Rapid eye movement (REM) sleep is characterized by bizarre dreaming consciousness upon lack of external input (Hobson, 2009; Nir and Tononi, 2010), which is accompanied by specific neurophysiologic signatures including wake-like low frequency desynchronized electroencephalogram (EEG) dominated by theta and gamma EEG oscillation, swift occurrence of REMs and pontine-geniculate-occipital waves and absence of muscle tone (Rechtschaffen and Kales, 1968; Hobson and Pace-Schott, 2002; Hobson et al., 2014). Remarkably, amongst all sleep-wake stages, REM sleep has a prominent role in enabling neuronal plasticity, increased synaptic connectivity, and immediate early genes synthesis (Ribeiro et al., 2002; Grosmark et al., 2012) and is signified by a strong cortical activation (De Gennaro et al., 2004; Massimini et al., 2010). Thus, the functional roles of REM sleep and dreaming have been of sustained interest. In healthy adults, REM sleep has been shown to support many adaptive functions. These include consolidation of emotional memory (Wagner et al., 2001; Nishida et al., 2009), resolution of affect (van der Helm et al., 2011), further transformation of previously consolidated during non-REM (NREM) sleep memories (Walker and Stickgold, 2010; Rasch and Born, 2013; Llewellyn and Hobson, 2015), consolidation of procedural or implicit memory and motor learning (Yordanova et al., 2008; Diekelmann and Born, 2010), and promoting human heuristic creativity (Cai et al., 2009; Brand et al., 2010). REM sleep’s physiological and psychological features also have been associated with more complex functions. For example, REM sleep has been suggested to heighten autobiographic memory (Horton and Malinowski, 2015; Malinowski and Horton, 2015) and to render previously encoded memories more distinct through its hyper-associative dreaming state (Llewellyn, 2013; Llewellyn and Hobson, 2015). Also, it is thought to incorporate these previously encoded memories into a broader vital context, thus embedding them in consolidated residuals of hypotheses, emotions, basic needs, and individual genetic traits (Kirov, 2013). Next, REM sleep has been proposed to generate an innate virtual model of the world, thus modulating predictive coding (Hobson and Friston, 2012; Hobson et al., 2014; Hopkins, 2016). In view of REM sleep deviations in almost all psychiatric conditions (Benca et al., 1997; Gottesmann and Gottesman, 2007; Baglioni et al., 2016), REM sleep also has been regarded as a mechanism mediating brain adaptation in normal and pathological conditions (Benca et al., 1992; Horne, 2013, 2015; Goldstein and Walker, 2014; Hobson et al., 2014; Hopkins, 2016; Mota et al., 2016).

From a developmental perspective, infants have much more REM sleep quantity, which descends through childhood and adolescence, than adults (Roffwarg et al., 1966; Brand and Kirov, 2011). It has been proposed that this developmental decrease in REM sleep sub-serves brain maturation through synaptic reorganization and/or pruning, internally generated stimulation of neuronal assembles, or genetic programming (Marks et al., 1995; Jouvet, 1998; Feinberg and Campbell, 2010; Hopkins, 2016). In this regard, an insufficient decline or variations of the normal REM sleep decline during development is proposed to underpin a broad spectrum of child and adolescent psychiatric disorders (Partonen, 1998; Kobayashi et al., 2004; Garcia-Rill et al., 2008; Brand and Kirov, 2011; Kirov and Brand, 2014). Given that both the hypothalamus-pituitary-adrenocortical axis activation and REM sleep overdrive are closely associated in psychiatric conditions (Steiger et al., 2013), this notion has received an indirect support by documenting an existence of elevated cortisol levels in association with disturbed sleep and impaired neurobehavioral functions in a cohort of children with various psychiatric symptoms (Hatzinger et al., 2012), and thus, probably, with deviant stress sensitivity (Brand and Kirov, 2011; Gruber, 2014). However, whether and how REM sleep in common developmental child psychiatric disorders may be linked to neurobehavioral functioning is still less well understood.

We have shown previously a REM sleep overdrive in children with attention-deficit/hyperactivity disorder (ADHD), Tourette syndrome/chronic tic disorder (TD) and ADHD/TD comorbidity, with this REM sleep overdrive being associated mostly with ADHD core symptoms (Kirov et al., 2007). More recently, we have demonstrated the following pattern of associations between REM sleep quantity and neurobehavioral functions in youth ADHD: (1) In children with ADHD, REM sleep proportion correlated positively with inattention and negatively with performance intelligence quotient (IQ). (2) In opposition, the proportion of REM sleep in typically developing children (TDC) correlated negatively with inattention and positively with performance IQ (Kirov et al., 2011). Similarly, another recent study showed that whereas in youths with ADHD, theta (4–8 Hz) EEG power during REM sleep correlated negatively with emotional memory consolidation, in healthy individuals, this correlation was positive (Prehn-Kristensen et al., 2013). Collectively, these latter findings suggest at least a bi-directional role of REM sleep and its physiology, depending on presence or absence of ADHD psychopathology.

