The effect of REM-sleep disruption on affective processing: A systematic review of human and animal experimental studies

Evidence on the importance of rapid-eye-movement sleep (REMS) in processing emotions is accumulating. The focus of this systematic review is the outcomes of experimental REMS deprivation (REMSD), which is the most common method in animal models and human studies on REMSD. This review revealed that variations in the applied REMSD methods were substantial. Animal models used longer deprivation protocols compared with studies in humans, which mostly reported acute deprivation effects after one night. Studies on animal models showed that REMSD causes aggressive behavior, increased pain sensitivity, reduced sexual behavior, and compromised consolidation of fear memories. Animal models also revealed that REMSD during critical developmental periods elicits lasting consequences on affective-related behavior. The few human studies revealed increases in pain sensitivity and suggest stronger consolidation of emotional memories after REMSD. As pharmacological interventions (such as selective serotonin reuptake inhibitors [SSRIs]) may suppress REMS for long periods, there is a clear gap in knowledge regarding the effects and mechanisms of chronic REMS suppression in humans.


Introduction
Sleep is fundamentally important in preserving and modulating affective representations and thus mental well-being.Disturbed sleep, such as insomnia, is associated with an up to 10-and 17-fold increased likelihood of having depression and anxiety disorder, respectively (Taylor et al., 2005).Non-depressed individuals with insomnia have a two-fold increased risk of developing depression over time (Baglioni et al., 2011).However, the underlying mechanisms between disturbed sleep and mental health problems are poorly understood.One emerging possibility is offline affective processing during sleep.
Sleep consists of cyclic rotation of non-rapid-eye-movement sleep (NREMS) and REM-sleep (REMS).In humans, NREMS constitutes about 75-80% of total sleep time and consists of three sleep stages (N1-N3) that represent a continuum of relative sleep depth towards high-voltage, slow-wave sleep (SWS) in N3.REMS constitutes the remaining 20-25%.Dreaming can occur both in REMS and NREMS (Siclari et al., 2017).The fundamental sleep pattern involving two alternating states (NREM and REM) is preserved in rodents.Unlike humans, rats and mice are nocturnal and show multiple periods of sleep-wake activity, and their individual sleep episodes and NREM-REM cycles are much shorter than in humans.
Sleep-related affective processing is considered to occur during REMS due to its wake-like physiology.The underlying mechanisms of this processing have been investigated since the discovery of paradoxical sleep in the early 1960 s (Jouvet, 1965).REMS is the most aroused sleep state, with a desynchronized electroencephalogram (EEG) with theta and alpha waves, fast oscillations in the 20-50 Hz range beta and gamma activity typical for wakefulness, muscle atonia, vivid dreaming, and varied and mental event-bound autonomic nervous system (ANS) activity.Indeed, brain activity in REMS converges with waking brain activity in terms of heightened activity in emotion-related brain regions such as the amygdala, striatum, hippocampus, insula, and medial prefrontal cortex (mPFC) (Nofzinger, 2005, Dang-Vu et al., 2010).REMS is associated with specific brain-activity patterns, simultaneous activation of the limbic regions, and the acetylcholine and gamma-aminobutyric acid (GABA)-mediated suppression of cortical arousal (Pace-Schott and Hobson, 2002), which is permissible also for offline processing of emotions (Goldstein and Walker, 2014).
As comprehensively reviewed (Palagini et al., 2013), aberrant REMS, such as shorter latency from sleep onset to first REMS period, and an increase in total REMS time and in the frequency of rapid eye movements per REM period (REMS density) are common findings in clinical depression.These findings have also been suggested as an endophenotype feature for subsequent depression (Hasler et al., 2004, Modell and Lauer, 2007, Palagini et al., 2013).However, REMS alterations are not observed in all depressive patients (Riemann et al., 2001), and the theories postulating the underlying mechanisms between depression and REMS have failed to gain wide support (see (Riemann et al., 2001)).In contrast, the role of REMS in maintaining well-being is not specific to any single psychiatric disorder.Consequently, the focus of studies and discourse on REMS has shifted from purely psychiatric questions to a more universal understanding on the role of REMS in information and emotion processing, including affective memory consolidation and modulation.

REMS and affective processing
The sleep spindles are thalamocortical activation bursts occurring in N2 and N3 at a frequency rate of 9-16 Hz.Especially those spindles that are temporally coordinated (coupled) with upstate phases of slow oscillations are involved in the consolidation and later recall of declarative memory contents (Klinzing et al., 2019).However, the consolidation processes may not be independent of emotional factors.Sleep may selectively consolidate contents with emotional valence over neutral memories (Wagner et al., 2001, Hu et al., 2006, Wagner et al., 2006, Nishida et al., 2009) and both the emotional aspects of the memories (Payne et al., 2008, Payne et al., 2012) and the value one places on them (Saletin et al., 2011, Perogamvros andSchwartz, 2012) may influence their consolidation vs forgetting during sleep.
The most compelling evidence for REMS-related affective processes in both humans and other animals comes from experiments with conditioned fear.In particular, cued fear memory formation is dependent on amygdala activity (Phelps and LeDoux, 2005).While REMS activates brain regions that are central for fear processing (mPFC, amygdala, hippocampus), the amount of REMS may contribute positively to fear consolidation (Menz et al., 2013) but may also serve fear extinction by consolidating the extinction memory (Pace-Schott et al., 2015).
The bi-directional influence of sleep on affective processing has been conceptualized in the "Sleep to forget and sleep to remember" (SFSR) hypothesis (Walker and van der Helm, 2009), which suggests that the decoupling of the emotional aspect of a memory (associated with amygdala activity) from the strength of the memory (hippocampal-associated activity) occurs during sleep.In other words, the emotional tone may decay in sleep, while the tagged memory of the episode becomes consolidated.REMS may represent an optimal brain state for this process (Walker andvan der Helm, 2009, van der Helm et al., 2011).However, in the case of hyperarousal or heightened amygdala activation, disturbed mPFC function, or both, as often observed in insomnia (Shao et al., 2020, Ma et al., 2021) and PTSD (Bremner et al., 2005), REMS can be more fragmentary, the emotional dissipation process may become impaired (Bottary et al., 2020).In these cases, sleep may function as the maintaining factor of the affective load (Germain et al., 2008, Murkar andDe Koninck, 2018).Also, in healthy participants, the evidence for the dissipation of affective load in sleep is not without controversies, as decreases in emotional reactivity have not been observed in all studies (Wagner et al., 2002, Baran et al., 2012).Of note, the emotional reactivity component has rarely been measured with physiological methods in human studies.
REMS may also have an important function in qualitative reorganization of memories, as formulated by Stickgold and Walker (Stickgold and Walker, 2013).Sleep-dependent memory evolution occurs in both NREMS and REMS.The evolution can take many forms, including multi-item integration, where reactivation of a newly acquired memory in sleep is accompanied by a parallel activation of an existing associative memory network, thus enriching both the network and the new memory item, with a potential to create new knowledge (Stickgold and Walker, 2013).Based on empirical evidence, the model assumes that associative inter-item memory processing is specifically linked to REMS (Stickgold and Walker, 2013), whereas NREMS is more prone to probabilistic learning, such as rule extraction (Wagner et al., 2004).Landman et al. (Landmann et al., 2014) further proposed that REMS potentiates disintegration and reorganization of existing schemas, which in turn leads to associative thinking, creativity, and shaping of emotional memory representations (Landmann et al., 2014).This model provides perspectives into how REMS could be harnessed towards therapeutic aims in mental health problems (Landmann et al., 2015).

