage of (arbitrary) input patterns in the hippocampus with low interference (Jeffery, 2011). In attractor networks, the recall of a stored memory entry from a degraded version of the original input is possible, in a process called pattern completion. At the same time, guarding this recalled memory from interference by other memories stored in the network is done by pattern separation (Moser et al., 2008). Attractor dynamics as a mechanism for storing memories in the hippocampus has recently received some empirical support (Wills et al., 2005). 13

Figure 4 – Schematic illustration of the proposed means of encoding of spatial trajectories in the place cell system in the hippocampus. This figure shows place cells in the hippocampus firing at their respective place fields. As the animal moves through an environment, the place fields fire in a sequential pattern (represented by place A, B and C), exhibiting partial or entire place field overlap. At the same time, the hippocampal networks are submitted to strong theta modulation. The overlapping place fields fire within the same theta cycle, which is thought to establish the conditions necessary for synaptic plasticity to occur. (Adapted from Sadowski et al., 2011)

4.2 Replay in place cell assemblies during sleep: consolidation at work?

It is here that we (finally) turn to the relation between sleep and (systems) consolidation of spatial memories. In paragraph 2.2.3, the two-stage model of memory consolidation was discussed as a currently influential model of memory. Because encoding and consolidation are thought to depend on overlapping neuronal resources in the hippocampus, a period of reduced interference in which no new memories are encoded is hypothesized to be beneficial for the consolidation process. Sleep, a state of greatly reduced external input, may thus provide the optimal time window for consolidation, as reactivation of newly encoded memory traces could take place. This repeated reactivation could train the neocortex’s long-term store, and enable the strengthening and integration of new representations into pre-existing long-term memories. The first evidence for the reactivation of place cells during sleep came from a study by Pavlides and Winson (1989). By recording the spiking activity from pairs of cells with non-overlapping place fields, and not allowing rats to enter the place field of one of those cells, they showed that the ensuing firing pattern during the waking period reoccurred during SWS and REM sleep. Subsequently, Wilson and McNaughton (1994) recorded the activity of CA1 place cells with overlapping place fields, which showed correlated activity while the rat ran along a track in order to acquire a food reward. In the sleep period that followed, the correlation pattern of the recorded place cell assemblies during SWS was remarkably similar to the pattern that had been observed during the awake task performance. Importantly, this similar firing pattern was not observed in the sleep period prior to running along the track. Some years later, it was estimated that the activity pattern observed during task performance explained between 10-30% of the variance of spiking activity during the succeeding sleeping period (Kudrimoti et al., 1999). Further studies by Skaggs and McNaughton (1996) showed that cells in the recorded assembly were not only reactivated during sleep, but that the temporal order in which the place cells had been activated during waking was also preserved (fig. 5). In fact, entire spike trains that were reliably observed during the performance of a task re-emerged in the same temporal order during subsequent sleep (e.g. Lee & Wilson, 2002; Nádasdy et al., 1999; O’Neill et al., 2008). This “replay” of previous activity appears in a time compressed form, running at a much faster rate than during the online state, and remains a reliably observed phenomenon. 14

Initial studies that demonstrated the reactivation of place cell assembly firing patterns during sleep were criticized because the animals were exceedingly over-trained in the experimental spatial tasks (Rasch & Born, 2013; Tononi & Cirelli, 2006). This meant that when the rats frequently performed stereotypical behaviours in an environment they were well acquainted with, the activity patterns of pre-training sleep became increasingly similar to post-sleep periods, thus skewing the data. This discussion drove several researchers to examine the reactivation patterns of rats during sleep after exploratory behaviour in new environments. These later studies established that reactivation occurs during sleep following exploration, with the neurons linked to more thoroughly explored (i.e. longer or more frequently) place fields showing stronger reactivation during subsequent sleep (e.g. O’Neill et al., 2008; O’Neill et al., 2006). Many studies have since confirmed replay in hippocampal place cells. A possible mechanism for replay during sleep is pattern completion in the CA3 region (see paragraph 3.1). According to O’Neill et al. (2010), if a number of place cells have formed a memory trace by jointly firing and thus forming associations through alterations in synaptic weights, the spontaneous activation of only a subset of these cells (“initiator cells”) would tend to recruit the whole group. Because of the strong recurrent connection in CA3, the entire pattern would be recalled in a rehearsal-like reactivation, potentially aiding the (re)consolidation of memory traces.

