Fig. 21.1
Cerebral glucose metabolism (CGM) and regional cerebral blood flow (CBF) during deep NREM sleep (first column), REM sleep (second column), and wakefulness (third column). Row a CGM quantified in the same individual at 1-week interval, using FDG and PET. The three images are displayed at the same brain level using the same color scale. The average CGM during deep NREM sleep (versus wakefulness) is significantly decreased. During REM sleep, the CGM is as high as during wakefulness. Row b1 Distribution of the highest regional brain activity, as assessed by CBF measurement using PET, during wakefulness and REM sleep. The most active regions during wakefulness are located in the polymodal associative cortices in the prefrontal and parietal lobes (both on the medial wall and convexity). During REM sleep, the most active areas are located in the pontine tegmentum, thalami, amygdaloid complexes, and anterior cingulate cortex. Other data (not shown) have shown a large activity in the occipital cortices, insula, and hippocampus [15]. Row b2 Distribution of the lowest regional brain activity, as assessed by CBF measurement using PET, during NREM and REM sleep. In both sleep stages, the least active regions are located in the polymodal associative cortices in the prefrontal and parietal lobes (convexity). During NREM sleep, the brainstem and thalami are also particularly deactivated
Regional brain activity in subcortical mesopontine and thalamic regions during human REM sleep [14, 31] is in keeping with our current understanding of sleep generation in animals. REM sleep is generated by neuronal populations of the mesopontine reticular formation that activates the thalamic nuclei, which in turn activate the cortex [28].
Functional connectivity between remote brain areas is also modified during human REM sleep. The functional relationship between striate and extrastriate cortices, usually excitatory, is reversed during REM sleep [15, 33]. Likewise, the functional relationship between the amygdala and the temporal and occipital cortices is different during REM sleep than during wakefulness or NREM sleep [34]. This pattern suggests that functional interactions between neuronal populations are different during REM sleep than during wakefulness. A recent resting-state fMRI study demonstrated that REM sleep connectivity is also different from NREM sleep connectivity, specifically in the set of brain areas known as the default-mode network [35]. The default-mode network consists of regions activated when the brain is not engaged in externally oriented behavior [36]. The functional connectivity within the default-mode network is diminished in NREM sleep, but in REM sleep, it is comparable to that in wakefulness [35]. Moreover, in REM sleep, the higher-order association cortices (encompassing the default-mode network) engage in a rapid oscillatory fluctuation of anticorrelated activation with the sensorimotor areas [35]. These characteristic patterns of activation may underlie neuroplastic and phenomenal aspects of REM sleep.
Pontine waves, or ponto-geniculo-occipital (PGO) waves, are also primary features of REM sleep. In rats, the generator of the pontine waves projects to a set of brain areas shown to be active in human REM sleep: the occipital cortex, the entorhinal cortex, the hippocampus, and the amygdala as well as brainstem structures participating in the generation of REM sleep [37]. Although most easily recorded in the pons [38], the lateral geniculate bodies [39], and the occipital cortex in cats [40], PGO waves are also observed in many parts of the brain, including limbic areas (amygdala, hippocampus, cingulate gyrus) [41]. Several observations suggest that PGO waves also occur during human sleep. In epileptic patients, direct intracerebral recordings in the striate cortex showed monophasic or diphasic potentials during REM sleep, isolated or in bursts [42]. In normal subjects, surface EEG revealed transient occipital and/or parietal potentials time-locked to the REMs [43]. Source dipoles of MEG signal were localized in the brainstem, thalamus, hippocampus, and occipital cortex during REM sleep [44, 45]. Using PET, we showed that the rCBF in the lateral geniculate bodies and the occipital cortex is tightly coupled to spontaneous eye movements during REM sleep, but not during wakefulness [46]. This finding has been confirmed by fMRI studies [47, 48]. Although fully conclusive components are still awaited, these elements support the hypothesis that PGO-like activities participate in shaping the distribution of regional brain activity during human REM sleep.
Many subjective aspects of dreams may be related to changes in brain activation during REM sleep [49], as evidenced by a series of PET studies. Increased perfusion in the occipital and temporal lobes, in the absence of afferent sensory input, may relate to the visual and auditory experiences that characterize dreams [33]. Indeed, lesions in the occipital cortex are associated with suppressed visual mentation during dreams. Increased perfusion in the motor and premotor areas during REM sleep may relate to movement perception in dreams [8, 15]. These intended movements are not physically enacted because efferent motor signals are suppressed in REM sleep by muscle atonia. This suppression can be inhibited by pontomedullary lesion, as has been shown in cats. Lesioned cats then engage in coordinated movement during REM sleep, ostensibly acting out their dreams. Similar behavior is also observed in humans with REM sleep behavior disorder (RBD) (see section below). The amygdala is particularly active during REM sleep [15, 31], perhaps underlying the potent emotions experienced in dreams. Regional deactivations during REM sleep are also related to certain elements of dreams. The lateral prefrontal cortex, an area involved in the monitoring of complex cognitive processes during waking, shows marked hypoperfusion during REM sleep [8]. Accordingly, dreams are often characterized by a bizarreness and incoherence of which the dreamer is unaware. The failure to coherently access episodic memory during dreams is also attributable to lateral prefrontal deactivation. In contrast, the medial prefrontal cortex remains activated in REM sleep. The representation of others’ minds, known as theory of mind, is a function thought to be subserved by this area, and indeed, dreams contain characters with thoughts, emotions, and intentions of their own. However, the delimitation of the dreamer’s mind is less distinct. Dreamers commonly move between first-person and third-person perspectives over the course of a dream. The inability to differentiate first- and third-person perspective may be associated with decreased perfusion in the inferior parietal cortex [50].
Recently, Dresler et al. [51, 52] were able to image lucid dreaming, a rare and intriguing feature of REM sleep. Unlike in normal dreams, lucid dreamers can gain awareness of the fact that they are dreaming, can access long-term memory stores, and can willingly control the events of their dream. Moreover, lucid dreamers can, to a slight extent, control their real eye and hand movements while maintaining polysomnographic sleep. By instructing the dreamer to provide gestural signals at the onset of a lucid dream, researchers determined that several areas normally deactivated in normal dreaming become activated during lucid dreaming, including the dorsolateral prefrontal cortex [51]. In a related experiment, Dresler et al. [52] asked participants to perform a pretrained motor task during lucid dreaming. They found that similar areas were activated during the dreamed mentation of a motor task and during its execution in wakefulness. It should be noted that both of these studies are limited by very small samples of 1 or 2 participants, owing to the rarity of lucid dreaming.
Brain Reactivity to External Stimulation During Sleep
Electrophysiologic studies have demonstrated that sleep is not a state of complete unresponsiveness to external stimuli (e.g., Perrin et al. [53]). Early studies have shown that external stimuli can induce an autonomic or electrophysiologic response during human sleep, in particular after a relevant or meaningful stimulus presentation [54]. Available PET and fMRI data globally suggest that the processing of external stimuli can proceed beyond the primary cortices during NREM sleep. However, the mechanisms by which salient stimuli can recruit associative cerebral areas during sleep remain unclear. A pioneering fMRI study found that, during NREM sleep as during wakefulness, several areas continue to be activated by external auditory stimulation: the thalamic nuclei, the auditory cortices, and the caudate nucleus [55]. Moreover, the left amygdala and the left prefrontal cortex were found to be more activated by subjects’ own names than by pure tones, suggesting the persistence during sleep of specific responses for meaningful or emotionally laden stimuli.
Other groups observed that auditory stimulation induced a decreased response in the auditory cortex [56, 57]. Intriguingly, visual stimulation during SWS in adults elicited a decrease in activity in the occipital cortex [58]. This decrease was more rostral and dorsal compared to the relative rCBF increase along the calcarine sulcus found during visual stimulation in the awake state. The origin of this negative blood oxygenation level is still unclear [59]. One fMRI study investigated responsiveness to sounds during REM sleep, and detected auditory cortex activation during tonic but not phasic REM sleep. The non-responsiveness observed in phasic REM sleep was interpreted as a state of functional isolation [60].