Oscillatory activity in REMS
Understanding of the mechanisms underlying the REMS-related affective processes is emerging.Similar to emotional processing in the wake state, limbic circuits are active during REMS (Maquet et al., 1996), and expression of immediate early genes associated with brain plasticity is increased during REMS in these brain regions (Luppi et al., 2017).The driving oscillatory activity related to offline affective processing is often attributed to REMS theta (approximately 4-8 Hz) activity, as indicated from intracranial recordings in rodents.For instance, REMS theta coherence between the amygdala, hippocampus, and mPFC predicted affective memory consolidation in fear learning (Popa et al., 2010).Following an avoidance task training, REMS theta power increased during the subsequent night (Fogel et al., 2009).In this context, pontine-geniculo-occipital (PGO) waves (Hutchison and Rathore, 2015), which are linked with theta wave coherence between the hippocampus and amygdala (Karashima et al., 2010) and which promote synaptic plasticity (Datta et al., 2008) are especially interesting.Accordingly, evidence from rodent models indicates that PGO waves are linked with emotional learning (Datta et al., 2008).
In humans, PGO waves are not measurable with scalp surface EEG.However, the association between generation of PGO waves and theta oscillations (Karashima et al., 2005) implies that measuring REMS theta activity may depict offline processing of emotional information.Indeed, human studies on emotional memory have shown that REMS theta power, or its frontal lateralization, predicts emotional over neutral declarative memory performance (Nishida et al., 2009, Sopp et al., 2018).In sum, suppressing REMS would thus disrupt theta-driven processing of emotional memories and their attached affective charge.

Motivation for the current review
While models and theories of the role of REMS in affective processing are evolving, the acquired scientific evidence is still controversial (Landmann et al., 2015).The methods used to manipulate REMS, both in animal and human studies, are also highly varied, limiting holistic syntheses of their outcomes.Moreover, the models of REMS in human affective processing are often based on data from animal models and have not yet been tested experimentally in humans.
The operationalization of "affective processing" has also been challenging in human studies.The term "affective" usually refers in this context to both memories with affective content, thus having a declarative memory component, and to subjective evaluations or physiological reactions to affective stimuli.The term "processing" refers to structural, functional, and behavioral changes occurring offline after the encoding or exposure to affective stimuli.In human studies, affective processing related to memory outcomes has frequently been operationalized with pictures of affective content, where both the recognition of the pictures and the subjective evaluation of their emotional valence have been assessed.Some studies aimed to measure the change in physiological or brain-related responses (or both) by repeated exposure to stressful A.-K. Pesonen et al. experiences.
The concepts used in human studies are not directly translatable to animal models due to difficulties in measuring and evaluating emotions.Animal models, however, present an invaluable tool to elucidate underlying mechanisms of behavior, and many behavioral components are still accessible through animal models.For example, anxiety and fear (threat detection) are evolutionarily conserved responses given their necessity for survival.Indeed, the concept of affective memory processing has been mostly studied in rodent models using fear conditioning and fear extinction paradigms.
The aim of the current systematic review is to synthesize information on REMS in affective processing, including reactivity to pain, as presented in animal model and human experimental studies, focusing also on the methodological aspects.While there exist a number of nonexperimental observations of REMS length, density of eye movements, and quantification of the REM fragmentation, in this review, we selectively concentrated on experimental REMS deprivation and suppression methods and their influence on emotional processing.As the outcomes of this research are inconclusive both in animal and human studies, the current systematic review will synthesize and critically discuss the existing evidence and provide suggestions for future research.

Study identification and eligibility criteria
PubMed and Web of Science (WoS) databases were used to identify eligible studies.Duplicate references were removed before screening.Retrieved articles were screened according to the following inclusion criteria: a) experimental study design such that REMS was deprived or suppressed as the independent variable in the experimental condition; b) emotion-related outcome; and c) the record language was English.The following search command was used:

Data extraction
Key information from the identified studies were collected using a standard Microsoft Excel spreadsheet.The following information was extracted from each included study: 1) first author name; 2) publication year; 3) mean age and standard deviation/standard error or alternatively age range (for studies with human subjects); 4) sex of participants; 5) sample size (cases, controls); 6) study design; 7) sleep measurement method; 8) main outcome variables and measures; 9) primary findings (REMS deprivation -> affect outcome); 10) circadian timing (napping, overnight); 11) the amount of sleep on experimental night (minutes; TST and REMS); 12) percentages of different sleep phases on REMSD night (N1, N2, SWS, REMS); and 13) whether the study reported a REMS rebound effect or not.

Study selection process and risk of bias/quality assessment
Two independent reviewers assessed the 1919 records (duplicates excluded) retrieved by title and abstract.Articles that met the eligibility criteria were retrieved for further assessment.These articles were rechecked by five other reviewers and any discrepancies were resolved by consensus.A flow chart of the study selection process is presented in Fig. 1.

Human studies
The literature search yielded 14 eligible studies summarized in Table 1.Sample sizes in the human studies included in this review ranged from 6 to 62, with an average of 25 participants.Eight of the 14 studies included healthy volunteers as participants.Four studies did not report information on the participants' age or sex (Vogel et al., 1973, Vogel et al., 1975, Vogel et al., 1977, Vogel et al., 1980).In the remaining studies, participants were mostly young adults (age range 19-43 years) except for one study with participants aged over 60 years (Buysse et al., 1992).Four studies had only male participants (Onen et al., 2001, Azevedo et al., 2011, Rosales-Lagarde et al., 2012, Kaida et al., 2015); in the remaining studies, the percentage of male participants ranged from 14% to 48%.

Animal studies
The literature search yielded 66 eligible studies summarized in Table 2.All animal studies were conducted in mice or rats except for one   study involving prairie voles (Jones et al., 2019) and a study conducted in domestic dogs (Bolló et al. 2020).Most studies only included males, except for 11 studies that included either both sexes or only females.While most studies investigated the effects of REMSD in adults, five studies included young (adolescent) animals (Feng and Ma, 2003, da da da Silva Rocha-Lopes et al., 2018, Jones et al., 2019, Jung and Noh, 2021, Simionato et al., 2022).

