Microsleep disturbances are associated with noradrenergic dysfunction in Parkinson’s disease

Abstract

Study Objectives

Parkinson’s disease (PD) commonly involves degeneration of sleep-wake regulating brainstem nuclei; likewise, sleep-wake disturbances are highly prevalent in PD patients. As polysomnography macroparameters typically show only minor changes in PD, we investigated sleep microstructure, particularly cyclic alternating pattern (CAP), and its relation to alterations of the noradrenergic system in these patients.

Methods

We analyzed 27 PD patients and 13 healthy control (HC) subjects who underwent overnight polysomnography and ¹¹C-MeNER positron emission tomography for evaluation of noradrenaline transporter density. Sleep macroparameters, as well as CAP metrics, were evaluated according to the consensus statement from 2001. Statistical analysis comprised group comparisons and correlation analysis of CAP metrics with clinical characteristics of PD patients as well as noradrenaline transporter density.

Results

PD patients and HC subjects were comparable in demographic characteristics (age, sex, body mass index) and polysomnography macroparameters. CAP rate as well as A index differed significantly between groups, with PD patients having a lower CAP rate (46.7 ± 6.6% versus 38.0 ± 11.6%, p = 0.015) and lower A index (49.0 ± 8.7/hour versus 40.1 ± 15.4/hour, p = 0.042). In PD patients, both CAP metrics correlated significantly with diminished noradrenaline transporter density in arousal prompting brainstem nuclei (locus coeruleus, raphe nuclei) as well as arousal propagating brain structures like thalamus and bitemporal cortex.

Conclusions

Sleep microstructure is more severely altered than sleep macrostructure in PD patients and is associated with widespread dysfunction of the noradrenergic arousal system.

microsleep, cyclic alternating pattern, Parkinson’s disease, positron emission tomography, noradrenaline

Statement of Significance

Parkinson’s disease is the second most common neurodegenerative disorder and characterized by a widespread degeneration of brainstem nuclei. We found that noradrenergic system demise in Parkinson’s disease is associated with disruption of microsleep structure assessed as metrics of cyclic alternating pattern. Our study provides evidence for the involvement of a specific neurotransmitter system in modulation of cyclic alternating pattern occurrence in humans.

Introduction

More than 200 years ago, intriguing sleep phenomena were described as a remarkable feature of Parkinson’s disease (PD) in James Parkinson’s essay on the “Shaking Palsy” [1]. In a more recent survey from 1988, up to 98% of PD patients reported sleep-wake disturbances [2]. Even though PD is still primarily defined by slowing of movements resulting from dopaminergic cell death in the substantia nigra [3], the abundance of patients suffering from disordered sleep is unsurprising: neuropathological changes are present in virtually all key neurotransmitter systems of sleep-wake regulation early in the course of the disease [4]. Surprisingly, despite these dramatic neuropathological changes, assessment of classic sleep macrostructure parameters with polysomnography, i.e. sleep stages, sleep efficiency, and sleep-related breathing events, has typically yielded heterogenous results, showing no explicit differences between PD patients and control populations [5, 6]. In the most recent meta-analysis of 63 case-control studies using polysomnography, differences in polysomnography variables were relatively modest, i.e. less than a 2% difference in the amount of slow-wave sleep, which cannot explain patients’ high burden of disordered sleep [5]. Thus, a closer look on sleep microstructure might elucidate alterations of sleep in PD patients.

The cyclic alternating pattern (CAP) is defined as a periodic electroencephalographic (EEG) activity during non-rapid eye movement (NREM) sleep and represents an essential component of physiological sleep microstructure [7]. CAP is a physiological oscillatory EEG phenomenon involving sequences of salient transient events (A phases) in alternation with background EEG activity (B phases). Depending on the EEG rhythms during the A phase, 3 subtypes can be defined: A1, with EEG synchrony as predominant activity and EEG desynchrony occupying less than 20%; A2, with EEG activity as a mixture of slow and fast rhythms with 20%–50% of the A phase occupied by EEG desynchrony; and A3, with EEG activity of predominantly rapid low-voltage rhythms occupying more than 50% of the A phase [7]. CAP can be provoked by internal and external stimuli of arousal, i.e. auditory perturbations [8] or apnea events [9]; thus, CAP appearance is thought to be associated with arousal instability during sleep [7, 10].