Processing of external stimuli during sleep may be dependent on spontaneous brain activity, such as sleep spindles. In order to explore this interaction, we conducted an EEG/fMRI study involving the presentation of pure tones to participants sleeping in the scanner [61]. When comparing blood-oxygen-level-dependent (BOLD) activation from tones presented during and outside of sleep spindles, we found a stark difference in global activation. Whereas tones played outside of spindles elicited activation in the thalamus and auditory cortex, replicating earlier results [55], tones played during spindles produced no significant activation, save for a small area in the brainstem [61]. These results show that sleep spindles effectively hinder the transmission of external auditory stimulation to the cortex during NREM sleep. This finding may explain variable susceptibility to noise disturbance during sleep, within and between subjects. Indeed, in another study, sleep stability in the presence of auditory disturbance was correlated with spindle density [62]. The “noise-canceling” function of sleep spindles is also consistent with its hypothesized role in sleep-dependent memory consolidation, in that it may contribute to brain functional isolation necessary for endogenous information processing in the absence of contaminating inputs. In support of this notion, hippocampal connectivity with neocortical areas was increased during sleep spindles, suggesting information transfer akin to systems consolidation [63]. In the same EEG/fMRI study, we also investigated the interaction between incoming auditory stimulation and K-complexes. It was found that sound-induced K-complexes were associated with increasing activation in the auditory cortex [61]. This finding suggests that K-complexes reflect a facilitated processing of sensory information at the cortical level during NREM sleep. Finally, in a complementary study, we examined the effect of slow-wave phase on the processing of auditory stimulation in NREM sleep [64]. We found a larger activation of the superior temporal gyrus with sounds presented during the up state of the slow wave compared to those presented during the down state. Therefore, the processing of sensory stimulation during NREM sleep is closely regulated by the underlying state of neural synchronization and firing, as reflected by slow waves and spindles.
Sleep and Brain Plasticity
Evidence accumulates suggesting that sleep participates in the consolidation of recent memory traces [65]. Accordingly, PET studies have shown that waking experience influences regional brain activity during subsequent REM and NREM sleep. Several brain areas, activated during procedural motor sequence learning (using a serial reaction time task) during wakefulness, have been found to be significantly more activated during subsequent REM sleep in subjects previously trained on the task than in non-trained subjects [66]. Furthermore, this effect is not observed in subjects trained to a task with similar practice requirements but devoid of any sequential content [67]. These findings speak against use-dependent changes in regional brain activations. Additionally, functional coupling between learning-related areas was found to be enhanced during post-training REM sleep [68]. Another PET study demonstrated that hippocampal and parahippocampal areas, which are activated during a spatial memory task, can be reactivated during post-training NREM sleep and that the amount of hippocampal activity during SWS positively correlated with overnight improvement in the memory for spatial locations [69]. Collectively, these findings suggest that reactivations of regional activity and modifications of functional connectivity during post-training sleep reflect the off-line processing of recent memory, which eventually leads to improved performance the next day. Moreover, these results are in line with behavioral data suggesting that NREM sleep and REM sleep differentially modulate the consolidation of declarative and non-declarative memories, respectively [70, 71]. However, they do not rule out an alternative hypothesis that natural succession of NREM sleep and REM sleep is also mandatory for memory consolidation.
Functional MRI studies demonstrated that sleep deprivation hinders the plastic changes that normally would occur during post-training sleep [72]. In this study, the effects of normal sleep or sleep deprivation on learning-dependent changes in regional brain activity were assessed after the subjects were trained on a pursuit task, in which they had to hold a joystick position as close as possible to a moving target, whose trajectory was predictable on the horizontal axis but not on the vertical axis. The time on target was used as the behavioral performance parameter. In the first group, subjects were totally sleep-deprived during the first post-training night, while in the second group, they were allowed to sleep. Both groups were then retested after 2 more nights of normal sleep in order to recover a similar state of arousal across the two groups and between the training and retest sessions. The fMRI scanning session was recorded during the retest, while subjects were exposed to the previously learned trajectory and also to a new one in which the predictable axis was vertical. Behavioral results showed that the time on target was larger for the learned trajectory than for the new one in both groups during the retest and that this performance gain was greater in the sleeping group than in the sleep deprivation group. The fMRI data showed a significant effect of learning, irrespective of the group, in two regions: the left supplementary eye field and the right dentate nucleus. A region of the right superior temporal sulcus, close to regions coding for motion processing (biologic motion, smooth pursuit, etc.), was found to be more active for the learned than for the new trajectory, and more so in the sleeping group than in the sleep deprivation group. The functional connectivity also showed that the dentate nucleus was more closely linked to the superior temporal sulcus, and the supplementary eye field to the frontal eye field, for the learned than for the new trajectory, and more so in the sleeping group. Moreover, interactions between the temporal cortex and cerebellum, as well as between the frontal eye field and the supplementary eye field, are both known to be implicated in the standard pursuit eye movement pathways [73]. These results therefore suggest that the performance on the pursuit task relies on the subject’s ability to learn the motion characteristics of trajectory in order to program optimal motor pursuit execution. Sleep deprivation during the first post-training night would disturb the slow processes that lead to the acquisition of this procedural skill and alter related changes in functional connectivity that were reinforced in subjects allowed to sleep [72].
Orban et al. [74] used fMRI in order to map regional cerebral activity during place-finding navigation in a virtual town, immediately after learning and 3 days later, in subjects either allowed regular sleep (RS) or totally sleep-deprived (SD) on the first post-training night. Results showed that, at immediate and delayed retrieval, place-finding navigation elicited increased brain activity in an extended hippocampo-neocortical network in both RS and SD subjects. Moreover, behavioral performance was equivalent between groups. However, striatal navigation-related activity increased more at delayed retrieval in RS than in SD subjects. Furthermore, correlations between striatal response and behavioral performance, as well as functional connectivity between the striatum and the hippocampus, were modulated by post-training sleep. Overall, these data suggest that brain activity is restructured during sleep in such a way that navigation in the virtual environment, initially related to a hippocampus-dependent spatial strategy, becomes progressively contingent in part on a response-based strategy mediated by the striatum. Interestingly, both neural strategies eventually relate to equivalent performance levels, indicating that covert reorganization of brain patterns underlying navigation after sleep is not necessarily accompanied by overt changes in behavior [74]. Further studies have also evidenced a reorganization of brain activity when post-training sleep is allowed both for neutral [75] and emotional [76] verbal material, as well as for visual face-to-location associations [77]. In addition, exposure to an odor during SWS that had been presented as context during prior learning improved the retention of hippocampus-dependent declarative memories and elicited a significant hippocampal activation during SWS [78]. In a similar experiment using sounds to cue object-location memory, it was also shown that functional connectivity between the parahippocampal region and the visual cortices was enhanced upon sound presentation during SWS [79].
In addition, EEG and MEG studies have provided robust evidence for the “sleep and memory consolidation” hypothesis by focusing on more specific sleep features and mechanisms that are regarded as important for different types of memory, including sleep spindles [80–83], slow waves [84], or the actual number of rapid eye movements [85]. For instance, sleep homeostasis has a local synaptic component, which can be triggered by a learning task involving specific brain regions. The local increase in slow-wave activity after learning correlated with improved performance of the task after sleep [84]. Moreover, the induction of slow oscillation-like potential fields by transcranial application of slowly oscillating potentials (0.75 Hz) during early nocturnal NREM sleep (i.e., a period of emerging SWS) enhanced the retention of hippocampus-dependent declarative memories in healthy humans. This stimulation induced an immediate increase in SWS, cortical slow oscillations, and slow spindle activity in the frontal cortex [86].
Sleep spindles appear to play a central role in NREM sleep-dependent memory consolidation. Specifically, spindles appear to organize the concerted reactivation of neocortical areas involved in a memory with its hippocampal trace. In this way, spindles are believed to facilitate the transfer of newly consolidated memories from the hippocampus to the neocortex. Accordingly, Bergmann et al. [87] demonstrated in an EEG-fMRI study that sleep spindles were temporally coupled with concurrent activation in the hippocampus and neocortical areas relevant to a previously learned declarative memory task. Notably, this spindle-hippocampo-neocortical synchrony was greater after declarative learning than after a visuomotor control task and was correlated with performance on the declarative task. Furthermore, the magnitude of hippocampo-neocortical activation was proportional to variations in spindle amplitude. The key role of sleep spindles in memory consolidation is consistent with its aforementioned ability to block out external stimulation during NREM sleep [61].
Alertness, Performance, and Sleep Deprivation
Sleep deprivation or fragmentation is increasingly common in industrialized societies (noisy environments, shift work). Likewise, many sleep disorders are common in the general population (e.g., insomnia, anxiety disorders). The considerable proportion of vehicle accidents related to sleep loss is now viewed as a serious concern for public health [88]. The impact of sleep deprivation on cognition and brain functions has been assessed mainly in healthy subjects. By comparison, studies on the consequences of sleep disorders on behavior and cerebral activity remain scarce.
Cognitive Challenges
Sleep deprivation is known to alter alertness and performance in a series of cognitive tasks. Several neuroimaging studies have tried to determine the underlying patterns of cerebral activity during different cognitive tasks. The cerebral responses to sleep deprivation seem to depend on the type of task and also on its level of difficulty. Both decreases and increases in responses were reported. The former were interpreted as metabolic impairments related to sleep deprivation, whereas the latter were viewed as compensatory responses.