REM sleep deprivation methods and control conditions in human studies
The non-pharmacological REMSD procedures lasted for one night in six studies (Roehrs et al., 2006, Lara-Carrasco et al., 2009, Rosales-Lagarde et al., 2012, Kaida et al., 2015, Wiesner et al., 2015, Glosemeyer et al., 2020).Two studies conducted awakenings for two consecutive nights (Buysse et al., 1992, Onen et al., 2001) and one study for four consecutive nights (Azevedo et al., 2011).The early studies reported conducting the REMSD procedure for several consecutive nights until they reached 30 awakenings per night or for six nights, whichever came first (Vogel et al., 1973, Vogel et al., 1975, Vogel et al., 1980).This procedure amounted to an average of 4.1-4.2(Vogel et al., 1973, Vogel et al., 1975) consecutive nights of REMSD, after which the subjects had one recovery night.After the recovery night, the REMSD protocol proceeded from the start.The entire procedure continued for several weeks.Similar protocols have not been used since these early studies.
The most common auditory stimulus to disturb REMS was addressing the subject by name through an intercom.Other auditory stimuli were also used, such as acoustic beeps (Glosemeyer et al., 2020) or a set of different alarm tones (Kaida et al., 2015).Not all studies reported the type of REMS-disturbing stimuli they used.Six studies reported intensifying the stimuli if REMS continued, such as increasing the volume of auditory stimuli until the subject was fully awake (Rosales-Lagarde et al., 2012, Glosemeyer et al., 2020).Additionally, some studies reported other means if auditory stimulus was not adequate to wake the subject; these included the experimenter entering the subject's room and addressing them directly (Wiesner et al., 2015), and in the next step lights would be switched on, and the participant was asked to solve simple mathematical problems to avoid immediate relapse to REMS (Wiesner et al., 2015).Other methods included asking questions from the participant during awakenings (Lara-Carrasco et al., 2009, Rosales-Lagarde et al., 2012) or giving the participant a reaction-time task (Roehrs et al., 2006).One study reported calling the patient's name through an intercom or alternatively "going in person" to arouse the subject (Buysse et al., 1992).Two studies reported "gently shaking" the sleepers until they showed an awakening response (Onen et al., 2001, Azevedo et al., 2011).The frequency of applying this method was not reported.
In one study (Wyatt et al., 1971), REMS suppression was performed by administration of monoamine oxidase inhibitors (MAOI), a class of drugs that are capable of completely suppressing REMS.The MAOI phenelzine was given to 6 anxious-depressed patients while daily behavioral and EEG sleep records collected.The protocol lasted from 14 to 40 nights until the REMS was completely suppressed for a variety of periods (Wyatt et al., 1971).

Quality of awakenings in human studies
While most studies aimed to wake the participants, for some the
Single platform.
Emotional and fear reactivity: OFT.
Single platform.
Single platform.
↓ Fear reaction (number of boluses and urination).
Single platform.
↓ Depression-like behavior (increased swimming activity, no difference between control groups).(de Oliveira et al.
Multiple platform.Social environment.
Single platform.
Anhedonia-like behavior with and without CCI: Sucrose consumption.
Single platform.
Aggressive behavior: Latency to first attack, frequency of attacks, rate of muricide.

Single
platform.
Single platform.

Contextual and cued fear conditioning and extinction
↓ Impaired cued fear extinction.
Single platform.
Contextual and cued fear conditioning and extinction training.
↓ Impaired recall of cued extinction memory.
Contextual and cued fear conditioning.
(continued on next page) A.-K. Pesonen et al. discontinuation of REMS and shifting to another sleep stage was sufficient.In seven studies, the length of awakenings was 3-5 minutes (Vogel et al., 1973, Vogel et al., 1975, Buysse et al., 1992, Lara-Carrasco et al., 2009, Rosales-Lagarde et al., 2012, Wiesner et al., 2015, Glosemeyer et al., 2020).In one study, participants had to be awake for 15 minutes before being allowed to reach sleep again (Roehrs et al., 2006).The range in the number of awakenings varied between 12 and 30 per night in the studies that reported them, except for one study that limited the maximum number of awakenings to six (Lara-Carrasco et al., 2009).

Time in bed in human studies
The subjects' time in bed (TIB) during the REMSD night was not restricted in any of the studies that reported either TIB or lights-on/ lights-off times.Conversely, TIB was even extended to compensate for the lost sleep time caused by the REMSD.The subjects were allowed to sleep (TIB or lights-off to lights-on) between 6.1 and 8 hours per night in four studies (Onen et al., 2001, Kaida et al., 2015, Wiesner et al., 2015, Glosemeyer et al., 2020) (Lara-Carrasco et al., 2009) and between 9 and 9.5 hours per night in two (Roehrs et al., 2006, Azevedo et al., 2011).Other studies did not report the time during which the subjects were allowed to sleep.

Effects of REMS disruption on sleep architecture in human studies
Half of the human studies reported sleep architecture of the REMSD nights (Onen et al., 2001, Roehrs et al., 2006, Lara-Carrasco et al., 2009, Rosales-Lagarde et al., 2012, Kaida et al., 2015, Wiesner et al., 2015, Glosemeyer et al., 2020), some reported only the percentage in REMS (Vogel et al., 1973, Vogel et al., 1975, Azevedo et al., 2011), and the remainder did not provide these measures in detail or were reported elsewhere.The total sleep time in REMSD groups on experimental nights varied between 293 and 399 minutes.In five studies, there was a decrease in total sleep time during the REMSD nights (Vogel et al., 1975, Onen et al., 2001, Roehrs et al., 2006, Lara-Carrasco et al., 2009, Rosales-Lagarde et al., 2012).REMS suppression or deprivation was successful in those studies that reported stage durations, as the amount of REMS decreased significantly compared to a baseline night, control condition, or both.In most studies, the amount of REMS varied between 0.70% and 8.3%.In two studies, the amount of REM sleep on the deprivation night was notably greater (12.76-14.28% of REM (Lara-Carrasco et al., 2009, Glosemeyer et al., 2020)) yet significantly less compared with habituation night and control conditions.
Other changes in sleep architecture also occurred, such as increase of N1 (Buysse et al., 1992, Roehrs et al., 2006, Rosales-Lagarde et al., 2012, Kaida et al., 2015) and N2 (Buysse et al., 1992, Rosales-Lagarde et al., 2012) but also decrease of N2 sleep (Kaida et al., 2015).In one study, the amount of both N1 and N2 sleep decreased during the REMSD night compared with baseline, but this may be due to SWS rebound after total sleep deprivation (Onen et al., 2001).The amount of N1 sleep varied between 1.72% and 23.12%, the amount of N2 sleep between 44.24% and 68.66%, and the amount of SWS between 15.7% and 46.11%.