Little is known about changes of CAP metrics in PD patients, but a recent report on subjects with idiopathic REM sleep behavior disorder (iRBD)—representing a prodromal stage of PD [11]—demonstrated a reduction of CAP rate, which was associated with earlier phenoconversion to motor PD [12]. Interestingly, iRBD subjects already show degeneration of lower brainstem nuclei including the arousal-prompting noradrenergic locus coeruleus and its thalamic and cortical terminals comparable to PD [13–15].

In this study, we aimed to analyze CAP metrics in PD patients compared to matched healthy control (HC) subjects and correlate findings with measures of noradrenergic integrity, i.e. noradrenergic terminal density as measured with ¹¹C-MeNER positron emission tomography (PET). We postulated that PD patients would show a reduced CAP rate similar to iRBD subjects, and this would be associated with altered noradrenergic terminal density.

Methods

Participants

We evaluated 30 PD patients and 12 HC subjects from a previous study [15], and additional five subjects (three PD patients and two HC subjects) from an ongoing study at our institution who underwent overnight polysomnography and ¹¹C-MeNER PET. To reduce a potential bias on CAP metrics due to extraordinary high rates of apnea/hypopnea or periodic limb movement (PLM) events [9], we only included subjects with apnea-hypopnea index (AHI) in the range of 0–25/hour and PLM index in the range of 0–80/hour, resulting in 27 PD patients and 13 HC subjects for final analysis. All subjects were between 50 and 85 years old, non-demented (Montreal Cognitive Assessment > 22), non-depressed (Geriatric Depression Scale 15 < 6), and without known sleep-related breathing disorders. Diagnosis of PD was undertaken according to the current consensus criteria [3], differential dopaminergic therapy was ascertained, levodopa equivalent daily doses (LEDD) were estimated as previously recommended [16], disease severity was judged according to Hoehn and Yahr stage, and motor symptoms were quantified according to the Movement Disorder Society Unified Parkinson’s disease rating scale part III (MDS-UPDRS III) after 12 hours of medication withdrawal. Patients were recruited through advertisements in the Danish Parkinson’s disease magazine and from collaborating neurological clinics. HC subjects were recruited through newspaper advertisements. All subjects gave written informed consent prior to study inclusion.

Polysomnographic data and CAP analysis

We used a mobile SOMNOscreen plus device for overnight video-polysomnography including 10 EEG recordings (according to the international 10/20 system: F3, F4, C3, C4, O1, O2, A1, A2, Fpz as grounding, and Cz as reference), electrooculography, surface electromyography of the submental muscle and the tibialis anterior muscles, electrocardiography, nasal pressure monitoring, thoracic and abdominal respiratory effort belts, finger pulse oximetry, and synchronized audio-visual recording. Visual PSG scoring was performed on 30-second epochs, including sleep efficiency, sleep cycles, total sleep time, relative amount of stage 1 (N1), stage 2 (N2), stage 3 (N3) sleep, and REM sleep, arousal index (number of arousals per hour of sleep), apnea-hypopnea index (AHI, number of apnea plus hypopnea events per hour of sleep) and periodic limb movement index (PLMI, number of periodic leg movements per hour of sleep) according to the American Academy of Sleep (AASM) Manual for the Scoring of Sleep and Associated Events Version 2.5. An arousal was defined as a visible and abrupt shift to higher frequencies in any of the electroencephalogram leads, with a duration of 3 to 15 seconds [17]. Movement- and breathing-related arousals were included in the further analysis. Patients were allowed to take their regular evening medication during overnight polysomnography, motor assessments were performed in the morning after 12 hours of medication withdrawal.

CAP scoring followed the rules of the consensus criteria published in 2001 [7]. Scoring was performed using DOMINO software by a board-certified sleep expert (MS) who was blinded to clinical features during analysis by anonymizing of the PSG and blocked video recording. EEG channels were presented referenced to the contralateral ear as well as bipolar in a longitudinal montage to optimize A and B phase detection. A phase subtypes (A1, A2, A3) during NREM sleep were tagged directly on the EEG data and exported from DOMINO software into the American Standard Code for Information Interchange (ASCII). Similarly, sleep stage scorings were exported and a custom-made Python script identified NREM episodes including CAP by applying the consensus criteria to the exported data [7]. The following CAP parameters were calculated: CAP rate (CAP time / total NREM time × 100), occurrence rate A1-A3 (amount of A1(-A3) phases / all A phases x 100), mean duration of A and B phases, and average duration of CAP sequences.