Figure 5 –Schematic illustration of replay in hippocampal place cells during sleep before and after navigating a spatial track. The top panel shows the firing probability of six hippocampal place cells (A-F) as a function of the location of the animal’s position on the track. The bottom panels show two raster plots of the firing times of the same six cells during sharp-wave ripples (SWRs) before and after the track running. Note that in the sleeping period before track running, the place cells fire during the SWRs in an order that is not related to the firing patterns during subsequent track running. However, in the sleeping period following track running, the order of cell firing during SWRs does reflects the order in which the cells fired during track running, indicating the occurrence of replay. (Adapted from O’Neill et al., 2010)

15

4.2.1 Replay during slow wave sleep Replay in place cell assemblies is typically most prominent during the beginning of sleep (the first 20-40 minutes), decaying over time (Kudrimoti et al., 1999). The early period of sleep is dominated by non-REM slow-wave sleep (SWS) and this is no coincidence, as replay of hippocampal place cell assemblies is mostly observed in association with sharp-wave ripples (SWR), which occur mainly during SWS (Kudrimoti et al., 1999; O’Neill et al., 2008). It is suggested that these SWR events reflect communication between the hippocampus and cortex (Buzsáki, 1989). SWR events are thought to coordinate the sporadic bursts of synchronous firing in the CA3 and CA1 regions. SWRs are likely initiated within the CA3 region (Csicsvari et al., 2000), which according to O’Neill et al. (2010) would mean that ensemble activity during these events is principally governed by the established synaptic weights within the CA3 area, and between the CA3 and CA1 regions. As such, the patterns that occur during SWR could well be the result of the pattern completion mechanism, and facilitate consolidation of stored spatial memories. Indeed, ripple density has been associated with learning experience (Eschenko et al., 2008). Other evidence for the contribution of SWR-related reactivation comes from research in which auditory or olfactory cues are administered with visual stimuli during the awake learning phase. When these cues are presented during subsequent SWS, they selectively strengthen memory for the visual stimuli (Rasch et al., 2007; Rudoy et al., 2009). Furthermore, when hippocampal SWR events were blocked by electrical stimulation of afferents from the hippocampus (85% of which occurred during SWS), the acquisition of spatial memories was impaired (Girardeau et al., 2009). In another study, hippocampal CA3 output was temporarily blocked using a genetic approach in transgenic mice. This strongly decreased the incidence of SWR events, and accordingly a reduction in hippocampal reactivation activity was observed. During subsequent testing in the context of a fear-conditioning task, the mice showed diminished memory retrieval performance (Nakashiba et al., 2009). We have already seen that spiking sequences during waking movement occur in a compressed timescale as a result of phase precession (see paragraph 4.1). During ripple events in SWS, firing sequences happen at a greatly accelerated timescale as well, about 20 times faster than the speed at which an animal moves through a series of place fields (Nádasdy et al., 1999). According to Sadowski et al. (2011), this puts the synaptic activity during replay within the timeframe of long-term potentiation (LTP), thus providing the right circumstances for plasticity to take place and memory traces to be strengthened and/or transferred (i.e. cellular and synaptic consolidation). However, Sadowski et al. also note that replay during SWS may have at least one notable disadvantage, namely the attenuated levels of cholinergic tone in this sleep stage. During waking, it hast been shown that place cell activity does not induce plasticity in the absence of a cholinergic tone. Whether or not the temporal compression observed during ripple-associated replay is sufficient for inducing plasticity during SWS remains to be seen (see Sadowski et al., 2011 for a discussion). The instances of reactivation during SWS discussed above occur in the hippocampus, but replay has also been observed in several other brain regions including the , the ventral striatum and the primary and secondary visual cortices (V1 and V2) (Buhry et al., 2011). The two-stage memory model predicts that the hippocampus will lead in replay, reflecting the transfer of memory traces to a long-term neocortical store. As it stands, it is not yet clear if replay in these other brain areas indeed reflects the transfer of memories to a long-term (neocortical) store. In line with the two- stage memory hypothesis are the observations of hippocampal replay preceding replay in the ventral striatum and prefrontal cortex. However, there is also evidence pointing to an opposite flow of information, from the neocortex to the hippocampus. For instance, distinct periods of high spiking activity in V1 during SWS precede high activity states in the hippocampus, and they appear to be involved in the generation of SWR events. Additionally, synchronous slow oscillations (“up-down states”) in the prefrontal cortex are suggested to influence the probability of replay events in the hippocampus (see Buhry et al., 2011 and articles reviewed therein). These experimental observations lead Buhry et al. (2011) to conclude that the interactions between the hippocampus and other (neocortical) brain regions during consolidation are more complex than is suggested by the simplified 16 view of the two-stage memory model. Rather, they write, interactions between the hippocampus and other areas might be bidirectional, and the direction of the flow of information might change dynamically throughout the consolidation process (Buhry et al., 2011). At any rate though, the reviewed evidence above (strongly) suggests a role for slow-wave sleep in the offline consolidation of spatial memories, by promoting replay in place cell assemblies.