An early PET study with [18F]2-fluoro-2-deoxy-d-glucose (18FDG) investigated the effect of total sleep deprivation (about 32 h) on brain metabolism [89]. Although global brain metabolism was not affected by sleep deprivation, regional glucose metabolism significantly decreased in the thalamus, basal ganglia, and cerebellum. A significant reorganization of regional activity was observed after sleep deprivation, with relative decreases in the cerebral metabolic rate of glucose (CMRglu) within the temporal lobes and relative increases in the visual cortex [89]. Additionally, sleep deprivation significantly reduced performance in an attentional continuous performance test, and this decrease was significantly correlated with reduced metabolic rate in thalamic, basal ganglia, and limbic regions [89].
One study showed that, even after as little as 24 h of continuous wakefulness, significant decreases in global CMRglu are observed with 18FDG PET [2]. When subjects performed a sleep deprivation-sensitive serial addition/subtraction task (which combines arithmetic processing and working memory), significant decreases in absolute regional CMRglu were found in several cortical and subcortical structures, whereas no areas of the brain showed any significant increase in regional metabolism. Alertness and cognitive performance scores declined in parallel with deactivations in the thalamus and in the prefrontal and posterior parietal cortices [2].
The same group of researchers characterized the cerebral effects of 24, 48, and 72 h of sleep deprivation during the same task performance in 17 healthy subjects using correlations with performance measures outside of the scanner and metabolism during resting state assessed by 18FDG PET [90]. Results showed that absolute CMRglu and relative regional CMRglu decreased further at 48 and 72 h of sleep deprivation primarily in the prefrontal and parietal cortices and in the thalamus, the same areas that showed decreases at 24 h of sleep deprivation. The authors proposed that the decreases in CMRglu induced in the prefrontal-thalamic network by sleep deprivation underlie the progressive impairment in cognitive performance and alertness and the progression toward sleep onset. In contrast, increased activity in visual and motor areas would reflect voluntary attempts to remain awake and perform despite a continuing decline in prefrontal-thalamic network activity [90].
In these PET CMRglu studies, metabolism during resting state was correlated with performance measures obtained outside of the scanner. However, a different picture emerges when subjects are scanned during task performance using fMRI.
Drummond and colleagues used fMRI on normal subjects while those subjects performed different cognitive tasks after a normal night of sleep or following 35 h of sleep deprivation. In a first report [4], the study used a serial subtraction task. Bilateral activations in the prefrontal, parietal, and premotor cortices were found during task practice after a normal night of sleep, whereas activity in these regions declined markedly after sleep deprivation, mainly in the prefrontal cortex [4], which is in agreement with the hypothesis of prefrontal cortex vulnerability to sleep deprivation [91]. Likewise, Mu et al. [92] found reduced activations in several frontal and parietal regions (left dorsolateral prefrontal cortex, right ventrolateral prefrontal cortex, supplementary motor area, Broca’s area, and bilateral posterior parietal cortices) during practice of the Sternberg working memory task after 30 h of sleep deprivation compared to normal sleep. However, a very different pattern emerged when using other types of tasks. For example, the effects of 35 h of continuous wakefulness on cerebral activation during verbal learning (memorizing a list of words) were also investigated using fMRI [93]. The authors found that the prefrontal cortex and parietal lobes were more activated during verbal learning after 1 night of sleep deprivation than after normal sleep. In addition, increased subjective sleepiness in SD subjects correlated significantly with the amount of prefrontal cortex activation, while stronger parietal lobe activation was linked to less impairment in the free recall of words. Neurobehavioral (fMRI) effects of 24 h of continuous wakefulness were assessed using two verbal working memory tasks of different difficulty levels, known to induce responses in frontal-parietal networks in normal, non-sleep-deprived conditions. After sleep deprivation, activity was reduced in the medial parietal, anterior medial frontal, and posterior cingulate regions in both tasks, and disproportionately greater activation of the left dorsolateral prefrontal cortex and bilateral thalamus was observed when additional manipulation of information in working memory was required [94] (see also [95]). It has been suggested that these results reflect dynamic, compensatory changes in cerebral activation during verbal learning after sleep deprivation [93]. Compensatory mechanisms may lead to stronger responses within regions typically underlying task performance, as well as to activation in regions that do not show significant responses to task demands in the well-rested condition [96]. Indeed, the thalamus was also found to be hyperactivated in sleep deprivation during a visual attention task, in the absence of performance deficits [97]. In another fMRI study, performance on a higher-order attention task was unaltered by sleep deprivation, while greater thalamic activation was recruited in the SD group [98]. The thalamus may thus play an important role in cognitive compensation in SD states.
These data suggest that decreases in regional brain activity could contribute to cognitive impairment after sleep deprivation and that increased prefrontal and thalamic activation may represent compensatory adaptation. In a similar attempt to better understand how sleep deprivation might interact with task difficulty, an fMRI study found stronger correlation between difficulty in a logical reasoning task and increased activity in the bilateral inferior parietal lobes, bilateral temporal cortex, and left inferior and dorsolateral prefrontal cortex following 35 h of continuous wakefulness than after normal sleep [96]. In a second study by the same group, the effects of normal sleep and 36 h of total sleep deprivation were assessed by fMRI during a verbal learning task with two levels of difficulty (easy and difficult words) [99]. A set of regions showed increased response to difficult words after sleep deprivation compared with normal sleep (inferior frontal gyrus, dorsolateral prefrontal cortex, and inferior parietal lobe, bilaterally). While better free recall performance on the difficult words following sleep deprivation was positively related to activation within the left inferior and superior parietal lobes and left inferior frontal gyrus, it was negatively related to activation within the right inferior frontal gyrus. Consequently, the performance relationships are thought to be both beneficial (as a compensatory function) and deleterious (as an interference with task performance), depending on the brain regions implicated. In addition, these studies suggest that increased recruitment of compensatory brain regions is yoked to rising task difficulty.
The default-mode network appears to be compromised by sleep deprivation. Disengagement of these brain regions is thought to be necessary for optimal externally oriented cognitive performance, and disturbances in default-mode function are observed in mental disorders [100]. According to a few recent studies, typical task-related default-mode network suppression was disrupted after sleep deprivation [101] and functional connectivity between default-mode areas was compromised [102, 103]. Furthermore, the anticorrelation between the default-mode network and its externally oriented counterpart was weakened in the SD condition [102].
Other cognitive domains seem to be impaired by sleep deprivation. For instance, competent decision making was impaired after sleep deprivation, which induced a modulation of activation in the nucleus accumbens and insula, brain regions associated with risky decision making and emotional processing [104]. Likewise, SD healthy adults were more likely to adopt risky strategies in a gambling task, and this change correlated with ventromedial prefrontal activation and anterior insular deactivation [105]. Furthermore, the ventromedial prefrontal and ventral striatum showed greater activation during wins, and the anterior cingulate displayed smaller deactivation during losses. These data indicate that the pursuit of gain is amplified in sleep deprivation, while losses are perceived more optimistically. Another fMRI study deprived participants of REM sleep only or NREM sleep only and evaluated emotional reactivity in a visual emotional reactivity task [106]. They found emotional reactivity to be enhanced in the REM-deprived group only, concurrently with increased activity in occipital and temporal areas, compared to NREM-deprived controls. The authors conclude that emotional reactivity is modulated by REM sleep.
Since prefrontal cortex functioning appears to be affected by sleep loss, processes mediated by this region should be altered after sleep deprivation (e.g., attention, emotion, motivation, feeding, and olfaction). In order to assess the effects of sleep deprivation on olfaction, which is mediated by the orbitofrontal cortex, a region known to have decreased activity after sleep deprivation [2], Killgore and McBride [107] studied 38 healthy subjects at rest and after 24 h of sleep deprivation. Relative to rested baseline performance, SD subjects showed a significant decline in the ability to identify specific odors on the Smell Identification Test. In relation to effects on feeding, an fMRI study examined BOLD response to the presentation of images of food items after a night of sleep deprivation [108]. Activation of the anterior cingulate cortex in response to food was increased in sleep-deprived participants with respect to non-sleep-deprived controls, independent of prescan blood-glucose levels. Moreover, this activation correlated significantly with increased post-scan hunger ratings. Sleep deprivation thus seems to enhance hedonic response to food stimuli.
