Introduction

Parkinson’s disease (PD) is a complex neurodegenerative disorder. Clinically defined by rigidity, bradykinesia and tremor, PD symptoms also include numerous sensory, autonomic, sleep/wake symptoms, and other non-motor symptoms. Some of these symptoms likely result from dopamine cell loss, while others do not. There is actually relatively little pathologic data that directly compare PD patients with any specific sleep/wake disorder versus PD patients without that problem. However, the pathophysiology of sleep/week disorders is increasing elucidated by epidemiologic data, physiological assessments, and a better understanding of the pathology sleep/wake disorders in the general population.

This article will first explore alterations in sleep/wake function and its two main consequences, insomnia and sleep fractionation, then at excessive daytime sleepiness, including alterations in circadian control, and then review the pathophysiology of specific common sleep issues including REM behavioral disorder (RBD), sleep apnea (OSA), restless legs syndrome (RLS), and periodic limb movements of sleep (PLMS).

Alteration of sleep architecture and sleep fragmentation

Sleep is divided into different stages, the composite of which is termed architecture. We usually enter light sleep (manifest by slow eye movements and loss of occipital alpha activity). Stage 2 sleep is defined by the presence of sleep spindles and usually constitutes the majority of the sleep epoch. Slow-wave sleep (formal stage 3 or 4) is the “deepest” sleep manifest by delta activity on EEG. Patients often then cycle into rapid eye movement (REM) sleep. This peculiar state includes EEG activity similar to the wake state, loss of muscle activity (atonia), and rapid jerky eye movements. They may then cycle back to light sleep and imperfectly cycle through these states. REM sleep is relatively more prevalent in the second half of the sleep event.

Some studies do not show any marked or consistent abnormalities in sleep architecture in PD, other than reduced REM atonia (Happe et al. 2003; Diederich et al. 2013). The most consistently observed abnormalities are reduced sleep efficacy, sleep fragmentation and frequent arousals (Tandberg et al. 1998; Porter et al. 2008). In other studies, PD patients had less total sleep time, lower sleep efficiency, increased sleep latency, longer stage 2 of sleep and shorter or longer REM sleep stage compared to healthy controls (Young et al. 2002; Yong et al. 2011; Antczak et al. 2013). There are some reports showing increased alpha activity during REM sleep (Brunner et al. 2002) and less sleep spindles during non-REM sleep (Friedman 1980). That said it is somewhat surprising that basic sleep architecture is not more consistently abnormal, and overall sleep efficiency is independent of motor functions in many studies (Young et al. 2002; Diederich et al. 2005).

The specific causes of these documented abnormalities are multifactorial and direct pathological assessments are limited. Sleep fragmentation has been associated with longer disease duration, female gender and depression (Gjerstad et al. 2007). They occur more frequently in light sleep (stage 1 and 2), but can happen during any sleep stage (Porter et al. 2008). Sleep fragmentation in PD can be primary, i.e. the consequence of a disruption of sleep architecture, or secondary to nocturnal recurrence of PD symptoms (tremor, dystonia, rigidity), effect of medications, OSA, PLMS, or nocturia.

The effect of dopaminergic treatments on sleep in PD is a complex subjects and will not be discussed in detail. In general, dopaminergic drugs and STN DBS usually improve sleep (Amara et al. 2012); however, higher doses of dopaminergics, especially L-dopa, have been associated with sleep fragmentation and insomnia (Chahine et al. 2013).

Only one autopsy study has primarily compared PD and PD/dementia patients with, versus without, “sleep problems” (Kalaitzakis et al. 2013). There was a significant association between disturbed sleep in PD and α-synuclein pathology in the locus coeruleus (most robust) and raphe nuclei; hypothalamic areas including the paramammillary nuclei and posterior nucleus; subcortical/limbic system including the amygdala, thalamus, and entorhinal cortex regions. A statistically significant increase of tau pathology was observed in the amygdala, CA2 sector of the hippocampus and entorhinal cortex. In general, there was more widespread pathology in PD cases that reported more sleep problems.