4.2.2 Reactivation during REM sleep For the most part, research has focussed on hippocampal place cell assembly reactivation and replay in slow-wave sleep, with the role of REM sleep often overlooked. While this is reflected in the focus of this thesis, it must be noted that reactivation in the hippocampus is sometimes observed during rapid eye movement (REM) sleep (e.g. Louie & Wilson, 2001). However, while temporal compression of firing sequences does occur in REM sleep, it is much less than the compression observed in SWS. Moreover, there is little evidence pointing towards an active role for REM sleep in the consolidation of long-term spatial memories (Siegel, 2001). Nevertheless, it has been suggested that REM sleep reactivation might serve other functions in memory consolidation, including facilitating stabilization of memory traces at the synaptic level (Diekelmann & Born, 2010) and/or the processing and integration of memory traces for fluent recall during conscious thought and activity in humans (Sadowski et al., 2011).

4.3 Replay during waking: opportunistic consolidation?

The findings that spatial memory benefits from slow-wave sleep (SWS) and that there is currently no strong evidence for a role for REM sleep seem to suggest that SWS is crucial for the consolidation of spatial memories, perhaps through neural mechanisms unique to this sleep stage (such as SWR related replay). This is in line with the two-stage model of memory consolidation. However, SWR events have also been reported to occur when the animal is awake, both during extended periods of immobility (referred to as quiet wakefulness), as well as during brief pauses in locomotion while the animal explores or otherwise navigates an environment (Foster & Wilson, 2006; O’Neill et al., 2006). SWR events in the latter (exploratory) context, the so-called “eSWR” events, have been correlated with the reactivation of place cell assemblies, and also with subsequent spatial memory performance (Dupret et al., 2010). Sadowski et al. (2011) write that although it can be argued that eSWR events transpire during offline periods as the animal pauses, their presence in the active and awake state blurs the boundaries between offline and online processing. It must be noted however, that eSWR events as of yet have only been linked to the reactivation of specific place cells assemblies, and that this reactivation does not necessarily reflect the replay of extended sequences. As such, it is thought that eSWR events could enable spatial processing functions in the hippocampus by establishing and/or remapping existing place cell assemblies instead of stimulating the consolidation of specific (spatial) memory traces (Dupret et al., 2010a). It is noteworthy that this reactivation and/or reorganisation of place cell assemblies has only been observed in CA1, and not in CA3, suggesting a dissociation in the functions of these areas, at least during waking activity (Dupret et al., 2010b). As mentioned above, online replay does occur during extended periods of quiet wakefulness. While it is certainly possible that this reflects some aspect of information processing such as decision- making or behaviour-reward associations (Sadowski et al., 2011), it has also been suggested that these instances of replay assist consolidation in much the same way as SWS does. Accordingly, Mednick et al. (2011) have recently proposed that the processes that underlie (systems) consolidation might not be tied to sleep specifically, but instead could unfold whenever the hippocampus is not currently preoccupied with the encoding of new memories. Potentially, any period of reduced input (and thus reduced retroactive interference) could be used for (replay-related) consolidation; this includes SWS, but is not necessarily limited to it. Findings in humans seem to underpin this opportunistic account of consolidation. A period of quiet waking, in which encoding and interference are reduced, has been shown to have comparable behavioural enhancing effects as sleep. For instance, the learning profiles of the quiet wake and nap groups in a study using a hippocampus dependent visual search task were 17 very similar, while the active wake group showed reduced learning (Greene, 2007). In other studies, memory gain was comparable for both the sleep group and the quiet waking group (see Mednick et al., 2011). In a study using blood oxygen level-dependent (BOLD) fMRI, test subjects showed enhanced hippocampal-cortical connectivity during a period of quiet wakefulness after a learning task compared to a resting baseline (Tambini et al., 2010). Additionally, individual differences in post-task connectivity were shown to be predictive of subsequent memory performance (Gottselig et al., 2004). The occurrence of SWR-associated replay during quiet wakefulness, taken together with the other evidence discussed above suggest that the consolidation process of (spatial) memories might not depend on SWS per se, but may benefit from any (extended) period of reduced interference in which no new memories need to be encoded. This period of reduced interference is suggested to actively allow the consolidation process to proceed, and must occur within a limited temporal window after a new memory is acquired (Mednick et al., 2011).

