Schematic representation of the AIM model. (
) During REM sleep dreams, cholinergic neurons in the LDT/PPT are involved in the generation of internal input PGO waves, while serotonergic dorsal raphe nucleus and noradrenergic locus coeruleus neurons and external inputs are inhibited. These internal inputs activate the LGN and occipital cortex, and produce dreams. The LDT/PPT–thalamus–cerebral cortex pathway causes the desynchronization of the cerebral cortex. The LDT/PPT–amygdala pathway may be involved in mood regulation during dreams. (
) The cubic 3-dimensional model shows normal transitions within the 3-dimensional parameters (activation, input, and modulation) from wakefulness to NREM and then to REM sleep. Abbreviations: 5–HT, serotonergic; ACh, cholinergic; DR, dorsal raphe nucleus; LC, locus coeruleus; LDT/PPT, laterodorsal tegmental and pedunculopontine nuclei; LGN, lateral geniculate nucleus; NA, noradrenergic; NREM, non-rapid eye movement; REM, rapid eye movement; W, wakefulness.
REM sleep is generally thought to be mediated by a neural network located primarily in the brainstem. More recently, the concept of REM sleep regulation has evolved. Several hypothalamic and forebrain networks, including newly identified neuropeptides such as orexin and melanin-concentrating hormone (MCH), have been implicated in both the control and the final expression of this behavioral state
[14,16,17,49,50,51,52,53,54][14][16][17][42][43][44][45][46][47]. Firstly, a reciprocal interaction between REM on and REM off states has been proposed to occur in the brainstem. In this model, the cholinergic laterodorsal and pedunculopontine tegmental neurons (LDT/PPT) are REM-on cells. The serotonergic dorsal raphe nucleus and noradrenergic locus coeruleus neurons are REM-off cells
[22,55,56][22][48][49] (
Figure 3A). However, this reciprocal interaction may not be sufficient to produce REM sleep, as suggested by experimental evidence over the past decade
[16,53][16][46]. The glutamateric sublaterodorsal tegmental nucleus (SLD) and the GABAergic lateral paragigantocellular nucleus (LPGi) have been subsequently found to act as REM-on neurons, whereas GABAergic ventrolateral periaqueductal gray matter (vlPAG) and lateral pontine tegmentum (LPT) act as REM-off neurons
[16,53][16][46].
2.5. Other Recent Mechanisms of Dreaming
Solms
[59][50] has hypothesized that dreaming is controlled by forebrain mechanisms. It is suggested that the cholinergic brain stem mechanisms that control the REM state are only able to produce the psychological phenomena of dreaming through the mediation of a second, presumably dopaminergic, forebrain mechanism. The dopaminergic forebrain circuits arise from the neurons in the ventral tegmental area (VTA) and terminate in the amygdala, anterior cingulate gyrus, and frontal cortex. This neural circuit of the mesocortical–mesolimbic dopamine system has been implicated in dream generation, and has been described as the “SEEKING” or “wanting” command system to subserve emotional drive and motivation
[59][50].
The amygdala, a limbic structure associated with emotions, memory, and dreams, receives the dopaminergic projection. It serves as a node to integrate the regulation of REM sleep and causes the intermittent appearance of REM sleep-related dreams and REM sleep behavior disorder (RBD)
[58,61,62][51][52][53]. The amygdala is very active in REM sleep, especially in humans
[63,64][54][55]. It can influence the frequency of PGO waves during REM sleep. This suggests that it also plays a key role in setting the “emotional tone” for PGO activity
[65][56].
3. The Physiological Functions of Dreaming: The Involvement of PGO Waves during REM Sleep
3.1. Memory Consolidation
Oneiric production is a form of mental sleep activity that appears to be closely related to memory processes and cognitive elaboration
[48,66,67][57][58][59]. Converging evidence suggests that dreaming is influenced by memory consolidation during sleep
[68][60]. A number of studies have reported that PGO waves during REM sleep in rodents have been repeatedly associated with memory consolidation
[69][61]. For example, PGO waves increased during REM sleep following learning tasks
[70,71,72,73,74][62][63][64][65][66]. Artificially enhancing PGO waves by injecting carbachol prevented avoidance memory deficits during a period of REM sleep deprivation
[75][67], while suppressing PGO wave generation in rats impaired avoidance memory retention during sleep
[76][68]. It appears that the density of PGO wave activity is directly related to memory processes. A number of studies have reported an increase in PGO wave density following fear memory training in rats, which predicted overnight memory consolidation
[70,71,72,73][62][63][64][65]. The success of fear extinction was recently shown to be predicted by PGO wave density during REM sleep
[74][66].
In general, PGO waves in the transition from NREM to REM sleep are considered to be the physiological signals that initiate and maintain REM sleep, constituting a state with characteristics distinct from both the preceding NREM sleep and the following REM sleep
[77,78][69][70]. Pontine caudolateral parabrachial neuronal discharge has been found to contribute to the shift toward the two PGO-related states, the transition from NREM to REM sleep and REM sleep
[34,77,79][34][69][71].