The pedunculopontine tegmental nucleus (PPN) has specifically been proposed to be involved in PD-associated sleep disruption, as it is involved in the control of sleep cycling and REM sleep via cholinergic ascending thalamic pathways (Monderer and Thorpy 2009). DBS stimulation of this area, used for severe gait disorders, has improved subjective sleep and increased observed REM sleep (Amara et al. 2011; Peppe et al. 2012), but specific pathologic confirmation is lacking.

Excessive daytime sleepiness and circadian function

In 1999, an article first reported “sleep attacks” while driving in patients taking either pramipexole or ropinirole (Frucht et al. 1999), drawing attention to this subject. Sleep attacks correlate well with general excessive daytime sleepiness (EDS) and people are usually amnesic of sleepiness leading to short sleep episodes, so the term “sleep attack” has given way to EDS. A large number of studies subsequently confirmed “sleep attacks” and associated EDS. Overall, EDS frequency, usually defined by elevated Epworth Sleep Scale scores (>7 or >10 depending on the study), ranges from 15 to 50 % of PD patients (Ondo et al. 2001; Gjerstad et al. 2002; Hobson et al. 2002; Razmy et al. 2004; Ferreira et al. 2006; Suzuki et al. 2011). Major and consistent risk factors for EDS include older age, more advanced PD, dopaminergic use and male sex (Ondo et al. 2001; Hobson et al. 2002; Roth et al. 2003). Male sex is also a risk factor in the general population. Less consistent risk factors include cognitive decline, hallucinations, rapid progression of PD, onset of PD on the left body (right brain), anxiety, depression, and postural instability (Ondo et al. 2001; Gjerstad et al. 2002; Razmy et al. 2004; Ferreira et al. 2006; Boddy et al. 2007; Ghorayeb et al. 2007; Stavitsky et al. 2008; Suzuki et al. 2011).

The main physiologic controversy is whether these sleep episodes truly represent early onset REM sleep, which is a characteristic of narcolepsy. Although data are mixed, a minority of PD patients do seem to have early onset REM (Rye et al. 2000; Ulivelli et al. 2002; Razmy et al. 2004; Ondo et al. 2005; Monaca et al. 2006; Arnulf et al. 2008; Poryazova et al. 2010) (Table 1). It should be noted that determining REM state is difficult in the PD population due to the lack of REM atonia, which may account for some of the variance among studies.

Table 1 Summary of sleep latency testing in patients with PD
Full size table

The mechanisms by which PD is associated with EDS are likely multifactorial. Potential explanations may include exogenous medications, loss of dopamine, loss of norepinephrine and serotonin (alerting monoamines), loss of hypocretin pathways, loss of autonomic function, and loss of circadian control, which has been demonstrated physiologically, and at the molecular level.

Initial investigations into EDS assumed that this resulted from poor nocturnal sleep. Counter-intuitively, this did not seem to be culpable. Several studies using polysomnography and MSLT suggest that patients with more total sleep at night actually have similar or more EDS compared to patients who slept less well (Rye et al. 2000; Arnulf et al. 2002; Roth et al. 2003). Other specific sleep abnormalities such as RBD, RLS, PLMS, and OSA generally fail to correlate with EDS. (Ondo et al. 2001; Cochen De Cock et al. 2010; da Silva-Junior et al. 2014; Plomhause et al. 2013).

The loss of monoamines neurotransmitters through neurodegeneration may contribute to EDS. Endogenous dopamine is generally considered a wake-promoting neurotransmitter, especially when it interacts with D1 receptors. Various lesioning studies of nigral and ventral tegmental areas of non-human primates usually result in sedation and in some cases alter circadian function (Barcia et al. 2004; Barraud et al. 2009). Dopamine receptor type 2 knockout mice are also sleepy (Qu et al. 2010). Exogenous dopamine studies in animals, however, have yielded a spectrum of often contradictory results, seemingly very dependent on dose and animal species and dose. No specific pathologic study has attempted to correlate dopamine cell loss specifically with EDS.

Both serotonin and norepinephrine, which potentiate wakefulness, are also markedly reduced in PD, although no pathologic assessment has specifically tested this.