5. Discussion & Conclusion

The aim of this thesis was to investigate the role of sleep in spatial memory consolidation. To this end, we have discussed the brain’s spatial navigation system, the way spatial memories might be encoded and stored within the brain, and the reactivation and replay of hippocampal place cells during sleep. In summary, the findings reviewed in this thesis suggest a spatial navigation system which allows for the rapid encoding of salient spatial information in the CA3 and CA1 areas of the hippocampus. We have seen that place cells in these areas appear to code for the position of the animal in space by firing whenever the animal enters into a specific spatial field (i.e. a place cell’s place field). When an animal runs along a path, its place cells fire in a specific sequence, and this specific sequence is preserved when the animal runs down the same path at different times. This indicates that the firing pattern is related to the memory of spatial environment, and the formation of a dynamic, allocentric and continuously updating spatial map. Theta phase precession provides an additional phase code for location, and other cells that contribute to spatial navigation are grid cells, head- direction cells, and border cells. Together, these different cell types could provide the animal with a metric navigation system that is able to store paths and map distances, directions, and boundaries in the local environment. Current theoretical models point towards a role for attractor dynamics as a means to store and recall spatial memories without interference in the place cell system of the hippocampus. Attractor dynamics also provide a possible explanation for the reactivation of place cell sequences during sleep. Place cell assemblies that are activated in an animal’s hippocampus during the performance of some kind of spatial navigation or exploration task reactivate in a conserved temporal order during the slow-wave sleep (SWS) stage in the subsequent sleeping period. This repeated reactivation takes place in association with hippocampal sharp-wave ripples (SWRs), and is thought to reflect the replay of newly encoded memories. During SWR events, firing sequences in place cell assemblies occur at greatly accelerated timescales, putting synaptic activity during replay within the timeframe of long-term potentiation. Thus, replay is believed to reflect the consolidation process, as it may enable the strengthening and integration of newly encoded memories into pre-existing long- term memories in the neocortex’s long-term store without overwriting older memories. The two guiding questions throughout this thesis have been (1) whether replay during slow- wave sleep might reflect spatial memory consolidation, and (2) if insights from spatial memory and sleep research are in line with the two-stage model of memory consolidation. Although this thesis by no means captures all the available evidence, we can at least answer the first question with a tentative “yes”. The evidence reviewed above suggests that replay of place cells during slow-wave sleep may indeed reflect at least some aspect of the consolidation process of spatial memories. The finding that spatial memory benefits from SWS points to an important role for this sleep stage in the consolidation of spatial memories, as is predicted by the two-stage model of memory consolidation. So far, the answer to the second question also appears to be positive. However, as we have seen, sharp-wave 18 ripple associated replay does not occur during SWS alone. Instead, replay is also observed during extended periods of immobility (i.e. quiet wakefulness), and it is accordingly thought that these instances of replay reflect (or assist) consolidation as well. As such, it is suggested that the consolidation of spatial memories does not depend on SWS specifically, but may rather unfold whenever the hippocampus is not engrossed with the encoding of new information (Mednick et al., 2011). This opportunistic account is not directly at odds with the two-stage model of memory consolidation, as quiet wakefulness can be interpreted as an extended period of “downtime” in which input is reduced and interference is minimal. It does however provide and important addition to the current theoretical framework, and deepens our understanding of the consolidation process and the role for sleep in spatial memory consolidation. Concluding, while much research still needs to be done (some of the remaining questions are listed in box 1), the theoretical and empirical insights discussed in this thesis point to an active role for sleep in the consolidation of spatial memories. However, an active role does not imply a privileged role: emerging evidence suggest that consolidation is an opportunistic process, taking place whenever the brain is not primarily preoccupied with encoding new memories. Whether or not there are salient differences between replay during waking and replay during sleep remains a question for future research.

Box 1 - Open questions in sleep and spatial memory research

- Is there a difference between the believed consolidation processes during sleep and during quiet wakefulness? Are there different roles for systems and synaptic consolidations within these processes? - For the consolidation of spatial memories, is it the stage of sleep (i.e. SWS) that is important or SWR events across waking and sleeping? Why are SWRs during SWS associated with replay, but eSWR during short pauses in locomotion not? What is the mechanism that triggers SWR during sleep or quiet wakefulness? - To what extent and how do the mechanisms of systems consolidation (i.e. replay) and synaptic consolidation (i.e. long term potentiation) interact? Are they intimately connected and do they occur in a specific temporal order? - What exactly is the relation between spatial memory in rodents and episodic memory in humans? To what extent can we truly speak of a “map”-function, especially in humans and primates? - Are memories truly stored in the hippocampus through attractor dynamics? What about long term storage in the neocortex? - Does the hippocampus remain involved when spatial memories are remembered? Is there a difference between species in the involvement of the hippocampus? - What is the relation between reactivation in the hippocampus and reactivation in other (neocortical) areas? Can LTP be readily induced in the cortex during SWS? What about under conditions of quiet wakefulness?

