|Conquerors of Sleep?|
|These animals are currently believed to either require
very little or no sleep at all during migrations,
after giving birth, or throughout their lives.
a) Commerson's Dolphin. b) Tursiops truncatus, the bottlenose dolphin.
c), d) Orcinus orca, the killer whale. e) Rana catesbeiana, the bullfrog.
f) Zonotrichia leucophrys, the white-crowned sparrow.
Photo from Siegel 2008 
In land mammals and birds, a phenomenon called REM (rapid eye movement) sleep has puzzled scientists for decades. It is putatively unique to land mammals and birds alone; reptiles, invertebrates, amphibians, and the like, do not seem to have REM sleep . REM sleep is also called “paradoxical” sleep in the field because its properties are counter-intuitive. In slow wave, non-REM sleep, brain activity is greatly reduced. This matches the organism’s outward inactive behaviour and conserves energy. However, during REM sleep brain activity increases to match or even exceed that during waking. At the same time, motor impulses are inhibited resulting in muscle atonia, a property not seen in slow wave sleep, hence its paradoxical nature .
Despite decades of research, the purpose of REM sleep is unclear, especially in humans. Although past research has suggested REM is important for memory consolidation, recent findings have cast doubt on such results. The generalization of REM sleep benefiting procedural and non-REM sleep benefiting declarative tasks is now considered too simple . Even the latest research is conflicted on the functions of REM sleep. For example in rodents, low REM may be associated with high emotional reactivity  and increased cortical excitability . On the other hand perhaps REM has functions that are too subtle for us to detect reliably, especially in humans compared to rats or mice . A major difficulty in finding the putative functions of REM is the diverse heterogeneity in sleep characteristics amongst animals. Thus, properties found in rats may not be generalizable to humans or even mice.
Table of Contents
REMS as "internalized" sensory input: the Ontogenic Hypothesis and the Atonia/Pseudo-Locomotion hypothesis.
REM sleep physiology
What defines REM sleep?
As the name suggests, Rapid Eye Movement sleep (REMS) was first characterized in 1969 as "a stage of sleep characterized by rapid movements of the eye and low voltage fast pattern EEG" . REMS is the period during which dreaming occurs and EEG activity becomes unsynchronized, resulting in small, rapid spikes as opposed to the larger, more summative spikes observed in slow-wave sleep . Three important traits make it very curious for study. Firstly, brain activity in REM sleep closely resembles and sometimes even exceeding the activity observed during waking  . Secondly, during REM organisms undergo muscle atonia     . As a result, despite ample corticohippocampal activation, the postural muscles are unable to move to prevent accidental self-harm to the organism . Despite this, arousal thresholds are lower during REMS compared to slow-wave sleep, although they appear to be selective in nature, i.e. stimulus perceived as benign do not awaken the animal . Thirdly, REM sleep is not under tight homeostatic regulation. In adult mammals, it accounts for 5-20% of total sleep, is not readily recovered following REM deprivation, and can also be taken to excess: during naps taken by sleep-satiated individuals following a night of sufficient sleep, REM accounts for roughly 80% of the total sleep . Antidepressants in humans can reduce or abolish REMS directly with apparently no noticeable consequence  .
Often we associate sleep with rest, restoration, and inactivity. Indeed, during non-REM sleep, brain activity is slow and synchronized, with activity much below that of waking. Combined with quiescence, organisms have a much lower basal metabolic rate, and expend much less energy during this state. However, the brain is a metabolically expensive organ to maintain. In humans especially, where the waking brain accounts for roughly 20% of metabolic demand, it seems paradoxical that so much energy should be spent during a resting state. In addition, the purpose and effects of REM are not immediately obvious, since deprivation and excess apparently do not impair normal brain function.
Muscle Atonia and REM behavioural disorder
During REMS, somatic motoneurons in the brainstem are deactivated to prevent postural muscle movement. The exact mechanism for this inhibition is still under intense debate. Originally, it was hypothesized that these motoneurons are strongly inhibited by glycine  , and that glycine alone is sufficient and necessary for muscle atonia. Some researchers adhere to the sufficiency of glycine , some believe it may be due in part to a loss of serotonergic and adrenergic input  , and still others think that neither glycine nor excitatory input is sufficient alone to explain REM atonia, and that GABA inhibition may play a role as well .
