Building the dream: uncovering the development of REM sleep’s features

Imagine a factory that builds cars. Raw materials come in one end, the individual components are progressively assembled, and complete cars come out the other. It would be difficult—and somewhat arbitrary—to identify the exact moment that adding another part makes the partially assembled mass a “true” car. Likewise, REM sleep, also known as paradoxical sleep, consists of many different components, or features, which are incorporated into the larger whole at their own pace across development. But unlike the car, sleep goes by different names throughout infancy—for example, active sleep and REM sleep—and there have been debates (see e.g. [1, 2]) about what point in time this nascent, developing behavioral state has enough features of REM sleep to unambiguously bare its name.

The ontogenetic hypothesis of REM sleep suggests that—unlike a half-assembled car—“half-assembled” REM sleep serves an important function [3], likely by promoting activity-dependent development [4]. Still, the development of specific features of REM sleep likely enables the development of additional functions. For this reason, many have argued that the field should focus less on terminology and more on the development of these individual features [5], and this is exactly the approach taken by Gao et al. in their manuscript in this edition of SLEEP [6], entitled “Ontogenetic development of ponto-geniculo-occipital waves during paradoxical sleep in kittens.”

As it is typically conceived, REM sleep consists of many individual features. Of course, there are the rapid eye movements that give it its name. These rapid eye movements are typically accompanied by muscle atonia, myoclonic twitches, hippocampal theta oscillations, paradoxical “wake-like” cortical activity, and, relevant to the work of Gao et al., ponto-geniculo-occupital (PGO) waves. Every one of these features develops on its own timeline. But many of these developmental timelines remain poorly understood or, prior to Gao et al.’s recent work, wholly undescribed, a fact made even more surprising by the predominance of REM sleep as a behavioral state throughout early life.

Gao et al. answer many of these fundamental questions, including establishing when PGO waves first appear and describing their developmental trajectory. Their work in kittens reveals a somewhat discrete developmental onset, with PGO waves first appearing within a day or two of postnatal day (P) 19. Extrapolating these findings through a cross-species model of neurodevelopment suggests that PGO waves likely emerge in mice and rats around P11–P13 and, in humans, around the first or second postnatal month [7]. Gao et al. additionally show that PGO waves emerge as sparse, low-amplitude signals, but rapidly increase in both density and amplitude, ultimately adopting an adult-like pattern by P50. This timeline places PGO waves as the last major feature of REM sleep to develop (Figure 1A), underscoring their distinct developmental timeline relative to other features of REM sleep. These results suggest that the evolving role of REM sleep—particularly in promoting neural plasticity, learning, and memory—is likely intimately linked to the precise assembly and refinement of REM sleep’s constituent parts.

Figure 1.

(A) Across development, different features of REM sleep emerge at different times. Muscle atonia and myoclonic twitches are the first features of REM sleep to develop, while PGO waves are among the last to develop. (B) The features of REM sleep can map onto the functions of REM sleep in many different ways, and the mapping of features to functions is still largely unknown.

Perhaps because of the historically close association between PGO waves and dreaming [8], many have considered the developmental emergence of PGO waves as an unambiguous marker of the developmental onset of “mature” REM sleep [9, 10]. However, this semantic argument of when, in development, REM sleep truly begins obfuscates a more informative question: Given what is known about PGO waves in adults, how might the functions of REM sleep differ before and after they develop?

Mapping individual features of sleep to their functions is emerging as a critical frontier in sleep research, and several studies have begun to tease apart these relationships. For example, showing that REM sleep twitches drive activity throughout infancy [11] and into adulthood [12, 13] suggests that they may serve to build and calibrate sensorimotor circuits, ensuring precise integration between motor commands and sensory feedback. Similarly, the association of hippocampal theta oscillations with memory consolidation shows a direct mechanistic link [14]. However, the way these features map onto functions is not necessarily straightforward (Figure 1B). Like the preceding examples, some may follow a one-to-one mapping, where a single feature is tied to a specific function. But it is more likely that they display a one-to-many relationship, where a feature serves multiple functions, or a many-to-one relationship, where several features—or maybe even all features—contribute to a single, unifying function.

At its heart, the many unknown mappings of REM sleep features to REM sleep functions underlie the arguments of when, across development, REM sleep truly begins. If all features of REM sleep are necessary for a specific function, then it is worthwhile to argue that REM sleep begins only after the development of its last feature. But if—as is more likely in a complex biological system—the mapping of features to functions is less straightforward, the requirement for REM sleep to have every feature feels rigid and arbitrary. Indeed, birds, and even lizards, are now widely believed to exhibit REM sleep, despite possessing only a subset of the typical mammalian features [15, 16]. Consequently, these unknown mappings underscore the need for a more comprehensive understanding of how REM sleep features and functions evolve over time and map onto brain development, and developmental and evolutionary variability provide excellent opportunities to address these issues. Documenting the development of an individual feature, as was done by Gao et al., is a critical first step as the field prepares for a deeper exploration of features and functions.

Integrating these insights is not only essential for advancing our understanding of sleep, but it also holds significant clinical implications. Mapping REM sleep features onto functions would likely identify novel biomarkers for many disorders—including Parkinson’s and many neurodevelopmental disorders—where disrupted sleep can predict cognitive or motor deficits [17–19]. Such an integrative approach may eventually inform targeted interventions designed to harness or restore the beneficial effects of REM sleep. As the field moves forward, bridging the gap between sleep, development, and neural activity will be key to unraveling the complex, multifaceted functions of REM sleep.

Disclosure Statement

Financial disclosure: none.

Non-financial disclosure: none.