19

References

Beersma, D. G. M. (1998). Models of human sleep Buzsáki, G. (2006). Rhythms of the Brain. Oxford regulation. Sleep medicine reviews, 2(1), 31– University Press. 43. doi:10.1093/acprof:oso/9780195301069.001 .0001 Bird, C. M., & Burgess, N. (2008). The hippocampus and memory: insights from spatial Cirelli, C., & Tononi, G. (2008). Is sleep essential? processing. Nature reviews. , PLoS biology, 6(8), e216. 9(3), 182–94. doi:10.1038/nrn2335 doi:10.1371/journal.pbio.0060216

Boccara, C. N., Sargolini, F., Thoresen, V. H., Colgin, L. L., Moser, E. I., & Moser, M.-B. (2008). Solstad, T., Witter, M. P., Moser, E. I., & Understanding memory through Moser, M.-B. (2010). Grid cells in pre- and hippocampal remapping. Trends in parasubiculum. Nature neuroscience, 13(8), , 31(9), 469–77. 987–94. doi:10.1038/nn.2602 doi:10.1016/j.tins.2008.06.008

Borbély, A. A., & Achermann, P. (1999). Sleep Collingridge, G. L., Peineau, S., Howland, J. G., & homeostasis and models of sleep Wang, Y. T. (2010). Long-term depression in regulation.pdf. Journal of Biological Rhythms, the CNS. Nature reviews. Neuroscience, 14(6), 559–568. 11(7), 459–73. doi:10.1038/nrn2867

Buhry, L., Azizi, A. H., & Cheng, S. (2011). Corkin, S. (2002). What’s new with the amnesic Reactivation, replay, and preplay: how it patient H.M.? Nature reviews. Neuroscience, might all fit together. Neural plasticity, 2011, 3(2), 153–60. doi:10.1038/nrn726 1–11. doi:10.1155/2011/203462 Csicsvari, J., Hirase, H., Mamiya, A., & Buzsáki, G. Burgess, N. (2008). Spatial cognition and the brain. (2000). Ensemble patterns of hippocampal Annals of the New York Academy of Sciences, CA3-CA1 neurons during sharp wave- 1124, 77–97. doi:10.1196/annals.1440.002 associated population events. Neuron, 28(2), 585–94. Retrieved from Burgess, N., Maguire, E. a, & O’Keefe, J. (2002). The http://www.ncbi.nlm.nih.gov/pubmed/1114 human hippocampus and spatial and 4366 episodic memory. Neuron, 35(4), 625–41. Retrieved from Derdikman, D., & Moser, E. I. (2010). A manifold of http://www.ncbi.nlm.nih.gov/pubmed/1219 spatial maps in the brain. Trends in cognitive 4864 sciences, 14(12), 561–9. doi:10.1016/j.tics.2010.09.004 Burgess, N., & O’Keefe, J. (2011). Models of place and grid cell firing and theta rhythmicity. Diekelmann, S., & Born, J. (2010). The memory Current opinion in neurobiology, 21(5), 734– function of sleep. Nature reviews. 44. doi:10.1016/j.conb.2011.07.002 Neuroscience, 11(2), 114–26. doi:10.1038/nrn2762 Buzsáki, G. (1998). Memory consolidation during sleep: a neurophysiological perspective. Dudai, Y. (2004). The neurosciences: the danger Journal of sleep research, 7 Suppl 1, 17–23. that we will think that we have understood it Retrieved from all. In S. Rose & R. Dees (Eds.), The new brain http://www.ncbi.nlm.nih.gov/pubmed/9682 sciences: perils and prospects (pp. 167–180). 189 Cambridge, UK: Cambridge University Press.

Buzsáki, G. (1989). Two-stage model of memory Dudai, Y. (2012). The restless : trace formation: a role for “noisy” brain consolidations never end. Annual review of states. Neuroscience, 31(3), 551–570. neuroscience, 35, 227–47. doi:10.1146/annurev-neuro-062111-150500 20