Changes in dopamine neurotransmission may underlie the cognitive challenges observed in sleep deprivation. Dopamine’s involvement in sleep/wake regulation remains unclear, but degeneration of dopamine pathways is associated with sleep disorders such as REM sleep behavior disorder and excessive daytime sleepiness in Parkinson’s disease (PD) [109, 110]. In a series of studies, Volkow et al. [111–113] studied the effect of sleep deprivation on dopamine using PET with 11C-raclopride, a dopamine D2/D3 receptor radioligand. They found dopamine receptor occupancy to be increased in the ventral striatum. Furthermore, this increase does not seem to be due to changes in dopamine transporter (DAT) density [114]. Increases in dopaminergic activity were associated with worse performance on a visual attention task, concurrent with BOLD changes in known dopamine-modulated areas (smaller deactivation of anterior cingulate cortex and insula) as well as in other areas (greater deactivation of inferior occipital cortex and cerebellum) [112]. These findings suggest that striatal dopamine hyperactivity may interfere with attentional processes, or instead, may represent a compensatory mechanism in order to maintain arousal after sleep deprivation.
Personal Vulnerability to Sleep Deprivation
People may be differently affected by the same sleep-depriving environmental conditions. Studies suggest that brain responses to sleep deprivation for a given task are modulated by individual vulnerability to sleep deprivation [115]. For example, in one fMRI study, subjects were divided into two groups, a sleep deprivation-resilient group and a sleep deprivation-vulnerable group, according to their performance on a working memory task after sleep deprivation. In the sleep deprivation-resilient group, significant activations were found in several cortical areas (left dorsolateral prefrontal cortex, left ventrolateral prefrontal cortex, left supplementary motor area, and left posterior parietal cortex) during practice of the working memory task after sleep deprivation. By contrast, in the sleep deprivation-vulnerable group, only the left dorsolateral prefrontal cortex was activated after sleep deprivation. The patterns of brain activation after sleep deprivation may therefore differ as a function of the subjects’ individual vulnerability to sleep deprivation [115]. The same group conducted another fMRI study on fatigue vulnerability in military pilots. Pilots were scanned during the working memory task under non-sleep-deprived conditions, and individual fatigue vulnerability was quantified using performance on a flight simulation during 37 h of continuous wakefulness. Analyses revealed that global cortical activation during the working memory task was positively correlated with fatigue resistance in flight-simulator performance. The authors therefore proposed that baseline fMRI activation during the working memory task may provide a good index of individual fatigue susceptibility [116].
Individual differences in the effect of sleep deprivation on working memory performance were assessed by fMRI in 26 healthy volunteers, in a rested condition and after 24 and 36 h of sustained wakefulness [117]. In both sleep deprivation conditions, task-related activation was significantly decreased with respect to the rested condition in the superior parietal regions and left thalamus. There was also an inverse correlation between activation of left parietal and left frontal regions during the rested condition, and individual decline in working memory performance from the rested condition to the 24-h sleep deprivation condition. In this way, frontoparietal activation in the rested state may distinguish individual cognitive vulnerability to sleep deprivation. Similarly, another fMRI study showed that lower task-related activation of the ventral prefrontal cortex during rested wakefulness was predictive of greater individual ability to maintain inhibitory efficiency in a go/no-go task [118]. Low-vulnerability individuals also showed reduced activation in this region as well as in the right insula after sleep deprivation, compared to highly vulnerable individuals.
In addition to functional imaging, structural neuroimaging techniques were also used to investigate interindividual differences in responses to sleep deprivation. For instance, studies of white matter anatomy with fractional anisotropy (FA) may provide one such physiologic marker of vulnerability to sleep deprivation [119]. In that study, West Point cadets first completed a simple visual-motor task before and after 24-h sleep deprivation and their change in performance was assessed. Taking these change scores as indices of sleep deprivation vulnerability, cadets were separated into a more vulnerable and a less vulnerable group by median split. They were then scanned with MRI using diffusion tensor imaging (DTI), and FA was calculated for each cadet as an indicator of fiber number, density, and myelination in white matter tracts. FA values were significantly greater in the low-vulnerability group than in the high-vulnerability group, predominantly in ascending and longitudinal pathways of the right hemisphere and the genu of the corpus callosum. Furthermore, greater FA values correlated significantly with smaller decreases in the performance on the task. These findings are consistent with those of aging studies linking greater white matter integrity to improved cognitive performance [120].
Sleep Deprivation in Depression
Sleep deprivation has profound effects on brain metabolism in both normal and depressed subjects. When used therapeutically (i.e., wake therapy), sleep deprivation relieves acute depressive symptoms in 60 % of patients. In depressed patients responding favorably to sleep deprivation, 18FDG PET [121–123], technetium-99m–labeled hexamethylene-propyleneamine oxime (99mTc-HMPAO) SPECT [124, 125], and fMRI [126] revealed greater baseline brain activity during wakefulness in responders than in non-responders in the anterior cingulate cortex and/or the nearby mesial frontal cortex. This activity was significantly decreased after sleep deprivation. A similar profile of brain metabolism was observed with 18FDG PET in elderly depressed patients, including normalization after total sleep deprivation associated with antidepressant treatment [127]. Moreover, the normalization of anterior cingulate metabolism persisted even after recovery sleep. However, these results were not replicated with placebo control [128].
MRI spectroscopic studies have shown glutamate [129, 130] and serotonin [131] levels to be globally increased after sleep deprivation in depressed patients. These neurotransmitters may then underlie the antidepressant effect of sleep deprivation. This is consistent, in the case of serotonin, with the widespread use of selective serotonin reuptake inhibitors as antidepressants. With PET and SPECT, it was also shown that sleep deprivation responders exhibit a significant decrease in relative basal ganglia D2 receptor occupancy after sleep deprivation, as compared to non-responders [132]. Sleep deprivation also resulted in lower binding of 11C-raclopride, a D2/D3 receptor radioligand, in the striatum and thalamus of healthy subjects [111]. These results suggest that the antidepressant benefits of sleep deprivation are correlated with enhanced endogenous dopamine release in responders, although a later study suggests DAT downregulation may instead be responsible for observed increases in dopamine receptor occupancy [113]. Nonetheless, these results corroborate previous hypotheses about the role of dopaminergic response in the therapeutic action of sleep deprivation and indirectly support a dopamine hypothesis of depression [132]. In relation to this hypothesis, sleep-deprived healthy subjects reacted more intensely to pleasure-evoking stimuli in dopaminergic mesolimbic brain networks associated with reward [133]. In addition, gains produced higher ventromedial prefrontal and ventrostriatal activation in healthy subjects, and losses produced less anterior cingulate deactivation [105]. Together, these findings suggest that the antidepressant effect of sleep deprivation may operate through reward-enhancing increases in dopaminergic activity.
31Phosphorous magnetic resonance spectroscopy (31P-MRS) has also been applied to healthy SD participants to understand the metabolic and bioenergetic changes that may underlie the antidepressant effect of wake therapy. Within-subject designs with healthy participants showed no change in phosphate brain chemistry after a night of sleep deprivation [134, 135], although one study found elevated β-nucleoside triphosphate and reduced phospholipid catabolism after a subsequent recovery night [134]. In a sample of depressed women, higher baseline levels of choline compounds in the pons were associated with the improvement in mood after sleep deprivation, indicating a role of pontine choline metabolism in the antidepressant response to sleep deprivation [136]. Further studies are needed scanning both depressed patients and healthy controls before and after sleep deprivation.
Sleep deprivation data suggest a tight link between mood alteration and activity in limbic and paralimbic structures. The data suggest that anterior cingulate hyperactivity in depressed patients during wakefulness may hinder further increases during REM sleep. Hence, sleep deprivation may alleviate depression symptoms by decreasing abnormally elevated activity in the anterior cingulate cortex during wakefulness. However, further studies are needed to understand the causes and consequences of these mesial frontal metabolic disturbances.
Instrumental Manipulation of Sleep Deprivation Effects
Repetitive transcranial magnetic stimulation (rTMS) can be used to artificially stimulate cortical brain areas, enabling true experimental designs in human neuropsychology. Luber et al. [137] used rTMS to artificially relieve impairments in working memory performance from sleep deprivation. Fifteen participants were first scanned with fMRI before and after sleep deprivation, while completing a working memory task. The brain areas that showed deactivation under sleep deprivation were then used as guides for rTMS application after a second night of sleep deprivation, two weeks later. Of the three identified sites, stimulation of the upper middle occipital site produced improvements in performance, whereas no improvement was accrued from the other sites or from sham rTMS. Furthermore, the degree of performance enhancement was directly proportional to the magnitude of each participant’s sleep-deprivation-induced deactivations. Hence, cognitive deficits from sleep deprivation can be corrected with rTMS by targeting the affected brain area. The authors replicated these findings in a recent experiment [138].