REM-sleep deprivation/disruption methods in animal studies
A common method of REMSD performed with rodents was the platform method, which is also referred to as the flowerpot method, water tank method, and pedestal-over-water technique.This method is mainly based on the muscle atonia characteristic of REMS.The animal is placed on a small platform surrounded by water, and upon entering REMS, muscle relaxation prompts the animal in contact with the water, subsequently waking it up.This method, or a modified version of it, was used in most of the eligible articles.The modified version included, for example, use of multiple platforms instead of one, allowing social interaction and some locomotor activity.
Techniques apart from the platform method were also applied.One study used direct electrical stimulation of the reticular midbrain formation to suppress REMS (Kovalzon and Tsibulsky, 1984).In addition, gentle cage shaking (Feng and Ma, 2003), cage tilting (Ravassard et al., 2016), or gentle handling of the animals (Zhou et al., 2020) was also performed to disrupt REMS.In all these studies, EEG recordings were conducted to detect onset of REMS.In one study conducted in prairie voles, cages were housed on orbital shakers to induce cage movement and sleep disruption (Jones et al., 2019).
A study with domestic dogs performed deprivation by awakenings performed by the experimenter by entering the room after the onset of REMS was detected and keeping the dog awake by talking and petting for 2 minutes to prevent immediate relapse to REMS (Bolló et al. 2020).
REMSD intensity also varied across studies.Most of the animal studies included in this review performed total REMS deprivation, that is, animals remained in deprivation conditions throughout the wakesleep cycle.Partial deprivation, during which animals were partially housed in home cage conditions, was also performed in some of the eligible studies (Sloan, 1972, Wang et al., 2015, da da da Silva Rocha-Lopes et al., 2018, Rabelo-da-Ponte et al., 2019, Bolló et al. 2020, Cai et al., 2022, Li et al., 2022, Simionato et al., 2022, Zhu et al., 2022).

Effects of REM disruption on sleep architecture in animal studies
The effects of REMSD on sleep architecture were validated only in a subset of studies.One study in mice using the multiple platform method assessed sleep architecture and observed an effective suppression of REMS, which was accompanied with a reduction in NREMS (Zager et al., 2009).As significant suppression of NREM following REMSD using the platform method has been reported (Grahnstedt andUrsin, 1985, Machado et al., 2004) disruption of NREMS is expected to have occurred in studies using the platform method.REMSD by midbrain reticular stimulation reduced REMS by 70%, while NREMS was reduced by 10% (Kovalzon and Tsibulsky, 1984).In adolescent rats, REMSD by automated cage shaking reduced REMS by 50-70%, which was accompanied by increased NREMS.However, an increase in REMS was found later in adulthood (Feng and Ma, 2003).In adolescent prairie voles, REMSD by cage shaking reduced REMS by 25%, which was accompanied with reduced gamma power and shortened average sleep bouts of both REM and NREM (Jones et al., 2019).In mice, automated cage shaking effectively suppressed REMS while leaving NREMS intact (Rosier et al., 2018).Moreover, gentle handling upon detection of REMS also effectively suppressed REMS and decreased NREMS (Zhou et al., 2020).Cage tilting upon detection of REMS specifically reduced the duration of REMS episodes, with no effect on NREMS (Ravassard et al., 2016).
Please see Appendix 1 for the explanations of the used outcome measures in human and animal studies.

Outcomes related to affective problems
Table 1 summarizes the findings from the human studies.The early studies by Vogel and Wyatt (Wyatt et al., 1971, Vogel et al., 1973, Vogel et al., 1975, Vogel et al., 1977, Vogel et al., 1980) concentrated exclusively on clinical populations and found evidence of beneficial effect of REMSD on "endogenous" depressive symptoms.Unfortunately, these studies did not control for SWS deprivation; later studies provided evidence that the effect may not be REMS-specific, as similar observations were made from studies using a regular sleep deprivation period of one night (Leibenluft and Wehr, 1992, Wirz-Justice and Van den Hoofdakker, 1999, Kuhn et al., 2020).
Table 2 summarizes findings from the animal studies.In animal models, the effects of REMSD on anxiety-and depression-like behavior were investigated in several studies using classical behavioral tests, such as the open field, elevated plus-maze, and forced swim test.A prevalent finding in many of these studies in which the platform method was used was altered locomotor activity following REMSD, with hyperactivity being more common (Hicks and Moore, 1979, Oniani, 1984, de Oliveira et al., 2004, Patti et al., 2010, Streck et al., 2015, Dal-Pont et al., 2019, Rabelo-da-Ponte et al., 2019, Jung and Noh, 2021, Kim et al., 2022, Zhu et al., 2022) than hypoactivity (Gonzalez-Castañeda et al., 2016, Wang et al., 2017).Given that the behavioral tests for the assessment of anxiety-or depression-like behavior are highly confounded by changes in locomotor activity, interpretation of the results is obscured.When activity changes remained unaltered, increased anxiety-and depressive-like behavior was reported (Wang et al., 2017, Xie et al., 2018, Cai et al., 2022, Saadati et al., 2022).However, it should be noted that the behavioral outcomes can be driven primarily by the stress response rather than REMSD itself.The platform method may induce stress due to restricted locomotion and the loss of environmental enrichment, which encompasses both social and material aspects that are essential for the animals' wellbeing.Moreover, sudden contact with water can also trigger a stress response.Two studies employed more gentle deprivation methods and assessed anxiety-related behavior but did not reveal any changes in anxiety-like behavior (Feng andMa, 2003, Jones et al., 2019).Altogether, studies reporting altered affective-related behavior in rodents often fail to sufficiently consider the stress-inducing nature of the common REMSD method.Therefore, systematic effects of REMSD on anxiety-and depressive-like behavior remains unclear.

Outcomes related to aggressive behavior
A consistent finding in rodent studies with emphasis on males was increased aggression following REMSD, which was observed both after a short (Hicks, 1979;Trotta, 1984) and chronic REMSD (Sloan, 1972, Peder et al., 1986).Methodological aspects should also be considered here, given that stress can increase aggressive behavior in rodents (Walker et al., 2018).Moreover, chronic REMSD induced by gentle cage shaking in neonatal rats was shown to alter aggressive behavior (increased defensive and decreased offensive behavior) in adulthood during shock-induced fighting, demonstrating lasting impact of early-life REMSD (Feng and Ma, 2003).Studies in humans did not measure aggression.