Hypocretin controls sleep/wake cycling and is markedly reduced in narcolepsy and cataplexy. EDS may result from degeneration of hypocretin cells in the hypothalamus, which are lost in parallel to symptomatic PD progression (Fronczek et al. 2007; Thannickal et al. 2007). However, CSF hypocretin levels, which are markedly reduced in cataplexy/narcolepsy, are usually not very reduced in PD when obtained via lumbar puncture (Yasui et al. 2006; Baumann et al. 2008; Compta et al. 2009). A single ventricular CSF measure done during DBS surgery did show marked CSF hypocretin reduction, and one would speculate that neuron count would correlate to the physiology better than CSF levels (Maeda et al. 2006). Loss of hypocretin function is an attractive model, as it results in sleep/wake instability and could thus explain both sleep fragmentation and frequent daytime sleep episodes.

There are no robust findings from genetic studies that have looked for sleep/wake disorder associations. “Sleep attacks” in PD have been weakly associated with the dopamine D2 receptor gene polymorphism Taq1A (Rissling et al. 2004) and a polymorphism of prehypocretin (Rissling et al. 2005).

Others attribute EDS, and to some extent nocturnal sleep problems, to perturbations in the circadian cycle (Videnovic et al. 2013). Many normal circadian patterns are blunted or desynchronized in PD, including circadian motor activity, increased day and decreased at night (Niwa et al. 2011), R–R variability measured by Holter monitor (Pursiainen et al. 2002), and 24-h cortisol levels (Videnovic and Golombek 2013). There is mixed data on circadian melatonin levels in PD (Fertl et al. 1991; Bordet et al. 2003), but a recent very carefully controlled trial showed reduced melatonin in general and a marked attenuation of the nighttime melatonin spike, suggesting a blunting of normal circadian wake/sleep cycling (Videnovic and Golombek 2013).

There is no human pathological study that specifically correlates alterations circadian control center pathology to sleep/wake phenotype. Cellular loss in PD is seen in almost all circadian control areas including the suprachiasmatic nucleus (SCN). Peripheral autonomic dysfunction, commonly found in PD, may also mitigate circadian physiology responses (perspiration and skin temperature changes).

There is increasing understanding of the molecular basis of circadian function, alterations in which may also contribute to sleep/wake problems in PD. At the most superficial level, cellular endogenous circadian control transcriptional–translational feedback loops are centered in the SCN. The protein products of Bmal1 and CLOCK heterodimerize, forming complexes that stimulate transcription of the “negative feedback” genes Period (Per 1, 2, and 3) and Cryptochrome (Cry 1 and 2) (Lowrey and Takahashi 2011). These proteins then heterodimerize and translate back into the nucleus where they inhibit transcription of Bmal1 and CLOCK genes. Eventually, the Per and Cry proteins breakdown and Bmal1 and CLOCK translation begins again, starting the cycle over again. The process takes on average a little longer than 24 h, although individual SCN cells have periods that vary from 23 to 28 h. They are synchronized by a number of proteins including vasoactive intestinal polypeptide (VIP). The system is also modulated and entrained by physical activity, other environmental factors, and most robustly, by light. Interestingly, retinal dopaminergic cells control light input into the SCN and seem to have their own independent circadian clocks (Wirz-Justice et al. 1984). Perturbation of these has been speculated to blunt downstream circadian function, although no study has attempted to correlate this to empiric sleep/wake disorders in PD (Willis 2008).

Dopamine and CLOCK genes have a complex reciprocal interaction. Dopamine production, especially tyrosine hydroxylase gene activation, the rate-limiting step in dopamine production, is controlled by CLOCK genes (Yujnovsky et al. 2006). Conversely D2/D3 agonists inhibit mCLOCK and mPer1 gene expression, whereas a D1 agonist stimulates these genes (Imbesi et al. 2009). Most importantly, studies have demonstrated that PD patients have a decrease in the expression of the Bmal1 gene during night, suggesting a blunted CLOCK system (Cai et al. 2010). Further research is clearly needed to determine if this underlies many sleep/wake issues in PD.

REM sleep behavior disorder

REM sleep behavioral disorder (RBD) has been extensively studied as a precursor to Lewy body disease and is discussed in general terms elsewhere. In short, RBD is defined by the absence of the normal muscle atonia during the REM phases of sleep. This can lead to dream enacting with potential injury to the patient or the bed partner (Schenck and Mahowald 2002). REM sleep without atonia (RWA) lacks the manifest behavior but demonstrates abnormal muscle activation during REM sleep (Gagnon et al. 2002).