Summary
There is a marked heterogeneity in the functional findings relating to the cognitive effects of sleep deprivation. Deficits in cognitive function after sleep deprivation have been correlated with both activations and deactivations in several brain regions. Hypotheses explaining these findings range from compensatory mechanisms in order to maintain cognitive function, to homeostatic pressure diverting energy from certain regions, resulting in cognitive deficits. These hypotheses remain speculative. The wide range of tasks employed in these studies may in part explain the variability in results. Nonetheless, some interesting markers of individual vulnerability to sleep deprivation have been identified, including white matter integrity. In addition, the study of the therapeutic effect of sleep deprivation in depression has yielded some promising findings, particularly in relation to the dopaminergic system. Lastly, sleep deprivation effects can seemingly be rectified by the targeted use of rTMS.
Neuroimaging in Sleep Disorders
Sleep may be disrupted in a number of conditions ranging from medical diseases (e.g., endocrine disorders, chronic pain, brain lesions, and sleep apnea) and psychiatric disorders (e.g., anxiety, depression, and schizophrenia) to environmental situations (e.g., jet lag, shift work, and noisy environment).
In this section, we consider several primary sleep disorders (narcolepsy, periodic limb movement disorder, idiopathic insomnia, recurrent hypersomnia, and obstructive sleep apnea) as well as specific parasomnia syndromes (sleepwalking and RBD). We do not review sleep disorders due to disturbances from external environmental sources.
Idiopathic Insomnia
Idiopathic insomnia is a lifelong inability to obtain adequate sleep that is presumably due to an abnormality in the neurologic control of sleep–wake regulation systems [139]. This disorder is thought to reflect an imbalance between the arousal system and the various sleep-inducing and sleep-maintaining systems. Neuroanatomic, neurophysiologic, or neurochemical dysfunctions or lesions within the sleep–wake systems are suspected in some of these patients [139].
Theoretically, either hyperactivity within the arousal system or hypoactivity within the sleep system may cause idiopathic insomnia, but hyperarousal is believed to be the final common pathway of the disorder [139]. Increased arousal might be of a physiologic, cognitive, or affective nature; it is likely that these categories overlap [6, 140], since several studies have reported increased alertness on the Multiple Sleep Latency Test, increased heart rate during the sleep period, increased anxiety on rating scales, and increased tension during wakefulness [140–142]. In addition, poor sleep leads to altered mood and motivation, decreased attention and vigilance, low levels of energy and concentration, and increased daytime fatigue [139].
Quantitative EEG recordings suggest an overall cortical hyperarousal in insomnia [143]. However, hyperarousal in primary insomnia was also found to be associated with greater increase in beta/gamma activity at sleep onset, followed by a decline of high-frequency EEG activity leading to a period of hypoarousal [143]. This could explain why some neuroimaging studies showed a cortical hyperarousal pattern in insomnia while others reported a decrease in cortical functions. In the latter case, decreased metabolism might originate from time-window coincidence of the cortical hypoarousal period to neuroimaging acquisition and therefore does not discard the hyperaousal hypothesis of primary insomnia.
Only a few studies tried to characterize the functional neuroanatomy of idiopathic insomnia disorder during sleep (referred to as primary insomnia in these reports) (Fig. 21.2a1). rCBF was estimated using 99mTc-HMPAO, a gamma-emitting radionuclide imaging agent used in the evaluation of rCBF, in five insomniacs and four normal sleepers. Patients with insomnia showed major rCBF decrease in the basal ganglia, medial frontal cortex, occipital cortex, and parietal cortex. These results suggest that idiopathic insomnia is associated with an abnormal pattern of regional brain activity during NREM sleep that particularly involves a dysfunction in the basal ganglia [144]. Four of the insomnia patients from the Smith et al. study [144] were rescanned after they had been treated with cognitive behavioral therapy [145]. After this treatment, sleep latency was reduced by at least 43 % and there was a global 24 % increase in CBF, with significant increases in the basal ganglia. Smith and collaborators proposed that such increase in brain activity might reflect the normalization of sleep homeostatic processes.
Fig. 21.2
Findings from PET, SPECT, fMRI and MRI studies in sleep disorders. a1 Brain areas showing metabolic changes in idiopathic insomnia. Most studies show hypermetabolism in cortical and subcortical areas during NREM sleep, as well as smaller decreases in metabolism in the transition from wake to sleep. These results are consistent with the hyperarousal theory of insomnia. During wake, insomnia patients show decreased prefrontal activity. a2 Neuroanatomical alterations in idiopathic insomnia. While many studies found no anatomic changes in insomnia patients, some reported cortical gray matter loss or changes in hippocampal or rostral anterior cingulate volume. b1 Areas showing functional abnormalities in narcolepsy. All metabolic and hemodynamic changes are shown during wake, and one study imaged hypoperfusions during a cataplectic episode. Abnormalities in narcoleptics were reliably observed in the hypothalamus, consistent with a hypocretinergic dysfunction. b2 Neuroanatomical alterations in narcolepsy. Hypothalamic gray matter loss may relate to hypocretinergic dysfunction. Limbic and neocortical alterations may underlie mood disturbances in narcoleptic patients. c1 Functional studies on OSAS have focused mainly on activity changes during cognitive tasks and have mostly detected deactivations. A few resting-state studies using PET and SPECT have found decreases in activity during resting wakefulness. c2 Neuroanatomical alterations in OSAS. Gray matter loss in the hippocampus, parietal, and prefrontal cortices may underlie various neuropsychological deficits observed in OSAS, only some of which seem to be reversible by CPAP. d Brain areas showing structural or functional changes in RLS. Iron deficiencies in several areas, notably the SN, may contribute to RLS, possibly via dysregulation of the closely associated dopaminergic system. These changes may in turn sensitize sensorimotor pain sensation. e Areas showing functional and structural abnormalities in RBD. Changes have most consistently been shown in the pons, hippocampus, and superior frontal cortex. Adapted from Desseilles et al. [5], and from illustrations by Patrick J. Lynch and C. Carl Jaffe. http://creativecommons.org/licenses/by/2.5/
18FDG PET was used to measure regional CMRglu of 7 patients with idiopathic insomnia and 20 healthy age- and gender-matched subjects during waking and NREM sleep [146]. Insomniac patients showed increased global CMRglu during sleep as compared to healthy subjects, suggesting an overall cortical hyperarousal in insomnia. Moreover, insomniac patients had a smaller decline, compared to healthy subjects, in relative CMRglu from waking to sleep states in the ascending reticular activating system, hypothalamus, thalamus, insular cortex, amygdala, hippocampus, anterior cingulate, and medial prefrontal cortices (Fig. 21.3). During wakefulness, reduced relative metabolism, as compared to healthy subjects, was found in the prefrontal cortex bilaterally, in the left temporal, parietal, and occipital cortices, and in the thalamus, hypothalamus, and brainstem reticular formation. These findings confirm that regional brain activity does not normally progress from waking to sleep states in patients with insomnia. Additionally, it was proposed that daytime fatigue resulting from inefficient sleep may be reflected by decreased activity in the prefrontal cortex [146]. In another 18FDG PET study by the same group [147] examining 15 insomniac patients, CMRglu in the pontine tegmentum and thalamo-cortical networks during NREM sleep was found to be significantly correlated with self-reported wake time after sleep onset (WASO), based on a 7-day sleep diary. Hence, elevated cortical arousal during sleep is associated with insomnia symptom severity, lending support to the hyperarousal theory of insomnia.
Fig. 21.3
CMRglu assessed by 18FDG PET in insomniacs (versus healthy subjects) during waking and NREM sleep. a Brain structures that did not show decreased cerebral metabolic rate of glucose (CMRglu) from waking to sleep states in patients with idiopathic insomnia. b Brain structures where relative metabolism while awake was higher in healthy subjects than in patients with insomnia. Differences in all regions shown reached statistical significance (p < 0.05), corrected at the cluster level (Reproduced with permission from Nofzinger et al. [146]. Copyright 2004, American Psychiatric Association)
Additional studies have imaged idiopathic insomniacs during wakefulness, using fMRI (Fig. 21.2a1). In a first study [148], 21-year-old adult insomnia patients showed reduced medial and inferior prefrontal activation during the completion of a verbal fluency task, compared to 12 age-matched controls. In a second phase of the study after 8 weeks of non-pharmacological therapy, insomnia patients showed partial recovery of activation in these areas. These results corroborated similar findings by Nofzinger et al. [146] indicating prefrontal hypoactivation during wakefulness in insomnia. Another recent fMRI study [149] imaged 25 idiopathic insomnia patients and 25 age- and sex-matched controls during the performance of a working memory task. Interestingly, insomniacs failed both to recruit typical task-specific brain areas and to deactivate irrelevant, default-mode network regions. The authors suggested that the observed alteration in task-specific activation reflects compromised cognitive processing in insomnia, even though task performance in insomniacs was not different from controls. They attributed this to a masking effect of high perfectionism, a trait associated with insomnia.