Outcomes related to pain sensitivity
Two human studies on pain reactivity reported modestly increased pain sensitivity following REMSD (Roehrs et al., 2006) and one study reported no effect on pain perception or reactivity (Azevedo et al., 2011).
Several rodent studies investigated the effects of REMSD on pain sensitivity.A consistent finding across these studies was increased pain sensitivity following REMSD both in baseline conditions (Hicks et al., 1978, Hicks and Moore, 1979, Hakki Onen et al., 2001, Damasceno et al., 2009, Araujo et al., 2011, Skinner et al., 2011, Damasceno et al., 2013, Tomim et al., 2016, Hirotsu et al., 2018, Xue et al., 2018) and following a surgical procedure (Wei et al., 2007, Wang et al., 2015, Li et al., 2022).Hypersensitivity to noxious stimuli was typically observed after ≥2 consecutive days of REMSD.This effect was present in response to different stimuli, including mechanical, thermal, and electrical.However, assessing the impact of REMSD alone is challenging due to the potential pain sensitivity increase caused by stress.Studies that employed less stressful approaches did not measure pain sensitivity.

Outcomes related to emotional memory
The study of Kaida et al. (Kaida et al., 2015) in humans compared the encoding and memorizing of emotionally valenced pictures after a total sleep deprivation (TSD) and REMSD.They concluded that the ability to encode pictures deteriorates after TSD but not after selective REMSD.
Both the rate of correctly recognized pictures and of recalled frames associated with the pictures were significantly decreased after TSD but remained unchanged after REMSD.
REMSD in humans associated with both enhancing and decaying of affective memory content.Lara-Carrasco et al. (Lara-Carrasco et al., 2009) studied adaptation to emotional pictures and reported, contrary to their hypothesis, that a higher percentage of REMSD was associated with greater adaptation to negative pictures.They associated the finding with the pioneering work of Vogel (Vogel et al., 1975), paralleling the observations that REMSD temporarily alleviates dysphoric symptoms in depressed patients.
Wiesner et al. (Wiesner et al., 2015) compared the deprivation effects of REMS and SWS on emotional memory.Their data suggest that in contrast to wakefulness, sleep-including REMS fosters the consolidation of emotional memories but has no effect on the affective evaluation of the remembered contents, providing evidence for the sleep-to-remember hypothesis.They did not, however, find evidence for the sleep-to-forget part of the hypothesis.Neither the comparison of the affective ratings during encoding and recognition nor the affective ratings of recognized targets and rejected distractors supported the hypothesis that REMS dampens the emotional reaction to remembered stimuli and that this effect would be related to the consolidation of memories.Only new, negative pictures were rated slightly less negative by the REMSD group than the control group; this finding parallels the study of Lara-Carrasco (Lara-Carrasco et al., 2009).
In rodent models, the relationship of REMS and emotional memory has focused on the effects of REMSD on fear memory and extinction learning using the classical Pavlovian fear conditioning paradigm.For example, such studies report post-conditioning REMSD-induced impairment in recent fear memory recall both in the contextual (hippocampus-dependent) (Ravassard et al., 2016) and cued (amygdala-dependent) (Zhou et al., 2020) conditioning.However, in another study post-conditioning REMSD specifically impaired recall of remote, but not recent, contextual fear memory in mice (Rosier et al., 2018).Furthermore, one study assessed the role of REMS on the consolidation and recall of extinction memory and demonstrated impaired cued, but not contextual, extinction memory (Fu et al., 2007).In the aforementioned studies, REMSD was conducted using methods other than the platform technique and therefore stress effects are unlikely to be significant contributors to the outcome.Methodological advancements available in rodent models, such as use of optogenetics to activate or inhibit specific neuronal pathways, have the potential to elucidate the underlying mechanisms of REMS and its impact on behavioral processes without inducing stress or disturbing other vigilance stages.Indeed, recent studies using these techniques lend support to the importance of REMS in memory processes, such as encompassing the facilitation of consolidation and forgetting of hippocampus-dependent contextual memories, including emotional (fear) memories (Frazer et al., 2021).
Collectively, these findings support the involvement of REMS in the consolidation and extinction of fear memories.However, the evidence comes uniquely from animal models.In human studies, the evidence is fragmentary and controversial and are suggested to be confounded by the mood lift resulting from REMSD.

Outcomes related to social emotions and behavior
Glosemyer et al. (Glosemeyer et al., 2020) studied the effect of REMS suppression on emotions during social exclusion in humans.While they reported that lower amounts of REMS were associated with higher levels of general negative affect the next morning, there was no evidence for a direct link of REMS with the subjective emotional response to experimentally induced social exclusion.The ability to regulate negative emotions during social exclusion was also not affected by prior REMS, which was an unexpected finding.However, despite no changes in subjectively reported emotions, suppressed REMS led to higher activation of the right amygdala when participants passively experienced social exclusion.
A.-K. Pesonen et al.In rodents, sexual behavior is consistently decreased in adult males after exposure to REMSD (Alvarenga et al., 2009, Damasceno et al., 2009).One study also demonstrated that chronic REMSD early in life elicited reduced sexual behavior in adulthood (Feng and Ma, 2003).Although the effects of REMSD on females were minimally investigated, one study reported heightened proceptive response in proestrus females (Andersen et al., 2009).Furthermore, sexual behavior of male offspring from males who underwent REMSD prior to copulation was reduced, while female offspring demonstrated increased proceptive behavior in the estrous phase (Alvarenga et al., 2013).Collectively, these rodent studies suggest that REMSD influences sexual motivation, which can potentially have implications for reproductive outcomes.

Developmental outcomes
Rodent models have provided an opportunity to explore the effects of early-life REMSD.Based on these studies REMSD had enduring effects, including reduced social bonding and sexual behavior and altered aggression (Feng andMa, 2003, Jones et al., 2019).

Discussion
In this systematic review, we collected and evaluated the existing human and animal model data to create a systematic overview of the significance of REMS in affective processing.Towards this aim, we specifically focused on REMSD studies, as they represent the most common experimental method in the field and are used both in animal and human studies.Additionally, we acknowledged that there is a lack of a systematic view of the deprivation studies from a translational perspective, although current understanding has been greatly advanced by models and theories that have guided empirical research in the past 10-15 years (Walker and van der Helm, 2009, van der Helm et al., 2011, Stickgold and Walker, 2013).Our second aim was then to compare human and animal studies and to evaluate the existing explanatory models on the role of REMS in affective processing.