Damage to the REM pathway may result in REM Behavioural Disorder (RBD), when atonia is insufficient and patients' limbs undergo fast, jerking, and sometimes violent movements. These movements correspond to the actions undertaken during sleep, which often involve unpleasant violent dreams of being attacked, confronted, or chased . RBD behaviour is different from sleepwalking, in that very few patients stand up (3%), and most patients sleep with closed eyes . Interestingly, RBD seems to be closely linked with Parkinson's Disease. Firstly, patients with RBD are much more likely to develop Parkinson's disease, dementia with Lewy bodies, and multiple system atrophy (MSA), with a 25%-40% chance at 5 years, and 40-65% chance at 10 years . Additionally, During REM, patients with RBD and Parkinson's in the absence of L-dopa exhibit plenty of coordinated behaviour without tremors or akinesia. Other parkinsonian diseases such as MSA show improvements as well, albeit to a smaller magnitude . Could there be pathways that bypass damaged areas in Parkinson's that are revealed during REMS? Further research in this direction may reveal novel treatment options for these diseases.
REM sleep regulation
|Structures associated with REM sleep regulation|
|Diagram of the complex network associated with activating REMS. Of note is the location of the REM-on and REM-off neuron groups,
as well as their interactions with the substantia nigra pars reticula. Image taken from Mallick et al. 2012 
REMS is regulated by complex mechanisms located primarily in the midbrain and brainstem, and improper regulation may lead to pathologies such as narcolepsy. As with many aspects of sleep regulation, the exact mechanisms are still not fully understood. It is likely that a plethora of pathways are involved in regulating REMS, each of which contributes a small amount. Early studies of REMS concluded that the brain stem, specifically the pons, was necessary and sufficient to generate REMS  , or at least generate the electrical activity commonly associated with REMS. Current evidence suggests that the noradrenergic (NAergic) neurons of the locus coeruleus (LC) are vital to REMS regulation. In the brain stem, there are two important groups of neurons related to REMS: REM-on neurons, which are active or increase firing during REM and cease or significantly decrease activity during non-REM and the waking state, and REM-off neurons, which do the opposite. It is currently believed that these two systems of neurons are reciprocally linked, although through a variety of different pathways as opposed to a direct synaptic pathway. The LC-NAergic neurons previously mentioned are REM-off neurons, which have been shown to prevent REMS appearance. When these neurons are active, they undergo a rhythmic, pulsatile firing pattern due to collateral self-release of NA which negatively feeds back on neuronal firing. When the REM-off neuronal firing is at its minimum, the threshold for REM-on firing is lowered, and vice versa  .
Briefly, GABAergic neurons synapse onto the NAergic REM-off neurons and inhibit them to allow for REMS to begin. These GABAergic neurons are activated (at least in part) by acetylcholinergic (Ach-ergic) REM-on neurons, also in the pons. Recall that the REM-on and REM-off neurons reciprocally inhibit one another. GABAergic projections from the substantia nigra pars reticulata (SNrPr) in the midbrain presynaptically inhibit REM-off inhibition of REM-on neurons. How these SNrPr neurons are activated is still unexplained  . Once REM-on neurons are active, they are able to inhibit REM-off neurons via the GABA interneurons previously mentioned, and triggers REMS  .
It should be remembered that this model is only a portion of a much larger whole, the majority of which are still a mystery to us. For further reading, see here.
REM sleep across species
REMS differences in mammals
|Sleep across species|
|This picture demonstrates the wide variety of "normal" sleep and REM sleep observed in animals.
Of note is the disparity between even closely related species. Picture taken from Allada et al. 2008 
As previously mentioned, REMS phenotype varies a great deal between species, and even within species. For example, the domestic cat (Felis catus) sleeps for 12.5 hours a day, with REM comprising 3.5 hours. On the other hand, genets (Genetta genetta), a related feliform species, sleeps only 6.3 hours, with REM comprising 1.3 hours . On the other hand, humans sleep 8.0 hours with 1.9 hours as REM, similar to the Eastern American Mole (Scalopus aquaticus), which sleeps 8.4 hours with 2.1 hours of REM, despite the great phylogenetic differences . In humans, arousal threshold is lowest in REMS, but in rats the opposite is true . The flexible nature of the REMS phenotype makes it difficult to generalize.