It is now widely accepted that RBD is robustly associated with the development of Lewy body pathology (synucleinopathies, including PD, multiple systems atrophy and Lewy body dementia). Pathological examinations of subjects with RBD show that more than 90 % have Lewy body pathology (Boeve et al. 2013). A number of PD biomarker studies that have evaluated primary RBD in subjects without clinical motor signs of PD have shown abnormalities in between those of PD and normal controls, suggesting RBD heralds impending synucleinopathy diseases (Iranzo et al. 2010; Miyamoto et al. 2011; Shin et al. 2013a). Primary RBD is associated with other pre-motor features of PD including visual changes and olfactory deficits (Postuma et al. 2006; Iwanami et al. 2010; Iranzo et al. 2013). Most clinically important, longitudinal studies of RBD subjects show that a large number eventually develop clinical parkinsonism (Schenck et al. 1996; Iranzo et al. 2006; Kumru et al. 2007; Britton and Chaudhuri 2009; Postuma et al. 2009). In the longest followed series, 81 % of the original cohort, not lost to follow-up, developed parkinsonism, a mean of 14 years after the onset of RBD (Schenck et al. 2013). Within the PD population, active RBD is seen in about half, more commonly males, and the presence of current RBD symptoms has been associated with a worse prognosis, increased dementia, increased hallucination, increased falls, and increased autonomic dysfunction, suggesting a more widespread disease process (Vendette et al. 2007; Sixel-Doring et al. 2011; Postuma et al. 2012; Sorensen et al. 2013; Videnovic et al. 2013). Higher doses of dopaminergic drugs, especially L-dopa, are also associated with RBD, although the direction of causality, if any, is not known (Ozekmekci et al. 2005; Gjerstad et al. 2008; Nihei et al. 2012).

The general anatomy involved in RBD has been long known, but understanding of the specific culpable nuclei is recent and still in flux. The most accepted model of primary pathology of RBD involves the pontine sublateraldorsal (SLD) nucleus, aka “subcoeruleus”, which sends excitatory glutamate projections caudally to the ventrolateral medulla (including the nucleus raphe magnus and ventral and α-gigantocellular reticular nuclei) (Luppi et al. 2013). These send inhibitory GABA and probably glycine projections to the α-motor neurons in the spinal cord, resulting in atonia. Ascending projections from the SLD, thought to be responsible for cortical activation during REM, seem less effective. Other areas are also proposed to be involved with RBD but a primary dopaminergic pathology is unlikely (Luppi et al. 2013).

Pathologic studies of RBD both prior to and after the onset of parkinsonism show Lewy bodies and cell loss in these areas; however, this is not specific, as these areas are thought to be involved in PD in general, and studies to date have not quantified these areas to see if they specifically correlate with RBD within the PD population. This pathology correlates nicely with the concept that RBD is a form fruste of PD. Braak pathology suggests that Lewy body pathology within the CNS starts in the lower brainstem, affecting REM modulating centers before the substantia nigra, thereby potentially explaining the onset of RBD preceding motor features of PD (Braak et al. 2003).

Sleep apnea

When assessed with polysomnography, 20–50 % of PD patients are diagnosed with sleep apnea, usually obstructive (OSA). While some research shows that PD patients have a greater frequency of sleep apnea than healthy age-matched controls, the majority actually show similar or less OSA (Trotti and Bliwise 2010; Yong et al. 2011; Peeraully et al. 2012; Cochen De Cock 2013; Zeng et al. 2013), and when PD/OSA patients are compared to control OSA patients, they have less oxygen desaturations (Nomura et al. 2013). This likely relates to lower body mass index seen in PD (Nomura et al. 2013; Zeng et al. 2013). Moreover, no association has been found between the Apnea–Hypopnea index and multiple sleep latency tests, further suggesting that sleep apnea is not a major contributing factor to the severity of sleepiness in PD patients (Zeng et al. 2013). There may also be less medical consequences because the usual sympathetic discharge associated with apnea is blunted in PD (Valko et al. 2012).