3.2. Unlearning
One of the hypothesized functions of REM sleep is a process of “unlearning”
[20,88][20][72]. These authors proposed that the function of dream sleep is the removal of certain unwanted memories from the cerebral cortex. During REM sleep, the unconscious dream traces act to weaken rather than strengthen memory. It is noteworthy that “we dream to forget”
[20] is not the same as normal forgetting. Dreams are not simply forgotten; they are actively unlearned
[80][73]. The unlearning mechanism modifies the cerebral cortex by changing the strength of individual synapses. Because an increase in synaptic strength is necessary to consolidate memory, unlearning weakens synaptic strength.
The unlearning theory is supported by a growing number of studies. For example, during REM sleep, the postsynaptic dendritic spines of layer V pyramidal cells in the mouse motor cortex are eliminated during development and motor learning. On the other hand, critical spines are strengthened and maintained
[89][74]. The hypothalamic MCH neurons are known to be involved in the control of REM sleep and mood
[90][75]. An anatomical and functional study in cats found that MCH neurons project to cholinergic pontine neurons
[91][76]. When MCH was microinjected into the nucleus pontis oralis, there was a significant decrease in latency to REM sleep and a significant increase in the amount of REM sleep, accompanied by increased PGO wave activity and its duration. This suggests that the MCH system is involved in the regulation of REM sleep by modulating neuronal activity in cholinergic pontine neurons
[91][76]. However, it has recently been shown that MCH neurons are also involved in the unlearning mechanism of REM sleep
[92][77]. The activation or inhibition of MCH neurons by optogenetics and chemogenetics impaired or improved hippocampal-dependent memory, respectively. The activation of MCH nerve terminals in vitro reduced the firing of hippocampal pyramidal neurons by increasing inhibitory inputs. These results strongly suggest that the activation of MCH projections to the hippocampus during REM sleep actively contributes to forgetting
[92][77]. The activity of hippocampal neurons during REM sleep has been suggested to play a key role in unlearning
[93][78]. REM sleep serves to maintain or strengthen memories until they are transferred out of the hippocampus, whereupon they should be erased from this space-restricted short-term memory factory so that these synapses can be used to encode new associative memories
[93][78]. PGO waves that are phase-locked to the theta oscillation of the hippocampus during REM sleep may be involved in this process
[94][79].
3.3. Brain Development and Plasticity
REM sleep is known to be particularly abundant during early development. At birth, half or more of our sleep time is occupied by REM sleep, compared to <20% of sleep time in adults
[96][80]. PGO waves during REM sleep in development might be an important central nervous system (CNS) stimulator during a period when wakefulness is limited in time and scope and stimulation opportunities are few
[94,97,98][79][81][82]. The ascending impulses emanating from the brainstem during REM sleep may be required to promote neuronal differentiation, maturation, and myelination in higher brain centers
[17]; that is, REM sleep deprivation during the early life of animals has been used to understand some functional mechanisms of PGO waves in brain development.
Several lines of evidence have shown that PGO waves in REM sleep are associated with the regulation of neural plasticity
[109,110,111][83][84][85]. Brain plasticity allows for the preservation of the ability to change, adapt, and learn in response to different environmental experiences and new demands. These processes occur with sleep cycles throughout life and begin in response to REM sleep in late fetal and early neonatal life
[112][86]. PGO wave-associated cells are good candidates for generating or modulating plasticity in various brain structures
[71][63], as they discharge high-frequency spike bursts during pre-REM and REM sleep
[34]. Indeed, the activation of PGO wave-generating cells by cholinergic agonism induces changes to the electrical properties of PGO wave activity
[113][87], accompanied by prominent behavioral effects
[74,114][66][88].
3.4. Mood Regulation
Because REM sleep is thought to play an important role in emotional memory processing, disrupted REM sleep may be an important contributor to the pathophysiology of emotion-based disorders such as major depression and PTSD
[119][89]. Major depression is extremely common and is one of the leading causes of disability worldwide. Sleep disturbances are typical of most patients with major depression and are a core symptom of the disorder. Polysomnographic indices document objective changes in sleep continuity, slow-wave activity reduction, and REM sleep alterations, such as a shortening of REM sleep latency, prolongation of the first REM sleep period, and increased REMs density
[95,120][90][91].
It has been proposed that PGO waves enhance synaptic plasticity in the areas that they pass through
[70][62]. This includes the hippocampus and the amygdala
[69,122][61][92]. Several studies have found that high amygdala reactivity is associated with an increased risk for the development of major depression
[123][93]. The pons receives amygdala axonal projections
[124][94], and the electrical stimulation of the amygdala increases the density of the PGO wave during REM sleep
[125][95]. The antidepressant drugs (norepinephrine or serotonin reuptake inhibitors) reduce the density of the PGO waves
[126][96]. The inhibition of the generation of PGO waves may have an antidepressant effect
[126][96]. The hippocampus, which plays a central role in mood dysregulation and neurogenesis, appears to be associated with the behavioral symptoms of major depression. The negative effects of REM sleep disruption on hippocampus-dependent cognitive functions may be due to a decrease in adult hippocampal neurogenesis in humans
[127][97]. Decreased functional connectivity at limbic cortical levels, particularly in the prefrontal, anterior cingulate, and insula, altered amygdala microstructure, and decreased claustrum volume have been reported following major depression
[121,128,129,130][98][99][100][101].