Roles for sleep in memory: Insights from the fly

Abstract

Sleep has been universally conserved across animal species. The basic functions of sleep remain unclear, but insufficient sleep impairs memory acquisition and retention in both vertebrates and invertebrates. Sleep is also a homeostatic process that is influenced not only by the amount of time awake, but also by neural activity and plasticity. Due to the breadth and precision of available genetic tools, the fruit fly has become a powerful model system to understand sleep regulation and function. Importantly, these tools enable the dissection of memory-encoding circuits at the level of individual neurons, and have allowed the development of genetic tools to induce sleep on-demand. This review describes recent investigations of the role for sleep in memory using Drosophila and current hypotheses of sleep’s functions for supporting plasticity, learning, and memory.

Introduction

Sleep is vital for cognition, but the basic functions of sleep in the brain remain poorly understood. While consequences of sleep loss degrade systems throughout the body, cognitive deficits, including learning and memory impairments, are among the earliest to occur (For reviews, see [1-3]). Although neural architecture and physiology differ between vertebrates and invertebrates, the properties and functions of sleep are tightly conserved across evolution. Sleep-like rest was first described in a cockroaches roughly 35 years ago [4], and sleep has since been characterized in a variety of other invertebrates, including insects [5-7], nematodes [8], and mollusks [9-11]. In all of these species, sleep matches the behavioral criteria originally used to describe mammalian sleep patterns: quiescence, reversibility, postural change, decreased arousability, and homeostatic regulation [12,13]. Importantly, many drugs and genetic lesions influence sleep similarly in humans and invertebrates [5,14-17], providing strong evidence for broadly conserved mechanisms of sleep regulation and function.

Because sleep has been universally conserved across animal species, examining the functions of sleep in relatively small, simple invertebrate brains will likely provide a better understanding of why sleep is required for all animals. Sleep’s role in supporting learning and memory appears to be an evolutionarily ancient function – insufficient sleep degrades memory similarly in animals ranging from humans to invertebrates, including Drosophila melanogaster and Aplysia californica [3,18]. Because a role for sleep in memory has been broadly conserved across evolution, it is likely that sleep fulfills a basic function that promotes the consolidation of recent memories and maintains the capacity for new memory acquisition. To date, however, a mechanistic understanding of sleep’s functions in memory remain incomplete. Examining sleep’s effects within memory-encoding circuits of insect brains, therefore, will likely uncover functions of sleep that may be generalized across species. Due to a combination of genetic accessibility and a relatively simple nervous system, the fruit fly, Drosophila melanogaster, has become the focus of many investigations into sleep regulation and function.

Plastic regulation of sleep

Sleep is a homeostatic process that is influenced not only by the amount of time awake, but also by neural activity and plasticity. In vertebrates, slow wave activity during sleep is elevated in cortical areas that have been recently active or plastic [19-21], suggesting that sleep can be regulated in a use-dependent manner. Similarly, several studies have found an elevated need for sleep in invertebrates following novel experiences that drive plasticity. Housing Drosophila in an enriched social environment for several days, for instance, drives synaptic elaboration in visual and olfactory circuits [22-25]. After this social experience, flies sleep more than socially isolated siblings for up to three days [26]. The effect of waking experience on sleep scales with the size of the fly’s social group, and is eliminated by disrupting visual or olfactory perception or by mutating genes necessary for synaptic plasticity [24,26]. Similar increases in sleep are also temporarily observed after associative training protocols that produce protein synthesis-dependent long-term memories (LTM) [26]. Studies of honeybees have also identified changes in sleep time and architecture based on an individual’s social role in the hive [27,28]. Together, these findings suggest that insects provide models to understand why elevated neural plasticity/activity increase the need for sleep and how sleep might be altering recently plastic synapses.