¹¹C-MeNER PET imaging

All PET scans were acquired on a Siemens High Resolution Research Tomograph (HRRT) during daytime, less than 1 week apart of polysomnography recording. ¹¹C-MeNER is a high-affinity and highly selective radioligand to the noradrenergic transporter located presynaptically on brain neurons mostly originating from the locus coeruleus [18]. Ligand binding is thought to be unaffected by transient fluctuations of endogenous noradrenaline levels. ¹¹C-MeNER radiosynthesis and image acquisition were performed as described previously [19]. In brief, after a transmission scan and intravenous bolus injection of ¹¹C-MeNER, a 90 minutes PET scan was acquired. Images were analyzed with PMOD 4.0, including motion correction and normalization into Montreal Neurological Institute (MNI) space via rigid matching of subject’s PET to an additionally acquired anatomical T1-weighted MRI. To optimize kinetic modeling, we smoothed PET images with a 4 mm Gaussian filter prior to calculation of parametric maps of distribution volume ratios (DVRs) with the simplified reference tissue model 2 (SRTM2). We used voxel-wise calculations in PXMOD initiated with thalamus and caudate as ¹¹C-MeNER high- and low-binding regions, respectively. DVRs of the thalamus, locus coeruleus, median raphe, and dorsal raphe were extracted for volume of interest analysis for group comparison and correlation analysis as they show high ¹¹C-MeNER binding and represent brain structures tightly engaged in arousal propagation. Thalamus and caudate were defined from the PMOD built-in atlas, and the locus coeruleus as well as dorsal and median raphe were manually drawn accordingly to published MNI coordinates and anatomical landmarks [20, 21]. For correlation analysis of cortical ¹¹C-MeNER uptake, all individual parametric maps were non-rigidly transformed to an in-house ¹¹C-MeNER template and additionally smoothed with a 4 mm Gaussian filter prior to voxel-wise analysis [22, 23].

Statistical analysis

We analyzed the data with Statistical Package for the Social Sciences (SPSS) version 24. Group data are presented as mean ± standard deviation unless otherwise stated. Group comparisons were calculated using Student t, Mann-Whitney, and chi-square statistics as appropriate, univariate correlation analyses were calculated with Pearson’s r, and multivariate associations were analyzed with linear regression analysis. Normal distribution of data was assessed with the Shapiro-Wilk test, Q-Q plots, and box plots. Significance was accepted at p < 0.05. For volume of interest-based PET analysis, we calculated a repeated-measures analysis of variance (ANOVA) with group as between-subject and brain regions as within-subject factors. Statistical Parametric Mapping 12 (SPM12) software was used for voxel-wise correlation analysis of cortical ¹¹C-MeNER binding, including a cortical gray matter mask for explicit masking. The voxel-level analysis threshold was set at p < 0.005 uncorrected. Surface projections of the resulting voxels were overlaid on the rendered brain included in the SPM12 package.

Results

Demographic and clinical data

HC subjects and PD patients were comparable in age, sex, and body mass index (Table 1). Mean age of all included subjects was 66.1 ± 8.3 years, and 30% were female. PD patients had a mean disease duration of 6.3 ± 4.2 years and an average Hoehn & Yahr stage of 2.2 ± 0.4. Only one patient was treatment-naïve, nine patients were levodopa-naïve. On conventional polysomnographic parameters, PD patients and HC subjects showed largely similar metrics (Table 1), but PD patients exhibited a numerical but statistically nonsignificant decrease of sleep stage N1 (16.2 ± 4.9% versus 12.6 ± 6.2%, p = 0.079) and an increase of REM sleep amount (12.3 ± 4.6% versus 18.0 ± 9.3%, p = 0.045). Both groups were comparable according to arousal prompting events, i.e. AHI and PLMI, but PD patients showed a diminished arousal index compared to HC subjects (23.0 ± 4.9/hour versus 15.2 ± 6.1/hour, p < 0.001).

Table 1.