The concept of emotional processing in sleep
The premises of the SFSR hypothesis (Walker and van der Helm, 2009), which postulates that sleep consolidates initial emotional memories but dissipates reactivity towards them, was not properly tested in human REMSD studies.The challenge in studies aiming to test this hypothesis is that the modulation of affective responsivity to stimuli is not clearly independent from declarative memory components in most of these studies, as they used emotional (and narrative) pictures as experimental stimuli.Subjective evaluation of negative arousal elicited by affective pictures in the recall phase is not sufficient to understand the outcomes of affective processing because the evaluation depends on the current state (the value put on the task performance) expectations of the participant of the on-going study and how the participant memorizes their prior ratings.To properly test the SRSF hypothesis, it would be crucial to measure either brain activity with fMRI, physiological arousal, or both, to objectively assess the non-declarative emotional component as the outcome of affective processing.The few studies achieving this revealed controversial results.The most robust evidence for the SRSF hypothesis comes from pain sensitivity studies.Higher pain sensitivity after REMSD, both in animals (Hicks et al., 1978, Hicks and Moore, 1979, Hakki Onen et al., 2001, Damasceno et al., 2009, Araujo et al., 2011, Skinner et al., 2011, Damasceno et al., 2013, Tomim et al., 2016, Hirotsu et al., 2018, Xue et al., 2018) and humans (Roehrs et al., 2006, Azevedo et al., 2011), could be interpreted as losing the regulation potential of negative sensations.However, this does not provide evidence of the dual processing or modulation of the initial emotion and its declarative representation.We also acknowledge that not all changes in pain sensitivity are related to affective components.
Disentangling the affective memory consolidation from the emotional reactivity to affective stimuli is also a challenge from a translational research perspective, as these effects are difficult to dissect in animal models.

Methodological synthesis 4.2.1. Humans
Our first observation was that the number of experimental REMSD studies in humans is low.This was somewhat surprising given that total sleep deprivation studies (over the entire night[s]) are very common.We also observed that the sample sizes were mostly very small and allowed detection of mainly large effect sizes.
The deprivation protocols varied.After the pioneering studies in the 1970 s that applied highly stressful long-lasting clinical protocols, probably not meeting today's ethical standards (Wyatt et al., 1971, Vogel et al., 1973, Vogel et al., 1975, Vogel et al., 1977, Vogel et al., 1980), the contemporary deprivation protocol is mostly focused on one to two nights of deprivation.Although a wide range of antidepressants (such as certain tricyclic antidepressants and SSRIs) suppress REMS (Armitage, 2000), their application as an experimental method to understand the role of REMS in affective processing seems to be a rare exception (Wyatt et al., 1971), and their efficacy in improving the subjective quality of sleep may also constitute a confounding factor.
Most of the studies intervened with sleep using different repeated auditory stimuli, although some used activation prompts or even gentle shaking.As the REMS rebound pressure intensifies along successful deprivation episodes, most of the studies reported using intensifying stimulus protocols, either through increasing the volume or adopting harsher waking methods in a stepwise manner.These solutions imply that the stage-selective deprivation task may require more invasive methods than non-selective sleep deprivation, where the waking state collaboration with the participants facilitates the implementation.
Despite the severe experimental interruptions of sleep, the total sleep time mostly corresponded to an average regular night.All the reported protocols were successful in reducing REMS.Compared to an undisturbed sleep where REMS averages around 20%, the proportion of REMS was reduced to 0.7-14.3%depending on the study.A total REMSD was very difficult to attain without pharmacological intervention (Wyatt et al., 1971); most studies achieved a partial deprivation or aimed towards it.Importantly, in addition to REMSD, sleep architecture was altered for all sleep stages.

Animal models
Although there were more animal studies on REMSD than those in humans, yet an evident observation was that the literature on the topic is heterogeneous.A major caveat in these studies is that REMS was deprived, either acutely or chronically, using inherently stressful methods, during which activity is also significantly restricted.For example, the widely used platform-over-water method increases corticosterone levels (Nollet et al., 2020), which can induce depression-and anxiety-like behavior and interfere with memory processing.Thus, distinguishing REMSD effects from stress-related changes is very challenging, as stress may act as a major confounding factor.This could be a particular issue when home cage controls are used as the reference group.Apart from the platform method, more gentle REMS deprivation/suppression methods, achieved for example by manual or automated cage movement upon real-time detection of REMS, were used in a few studies.Indeed, intact corticosterone levels were reported after a 4-h REMSD by cage tilting (Ravassard et al., 2016) and after chronic REMSD by cage shaking (Jones et al., 2019), indicating a less stressful nature of these deprivation methods.
Another limitation in most animal studies was the lack of complementary EEG/EMG recordings, which limits the assessment of the manipulation effects on REMS, NREMS, and wakefulness.Only one study in the current review (in which the platform method, or a modified version thereof was used) included measures of sleep architecture and reported effective suppression of REMS accompanied with marked reduction in NREMS (Zager et al., 2009).Similarly, other studies have reported an almost complete loss of REMSD following deprivation by the platform method, which co-occurs with marked reduction in NREMS (Machado et al., 2004).Studies using more gentle deprivation methods reported varying effects on vigilance states.Housing cages on orbital shakers was reported to reduce REMS by approximately 25% in prairie voles, accompanied with an approximately 10% increase in waking and shorter REMS and NREMS bouts (Jones et al., 2019).A similar, but shorter, deprivation method in rats resulted in an approximately 50% reduction of REMS together with an approximately 10% increase in NREM and 5% waking (Feng and Ma, 2003).Automated cage shaking or the inclination method reduced REMS 80-90% with no effect on NREMS (Ravassard et al., 2016, Rosier et al., 2018).Gentle handling of mice upon REMS detection resulted in complete REMS suppression, together with substantial loss of NREM (Zhou et al., 2020).Overall, these findings suggest that automated REMS detection in combination with gentle cage movement is the most effective method to selectively suppress REMS, and, importantly, to prevent induction of major stress responses (Ravassard et al., 2016).Together, application of stressful deprivation methods and lack of validation of REMS-specific suppression in most studies obscures conclusive interpretation of the results from animal model studies.
Overall, it is important to acknowledge that variations in behavioral and physiological outcomes in animal studies may result from diverse factors, including species-, strain-, and sex-specific effects.Additionally, discrepancies in methodology, including factors such as platform and arena dimensions, along with social context, may account for the variability of the observed outcomes.

Synthesis
Both human and animal studies use sensory stimuli, mechanical disturbance or touching, and pharmacology to suppress REMS.While mechanical disturbance in animal models, such as gentle cage movement, is a valid method, shaking or touching may be considered too invasive in humans.However, human studies show that along with increasing REM pressure, more intensive intervention methods are needed, that may also increase the stressfulness of the experiment.Animal and human study paradigms also point to a likely partial suppression of REMS instead of total deprivation, which is very challenging to achieve.Changes in sleep architecture were also seen in both human and animal model studies.In humans, suppressing REMS led to an increase of NREMS, and in animals, both increase and suppression of NREMS was observed.
A marked difference across human and animal studies is the duration of the deprivation protocol.The contemporary studies in humans mostly use one-night paradigms whereas studies in the animal models extended up to 3 weeks.This difference is of major importance, as the question of the effects of acute vs accumulating/chronic REM suppression is clearly lacking from human studies.This creates a significant gap in the knowledge, given that some commonly used psychopharmacological compounds suppress REMS over long time periods and its effects are not known.This also hinders comparison of the outcomes of animal models and studies in humans, as long-term REMS disturbance is likely to have specific underlying mechanisms.Animal models emphasized the lasting impact of REMSD experienced in early life.Comparable study designs are not available in humans.
The reviewed studies in humans contribute to understanding the role of REMS in adaptive emotional regulation processes mostly in the immediate time perspective.In long-term suppression, the potential plasticity effects may turn into changes in brain structures, but this has not been investigated in the reviewed studies.