Dupret, D., O’Neill, J., Pleydell-Bouverie, B., & suppression of hippocampal ripples impairs Csicsvari, J. (2010a). The reorganization and spatial memory. Nature neuroscience, 12(10), reactivation of hippocampal maps predict 1222–3. doi:10.1038/nn.2384 spatial memory performance. Nature Neuroscience, 13(8), 995–1002. Gottselig, J. M., Hofer-Tinguely, G., Borbély, a a, doi:10.1038/nn.2599.The Regel, S. J., Landolt, H.-P., Rétey, J. V, & Achermann, P. (2004). Sleep and rest Dupret, D., Pleydell-Bouverie, B., & Csicsvari, J. facilitate auditory learning. Neuroscience, (2010b). Rate remapping: when the code 127(3), 557–561. goes beyond space. Neuron, 68(6), 1015–6. doi:10.1016/j.neuroscience.2004.05.053 doi:10.1016/j.neuron.2010.12.011 Greene, A. J. (2007). Human Hippocampal- Düzel, E., Penny, W. D., & Burgess, N. (2010). Brain Dependent Tasks : Is Awareness Necessary or oscillations and memory. Current opinion in Sufficient ? Hippocampus, 17, 429–433. neurobiology, 20(2), 143–9. doi:10.1002/hipo doi:10.1016/j.conb.2010.01.004 Hafting, T., Fyhn, M., Molden, S., Moser, M.-B., & Eichenbaum, H., Dudchenko, P., Wood, E., Shapiro, Moser, E. I. (2005). Microstructure of a M., & Tanila, H. (1999). The hippocampus, spatial map in the entorhinal cortex. Nature, memory, and place cells: is it spatial memory 436(7052), 801–6. doi:10.1038/nature03721 or a memory space? Neuron, 23(2), 209–26. Retrieved from Huxter, J. R., Burgess, N., & O’Keefe, J. (2003). http://www.ncbi.nlm.nih.gov/pubmed/1039 Independent rate and temporal coding in 9928 hippocampal pyramidal cells. Nature, 425(6960), 828–832. Ekstrom, A. D., Kahana, M. J., Caplan, J. B., Fields, T. A., Isham, E. A., Newman, E. L., & Fried, I. Jeffery, K. J. (2011). Place cells, grid cells, (2003). Cellular networks underlying human attractors, and remapping. Neural plasticity, spatial navigation. Nature, 425, 184–188. 2011, 182602. doi:10.1155/2011/182602 doi:10.1038/nature01955.1. Jeffery, K. J., & Burgess, N. (2006). A metric for the Eschenko, O., Ramadan, W., Mölle, M., Born, J., & cognitive map: found at last? Trends in Sara, S. J. (2008). Sustained increase in cognitive sciences, 10(1), 1–3. hippocampal sharp-wave ripple activity doi:10.1016/j.tics.2005.11.002 during slow-wave sleep after learning. Learning & Memory, 15, 222–228. Kandel, E. R. (2001). The molecular biology of doi:10.1101/lm.726008.15 memory storage: a dialogue between genes and synapses. Science (New York, N.Y.), Foster, D. J., & Wilson, M. a. (2006). Reverse replay 294(5544), 1030–8. of behavioural sequences in hippocampal doi:10.1126/science.1067020 place cells during the awake state. Nature, 440(7084), 680–3. doi:10.1038/nature04587 Knierim, J. J., Kudrimoti, H. S., & Mcnaughton, B. L. (1995). Place Cells, Head Direction Cells, and Frankland, P. W., & Bontempi, B. (2005). The the Learning of Landmark Stability. The organization of recent and remote Journal of neuroscience : the official journal memories. Nature reviews. Neuroscience, of the Society for Neuroscience, 15(3), 1648– 6(2), 119–30. doi:10.1038/nrn1607 1659.

Fyhn, M., Molden, S., Witter, M. P., Moser, E. I., & Kudrimoti, H. S., Barnes, C. a, & McNaughton, B. L. Moser, M.-B. (2004). Spatial representation (1999). Reactivation of hippocampal cell in the entorhinal cortex. Science (New York, assemblies: effects of behavioral state, N.Y.), 305(5688), 1258–64. experience, and EEG dynamics. The Journal doi:10.1126/science.1099901 of neuroscience : the official journal of the Society for Neuroscience, 19(10), 4090–101. Girardeau, G., Benchenane, K., Wiener, S. I., Retrieved from Buzsáki, G., & Zugaro, M. B. (2009). Selective 21

http://www.ncbi.nlm.nih.gov/pubmed/1023 http://www.ncbi.nlm.nih.gov/pubmed/1053 4037 1452

Lechner, H. A., Squire, L. R., & Byrne, J. H. (1999). Nadel, L., Hupbach, A., Gomez, R., & Newman- 100 Years of Consolidation −− Remembering Smith, K. (2012). Memory formation, Müller and Pilzecker. Learning & Memory, 6, consolidation and transformation. 77–87. doi:10.1101/lm.6.2.77 Neuroscience and biobehavioral reviews, 36(7), 1640–5. Lee, A. K., & Wilson, M. a. (2002). Memory of doi:10.1016/j.neubiorev.2012.03.001 sequential experience in the hippocampus during slow wave sleep. Neuron, 36(6), Nader, K., & Hardt, O. (2009). A single standard for 1183–94. Retrieved from memory: the case for reconsolidation. http://www.ncbi.nlm.nih.gov/pubmed/1249 Nature reviews. Neuroscience, 10(3), 224–34. 5631 doi:10.1038/nrn2590