Insomnia has been associated with high emotional reactivity [150]. An fMRI study investigated resting-state functional connectivity in the emotional system of 10 idiopathic insomnia patients and 10 matched controls [151]. Abnormalities were detected in functional connectivity between the amygdala and several cortical and subcortical areas in insomnia patients. While decreased connectivity was found for the striatum, insula, and thalamus, elevated connectivity was observed between the amygdala and the premotor cortex, sensorimotor cortex, a pathway associated with threat response. Increased motor activation is consistent with global cortical hyperarousal in insomnia.
Some recent efforts have been made to identify structural brain alterations in idiopathic insomnia, with the use of magnetic resonance imaging (MRI) (Fig. 21.2a2). While an early study found reduced hippocampal volumes in idiopathic insomnia patients compared to controls [152], subsequent studies were unable to replicate the finding [153–156]. One study found significant negative correlations between hippocampal volume and insomnia duration, as well as with a polysomnography-based index of arousal [154]. Consistent with these findings, significant inverse correlations have been reported between hippocampal volume and actigraphic WASO and sleep efficiency [153]. In one retrospective study, rostral anterior cingulate volumes were significantly greater in insomniacs [157]. The role of the anterior cingulate in insomnia is corroborated by earlier findings in insomniacs of attenuated deactivation in this area during the transition from waking to sleep [146], and of reduced levels of inhibitory neurotransmitters in this area [158]. Additional structural studies have inspected small-scale brain modifications using VBM. VBM is a neuroimaging analysis technique that allows the investigation of focal differences in tissue composition (gray and white matter) based on high-resolution MRI scans. Prefrontal gray matter concentrations were smaller in insomniacs than in good sleepers, specifically in the orbitofrontal [156, 159] and dorsolateral prefrontal cortices [159], but these findings were not replicated in a later study [155].
Magnetic resonance spectroscopy (MRS) has been applied to the study of idiopathic insomnia. According to proton MRS (1H-MRS) studies, relative concentrations of gamma-aminobutyric acid (GABA) appear to be reduced in insomniacs relative to good sleepers, globally (single-voxel) [160] and locally, in the occipital and anterior cingulate [158], although another study found occipital GABA levels to be increased instead [161]. This latter study also detected a significant inverse correlation between global GABA levels and polysomnography-based WASO. A 31P-MRS study investigating gray and white matter phosphocreatine levels found this cell metabolite in lower concentrations in the gray matter of insomniac patients relative to controls, indicating a possible increase in cortical energy demand in insomnia [162].
Together, neuroimaging studies of insomnia tend to support hyperarousal theory. Diminished inhibition in the transition from waking to sleep [146], increased connectivity in the emotional and threat response systems [151], decreased ability to inhibit irrelevant cognitive processes [149], depletion of inhibitory neurotransmission [158, 160], and increased cortical energy demands [162] all are consistent with an incapacity to modulate cortical arousal across the sleep–wake cycle. In addition, evidence of reduced hippocampal volume [152], reduced prefrontal gray matter concentrations [156, 159], and increased rostral anterior cingulate volume [157] provide possible neural correlates of cognitive and emotional dysfunctions experienced in insomnia. Still, these findings require replication in larger, clinically homogeneous samples.
Narcolepsy
Narcolepsy is a disorder characterized by excessive sleepiness that is typically associated with several manifestations of so-called dissociated or isolated REM sleep features, such as muscle atonia (i.e., cataplexy), sleep paralysis, and hallucinations [139, 163]. Human narcolepsy has been found to be associated with reduction in or loss of the hypothalamic peptide hypocretin (also called orexin) implicated in arousal systems [164–167].
Anatomic Neuroimaging Studies of Narcolepsy
Structural abnormalities in narcolepsy have been examined extensively (Fig. 21.2b2). The pontine tegmentum controls transitions between sleep states and was therefore first proposed as a possible main site of anatomic or functional impairments in narcolepsy. While Plazzi and coworkers had reported pontine tegmentum abnormalities in three narcoleptic patients [167], two other structural MRI studies [168, 169] found no pontine abnormalities (except in 2 of 12 patients who had long-standing hypertension [169]. The MRI abnormalities found in Plazzi et al.’s study could reflect non-specific age-related pontine vascular changes rather than a narcolepsy-related phenomenon [167].
VBM has been employed to find evidence for hypothalamic abnormalities in narcoleptic patients, reporting equivocal results. An early study found no structural change in brains of patients with hypocretin-deficient narcolepsy [170]. Subsequent studies did find cortical gray matter reduction predominantly in frontal brain regions [171–174], as well as in inferior temporal regions [175]. Interestingly, relative global gray matter loss was independent of disease duration or medication history, and there were no significant subcortical gray matter alterations [175]. Significant gray matter concentration decreases were found in the hypothalamus, cerebellum (vermis), superior temporal gyrus, and right nucleus accumbens in 29 narcoleptic patients relative to unaffected healthy controls [176] (Fig. 21.4). Given the widespread projection sites of hypocretin, the decreases in gray matter could thus reflect secondary neuronal losses due to the destruction of specific hypocretin projections. The results of this study were later corroborated by two other VBM studies [171, 177]. More recent MRI studies in narcolepsy with cataplexy identified volumetric decreases in the bilateral hippocampus [178] and amygdala [179]. Notably, amygdalar deficits may underlie emotional dysregulation in narcolepsy.
Fig. 21.4
Statistical parametric maps demonstrating the structural difference in gray matter between narcolepsy patients and healthy control subjects. Differences are shown superimposed in red on a normalized image of a healthy control subject. The left panel in A is the left side of the brain. A significant decrease in gray matter concentration was found in the hypothalamus (Hy) (a–c) and in the area of the right nucleus accumbens (Na) (a and c) (Reproduced with permission from Draganski et al. [176]. Copyright 2002, Nature Publishing Group)
Anatomic alterations have also been described using MRI cortical thickness measurements. A localized thinning of the cortex was detected in the prefrontal, cingulate, inferior parietal, and temporal areas in narcoleptic-cataplectic individuals [180]. Another study measured increased cortical volume and thickness in the dorsolateral prefrontal cortex of narcoleptic patients [181]. Furthermore, early-onset narcoleptic patients exhibited thinner cortex than late-onset narcoleptics in the precentral gyrus, inferior parietal cortex, and temporal regions. These results were interpreted as reflecting distinct pathological subtypes [181].
Recent studies have used DTI to detect alterations in fiber tract integrity, based on water diffusion in the brain. Using FA and mean diffusivity measures, it was shown that white matter tracts were disrupted in frontal, temporal, and anterior cingulate regions in narcolepsy [173]. Another DTI study found similar results, in addition to white matter abnormalities in the pons, right hypothalamus and left mesencephalon, consistent with a dysfunction in the hypothalamic hypocretin system in narcolepsy [182].
Proton MRS was also used to assess the N-acetylaspartate (NAA) and creatinine plus phosphocreatinine (Cr + PCr) content in specific brain areas of narcoleptic patients. A reduced NAA/Cr + PCr ratio indicates reduced neuronal function, which could reflect neuronal loss (i.e., fewer neurons) but could also be due to reduced activity of existing neurons. An analysis of spectral peak area ratios revealed a decrease in the NAA/Cr + PCr ratio in the hypothalamus [183] and the ventral pontine areas [184] of narcoleptic patients compared with control subjects. These results may indicate structural damage to these areas in narcolepsy.
Several factors can explain equivocal results across structural neuroimaging studies, such as inhomogeneous patient groups, history of treatment, or, for VBM, prestatistical image processing and limited sensitivity of this technique. VBM studies with larger samples of drug-naive patients are required to identify reliable structural abnormalities in narcolepsy.
Functional Neuroimaging Studies of Narcolepsy
A number of studies have examined brain metabolic differences in narcoleptic patients in the waking state (Fig. 21.2b1). Early functional observations using 133Xe inhalation showed that, during wakefulness, CBF in the brainstem was lower in narcoleptic patients than in normal subjects. However, after sleep onset (3 of 13 cases in REM sleep), the CBF increased in all regions, particularly in temporoparietal regions. This pattern was supposedly attributed to dreaming activity, in line with prior reports showing increased regional blood flow in temporoparietal areas during visual dreaming and hypnagogic hallucinations [185, 186].
A 99mTc-HMPAO SPECT study in six narcoleptic patients found similar HMPAO uptake in the waking state and REM sleep [187], suggesting a similar overall cortical activity. Data analysis using regions of interest additionally indicated an activation of parietal regions during REM sleep [187]. The latter result is intriguing given the parietal deactivation usually observed by PET studies during normal REM sleep [8]. Further studies during REM sleep are needed to confirm these results in a larger population.
