Upon discovery, rapid-eye-movement sleep (REMs) was met with curiosity regarding its function due to its association with dreaming and the seemingly paradoxical occurrence of ‘wake-like’ eye and brain activity during sleeping behavior. Six decades later, a substantial body of evidence linking REMs to memory formation has been described (Rasch and Born, 2013). In humans, increased REM amounts have been reported following procedural memory tasks, as well as declarative memory tasks incorporating complex or emotionally relevant material. Furthermore, depriving REMs after learning of such tasks produced memory deficits. Similar findings have been described in rodents using tasks such as the Morris water maze and active or passive avoidance (Abel et al, 2013).

Despite the cumulative body of evidence, the debate surrounding the role of REMs in memory formation has persisted for some time (Rasch and Born, 2013; Siegel, 2001). This has been due to the difficulty in experimentally isolating REMs, which occurs in multiple episodes of varying (seconds to minutes) duration throughout sleep interposed by periods of non-REMs (NREMs). Thus, traditional pharmacological techniques are not temporally precise enough to selectively target REMs. Additionally, as REMs is an integral component of mammalian sleep, selective REM deprivation inevitably results in physiological changes, such as increased metabolic hormone levels, making interpretation of data obtained difficult. Correlative approaches avoid these issues but cannot prove a direct relationship (Rasch and Born, 2013).

To overcome these caveats, we took advantage of the temporally precise control of specific neural circuits enabled by the use of optogenetic techniques. Our approach was to target a population of GABAergic neurons within the medial septum of mice (MSGABA) implicated in memory formation, yet not associated with the regulation of sleep itself. We therefore optogenetically silenced MSGABA neurons specifically during REMs in the period immediately following learning of either a spatial novel object place recognition test (n=6 mice) or standard fear conditioning paradigm (n=8 mice). Critically, the occurrence of REMs during MSGABA silencing was undisturbed as hallmark features of REMs—defined in the context of our experiments as a combination of sustained behavioral quiescence, muscle atonia, and absence of NREMs—and restful wakefulness-associated slow (1–4 Hz) EEG oscillatory activity—were indifferent from unsilenced control mice (Boyce et al, 2016). The following day mice in the test group demonstrated impaired spatial and fear-conditioned contextual memory. REMs was a critical factor as optical silencing of MSGABA neurons for similar durations non-specifically during NREMs and wakefulness had no effect on cognition, although this result does not preclude involvement of NREMs in memory formation as well. Indeed, the non-specific MSGABA inhibition occurring during NREMs did not influence the NREMs-associated neural activity patterns implicated in memory formation, including hippocampal sharp-wave ripples and neocortical spindles (Rasch and Born, 2013; Boyce et al, 2016).

Our study directly demonstrated for the first time that the activity of a specific population of neurons (MSGABA) selectively during REMs is required for normal memory formation. However, a precise mechanistic understanding of how REMs helps consolidate spatial and contextual memories requires further work. The ~7 Hz theta EEG rhythm that occurs in rodents during REMs has been implicated in the processing of spatial information at the neuronal level (Rasch and Born, 2013). Therefore, considering we found that a significant decrease in theta power accompanied MSGABA silencing selectively during REMs following learning in the test group (Boyce et al, 2016), it is possible that our experiments disrupted normal physiological processing of spatial information at the level of individual place cells, or perhaps that of more structured neuronal assemblies (Malvache et al, 2016). It is also possible that our optogenetic manipulation perturbed neural homeostasis during REMs (Grosmark et al, 2012;Tononi and Cirelli, 2014; but see Hengen et al, 2016). Fully investigating these potential mechanisms is an important future goal given the prevalence of sleep disruption in society and the connection between REMs disturbance and cognitive decline in aging and Alzheimer’s disease.

Funding and Disclosure

RB was supported by an Alexander Graham Bell Canada Graduate Scholarship (Natural Science and Engineering Research Council of Canada (NSERC)) while completing this research. AA is supported by the Human Frontier Science Program (RGY0076/2012), the Douglas Foundation, McGill University, Canadian Fund for Innovation (CFI), Canadian Research Chair (CRC Tier 2), CIHR, NSERC, the Swiss National Science Foundation, the Inselspital, and the University of Bern. The authors declare no conflict of interest.