Manipulation of neuronal pathways
As previously demonstrated, methods for REMSD have inherent limitations, as they can also disrupt other states of vigilance and induce stress.However, methodological advancements available in rodent models, such as application of optogenetics to activate or inhibit specific neuronal pathways, offer the potential to elucidate the underlying mechanisms of REMS and its influence on behavioral processes, including the processing of emotional memories.
Using optogenetics, Boyce et al. (Boyce et al., 2017) demonstrated that silencing of medial septum GABAergic neurons, resulting in specific suppression of REMS theta oscillations, impaired consolidation of contextual fear memory and non-emotional object memory (Boyce et al., 2017).Notably, this impairment was observed in contextual fear memory but not in amygdala-dependent cued fear conditioning.These findings provide compelling evidence for a causal role of REMS theta oscillations in the process of memory consolidation.Another study demonstrated the contribution of hippocampal adult-born neurons to REMS-dependent memory consolidation by showing that optogenetic silencing of this cell population during REM, but not NREM, impaired contextual fear memory.This effect was also specific for hippocampal-dependent contextual conditioning (Kumar et al., 2020).Recent work also showed that activation of REM-promoting melanin-concentrating hormone-producing neurons in the hypothalamus during REMS was impaired, while their inhibition improved contextual, but not cued, fear memory.This suggests that these REMS-active neurons contribute to the attenuation or suppression of fear memories (Izawa et al., 2019).Moreover, in a recent study, discrimination between conditioned safe and danger signals were regulated by decoupling of dendritic and somatic inhibition in cortical neurons during REMS (Aime et al., 2022).Notably, REM-specific optogenetic disruption of somatic inhibition enhanced responses to an aversive cue, while loss of dendritic disinhibition diminished the ability to differentiate between safe and aversive cues.Collectively, these findings provide strong evidence for the significant influence of REMS on the processing of emotional (fear) memories, encompassing the facilitation of consolidation and forgetting and the discrimination between safety and danger cues during associative learning.Additionally, these studies underscore the engagement of multiple distinct neural circuits, suggesting their distinct contributions to processing of emotional and object location memory during REMS.

The controversy related to use of antidepressants
While clinical depression is often followed by an increase of REMS, use of antidepressants has the opposite effect.Specifically, antidepressants suppress REMS even in healthy volunteers to a significant degree.The suggested mechanism is increased levels of synaptic serotonin, mediated by the 5-HT 1A receptors (Wilson and Argyropoulos, 2005), but may also include plasticity-related mechanisms important for their antidepressant effect (see e.g.(Castrén, 2023)).The commonly used antidepressants citalopram, fluvoxamine, paroxetine, and sertraline continuously suppress REMS by approximately 30-60%.Some MAOIs reduce REMS by >60% (for a review, see (Wilson and Argyropoulos, 2005)).
In accordance with the concept that REMS dysregulation has been suggested as an endophenotype of depression, a four-week treatment with antidepressants (SSRI or SNRI) was more efficient than psychotherapy in depression treatment if there was a higher REM density at the baseline, suggesting an overactive phasic REMS activity (Lechinger et al., 2021).However, some controversies remain.Not all depressive states are associated with REMS dysregulation, and even in the case of dysregulated REMS, its chronic suppression may bring along other unknown adverse effects.
Specific attention should be placed on youth.Already in 1960 s, Roffward et al. (Roffwarg et al., 1966) discovered the important role of REMS in brain development and in brain plasticity processes.He formulated the ontogenetic hypothesis of REMS, stating that REMS contains endogenous neural activity necessary for brain maturation towards normal adult brain function (Roffwarg et al., 1966, Frank, 2020).
Recent evidence from the adolescent population shows that SSRIs increase muscle activity during REMS, thus disturbing the muscle atonia characteristic of normal REMS (Ferri et al., 2021).The potential implications following the suppression or even deprivation of REMS due to SSRI treatment is clearly understudied; more discussion on the advantages and disadvantages of the effects of antidepressants on sleep is needed.A more detailed understanding of REMS in affective processing is then of major interest also from the psychopharmacological perspective.Further research is also warranted for the prenatal period, as maternal SSRI medication during pregnancy is associated with increased affective problems in the offspring (Malm et al., 2016) later in their development.A conservative approach would be to use medication with the fewest REMS effects whenever possible.

Limitations
We recognize that REMSD is only one experimental tool among others to study the role of REMS in affective processing.It is also evident from the reviewed studies that REMSD may be a source of stress, and depriving one sleep stage has side effects for the remaining sleep architecture.
Second, the concept of affective memory is not entirely clear from the perspective of its declarative and episodic components.Given that an experience may always contain declarative memories, the assessment of SWS in the offline processing of emotional content is likely to be equally important.The system consolidation hypothesis (Born and Wilhelm, 2012) argues that SWS serves as an offline period in which newly acquired hippocampus-dependent declarative memories are gradually transferred to neocortical networks.This process is not random but merely an active process with a selection of relevant memories (Born and Wilhelm, 2012).The role of REMS in this process is not clear, but arguments have been presented for the sequential hypothesis of memory consolidation.The hypothesis argues that memory processing during sleep requires the initial participation of SWS in addition to the subsequent involvement of REMS for optimal memory consolidation and related synaptic plasticity (Giuditta, 2014).The role of SWS is in sorting and retaining select memories, downscaling irrelevant or competing memory traces, and integrating retained memories with preexisting memories.SWS would then set the foundation for the information processes offline during REMS.REMSD clearly disturbs this sequential continuity.
As another limitation, we did not review targeted memory reactivation (TMR) studies, as they do not suppress or deprive REMS.However, this method is relevant for the affective processing, as it experimentally induces memory reactivation during sleep to boost the offline memory consolidation processes (Hu et al., 2020).While most of the TMR studies focus on SWS sleep, recent studies have reported memory replays also in REMS (Abdellahi et al., 2023), although their role for specifically affective processing remains unclear.
Additionally, some of the prior evidence regarding the role of REMS in affective processing is based on a split-night protocol, e.g.(Wagner et al., 2001).This protocol compares the SWS-rich sleep in the first half of the night with REM-sleep-rich sleep in the second half for the consolidation of emotional memories and is probably less stressful than undergoing an experimental deprivation.However, this method is usually applied in quasi-experiments not directly manipulating REMS, thus being the reason for exclusion from this review.