Clinical and sleep characteristics of included subjects

	Healthy controls, n = 13	Parkinson’s disease patients, n = 27	P value
Age [years]	67.4 ± 6.6	65.5 ± 9.0	0.669a
Sex [male / female]	9 / 4	18 / 9	0.871b
Body Mass Index [kg/m2]	25.3 ± 2.3	25.6 ± 3.4	0.805c
PD-related characteristics
Disease duration [years]		6.3 ± 4.2	-
Hoehn & Yahr stage		2.2 ± 0.4	-
MDS-UPDRS III, OFF state		35.9 ± 11.1	-
Levodopa equivalent daily dose [mg]		603.0 ± 418.6	-
Subjects without levodopa treatment		9
Levodopa daily dose [mg]d		525.0 ± 349.0
Polysomnography characteristics
Sleep efficiency [%]	82.3 ± 11.0	84.0 ± 9.3	0.613c
Sleep cycles	4.5 ± 0.7	4.0 ± 1.0	0.073a
Sleep latency N2 [min]	15.0 ± 14.5	16.0 ± 15.6	0.776a
Wake [%]	17.7 ± 11.0	15.9 ± 9.3	0.603c
N1 [%]	16.2 ± 4.9	12.6 ± 6.2	0.079c
N2 [%]	39.2 ± 9.9	39.3 ± 9.5	0.981c
N3 [%]	14.6 ± 6.0	13.9 ± 7.2	0.743c
REM [%]	12.3 ± 4.6	18.0 ± 9.3	0.045c
Apnea-Hypopnea Index [/hour]	10.1 ± 7.6	7.5 ± 6.8	0.264a
Periodic limb movement index [/hour]	17.6 ± 24.0	17.1 ± 22.5	0.955a
Arousal index [/hour]	23.0 ± 4.9	15.2 ± 6.1	<0.001a
REM sleep behavior disorder [n]	0	13	<0.001b

MDS-UPDRS III, Movement Disorder Society - Unified Parkinson’s disease Rating Scale part III; N1-3, non-rapid eye movement sleep stage 1-3, REM, rapid eye movement sleep. Values are given as mean ± SD.

^aNonparametric Mann-Whitney-Test.

^bChi-square test.

^cParametric Student’s t-test.

^dOnly patients taking levodopa were included.

CAP sleep microstructure analysis

PD patients exhibited an approximately 19% lower CAP rate than controls (46.7 ± 6.6% versus 38.0 ± 11.6%, p = 0.015; Table 2). Occurrence of CAP was diminished in PD patients in all NREM sleep stages by a similar amount but reached statistical significance in N2 and N3 only. Duration of averaged CAP sequences was shortened in PD patients (244.0 ± 40.7s versus 193.9 ± 39.5s, p < 0.001), despite similar duration for A and B phases between groups. This points to inclusion of less A and B phases in each CAP in PD patients. Accordingly, index of A phases was lower in PD patients compared to healthy controls (49.0 ± 8.7/hour versus 40.1 ± 15.4/hour, p = 0.042). Subtype analysis revealed that both, the A1 index and the combined A2+A3 index were numerically lower in PD patients, but only the difference seen in the A2+A3 index reached statistical significance (Table 2). The nine levodopa-naïve patients did not differ in any examined CAP measure compared to patients taking levodopa (all p > 0.05).

Table 2.

Metrics of cyclic alternating pattern in healthy controls and PD patients

	Healthy controls	PD patients	P value
CAP Rate [%]	46.7 ± 6.6	38.0 ± 11.6	0.015a
CAP Rate, N1 [%]	22.1 ± 9.7	16.4 ± 11.5	0.132a
CAP Rate, N2 [%]	48.4 ± 11.3	37.9 ± 13.0	0.017a
CAP Rate, N3 [%]	73.4 ± 10.2	56.8 ± 13.8	<0.001a
Sequence duration [s]	244.0 ± 40.7	193.9 ± 39.5	<0.001b
Duration A [s]	8.4 ± 1.0	8.6 ± 1.0	0.577a
Duration B [s]	26.1 ± 1.8	26.4 ± 2.6	0.754a
A index [/hour]	49.0 ± 8.7	40.1 ± 15.4	0.042b
A1 index [/hour]	36.8 ± 7.1	31.7 ± 13.4	0.204a
A2 + A3 index [/hour]	12.2 ± 3.8	8.5 ± 4.3	0.004b

Significant differences are indicated in bold and significance was accepted at P < 0.05.

CAP, cyclic alternating pattern; N1-3, non-rapid eye movement sleep stage 1-3; PD, Parkinson’s disease; s, seconds. Values are given as mean ± SD.

^aParametric Student’s t-test.

^bNonparametric Mann-Whitney-Test.

The combined A2+A3 index, but not the A1 index correlated with arousal index independently of group effects (β = 0.525, p = 0.002 for A2+A3 index, and β = −0.236, p = 0.169 for A1 index, respectively; multiple linear regression analysis).