Louie, K., & Wilson, M. a. (2001). Temporally Nakashiba, T., Buhl, D., McHugh, T., & Tonegawa, structured replay of awake hippocampal S. (2009). Hippocampal CA3 Output Is Crucial ensemble activity during rapid eye for Ripple-Associated Reactivation and movement sleep. Neuron, 29(1), 145–156. Consolidation of Memory Accessed Retrieved from Hippocampal CA3 Output is Crucial for http://www.ncbi.nlm.nih.gov/pubmed/1118 Ripple-Associated Reactivation and 2087 Consolidation of Memory. Neuron, 62(6), 781–787. Mackintosh, N. J. (2002). Do not ask whether they have a cognitive map , but how they find O’Keefe, J, & Recce, M. L. (1993). Phase their way about. Psicológica, 23, 165–185. relationship between hippocampal place units and the EEG theta rhythm. Marr, D. (1971). Simple Memory: A Theory for Hippocampus, 3(3), 317–30. Archicortex. Philos Trans R Soc Lond B Biol doi:10.1002/hipo.450030307 Sci, 262(841), 23–81. Retrieved from http://www.jstor.org/stable/2417171 O’Keefe, John. (1976). Place units in the hippocampus of the freely moving rat. McGaugh, J. L. (2000). Memory--a Century of Experimental Neurology, 51, 78–109. Consolidation. Science, 287(5451), 248–251. doi:10.1126/science.287.5451.248 O’Keefe, John, & Conway, D. (1978). Hippocampal Place Units in the Freely Moving Rat : Why Mednick, S. C., Cai, D. J., Shuman, T., Anagnostaras, They Fire Where They Fire. Experimental S., & Wixted, J. T. (2011). An opportunistic Brain Research, 34, 573–590. theory of cellular and systems consolidation. Trends in neurosciences, 34(10), 504–14. O’Keefe, John, & Dostrovsky, J. (1971). The doi:10.1016/j.tins.2011.06.003 hippocampus as a spatial map. Preliminary evidence from unit activity in the freely- Moser, E. I., Kropff, E., & Moser, M.-B. (2008). moving rat. Brain Research, 34, 171–175. Place cells, grid cells, and the brain’s spatial representation system. Annual review of O’Keefe, John, & Nadel, L. (1978). The neuroscience, 31, 69–89. hippocampus as a cognitive map. Oxford: doi:10.1146/annurev.neuro.31.061307.0907 CLarendon press. 23 O’Neill, J., Pleydell-Bouverie, B., Dupret, D., & Nádasdy, Z., Hirase, H., Czurkó, a, Csicsvari, J., & Csicsvari, J. (2010). Play it again: reactivation Buzsáki, G. (1999). Replay and time of waking experience and memory. Trends in compression of recurring spike sequences in Neurosciences, 33(5), 220–229. the hippocampus. The Journal of doi:10.1016/j.tins.2010.01.006 neuroscience : the official journal of the Society for Neuroscience, 19(21), 9497–507. Retrieved from 22

O’Neill, J., Senior, T. J., Allen, K., Huxter, J. R., & hippocampal pyramidal cells. Journal of Csicsvari, J. (2008). Reactivation of neuroscience, 21(5), RC134. Retrieved from experience-dependent cell assembly patterns http://www.ncbi.nlm.nih.gov/pubmed/1122 in the hippocampus. Nature neuroscience, 2672 11(2), 209–15. doi:10.1038/nn2037 Rolls, E. T. (1999). Spatial view cells and the O’Neill, J., Senior, T. J., & Csicsvari, J. (2006). Place- representation of place in the primate selective firing of CA1 pyramidal cells during hippocampus. Hippocampus, 9(4), 467–480. sharp wave/ripple network patterns in doi:10.1002/(SICI)1098- exploratory behavior. Neuron, 49, 143–155. 1063(1999)9:4<467::AID-HIPO13>3.0.CO;2-F

Pavlides, C., & Winson, J. (1989). Influences of Rolls, E. T. (2010). A computational theory of hippocampal place cell firing in the awake episodic memory formation in the state on the activity of these cells during hippocampus. Behavioural brain research, subsequent sleep episodes. The Journal of 215(2), 180–96. neuroscience : the official journal of the doi:10.1016/j.bbr.2010.03.027 Society for Neuroscience, 9(8), 2907–18. Retrieved from Rudoy, J., Voss, J., Westerberg, C., & Paller, K. http://www.ncbi.nlm.nih.gov/pubmed/2769 (2009). Strengthening Individual Memories 370 by Reactivating them During Sleep. Science, 326(5956), 1079–1081. Peigneux, P., Laureys, S., Delbeuck, X., & Maquet, doi:10.1126/science.1179013.Strengthening P. (2001). Sleeping brain, learning brain. The role of sleep for memory systems. Sadowski, J. H. L. P., Jones, M. W., & Mellor, J. R. Neuroreport, 12(18), A111–24. Retrieved (2011). Ripples make waves: binding from structured activity and plasticity in http://www.ncbi.nlm.nih.gov/pubmed/1174 hippocampal networks. Neural plasticity, 2260 2011, 1–11. doi:10.1155/2011/960389