Two 18FDG PET studies were conducted in the waking state to examine differences in narcoleptic brain activation. In the first study, a sample of 24 narcoleptic patients had reduced CMRglu in the bilateral posterior hypothalami, mediodorsal thalamic nuclei, and frontal and parietal cortices, compared to controls [188]. The second study instead found a CMRglu increase in the cingulate and visual association cortices in 21 narcoleptic patients also suffering from cataplexy [189]. A SPECT study revealed hypoperfusion during wakefulness in several areas including the bilateral anterior hypothalami, caudate nuclei, pulvinar, parahippocampal gyri, cingulate gyri, and prefrontal cortices [190].
There are very few data describing the neural correlates of cataplexy in narcoleptic patients (Fig. 21.2b1). One SPECT study was conducted on two patients during a cataplexy episode compared to REM sleep or a baseline waking period [191]. During cataplexy, perfusion increased in limbic areas (including the amygdala) and the basal ganglia, thalami, premotor cortices, sensorimotor cortices, and brain stem, whereas perfusion decreased in the prefrontal cortex and occipital lobe. Increased cingulate and amygdala activity may relate to concomitant emotional processing that is usually reported as a powerful trigger of cataplexy. However, such hyperperfusion in the pons, thalami, and amygdaloid complexes was not found in two subsequent studies [189, 192]. A more recent PET study imaged two patients during a cataplectic attack and found a large decrease in glucose metabolism in the hypothalamus, as well as increased metabolism in pre-postcentral gyri and somatosensory cortex [189].
Based on the clinical observation that cataplexy episodes are often triggered by positive emotions (e.g., hearing or telling jokes), an fMRI study was performed on narcoleptic patients and controls while they watched sequences of humorous pictures [193]. Group comparisons revealed that humorous pictures elicited reduced hypothalamic response together with enhanced amygdala response in the narcoleptic patients. These results suggest that hypothalamic hypocretin activity physiologically modulates the processing of emotional inputs within the amygdala and that suprapontine mechanisms of cataplexy might involve a dysfunction of hypothalamic–amygdala interactions triggered by positive emotions [193, 194]. Another fMRI study examined amygdalar response to emotional stimuli, by pairing visual stimuli with painful electric shock. Whereas healthy controls showed enhancement of amygdalar response to conditioned aversive stimuli, no such enhancement was observed in narcoleptic subjects [195]. In addition to this abnormal emotional circuitry, abnormal reward circuitry was evident in another report by the same group [196].
Neurotransmission in Narcolepsy
Given the role of acetylcholine as an important neurotransmitter in the generation of REM sleep [28, 197], it was hypothesized that disturbances in the cholinergic system might underlie narcolepsy. However, a PET study with 11C-N-methyl-4-piperidylbenzilate found no evidence for a change in muscarinic cholinergic receptors in narcoleptic patients [198].
Likewise, the dopamine system has been probed by PET and SPECT in narcoleptic patients because increased dopamine D2 receptor binding was shown in the brains of deceased narcoleptic patients [199, 200]. The results from these neuroimaging studies remain mostly inconsistent. One SPECT study showed elevated D2 receptor binding in the striatal dopaminergic system, correlating with the frequency of cataplectic and sleep attacks in seven patients with narcolepsy [201]. However, other PET [202–204] or SPECT [205, 206] ligand studies did not find such change in D2 receptor binding. A potential explanation for this discrepancy might be related to the drug treatment of narcoleptic patients. Indeed, considerable increase in the uptake of 11C-raclopride, a specific D2 receptor ligand, was observed in the putamens of narcoleptic subjects older than 31 years who underwent various regimens of prolonged treatment [207]. Likewise, despite the fact that the binding of iodobenzamide (IBZM, a highly selective CNS dopamine D2 receptor ligand), was similar in narcoleptic patients and normal controls, treatment by stimulants and/or antidepressants for 3 months significantly changed ligand uptake in four of five patients [206]. Therefore, elevated postmortem dopamine binding might be due to the long-term effect of prior treatment rather than intrinsic modifications.
Brain Response to Drug Probe in Narcolepsy
The effects of stimulant drugs on cerebral function in narcoleptic patients were assessed using functional imaging. The effect of amphetamines was evaluated using fMRI in two patients with narcoleptic syndrome [208]. The extent of the brain response to auditory and visual stimulation decreased after amphetamine administration in normal subjects. The reverse pattern was observed in the narcoleptic patients. These findings remain difficult to interpret, and larger samples of patients should be studied.
Modafinil is a wakefulness-promoting psychostimulant used to treat narcoleptic patients. Two fMRI studies have evaluated its effect on brain perfusion responses. A first study found that normal subjects showed larger brain responses to a multiplexed visual and auditory stimulation paradigm at 10:00 a.m. than at 3:00 p.m. in visual areas, but not in auditory areas, suggesting time-of-day influences [209]. Surprisingly, the reverse pattern of activity was observed in a group of 12 narcoleptic patients, with greater perfusion at 3:00 p.m. than 10:00 a.m.. Critically, modafinil administration did not modify the average level of activity either in normal subjects or in narcoleptics (n = 8), but post-drug activity level was inversely proportional to the predrug activity level. These findings are not easy to interpret but might suggest that modafinil can modulate brain activation in response to external stimuli. The second fMRI study tested the effect of modafinil on brain activation during a working memory task in healthy subjects, following a night of sleep deprivation. Modafinil at once improved working memory performance and enhanced activation in the executive network, specifically in the prefrontal and anterior cingulate cortices [210].
The metabolic effect of modafinil was studied in healthy individuals using 99mTc-ethylcysteinate dimer (99mTc-ECD) SPECT before and after the administration of the stimulant or a placebo [211]. Modafinil increased wakefulness and rCBF to areas in the brain associated with arousal, emotions and executive function, including the prefrontal, insular, cingulate, middle/inferior temporal and parahippocampal cortices and the pons. An 18FDG PET study imaged baseline CMRglu in awake patients and controls before and after treatment with modafinil. Decreased CMRglu was observed in the brainstem, hypothalamus, thalamus, and mesio-temporal areas in narcoleptics compared to controls. When comparing pre- and post-treatment conditions in narcoleptic patients, an increase in CMRglu was found in the left hippocampus post-modafinil [212]. A second 18 FDG PET study compared CMRglu levels in 14 narcolepsy-cataplexy patients treated with psychostimulants and anticataplectics, and 7 narcolepsy-cataplexy patients that were not treated with drugs. Treated narcoleptic had higher levels of CMRglu in their cerebellum and primary sensory motor cortex. However, these results were hard to interpret due to sample heterogeneity [189].
Summary
Structural and functional neuroimaging studies in narcolepsy with cataplexy show remarkable convergence. Primarily, the hypothalamus displays consistent abnormalities in narcoleptic patients, both structural and functional. Hypothalamic involvement in narcoleptic psychopathology is consistent with the characteristic loss of hypocretinergic neurons in this disorder. The limbic system also shows abnormal responding, including the amygdala, which may reflect emotional dysregulation seen in the disorder and triggering cataplectic episodes. Structural abnormalities were equally found in the hippocampus and various cortical areas, perhaps underlying dysfunctions in cognitive processing. Sparse imaging data acquired during cataplexy itself suggest a possible role of the hypothalamus and somatosensory cortex. As for treatment response, modafinil seems to alter metabolism in a variety of areas, but none have stood out from the few studies to date. Despite recent breakthroughs in the pathophysiology of narcolepsy, more studies using state-of-the-art technology of acquisition and analysis of functional neuroimaging data are needed to better characterize the functional organization of the narcoleptic brain during wakefulness and sleep.
Recurrent Hypersomnia
Recurrent hypersomnia is a disorder characterized by recurrent episodes of hypersomnia that typically occur weeks or months apart [139]. One SPECT study in a 24-year-old male with recurrent hypersomnia showed decreased blood flow in the left thalamus during the hypersomnolent period, but failed to report any abnormal activation during recovery or remission periods [213]. This case report neuroimaging study provides only limited information about possible pathophysiologic mechanisms of this disorder. By contrast, other clinical and electrophysiologic studies clearly point toward a hypothalamic rather than a thalamic dysfunction [214, 215].
A typical form of recurrent hypersomnia, KLS, is characterized by periodic hypersomnia as well as behavioral and cognitive abnormalities, mainly in male adolescents [216]. Neuroimaging studies of KLS patients have revealed normal brain anatomy [216–218]; however, functional abnormalities were observed in SPECT. In an early SPECT study with 99mTc-ECD, unilateral hypoperfusion of both thalami was found in a sample of 27 KLS patients. These hypoperfusions consistently occurred during the symptomatic period [218] and were confirmed in later studies using both SPECT and 1H-MRS [219–221]. During asymptomatic periods, perfusion of thalami returned to normal levels in all subjects, while other abnormalities in hypoperfusion persisted, mainly in the temporal lobe [218]. Other studies corroborated that hypoperfusion of the temporal lobe persists in asymptomatic patients [220]. Hypoperfusion of both thalami and basal ganglia [217, 220], as well as the frontal [222–224] and temporal lobe [216, 218, 220, 222, 225, 226], may provide insight to the pathophysiology of KLS, but the roles of the brain structures are not yet clear.