A stepwise multiple linear regression analysis showed an association between worsening motor symptoms (higher MDS-UPDRS III) and lower A index in PD patients (β = −0.402, p = 0.038). No associations were found for age, disease duration, Hoehn & Yahr stage, and LEDD. Similar analysis for CAP rate did not reveal any significant predictor (trend for MDS-UPDRS III: β = −0.339, p = 0.084). Results remained identical when AHI and PLMI were included in the analysis as potential confounders.

CAP metrics and noradrenaline transporter PET

Consistent with our previous publication [15], PD patients exhibited reduced ¹¹C-MeNER binding in arousal prompting brainstem nuclei and the thalamus (group effect, p = 0.002; pontine raphe, p = 0.115; dorsal raphe, p = 0.003; locus coeruleus, p = 0.004; thalamus p < 0.001; Figure 1A). CAP rate as well as A index correlated with ¹¹C-MeNER binding in these regions in PD patients but not in HC subjects (for PD patients: pontine raphe, r = 0.615, p = 0.001; dorsal raphe, r = 0.646, p < 0.001; locus coeruleus, r = 0.371, p = 0.057 (in pooled HC + PD analysis: r = 0.365, p = 0.021); thalamus, r = 0.630, p < 0.001; Figure 1B). The findings were robust when MDS-UPDRS III values and levodopa-treatment were included as covariates. CAP rate also correlated with bitemporal cortical ¹¹C-MeNER binding in PD patients (Figure 2A). Accordingly, A index correlated with ¹¹C-MeNER binding in similar regions (Figure 2B). CAP rate and A index did not correlate with ¹¹C-MeNER binding in the HC group.

Figure 1.

¹¹C-MeNER binding and its correlation with CAP metrics. (A) Left panel: coronal section of averaged ¹¹C-MeNER distribution volume ratio (DVR) of 13 healthy controls (HC, red border) and 27 Parkinson’s disease patients (PD, blue border) at the level of the dorsal brainstem as indicated on anatomical T1 image—images are leveled from 0.8 (dark blue) to 1.8 (red). Right panel: Averaged ¹¹C-MeNER DVR in noradrenaline transporter-rich regions (red HC subjects, blue PD patients)—all regions but pontine raphe differed significantly (p < 0.005). (B) Upper row: dot plots of cyclic alternating pattern (CAP) rate and ¹¹C-MeNER DVR in noradrenaline transporter rich regions; lower row: dot plots of A index and ¹¹C-MeNER DVR in noradrenaline transporter rich regions (red HC subjects, blue PD patients).

Figure 2.

Voxel-wise ¹¹C-MeNER correlation to CAP metrics in PD patients. (A) Correlation of cortical ¹¹C-MeNER binding with cyclic alternating pattern (CAP) rate and (B) with A index.

Discussion

We demonstrated that despite comparable polysomnographic macroparameters of sleep, PD patients exhibited significant alterations in sleep microstructure assessed as CAP metrics: patients showed reductions in CAP rate and A index and shortened CAP sequence duration. Occurrence of CAP was related to the integrity of the noradrenergic system in arousal-promoting and -propagating brain structures.

Even though arousals are determined as events of sleep interruption and fragmentation, their presence during sleep is a normal physiological phenomenon in all ages - even in the absence of an external or internal trigger [24]. The assessment of CAP expands the view on dynamic organization of sleep by considering physiological rhythms of stable (periods without CAP) and unstable sleep (phases showing CAP) [25]. Additionally, a three-stage hierarchy of arousal strength can be identified within the CAP methodology indicated by the different A phases with gradual decline of EEG synchrony [25]. The A3 phase with predominant EEG desynchrony—and to a lesser extent the A2 phase—are most comparable to the classical “arousal” definition according to the AASM scoring scheme and both parameters typically correlate, as they did in our sample [9]. However, the arousal index increases linearly with age [24], whereas CAP rate follows a U-shape with a nadir during young adulthood and an increasing propensity in the elderly [26], clearly indicating differences in both assessments—accounting for sleep fragmentation (classical arousal) and sleep instability (CAP) [25]. Accordingly, an increased CAP rate—but also arousal index—can occur upon pathologically elevated sleep disruptive elements such as periodic limb movements, apnea/hypopnea events [9], or auditory pertubations [8]. Nevertheless, CAP also represents physiological oscillations of the sleeping brain showing recurring periods of activation; hence, not only an increase of CAP occurrence but also a decrease might be indicative of defective sleep not capable to grant its full beneficial potential.