Given the frequent absence of OSA risk factors in PD patients with documented OSA, the pathophysiology in PD is not well understood. It does not correlate with PD duration, motor severity, or PD medications (Wetter et al. 2000). It has been postulated to result from pharyngeal narrowing secondary to degeneration of brainstem neurons that innervate the pharyngeal dilator muscles, which play an important role of maintaining tracheal patency during sleep (Veasey et al. 1996; Friedman and Millman 2008). However, there is no pathologic data to associate this, and brainstem PET imaging of dopaminergic and serotonergic ligands failed to correlate with OSA in PD subjects (Lelieveld et al. 2012).

Restless legs syndrome in PD

Restless legs syndrome is defined by the simultaneous presence of (1) desire to move the extremities, often associated with paresthesia/dysesthesia; (2) worsening of symptoms at rest; (3) transient improvement with movement; and (4) worsening of symptoms in the evening or night (Allen et al. 2003). No widely available biomarker or test corroborates the diagnosis, which is made exclusively via interview. RLS most closely resembles akathisia, also reported in PD, but differs in that (1) the urge to move is isolated in the limbs, rather than the entire body, (2) there is more dramatic relief with ambulation, and (3) there is a more robust worsening at night, with near complete cessation of symptoms in the early morning. RLS is very common, affecting about 10 % of Caucasian populations, although it appears less common in Asian and African populations (Ondo 2002). There is a strong genetic component (Winkelmann et al. 2007).

RLS and PD both respond to dopaminergic treatments and are both variably associated with PLMS. Therefore, a relationship between the two conditions has long been sought.

The pathology of idiopathic RLS involves CNS iron homeostatic dysregulation. CSF ferritin and other measures of iron are lower in RLS cases (Earley et al. 2001), MRI imaging and transcranial ultrasound show reduced iron stores in the striatum and substantia nigra (Allen et al. 2001), and pathologic data show reduced ferritin staining, reduced iron staining, increased transferrin stains, and reduced iron regulatory protein-1 activity (Connor 2008). In contrast, PD is associated with increased iron in dopaminergic areas.

Dopaminergic medications, especially dopamine agonists, robustly improve RLS, even at low doses. However, dopamine imaging studies of the striatum have been inconsistent and difficult to interpret, but do not show any clear dopamine deficiency. Most importantly, pathologic data does not suggest reduced dopamine in RLS. CSF studies and human brain studies of the nigro-striatal system generally suggest normal or even increased dopaminergic turnover (Earley et al. 2013). Specifically, substantia nigra dopaminergic cells are not reduced in number, nor are there markers associated with neurodegenerative diseases, such as tau or alpha-synuclein abnormalities (Pittock et al. 2004). PD, of course, exhibits reduced dopamine cells and multiple neurodegenerative markers. In summary, there are no clear pathologic similarities between PD and idiopathic RLS.

Nevertheless, most surveys that have queried the prevalence of RLS symptoms in PD populations show a twofold increase, compared to control populations (Table 2). Although the absolute number of people with both PD/RLS and idiopathic RLS is lower in Asian surveys, almost all still show relative increases compared to controls.

Table 2 Summary of RLS in PD Studies
Full size table

There is almost no evidence to support that RLS becomes PD, as the vast majority of PD/RLS cases report onset of PD motor symptoms prior to RLS. In fact, the few studies that have looked specifically at RLS populations or idiopathic RLS subjects found no clear increased risk of parkinsonism, and in fact showed a later onset PD in subjects with longstanding RLS (Dragan and Ondo 2010). In short, RLS in this population seems to be a consequence of the PD.

Aside from epidemiologic data, there is relatively little data exploring the pathophysiology of RLS in PD. Two Korean studies have evaluated CNS iron, as measured by sonography, in subjects with both PD and RLS, compared to PD without RLS and idiopathic RLS (Kwon et al. 2010; Ryu et al. 2011). Both found that the PD/RLS group had increased CNS iron similar to PD without RLS. Idiopathic RLS showed reduced CNS iron similar to other studies. It should be noted that patient with PD/RLS generally had PD first and later developed RLS. Similar studies on subjects with young onset RLS, who later developed PD, have not been done.