Obstructive Sleep Apnea Syndrome (OSAS)
OSAS is characterized by repetitive episodes of upper airway obstruction that occur during sleep, usually associated with a reduction in blood oxygen saturation [139]. Population-based epidemiologic studies have revealed a high prevalence (1–5 % of adult men) of OSAS. They also associate OSAS with significant morbidity, such as hypertension, cardiovascular disease, stroke, and motor vehicle accidents [227]. OSAS may lead to functional and structural brain alterations. Functional alterations such as sleep fragmentation are often associated with neuropsychological deficits that can be reversible after treatment of OSAS. Structural alterations may indicate irreversible consequences on brain integrity and suggest permanent cognitive impairment, although this proposal remains a matter of debate in the literature, especially given recent structural imaging data showing a reversibility after treatment [228].
The pathophysiology of OSAS is complex and not yet completely understood. Several studies suggest that OSAS across all age groups is due to a combination of both anatomic airway narrowing and abnormal upper airway neuromotor tone. Notwithstanding the known anatomic factors, such as craniofacial anomalies, obesity, and adenotonsillar hypertrophy, that contribute to OSAS, clear anatomic contributing factors cannot always be identified [139]. This suggests that alterations in upper airway neuromuscular tone also play an important role in the etiology of OSAS [229]. The pathophysiology of OSAS also includes enhanced chemoreflex sensitivity and an exaggerated sympathetic response during hypoxemic episodes [230].
Apnea episodes in OSAS patients have considerable hemodynamic consequences, which are mediated by a complex cascade of physiologic events. Repetitive episodes of apnea trigger marked fluctuations in both blood pressure and heart rate, with consequent effects on the estimates of cardiovascular variability [6]. Several important regulatory mechanisms in cardiovascular homeostasis seem to be impaired in OSAS patients. Specific chemoreceptors seem to be implicated in the pathophysiology of OSAS [231]. For instance, the ventilatory response to carbon dioxide is elevated in OSAS patients [231] due to an elevation of the partial pressure of carbon dioxide that delimits carbon dioxide ventilatory recruitment threshold. An altered autonomic balance has been suggested as one possible pathogenic factor. This autonomic dysfunction has been thought to be implicated in the subsequent development of cardiovascular diseases in patients with OSAS.
Functional Neuroimaging Studies of OSAS
Several fMRI studies have been conducted in OSAS patients to characterize the neural correlates of integrated afferent airway signals with autonomic outflow and airway motor response [232–234]. For instance, altered response after a Valsalva maneuver involves cerebellar, limbic, and motor area deactivation. Enhanced sympathetic outflow after a forehead cold pressor challenge results in both diminished response in the thalamus, hippocampus, and insula, and exaggerated response in the cingulate cortex, cerebellum, and frontal cortex. Mandibular advancement led to decreased fMRI response in the left cingulate gyrus and the bilateral prefrontal cortices in 12 healthy subjects during induced respiratory stress [235]. Simultaneously, the subjective effects of this treatment were assessed by a visual analog scale that confirmed successful reduction of respiratory stress.
OSAS has been associated with distinctive cognitive alterations in various domains. Both hypoxemia and fragmented sleep are proposed as the main factor leading to neurocognitive impairments during wakefulness [236–243]. Several studies emphasized the deterioration of executive functions in OSAS patients, including the inability to initiate new mental processes [244, 245] and deficits in working memory [244, 246], analysis and synthesis [244, 246], contextual memory [247], selective attention [248], and continuous attention [248]. A meta-analysis showed that untreated patients with OSAS had a negligible impairment of intellectual and verbal functioning but a substantial impairment of vigilance and executive functioning [249].
A number of functional imaging studies of OSAS patients have used fMRI to evaluate BOLD contrasts during performance on cognitive tasks (Fig. 21.2c1). Generally, these studies have found deactivations in various regions [250–254], while some instead showed increases in activation [255] or mixed results [252, 256, 257]. For example, Ayalon et al. [253] compared 14 OSAS patients with 14 healthy controls on a sustained attention task and found task-related reductions in activation in OSAS patients across parietal, cingulate, and frontal regions typically recruited in attention tasks. Archbold et al. [256] found that OSAS severity correlated with increased activation of the right parietal lobe during a working memory task, in a sample of 9 treatment-naïve male OSAS patients. Deactivations are consistent with OSAS-related structural deterioration (described below), whereas activation increases may represent compensatory recruitment. More recent studies have focused on resting-state functional connectivity changes and have found regional reductions in connectivity in the medial and dorsolateral prefrontal cortices, which make up part of the default-mode network [258, 259]. Concordantly, default-mode network deactivation was compromised in OSAS patients during a visuospatial N-back task, in parallel with performance deficits [252, 260].
Aside from fMRI studies during cognitive performance, functional correlates of OSAS have also been studied during restful wakefulness, using PET and SPECT. A 99mTc-ECD SPECT study observed decreased baseline rCBF in the parahippocampal and lingual gyri of OSAS patients compared to healthy volunteers [261]. Using 18FDG PET, another group found reduced CMRglu in the prefrontal, parieto-occipital, and cingulate gyri of OSAS patients, compared to healthy controls [262].
Long-term consequences of OSAS have been rarely assessed after nasal continuous positive airway pressure (CPAP) treatment. An early 99mTc-HMPAO SPECT study in 14 adult OSAS patients reported a marked frontal hyperperfusion in 5 patients [263]. In contrast, regional analysis showed reduced perfusion in the left parietal region. All these changes were reversed by effective CPAP therapy, suggesting that the main deleterious effects of OSAS on brain activity are reversible. According to the authors, there might be an apnea-associated effect of local vascular autoregulation mechanisms acting to compensate systemic blood flow alterations or blood gas changes in OSAS. Similar findings were obtained in an fMRI study, where hyperactivation of prefrontal and hippocampal areas was reversed with 3 months of CPAP treatment [257]. In a recent study, 2 months of CPAP treatment improved task-related default-mode network deactivation, concomitantly with improvements in behavioral performance [260].
Anatomic Neuroimaging Studies of OSAS
Structural modifications of brain morphology in OSAS have been studied extensively in several modalities (Fig. 21.2c2). In an early study, structural changes were assessed using VBM in 21 patients with OSAS and in 21 control subjects [264]. Gray matter loss was apparent in patients with OSAS in multiple brain sites involved in motor regulation of the upper airway as well as in various cognitive functions, including the frontal and parietal cortices, temporal lobes, anterior cingulate, hippocampus, and cerebellum. Additional VBM studies found gray matter loss in similar regions, mainly in the prefrontal, anterior cingulate, parietal, and hippocampus [228, 258, 262, 265, 266, 267, 268]. These anatomic changes were often [228, 266, 267] but not always [262] associated with cognitive deficits, notably memory impairment. Female gender [269] and depressive symptoms [270] were associated with exacerbated neural damage and neurocognitive symptoms from OSAS. Another MRI study compared both neuroanatomical and neuropsychological effects of hypoxia in patients with either carbon monoxide poisoning or OSAS and found a hippocampal atrophy in both groups [266]. Of note, a linear relationship between hippocampal volume and memory performance selectively in the OSAS group was found for a subset of tests (the delayed recall or the Rey-Osterrieth Complex Figure Design and Trail 6 of the Rey Auditory Verbal Learning Test among others). Moreover, hippocampal volume was related to performance on nonverbal information processing (Wechsler Adult Intelligence Scale–Revised Block Design) in both groups. In a more recent study, hippocampal volume was also negatively correlated with excessive daytime sleepiness [268]. Further investigation will be necessary to better delineate the specificity and contribution of hippocampal atrophy in OSAS. A set of interesting correlations between indices of OSAS severity and prefrontal cortex changes were observed with MRI. Apnea–hypopnea index was negatively correlated with frontal gray matter volumes [228]. Also in the frontal regions, prolonged arterial oxygen desaturation correlated with gray matter decreases [228], deactivations during working memory tasks [251, 252], and decreased cortical connectivity [259]. In one of these studies, 3 months of CPAP treatment in 17 treatment-naïve OSAS patients resulted in the recovery of gray matter concentrations in the hippocampus and frontal regions, alongside significant improvements in neurocognitive performance [228]. These findings underline the effectiveness of CPAP treatment in both physiologic and cognitive recovery from OSAS-related neural damage.
Related posts:
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