When investigating sleep across species, often the approach has been to perform comparative analysis in the wild as opposed to manipulating animals in laboratories. Ideally, through this approach we can isolate various behavioural and physiological variables and subsequently identify at least in part what sleep is responsible for. A comparative analysis done by J.M. Siegel suggests that in placental and marsupial mammals, REMS duration is positively correlated with total sleep time, negatively correlated with maturity at birth (i.e. animals born less mature have more REMS), and positively correlated with sleep site security . Unfortunately, studies correlating sleep time with these variables explain very little (30%) of the variance in REM sleep amongst placental and marsupial animals  . To further complicate matters, sleep studies often reach entirely opposite conclusions with respect to correlates, which presumably is due to post-hoc decisions regarding the data set and the methodology of statistical analysis . In summary, although we have hypotheses, no real consensus has been reached regarding why REM sleep differs across mammals.
REMS in birds and marine mammals
Another approach in investigating REMS is to probe the REMS phenotypes in other vertebrates to construct an evolutionary history of how REMS evolved. Reptiles and amphibians, for example, are much more evolutionarily ancient and conserved than the recently developed homo sapiens. Current evidence suggests that birds are the only taxonomic group next to mammals that clearly exhibit both REMS and non-REMS. Reptiles such as the box turtle do not exhibit the brain activity characteristic of REMS in their brainstem . Reptiles and birds are closely related groups, more so than either one is related to mammals. In fact, reptiles such as crocodiles are more closely related to birds than other lizards . The evolution of REMS and non-REMS in birds is therefore an example of convergent evolution. It stands to reason therefore that the differences shared by birds and mammals from non-REM-exhibiting groups are responsible for the development of REMS. The most important differences appear to be homeothermy (warm-bloodedness) and a complex brain allowing for advanced cognition , such as those seen in corvids and finches.
However, investigation into this idea is beset with difficulties as well. A recent quantitative comparative analysis of sleep in birds did not find correlates that matched those found in mammals, with the exception of a positive correlation between REMS duration and sleep site security . Furthermore, avian sleep states are not clearly defined, and thus are significantly more subjective. The main difficulty lies in determining when REMS begins. The transition in and out of REMS can be observed as an intermediate EEG trace between total REMS and non-REMS. These are by their nature subjective as well in mammals, but in birds REMS episodes last typically less than 10 seconds, greatly increasing the amount of subjective interpretation needed to define REMS duration .
Marine mammals exhibit very interesting sleep phenotypes, which will be elaborated on in the next section. Most famously, it is widely known that marine mammals sleep "with one hemisphere at a time". Cetaeans such as dolphins are perpetually submerged and do not exhibit REMS at all [((cite Siegel2011))]. Seals of the otariid family exhibit normal mammalian sleep patterns when sleeping on land, with REMS and non-REMS phases. Both eyes are closed, and the EEG is bilaterally synchronized. In contrast, when they are sleeping at sea, fur seals exhibit slow waves in only one hemisphere, with the contralateral eye open and the contralateral flipper actively maintaining body position . Similar to cetaceans, when they sleep at sea they exhibit no REMS.
REM sleep: What is it for?
It is generally accepted that REMS is necessary and confers some beneficial function, although some  believe it trivial. In addition to its convergent appearance in birds and mammals, REMS (and indeed, sleep in general) seems to place the animal into a vulnerable, inactive state. With such a high evolutionary cost, it stands to reason that despite its disadvantages REMS offers something important for the animal. As previously mentioned, marine mammals are unique in that they seem to have evolved out REMS entirely, at least in the water. Perhaps the open aquatic environment demands more vigilance and constant movement, and the costs of REMS have outweighed its benefits. Alternatively, the aquatic environment provides a compensatory factor not found on land, allowing marine mammals to eliminate REMS.