Our findings of diminished CAP presence are in line with earlier observations in iRBD subjects, which represent a prodromal stage of PD [12]. Therefore, disruption of microsleep structure may occur early in PD genesis due to degeneration of non-dopaminergic pathways. Interestingly, a lower CAP rate was also predictive of more rapid phenoconversion of iRBD subjects to overt motor PD, possibly indicating a more advanced stage of iRBD [12]. Growing evidence suggests that propagating α-synuclein pathology—probably driving neurodegeneration in PD—impacts lower brainstem nuclei in the medulla and the pons prior to dopaminergic neurons located in the mesencephalon [14, 27, 28]. This specifically holds true for the arousal promoting noradrenergic locus coeruleus [14, 29, 30], whereas the serotonergic raphe nuclei seem to be affected only in subpopulations of iRBD subjects [31–33]. These findings underpin the central role of noradrenergic signaling in the regulation of CAP occurrence. Accordingly, CAP rate as well as A index were associated with noradrenergic transporter density independently of motor deterioration as measured with the MDS-UPDRS III in our PD sample. Nevertheless, even though we identified nine PD patients as being levodopa-naïve and we did not observe significant differences between levodopa-naïve patients and patients on levodopa treatment, our study was not designed to properly assess treatment effects on CAP metrics. Levodopa as precursor for various monoamines could potentially also modify noradrenergic signaling, and we cannot rule out treatment effects altering sleep microstructure.

We also observed an association of thalamic and temporal cortical noradrenaline transporter density with CAP rate and A index. The thalamus is a key relay station, in which a plethora of thalamo-cortical connections are converging, which can be selectively enhanced and diminished, eventually modulating the state of consciousness and sleep [34, 35]. Strikingly, loss of connectivity in the temporal cortex is a hallmark of the transition from the wake state to sleep and vice versa [36, 37]. Hence, loss of noradrenergic function in the thalamus and temporal cortex together with reduced CAP in PD patients might represent an altered ability to (1) react to arousal prompting triggers and (2) to establish states of higher activation during sleep. In line with this, patients also showed a reduced arousal index compared to HC subjects, as shown previously [17].

Reports on patients with mild cognitive impairment (MCI) as a precursor of Alzheimer’s disease indicated similar changes of CAP metrics as seen in our study and in iRBD subjects [12, 38, 39]. Patients with MCI showed reductions in CAP rate and the A index, and a stronger reduction was associated with more rapid cognitive deterioration and conversion to clinical dementia [38, 39]. Strikingly, degeneration of the noradrenergic LC is—similar to PD—also an early insult in MCI and Alzheimer’s disease and LC cell loss could be associated with cognitive performance in these patients [40, 41].

However, limited functional imaging data has been published on brain circuitries engaged in CAP occurrence and modulation. A recent study using functional magnetic resonance imaging techniques reported enhanced activation of a network involving the insula, the middle cingulate gyrus, and the basal forebrain during A phases as recorded with simultaneous electroencephalography [42]. To the best of our knowledge, no other studies have utilized functional imaging data and investigated the relationship between CAP regulatory neurotransmitter systems and potentially involved networks - more work will hopefully shed light on this topic.

In summary, our data indicate that PD patients exhibit changes of microsleep parameters which exceed alterations seen with classic sleep analysis. Specifically, markers of sleep instability were reduced and associated with noradrenergic dysfunction. As noradrenergic function plays a crucial role in maintaining cognitive performance in AD and PD, microsleep disturbances might be another putative link to cognitive disturbances in these disorders.

Funding

This research was supported by the Lundbeck Foundation (grant number 6970), the Danish Council for Independent Research, Medical Sciences (grant number 0602-02700), Aarhus University Research Foundation, the Danish Association of Parkinson’s Disease (Parkinson-foreningen), the Swiss National Science Foundation (grant number P2SKP3_161812), the Hildegard Henssler-Stiftung, the Koeln Fortune Program / Faculty of Medicine, University of Cologne (grant number 453/2018), the Cologne Clinician Scientist Program (CCSP) / Faculty of Medicine / University of Cologne, funded by the German Research Foundation (DFG, FI 773/15-1) and the Else Kröner-Fresenius-Stiftung (grant number 2019_EKES.02).

Conflict of interest statement. None declared.

Author notes

Both authors contributed equally (shared last authorship).