While the function of sleep after novel experience is not clear, one recent hypothesis has raised the possibility that sleep acts as a mechanism for homeostatic synaptic scaling [29,30]. Waking increases the abundance of synaptic proteins [31], likely reflecting synaptic expansion and growth to encode newly formed associations. One function of sleep may be the homeostatic weakening or removal of these expanded connections to prevent synaptic saturation [29,30]. In support of this hypothesis, the overall abundance of several synaptic proteins in fly brain homogenates is elevated after sleep deprivation and reduced after a night of sleep [31]. Subsequent studies have also found that extended waking increases the size or number of pre-synaptic active zones from several cell types, including Mushroom body-intrinsic Kenyon cells, large-field visual interneurons in the medulla, and circadian ventral lateral neurons [24,25]. Similarly, recovery sleep following enriched social experience is required to reduce synaptic size and number back to baseline levels [24,25]. Recent publications have also found trends for increased synapse strength [32], synapse size [33], or dendritic spine retention [34] in rodent cortex following sleep deprivation, suggesting that similar scaling may occur in flies and mammals.

Although these studies are consistent with a role for sleep in homeostatic synaptic downscaling, the processes by which synapses are both tagged for pruning and weakened/removed have not been dissected. It is currently unknown, for example, which particular synapses are most likely to be pruned during sleep, or how a homeostatic set point for synapse number or strength might be maintained. The answers to these questions may lie in mechanisms that underlie homeostatic control of firing rate [35,36], neural excitability [37-39], and synaptic scaling [39-46]. Studies using rodents have begun to elegantly examine electrophysiological mechanisms of neural homeostasis across wake and sleep [36,47], and complementary investigations using the fly may also unravel the underlying molecular machinery.

The data described above support a role for sleep in homeostatic synapse downscaling, but plastic changes during sleep may vary across circuits or cell types. Sleep depriving young Drosophila during the first day after eclosion, for example, can cause long-lasting deficits in courtship and memory [48,49]. In flies, as in other species, sleep is elevated early in life as neural circuits undergo high levels of developmental plasticity and refinement [5,49-52]. During the initial days after eclosion, male flies require sleep for the expansion of olfactory glomeruli in the antennal lobe that process social pheromone cues [49]. This result parallels observations of ocular dominance plasticity in the visual cortex of cats and mice, in which sleep is required for neurons in primary visual cortex to weaken inputs from a closed eye and expand inputs from an open eye [52,53]. In the case of ocular dominance plasticity, sleep promotes the strengthening of inputs from an open eye to visual cortex via NMDA-dependent signaling mechanisms [53]. The sleep-dependent expansion of synapses in developing Drosophila olfactory glomeruli and mammalian visual cortex suggests that plasticity during sleep is not unidirectional, but the factors that determine how sleep alters synapses in different circuits or different stages of development have not been established. To address these issues, additional studies will be required to examine which rules govern synaptic scaling in different circuits or cell types during sleep, and to identify the molecular mechanisms that might modulate synaptic size or number during sleep.

Sleep and learning in the Mushroom body

Short-term memory deficits have been observed in wild-type flies that have been previously sleep deprived [54-57] or exhibit spontaneously fragmented sleep [55], in short-sleeping hyperkinetic [58] and crossveinless-c mutant flies [56], and in flies genetically selected for low sleep time [59]. Conversely, sleep is altered in a variety of mutants and neurodegenerative disease models that influence learning and memory [60,61] and promoting sleep in many of these genotypes is sufficient to improve learning [62,63]. Together, these findings point to a vital role for sleep before learning in maintaining the capacity to acquire new memories.