What exactly REMS provides for organisms is, unsurprisingly, still quite a mystery even today. A detailed survey of recent research reveals hints and theories, but no real consensus exists in the field regarding the purpose of REMS. Likely, its purposes are very subtle and multifaceted, contributing a little bit to a plethora of functions. This idea is supported by the sometimes conflicting results found in the literature: if REMS function is putatively subtle, even small variations in research methodology will mask or override the effects we are looking for. This caveat of REMS research will be discussed in more detail later on. Although REMS research is a field rich with debate, we present several theories and possibilities.
REMS regulates brain and cortical inhibition
Recent research in humans have suggested that REMS is antiepileptogenic, in that loss of REMS seems to increase the chances for synchronized cortical firing and subsequently, seizures. A study done by Placidi et al. in 2012 studied 10 humans and deprived them of REMS by waking them up following transition to REM. They found that this selective REMS deprivation reduced inhibitory action in the cortex and led to increased brain excitability . Other studies have found similar results: for example Huber et al. in 2012 found that the human frontal cortex increased in excitability as a function of the time awake . Regrettably, experimental studies targetting REMS itself in humans to examine brain excitability has been lacking.
REMS plays a role in regulating emotional responses
Keeping in mind the putative role of REMS in "topping off" the brain's capacity to inhibit activity, a recent study by Rosales-Lagarde in 2012 tested 20 (male) humans for emotional reactivity following REMS deprivation . Similar to the study above, they were woken every time they transitioned into REMS. The results suggest that REMS deprivation caused a greater reactivity to emotional stimuli, i.e. deprived individuals were more likely to indicate that they would defend themselves when shown a picture. The authors suggest this could be due in part to the role REMS plays in "dissipating" excess brain excitability.
REMS and memory
In basic neuroscience classes students have often been taught that REM sleep is responsible for the consolidation of procedural memories, whereas slow-wave sleep is responsible for consolidating declarative memories. With recent advances in research, it should come as no surprise that this generalization is now considered too simple  . Specifically, a relatively recent study done in humans found that pharmacological suppression of REMS in healthy men found a paradoxical increase in procedural memory tasks . A much more detailed examination on the role of sleep and memory consolidation can be found here.
REMS as "internalized" sensory input: the Ontogenic Hypothesis and the Atonia/Pseudo-Locomotion hypothesis.
REMS is often categorized into two states: the tonic state and the phasic state. In the tonic state of REMS, animals are able to distinguish between threatening and benign stimuli, suggesting that they maintain an altered, if not completely aware, state of consciousness. The phasic state is centered around pontine-geniculate-occipital (PGO) bursts which, as the name suggests, originate in the pons, travel to the lateral geniculate nucleus, and terminate in the occipital lobe. Although they have been studied in cats and rats, it is thought that they also exist in humans, although not detectable at the cortical level without invasive electrode recordings . During this state, arousal threshold is high; the PGO activity appears to be imagined novel stimuli that the animal responds to. In other words, PGO bursts are currently interpreted as sensory input to the cortices that is internally generated and sustained as opposed to input derived from sense organs, a phenomenon that is easily connected to dreams.
To continue along this line of thinking, we now consider the ontogenic hypothesis of REMS function, first posited by Roffwarg et al in 1966 . In humans, REMS is maximal in neonates at almost 50% of total sleep. At 6 months, this decreases to 30% of total sleep. Continued development increases the time spent awake apparently at the expense of REMS. Roffwarg hypothesized that this was due to the essential role of REMS in stimulating the brain during the critical period of central nervous system plasticity as a substitute for external stimuli, especially in utero. As we grow and mature, REMS becomes much less necessary due to the post-natal abundance of actual stimulus from the sense organs. As an extension to this idea, the brain may have a need for activity and stimulation which REMS provides, explaining why the phenomenon carries forward into adulthood .