The present study aimed at further investigating the impact of REM sleep on neurobehavioral functioning in children with a broader spectrum of common developmental psychiatric disorders. In an attempt to clarify if the associations between REM sleep proportion and neurobehavioral functions reported previously (Kirov et al., 2011) were only linked to ADHD psychopathology possibly reflecting a disorder-specific dysfunction of neural regulation leading to both daily symptoms expression and sleep disturbances (Kirov et al., 2007), we enrolled in the present study larger sample sizes of children with ADHD, TD and ADHD/TD comorbidity and compared them to healthy TDC, while testing associations not only between REM sleep parameters and psychopathological scores, but also considering children’s IQ. We hypothesized a differential impact of REM sleep on cognitive and behavioral functioning in the children with the continuum of common developmental psychiatric disorders (Gaze et al., 2006; Kirov et al., 2011) compared to the TDC. Also, we proposed that REM sleep may be associated dissimilarly with different psychopathological scores across the children’s specific psychiatric diagnoses.

The key findings of the present study were that among a sample of children diagnosed with ADHD, TD and ADHD/TD comorbidity, the more REM sleep amount was associated with higher scores of inattention and lower scores of performance IQ. By contrast, in TDC, the more REM sleep amount was associated with lower scores of inattention and higher performance IQ. Further, whereas in children with psychiatric disorders, the shorter REM sleep latencies were associated with higher scores of inattention, no such association was found for the TDC. The present findings add to the current literature in that we observed “double-edged” associations between REM sleep PSG parameters and neurobehavioral functioning, not only in ADHD (Kirov et al., 2011; Prehn-Kristensen et al., 2013). Notably, this study clearly showed that this opposite impact of REM sleep on daytime neurobehavioral functions is also observable in a broader spectrum of developmental child psychiatric conditions, as compared with TDC. Collectively, these findings give support to our hypothesis that REM sleep parameters may be associated dissimilarly with daytime neurobehavioral functions, depending on presence or absence of developmental psychiatric disorders. Although the relative REM sleep proportion in each of the three psychopathological groups was predicted consistently by inattention, some observations (Table 3) merit further attention. First, the observed positive association between REM sleep proportion and interference LOI scores in the TD group might underline the closeness of TD with subclinical obsessive-compulsive symptoms. Second, the modest differences in REM sleep predictors across the groups (Table 3 and the reported within text results) may be accounted for by variations in statistical power due to the relatively small sample sizes included in each group.

The present data do not allow a deeper introspection into the exact neurobiological and psychological mechanisms underlying the pattern of associations, as found and described above. It is notable, however, that currently reported REM sleep alterations in children with a broad spectrum of psychiatric disorders correspond to findings in adults where psychiatric conditions such as depression, major depressive disorder, and post-traumatic stress disorder are featured by enhanced amounts of REM sleep (Benca et al., 1997; Wang et al., 2015; Baglioni et al., 2016). In adults, two parallel ways with a common source in the monoaminergic or cholinergic systems have been assumed to regulate both REM sleep and symptoms of depression (Wang et al., 2015). Likewise, we have proposed previously that altered aminergic-cholinergic ratio may lead to a REM sleep overdrive in the spectrum of developmental psychiatric conditions, as shown in the present study (Kirov et al., 2004, 2014; Brand and Kirov, 2011; Kirov and Brand, 2014). Further, specific brain regions including the hippocampus, amygdala and medial pre-frontal cortex may contribute to both daily symptoms and REM sleep impairments in both depressive adults (Wang et al., 2015) and children with ADHD (Hart et al., 2013) While the precise neurobiological mechanisms of the co-existing deviations in daily behaviors and REM sleep remain to be established (Wang et al., 2015), our current findings on the relationships between REM sleep amount and daily behavior both in patients and healthy children may open a relevant new line of understanding by considering the contra-directionality of these relationships. If increased REM sleep amounts in children were expressions of co-impaired regulation of REM sleep and attention, increased REM sleep might not predict superior achievements in TDC. If, on the contrary, increased REM sleep in patients reflected compensation, it should have predicted less symptoms severity in patients. The currently found bi-directionality of associations suggests that functional efficiency of REM sleep might be critically impaired in children with psychiatric disorders. There is evidence that the functional efficiency of REM sleep may vary both in pathological and normal conditions. For example, chronic stress in rats has been found to synchronize the theta rhythm between the hippocampus and amygdala, which was accompanied by increased amounts of REM sleep (Hegde et al., 2011). Further, Pellicciari et al. (2013) provide evidence for functional inefficiency of REM sleep alpha activity in depression in humans, which can be remediated by repetitive trans-cranial magnetic stimulation. These observations imply that the functional rhythms and associated mechanisms during REM may be primarily impaired in children with psychiatric disorders. Hence, in pathology, the increased REM sleep amount may reflect an attempt to compensate for functional inefficiency. While functional inefficiency of REM sleep may not be compensated by increased REM sleep amount in patients, increase in functionally efficient REM sleep in TDC is associated with an improved attention and higher performance IQ.