Mushroom bodies (MBs) are an associative neuropil vital for the acquisition and storage of olfactory memories (See Figure 1 for schematic) [64,65]. The logic that underlies memory encoding in the MB has been the focus of intense investigation, and recently created Drosophila genetic tools have allowed the dissection of MB circuits at the level of individual neurons [66,67]. Drosophila form olfactory memories within the Mushroom bodies, an associative center that receives synaptic input from second-order olfactory projection neurons. MB-intrinsic Kenyon cells (KCs) encode olfactory cue information and can be divided into three subsets that project to distinct axonal lobes (α/β, α’/β’, and γ) [68]. Each of these lobes can be further subdivided into individual zones that provide output to individual MB output neurons (MBONs) and receive reinforcement signals from distinct types of dopaminergic neurons [67]. New olfactory memories are encoded in the MB when reinforcing dopamine signals modify the strength of synapses between active KCs and individual MBONs that encode either attraction or aversion [66,69,70] (See Figure 1B). Sleep is required for MBs to encode and retain new memories in several associative assays [55-57] (See Table 1). Two factors are most likely to contribute to learning and memory deficits following sleep loss: excessive synaptic potentiation (see description of the synaptic homeostasis hypothesis above), and dysregulated reinforcement from dopaminergic neurons in the MB. As shown by Seugnet et al [55], transcript levels of the dopaminergic receptor Dop1R1 are reduced after sleep loss. Expression of this receptor in the MB is required for associative learning [71], and learning impairments can be overcome in sleep-deprived flies either by pharmacologically increasing dopamine signaling or by overexpressing Dop1R1 in MB-intrinsic KCs [55]. Together, these data suggest that prolonged waking may degrade MB sensitivity to dopaminergic release that signals reinforcement. Sleep may, therefore, allow Drosophila KCs to re-sensitize to dopaminergic inputs by enforcing a period of low activity in both KCs and dopaminergic neurons [72,73]. Neuroimaging studies have also found that sleep deprivation reduces available D2/D3 receptors in the human striatum, suggesting that similar effects may be conserved in mammals [74,75].

Figure 1.

Organization of Mushroom body circuit that encodes olfactory memories and modulates sleep

(A)Cartoon of MB lobes formed by axons of MB-intrinsic Kenyon cells (KCs). KC axons are subdivided into α/β lobe shown in blue, α’/β’ in magenta, and γ in orange. (Illustration adapted from [89])

(B)Schematic of MB circuitry. Axons of odor-coding KCs can be subdivided into tiled zones that are each innervated by a single type of dopaminergic reinforcement neuron and provide output to a single MB output neuron (MBONs). Olfactory memories are encoded within the strength of synapses between KCs and MBONs. Modulatory DPM neurons project throughout the MB lobes and their recurrent activation after learning is required for consolidation of recent memories. (Based on findings from [67,69,81]).

Table 1.

Summary of behavioral assays used to test memory deficits associated with sleep loss. Learning/short-term memory loss was typically tested in flies sleep-deprived for one night prior to learning, while the role for sleep in long-term memory consolidation was tested using sleep deprivation that occurred after training.

	Short-term Memory	Long-term Memory
Aversive Phototactic Suppression	Impaired by SD before training Seugnet et al., 2008	No LTM in this assay
Courtship Conditioning	No published data	Impaired by SD after training Ganguly-Fitzgerald et al., 2006
Aversive Olfactory Conditioning	Impaired by SD before training Li et al., 2009	Impaired by SD after training Li et al., 2009
Aversive Taste Memory	Impaired by SD before training Seidner et al., 2015	LTM not characterized in this assay
Heat-box Spatial Memory	Impaired in short-sleeping hk mutants Bushey et al., 2007	No LTM in this assay

MB circuitry not only encodes memories, but also influences sleep regulation. Manipulating activity in distinct subpopulations of MB-intrinsic KCs or MBONs can strongly either increase or decrease sleep [76-78]. Dissection of this circuitry suggests that wake- and sleep-promoting circuits within the MB may be organized in parallel, with wake-promoting KCs exciting wake-promoting MBONs and sleep-promoting KCs exciting sleep-promoting MBONs [78]. Sleep-promoting KCs and MBONs exhibit increased activity in sleep-deprived flies [78], indicating that they may provide a representation of homeostatic sleep need. It is important to note that the MBONs that modulate sleep may also encode aspects of olfactory valence; wake-promoting MBONs are also activated by odors with a negative valence, and sleep-promoting MBONs are responsive to attractive odors [66,78]. Future studies will be required to untangle whether the coding of sleep/wake and positive/negative valence within these MB circuits are linked, or if each signal can be independently relayed to downstream effectors.