Continuing from this, REMS may thus be important in learning, as newer research suggests that adaptations to the environment requires cognition and learning, which depends strongly on active sensory engagement . For example, many students find they learn much better when doing a task hands-on as opposed to watching and listening to an instructor perform the same task. Kempermann's 2008 paper on neurogenesis takes this one step further: he puts forth the idea that locomotion and motor activity is akin to cognition, especially in rodents and other studied lab animals . A 2012 study in mice found that even a broad selection of environmental enrichment did not improve adult neurogenesis or learning without locomotion . Animals experience, learn, and think by exploring their environment. For example, mice and rats exposed to novel stimuli such as wood blocks or plastic chew bones will explore it with their paws, nose, and mouth. Humans, although more advanced cognitively, may still explore a newly acquired iPad by touching its surfaces and smelling the box that it came in. Kempermann's argument that locomotion and cognition are strongly linked may provide explanation into the phenomenon of adult neurogenesis, but it may also partly explain REMS function.
Following from Kempermann's reasoning, it is possible that REMS occurs as pseudo-locomotion that aids the animal in cognition and learning via REMS. Movement during sleep may be hazardous, but conveniently muscle atonia occurs to mask this movement. This hypothesis treats atonia as playing a vital role in REMS as opposed to a vestigial byproduct. Early studies in cats  found that pontine NAergic neuron lesions removed muscle atonia in REMS. Curiously, cats subsequently displayed stereotyped motor behaviours during REMS despite the absence of external stimuli. For example, they may stalk imaginary prey, even raising their head to search and grab for them. Real objects are ignored; during REM-atonia-disinhibited grooming, pieces of paper placed on the fur are ignored. Oddly, although aggressive and attack behaviours are common (i.e. piloerection, hissing, arching of the back), there is no sympathetic activation, as if the body is not aware of the cat's apparent mental state.
Perhaps REMS is involved in food-foraging rehearsal, hence the need for pseudo-motor activity . A more provocative interpretation is that REMS is maintaining brain plasticity, for example via brain-derived neurotrophic factor (BDNF) release, which has been found to be influenced by muscle feedback 
What naturally follows from this hypothesis is: if REMS is substituting for movement during sleep, can REMS be substituted by movement? A study on rats  deprived them of REM via a gentle rocking cage or "pendulum". This has the advantage of minimally stressing the animal (the importance of which will be elaborated on later), while producing REMS deprivations comparable to other "standard" methods. Briefly, during non-REMS sufficient postural muscle control remains to stabilize the animal in the rocking cage. However, upon entering REMS atonia sets in, causing the animal to roll over and wake up. Naturally, they will move around some before sleeping again, whereupon the cycle repeats. The authors found this method resulted in less REM rebound, learning impairment, and antidepressant effects (in rat depression models) than the standard method. They interpreted that the movement resultant in the rocking cage method compensated for the loss of REM. Similar results have been found in other studies in rats as well as cats. .
Most interestingly, marine mammals exhibit no REM sleep. At the same time, they must maintain locomotion throughout sleep; for example seals must use one flipper to stabilize themselves while sleeping at sea, while whales and dolphins must surface to breathe, and seals furthermore exhibit REMS on land . Could movement during sleep be compensating for REMS necessity? While tantalizing, we must keep in mind the numerous confounds involved in such a generalization, such as the role of REMS in thermoregulation . Even with blubber, atonia for homeotherms in cold water may cause hypothermia and death.
REMS research and the problem with methodology
A commonly cited argument for the necessity of sleep is a 1995 study done by Rechtschaffen & Bergmann , where the authors found that rats deprived of sleep developed skin lesions and ultimately died. This highlights a common problem with sleep studies in the field, especially with the study of REMS: the methods of sleep deprivation often cause the animals extreme stress, and may produce learned helplessness . The standard methods involve placing animals such as rats on a rotating platform over water, in continuous light, in an open area, physically restricted, etc. The effects of sleep deprivation, as studied by Rechtschaffen and Bergmann may simply be due to the extreme stress the rats were experiencing. Additionally, a role of REMS may be to cope successfully with stress . If so, stress controls in these experiments may be invalid, as deprivation of REMS removes the ability for animals to adequately handle stressful stimuli. If animals are not as stressed by the paradigm, REMS deprivation may not have as significant an effect, as shown by Coolen et al.'s 2012 study in tree shrews .
Keeping this in mind, it is hoped that future studies on sleep will avoid these confounds by studying REMS in more naturalistic, less stressful settings. The nuances of REMS are many and subtle in nature, and there is still much to learn.