Rasch, B., & Born, J. (2013). About Sleep’s Role in Siegel, J. M. (2001). The REM sleep-memory Memory. Physiological reviews, 93(2), 681– consolidation hypothesis. Science (New York, 766. doi:10.1152/physrev.00032.2012 N.Y.), 294(5544), 1058–63. doi:10.1126/science.1063049 Rasch, B., Büchel, C., Gais, S., & Born, J. (2007). Odor cues during slow-wave sleep prompt Skaggs, W. E., & McNaughton, B. L. (1996). Replay declarative memory consolidation. Science of neuronal firing sequences in rat (New York, N.Y.), 315(5817), 1426–9. hippocampus during sleep following spatial doi:10.1126/science.1138581 experience. Science, 271, 1870–1873.

Rattenborg, N. C., Martinez-Gonzalez, D., Roth, T. Solstad, T., Boccara, C. N., Kropff, E., Moser, M.-B., C., & Pravosudov, V. V. (2011). Hippocampal & Moser, E. I. (2008). Representation of memory consolidation during sleep: a geometric borders in the entorhinal cortex. comparison of mammals and birds. Biological Science (New York, N.Y.), 322(5909), 1865–8. reviews of the Cambridge Philosophical doi:10.1126/science.1166466 Society, 86(3), 658–91. doi:10.1111/j.1469- 185X.2010.00165.x Squire, L. R., & Schacter, D. R. (2002). of Memory (3rd ed.). New Rechtschaffen, A., & Bergmann, B. M. (1995). Sleep York: The Guilford Press. deprivation in the rat by the disk-over-water method. Behavioural Brain Research, 69, 55– Squire, L. R., & Zola, S. M. (1996). Structure and 63. function of declarative and nondeclarative memory systems. PNAS, 93(November), Redish, a D., Battaglia, F. P., Chawla, M. K., 13515–13522. Ekstrom, a D., Gerrard, J. L., Lipa, P., … Barnes, C. a. (2001). Independence of firing correlates of anatomically proximate 23

Steriade, M., & McCarley, R. W. (2005). Brain Walker, M. P. (2005). A refined model of sleep and Control of Wakefulness and Sleep (2nd ed., the time course of memory formation. The pp. 1–691). New Yorke: Kluwer Behavioral and brain sciences, 28(1), 51–64; Aademic/Plenum Publishers. discussion 64–104. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/1604 Tambini, A., Ketz, N., & Davachi, L. (2010). 7457 Enhanced brain correlations during rest are related to memory for recent experiences. Wills, T. J., Lever, C., Cacucci, F., Burgess, N., & Neuron, 65(2), 280–90. O’Keefe, J. (2005). Attractor dynamics in the doi:10.1016/j.neuron.2010.01.001 hippocampal representation of the local environment. Science (New York, N.Y.), Tononi, G., & Cirelli, C. (2006). Sleep function and 308(5723), 873–6. synaptic homeostasis. Sleep medicine doi:10.1126/science.1108905 reviews, 10(1), 49–62. doi:10.1016/j.smrv.2005.05.002 Wilson, M. A., & McNaughton, B. L. (1993). Dynamics of the hippocampal ensemble code Treves, a, & Rolls, E. T. (1994). Computational for space. Science, 261(5124), 1055–1058. analysis of the role of the hippocampus in Retrieved from memory. Hippocampus, 4(3), 374–91. http://www.ncbi.nlm.nih.gov/pubmed/8351 doi:10.1002/hipo.450040319 520

Tsanov, M., & Manahan-Vaughan, D. (2008). Wilson, W. A., & McNaughton, B. L. (1994). Synaptic plasticity from visual cortex to Reactivation of Hippocampal Ensemble hippocampus: systems integration in spatial Memories During Sleep. Science, 265(July), information processing. The Neuroscientist : 676–679. a review journal bringing neurobiology, neurology and psychiatry, 14(6), 584–97. Wixted, J. T. (2004). The and doi:10.1177/1073858408315655 neuroscience of forgetting. Annual review of psychology, 55, 235–69. Ulanovsky, N., & Moss, C. F. (2007). Hippocampal doi:10.1146/annurev.psych.55.090902.14155 cellular and network activity in freely moving 5 echolocating bats. Nature neuroscience, 10(2), 224–33. doi:10.1038/nn1829