Genes associated with RLS generally do not seem to be risk factors for PD (Vilarino-Guell et al. 2008). One large family with PD caused by Parkin mutations included a large number of members with RLS, both with and without concurrent PD (Adel et al. 2006). The RLS inheritance pattern was consistent with an autosomal dominant pattern; however, the authors did not find an association between RLS and Parkin mutations within the family. In a South Tyrolean population, Parkin status did not independently predict onset or severity of RLS but did synergistically interact with RLS to predict a younger age at RLS onset (Pichler et al. 2010).

Iatrogenic RLS is a possible consequence of PD treatment. One recent German study specifically evaluated for RLS in untreated PD subjects, and then followed them prospectively after initiation of dopaminergic treatments (Calzetti et al. 2014). After starting therapy, they found an incidence of new RLS of 4.7 and 3.7 %, if they excluded possible secondary causes of RLS. This is much higher than would be expected and suggested that the dopaminergic medications, which are also used to treat idiopathic RLS, may actually initiate the RLS symptoms. This is not entirely counter-intuitive as the use of these medications in idiopathic RLS often results in augmentation. Augmentation is defined by the gradual presentation of an earlier onset of symptoms and a general intensification of RLS symptoms (Garcia-Borreguero et al. 2007). In idiopathic RLS, this often eventually requires the discontinuation of dopaminergics, which results in an acute worsening of symptoms, but subsequent improvement in most cases. This is a compelling argument that RLS in PD is iatrogenic, but cannot control for progression of disease (perhaps RLS most commonly occurs around the same time that dopaminergic therapy is clinically indicated), and is not supported by retrospective trials that show no correlation to dose (Ondo et al. 2002), and a recent prospective trial that did show increased RLS prevalence in untreated PD patients (Shin et al. 2013b).

Interestingly, several reports also suggest that CNS surgeries for PD may also result in RLS symptoms. Kedia et al. (2004) reported the emergence of RLS after sub-thalamic nucleus (STN) deep brain stimulation (DBS) postoperatively in 11 of 195 patients. The mean reduction in antiparkinsonian medication was 74 %. Parra et al. (2006) also reported a case of RLS emergence after DBS. Both postulated that RLS may have resulted from reduction in PD medications. In contrast, several studies report improvement in RLS after both ablative and DBS procedures to both the globus pallidus internus, and STN (Chahine et al. 2011).

Finally, non-dopaminergic medications used to treat PD including anticholinergics/anti-histaminergics, serotonin re-uptake inhibitors, and anti-psychotics are recognized to exacerbate idiopathic RLS.

Periodic limb movements of sleep in PD

Periodic limb movements of sleep are defined by the American Academy of Sleep Medicine as “at least four periodic episodes of repetitive and highly stereotyped limb movements that occur during sleep”. The incidence in the general population increases with age and is strongly associated with RLS (Ancoli-Israel et al. 1991). The physiology of PLM is only partially understood but thought to result from disinhibition of the spinal cord (Bara-Jimenez et al. 2000). PLMS are associated with autonomic lability with transient hypertension and tachycardia accompany and K-complexes (Shneyder et al. 2013).

PD is associated with higher rates of PLMS in some (Trenkwalder 1998; Wetter et al. 2000), but not all reports (Yong et al. 2011; Peeraully et al. 2012). When present, PLMS correlate with the severity of PD, both clinically and on dopamine imaging studies (Happe et al. 2003; Covassin et al. 2012; Sixel-Doring et al. 2012) but have only modest subjective clinical correlates (Covassin et al. 2012). There are mixed data on whether PLMS in PD patients are associated with RLS (Loo and Tan 2008; Covassin et al. 2012).

Conclusion

With the possible exception of sleep apnea, nearly the entire cannon of sleep/wake disorders are more prevalent in PD compared to controls. This probably reflects the widespread pathology in PD. This broad pathologic scope, the confounders of treatment, and associated features, such as aging, all make isolating the exact pathophysiology for specific problems more challenging. Nevertheless, the past few years have afforded tremendous advances in our understanding of sleep/wake issues in PD. Future physiologic, pathologic, molecular, and genetic research will increase our understanding of sleep/wake disorders in PD, some of which will no doubt be extrapolated into a better understanding of sleep/wake physiology in general.