Sleep and Memory Consolidation

Sleep is also required after learning to consolidate recent associations into long-lasting memories. Initial studies examining the function of post-training sleep found that sleep was significantly increased following a spaced training protocol for Courtship conditioning, an associative assay during which male flies learn to suppress their courtship behaviors following unsuccessful mating attempts [26]. Sleep deprivation during the first several hours after training prevented memory consolidation, and sleep was not changed after training in flies with genetic defects that impair memory consolidation [24,26]. These studies suggested that sleep was necessary for LTM formation in MB-dependent assays (See Table 1), raising the possibility that acutely promoting sleep after training may be sufficient to enhance memory consolidation.

Drosophila provides a unique model for examining the beneficial effects of sleep – genetic screening has identified a cluster of neurons projecting into the dorsal fan-shaped body (dFB) that can be acutely activated to induce sleep on-demand [79]. To test the role of sleep in promoting memory consolidation, male flies experienced a single episode of Courtship training, then sleep was induced via dFB activation. While the single training episode was not sufficient to elicit LTM in control flies, the same protocol resulted in robust LTM when sleep was induced via dFB activation immediately following training [79]. Because this LTM enhancement was eliminated when flies were sleep deprived during dFB activation, it is likely that increased sleep after learning may actively promote the consolidation of recently acquired memories. Similar impairments in consolidation with sleep loss and consolidation enhancements with sleep induction have also been observed using classical olfactory conditioning [54,73], suggesting that the beneficial role for sleep in memory consolidation may be shared across MB-dependent assays.

To date, however, the consolidation mechanisms that occur during sleep have not been clearly described. It is known, however, that memory consolidation requires sleep and the recurrent activation of modulatory neurons within the MB during the few hours after training [26,73,80]. Synaptic release from the dorsal paired medial neurons (DPMs), which consist of 1 neuron in each hemisphere that innervate all MB lobes (See Figure 1B for schematic), is not required during memory acquisition or retrieval, but silencing DPM release during the hours after learning impairs LTM consolidation [80,81]. Similarly, slow oscillatory activity in dopaminergic MB neurons following learning is also correlated with memory consolidation [82]. While it is not known whether recurrent DPM activity during consolidation coincides with sleep, DPM activation has been shown to drive robust sleep increases [83]. While these studies suggest that DPM activation may gate memory consolidation by modulating sleep, no studies have yet confirmed whether DPM activity is required to promote sleep during consolidation. Recurrent activity of MB circuits during sleep may also contribute to the consolidation of recent memories in insects other than Drosophila. In an elegant study by Zwaka et al. (2015), honeybees were trained to associate a novel odor with a sugar reward. Bees that were re-exposed to the same conditioned odor during subsequent sleep episodes retained a stronger appetitive memory than bees that were exposed to the conditioned odor during post-training waking [84]. This study closely parallels human studies that used contextual odor cues during slow wave sleep to reactivate hippocampal circuits and increase declarative memory retention [85], suggesting that similar patterns of recurrent activation might support memory consolidation in insects and vertebrates.

Conclusion

Sleep is a physiological state that we share with all animals, including insects like the fruit fly. Because learning and memory impairments are closely shared from humans to Drosophila, the genetic accessibility and relative simple nervous system of the fly provide an ideal model to uncover the fundamental mechanisms of sleep function. While evidence suggests that sleep may play a role in synaptic scaling and calibrating the strength of dopamingergic reinforcement signals, the precise functions of sleep and their underlying mechanisms are still poorly understood. With recently developed tools to precisely dissect and control MB circuits at the level of single cells [67] and to control the timing of sleep in Drosophila [57,79,86-88], current and future research will have a precision to examine sleep’s role in plasticity that is unavailable in other species. Importantly, these studies open the possibility of studying not only the detrimental consequences of sleep loss, but also the benefits of promoting sleep [62].

Highlights.

• Insufficient sleep impairs the acquisition and consolidation of memories in the Drosophila Mushroom bodies

• Sleep may benefit memory-encoding circuits by homeostatically scaling synaptic connections

• Genetic tools to control sleep timing in Drosophila may provide unique opportunities to study the cognitive benefits of sleep