Future directions
The question of a critical threshold for a sufficient amount of REMS allowing appropriate offline processing of emotions remains open.In human studies, the variation in the duration of REMSD (partial vs entire suppression) was large, and the protocols not only decreased the amount of REMS but also induced REMS discontinuity.Evidence from REMS fragmentation studies (Pesonen et al., 2019, Wassing et al., 2019, Wu et al., 2021) indicates that the continuity of the REM episodes may also impact affective processing.Therefore, it is not clear whether the reported effects of REMSD in the studies published to date are due to suppression or fragmentation of REMS.Moreover, none of the studies examined power spectral densities and especially theta wave activity to investigate specific oscillation mechanisms in REMS, which are assumed to underlie offline processing of affects.
One important area requiring further investigation is the effect of antidepressants in suppressing REMS.This is especially important given the wide usage of these drugs and the emerging understanding of the significance of REMS on emotional processing.The mechanistic basis of this effect remains unknown, and in this context the recent discovery that SSRIs act by inducing neuronal plasticity through TrkB-BDNF signaling (Casarotto et al., 2021) is especially interesting, in addition to their effects on serotonin signaling.
Although it is evident that the complexity of human affective processes cannot be fully recapitulated in animal models, they are fundamental for dissecting the underlying mechanisms.Methodological advances, such as optogenetics, which allow cell type-and circuitspecific manipulation with high spatiotemporal control, are beginning to elucidate the underlying mechanisms of REMS, including its role in emotional memory processing.Importantly, such techniques enable selective manipulation of brain activity without triggering major stress responses or altering overall sleep architecture and are expected to continue to increase understanding of REMS-controlled circuits and behaviors.
One study examined the effect of REMSD (compared with SWSD or no deprivation) on general affect, emotional reactivity, and associated neural mechanisms of emotion regulation during the acute experience of social exclusion (Glosemeyer et al., 2020).In this combined polysomnography and fMRI study, the general affect was examined with Positive and Negative Affect Schedule (PANAS) that the participants completed before going to bed and after waking up.Emotion regulation during social isolation was examined using a game Cyberball during fMRI (Glosemeyer et al., 2020).One study used an Emotional Reactivity Task in the MR scanner to examine the effect of REMSD on reactions to threatening stimuli (Rosales-Lagarde et al., 2012).

Pain sensitivity
Three studies examined the effect of REMSD on pain perception and sensitivity.One study examined thermal nociceptive threshold and pain perception using subjective ratings of warmth and pain following a laser stimulus using a visual analog scale and laser-evoked potentials using EEG (Azevedo et al., 2011).One study examined changes in pain threshold following REMSD and TSD as finger-withdrawal latency to heat stimulus (Roehrs et al., 2006).One study used both thermal and pressure stimuli to assess changes in pain sensitivity after SWS or REMS interruption (Onen et al., 2001).
Animal studies

Depression-and anxiety-related behavior
Depression-or anxiety-like behavior (or both) were the affective outcomes in several studies (Table 2).Typically, tests for both depression-and anxiety-like behavior were included.Anxiety-like behavior was typically assessed with the open field, elevated plus maze, or lightdark box tests, all of which are based on the animal's exploratory activity.Depressive-like behavior was often assessed with the forced swimming or tail suspension test, in which active escape behaviors are measured.Furthermore, sucrose consumption was used in some studies to assess anhedonia-like behavior.In a few of the early rodent studies (Hicks et al., 1979, Kovalzon and Tsibulsky, 1984, Oniani, 1984), "emotionality" was reported as an outcome related to fear behavior.For example, defecation, urination, and freezing in an open field arena were interpreted to reflect emotional or fear reactivity.

Social interaction, sexual behavior, and aggression
The effects of REMSD on social behavior, apart from sexual and aggressive behaviors, were rarely explored.One study in prairie voles assessed social bonding using the partner preference and olfactory social recognition tests (Jones et al., 2019).Sexual behavior was assessed in several of the rodent studies, which included, for example, measures of mounting, genital exploration, and ejaculation in males (Feng and Ma, 2003, Damasceno et al., 2008, Alvarenga et al., 2009, 2013, Zhang et al., 2022) and proceptive and rejection responses in females (Andersen et al., 2009).In some studies, aggressive behavior was the main outcome, which was measured typically by observing behavior of paired rats (Sloan, 1972, Hicks et al., 1979, Trotta, 1984, Peder et al., 1986).

Emotion recognition
One study examined emotion recognition in domestic dogs (Bolló et al. 2020).The emotion recognition task was conducted after a 3-h nap.The dogs were presented simultaneously with a pair of emotional face images (positive and negative) among an emotional sound playback (Hicks et al., 1979, Kovalzon and Tsibulsky, 1984, Oniani, 1984).

Pain
Pain sensitivity was studied in several of the eligible articles (Table 2).The effects of REMSD on pain were assessed by using different stimuli, including mechanical, thermal, electric, and chemical.Mechanical pain sensitivity was typically assessed by mechanical paw withdrawal using the von Frey test.Thermal sensitivity was typically measured with the hot plate or hot water immersion, in which the threshold was determined by the paw/tail withdrawal latency.A quantitative analysis of the frequency of vocal responses or tail-arch response to electric stimulus was the measurement for electric pain threshold.A formalin test was applied to determine the threshold for a chemical stimulus.
Postsurgical pain response was studied in some of the included articles and was measured with a paw-withdrawal threshold in response to a mechanical stimulus and paw-withdrawal latency to hot and cold stimuli (Wang et al., 2015, Li et al., 2022).

Fear conditioning and extinction
The contextual and cued fear conditioning tests were used in several of the studies to investigate the impact of REMSD on fear learning and memory (Bueno et al., 1994, Dametto et al., 2002, Silvestri, 2005, Fu et al., 2007, Tiba et al., 2008, Ravassard et al., 2016, Hunter, 2018, Rosier et al., 2018, Zhou et al., 2020, Jung and Noh, 2021).During the acquisition phase, rodents were exposed to a conditioned stimulus (CS), either a context or a cue (auditory/visual), followed by an unconditioned stimulus (US), here an aversive foot shock.After an interval, which was typically 24 h but extended up to 30 days, the rodent was placed back to the same context (contextual fear memory) or the CS was presented in a novel context (cued fear memory) and freezing response was used to assess fear associations.Moreover, the influence of REMSD on fear extinction, elicited by repeated exposure to the conditioned context or cue, was assessed in some of the studies (Silvestri, 2005, Fu et al., 2007, Hunter, 2018, Jung and Noh, 2021).In addition to the classical conditioning task, the passive avoidance test, which assesses fear-related avoidance behavior, was used in some studies (Oniani, 1984, Bueno et al., 1994, Dametto et al., 2002, Patti et al., 2010, Lima et al., 2014).

Table 1
Human experiments by outcomes and chronological order.