The present results also imply effects on psychological functioning. REM sleep plays a role in the consolidation of negative emotional memories (Wagner et al., 2001; Nishida et al., 2009; Goldstein and Walker, 2014) and resolution of affect through dissipation of amygdala activity in response to previous emotional experiences, thus reducing next-day subjective emotionality (van der Helm et al., 2011). In this regard, the present results suggest that those REM sleep neuronal mechanisms sub-serving successful emotional processing may be intact in the TDC and may be insufficient or impaired in the children with the spectrum of psychiatric disorders. Thus, our results point tentatively to a possible contribution of emotional liability and related anxiety in these child psychiatric disorders (Banaschewski et al., 2012; Gruber, 2014), Since a wealth of empirical evidence show that children with common developmental psychopathologies display greater difficulties in coping with emotional problems, inappropriate behaviors and social interactions (Swain et al., 2007; Brand and Kirov, 2011; Kirov and Brand, 2014; Kirov et al., 2014), this may affect the natural functions of REM sleep and may lead to impaired attention and procedural skills, respectively. Further, from a view point of the aminergic-cholinergic reciprocal interaction model for NREM-REM sleep cycle regulation (Hobson et al., 1975), we have proposed previously that altered aminergic-cholinergic ratio may lead to a REM sleep overdrive in the spectrum of developmental psychiatric conditions, as shown in the present study (Kirov et al., 2004, 2014; Brand and Kirov, 2011; Kirov and Brand, 2014). Though speculative, changes in this ratio may lead to dissimilar functions of REM sleep for implicit motor memory. Indirect support to this assumption comes from two consecutive studies. Whilst suppression of REM sleep induced by noradrenergic agonists led to a slight improvement of procedural memory, suppressing REM sleep by application of the cholinergic antagonists have produced opposite effects (Rasch et al., 2009a,b). Thus, REM sleep physiological, neurochemical and psychological features may interact differently with pathological traits in psychiatric conditions and, on the other hand, in normative behavioral functions in healthy youths.

Our findings are in accord with previous research, observing an adverse impact of REM sleep overdrive or its EEG signatures on daytime neurobehavioral functions in children with ADHD compared to TDC (Kirov et al., 2007, 2011; Prehn-Kristensen et al., 2013). Yet, they are at odds with findings from other studies. An earlier study has found a less REM sleep proportion and prolonged REM sleep latency in a community-based sample of 5- to7-years-old children with ADHD relative to controls, and has demonstrated that the less REM sleep and the longer REM latency correlated positively with inattention and hyperactivity/impulsivity (O’Brien et al., 2003). Another study conducted among 6- to 16-years-old children with TD, ADHD and TD/ADHD comorbidity, diagnosed according to DSM-III-R (American Psychiatric Association [APA], 1987) did not find any changes in REM sleep, but did find modest correlations between movements in REM sleep and hyperactivity/impulsivity (Stephens et al., 2013). Recently, it has been shown a REM sleep overdrive in un-medicated children with ADHD which did not correlate to any psychopathologies or impaired daytime behaviors (Virring et al., 2016). Notably, however, our patients and controls differed from those in the above studies in age, applied diagnostic criteria and medication. Hence, as proposed by Kirov and Brand (2014), differences in age, diagnostic criteria used and medication status could significantly contribute to controversial sleep PSG findings in youths with ADHD and their effect on daytime behavior. Although further studies are needed to clarify the role of neurobiological variables, the present findings show that the delayed and/or deviant developmental decrease in REM sleep might represent a risk factor for developmental psychopathology.

Despite of the present findings, several considerations warrant against overgeneralization of the results. First, since periodic limb movements in sleep (PLMS) and sleep-disordered breathing (SDB) are among the most common sleep disorders in ADHD (Cortese et al., 2009), their presence was not reported here, because our focus was on the associations between sleep stages and daytime neurobehavioral functions. Thus, the role for PLMS and SDB in our findings mandates future investigations. Second, the sample sizes used are relatively small and heterogeneous in terms of age, gender and medication status before study. However, none of the multiple regression analyses extracted age, gender and medication status as predictors. Third, while it would be helpful to supplement our results by providing data about presence of parasomnias, sleep hygiene and circadian rhythm disorder, the lack of sleep diaries and actigraphy precludes such observations. Last, the present results might have emerged due to further latent neuroendocrinologic variables, which might have biased two or more dimensions in the same or opposite direction (Steiger et al., 2013).

Our results indicate an opposite impact of REM sleep on neurobehavioral functioning, depending on presence or absence of developmental psychiatric disorders. Thus, the development of REM sleep functions and the way how these relate to different kinds of children’s behavior later on seems to depend on different factors (e.g., genes, environment, adaptation, etc.) influencing the child’s brain maturation either to a normal or an aberrant neuronal system.

RK, TB, and AR: Substantial contributions to the conception and design of the work. RK, SB, and TB: Interpretation of data and drafting the manuscript. RK, SB, and TB: Statistical analysis. RK and TB: Data selection and matching the groups for age and gender. SB, TB, and AR: Final approval of the paper draft and agreement to be accountable for all aspects of the work.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.