Brain-first versus body-first Parkinson’s disease: a multimodal imaging case-control study

Abstract

Parkinson’s disease is characterized by the presence of abnormal, intraneuronal α-synuclein aggregates, which may propagate from cell-to-cell in a prion-like manner. However, it remains uncertain where the initial α-synuclein aggregates originate. We have hypothesized that Parkinson’s disease comprises two subtypes. A brain-first (top-down) type, where α-synuclein pathology initially arises in the brain with secondary spreading to the peripheral autonomic nervous system; and a body-first (bottom-up) type, where the pathology originates in the enteric or peripheral autonomic nervous system and then spreads to the brain. We also hypothesized that isolated REM sleep behaviour disorder (iRBD) is a prodromal phenotype for the body-first type. Using multimodal imaging, we tested the hypothesis by quantifying neuronal dysfunction in structures corresponding to Braak stages I, II and III involvement in three distinct patient groups. We included 37 consecutive de novo patients with Parkinson’s disease into this case-control PET study. Patients with Parkinson’s disease were divided into 24 RBD-negative (PD^RBD−) and 13 RBD-positive cases (PD^RBD+) and a comparator group of 22 iRBD patients. We used ¹¹C-donepezil PET/CT to assess cholinergic (parasympathetic) innervation, ¹²³I-metaiodobenzylguanidine (MIBG) scintigraphy to measure cardiac sympathetic innervation, neuromelanin-sensitive MRI to measure the integrity of locus coeruleus pigmented neurons, and ¹⁸F-dihydroxyphenylalanine (FDOPA) PET to assess putaminal dopamine storage capacity. Colon volume and transit times were assessed with CT scans and radiopaque markers. Imaging data from the three groups were interrogated with ANOVA and Kruskal-Wallis tests corrected for multiple comparisons. The PD^RBD− and PD^RBD+ groups showed similar marked reductions in putaminal FDOPA-specific uptake, whereas two-thirds of iRBD patients had normal scans (P < 10⁻¹³, ANOVA). When compared to the PD^RBD− patients, the PD^RBD+ and iRBD patients showed reduced mean MIBG heart:mediastinum ratios (P < 10⁻⁵, ANOVA) and colon ¹¹C-donepezil standard uptake values (P = 0.008, ANOVA). The PD^RBD+ group trended towards a reduced mean MRI locus coeruleus: pons ratio compared to PD^RBD− (P = 0.07, t-test). In comparison to the other groups, the PD^RBD+ group also had enlarged colon volumes (P < 0.001, ANOVA) and delayed colonic transit times (P = 0.01, Kruskal-Wallis). The combined iRBD and PD^RBD+ patient data were compatible with a body-first trajectory, characterized by initial loss of cardiac MIBG signal and ¹¹C-colonic donepezil signal followed by loss of putaminal FDOPA uptake. In contrast, the PD^RBD− data were compatible with a brain-first trajectory, characterized by primary loss of putaminal FDOPA uptake followed by a secondary loss of cardiac MIBG signal and ¹¹C-donepezil signal. These findings support the existence of brain-first and body-first subtypes of Parkinson’s disease.

PET, MRI, Parkinson’s disease, REM sleep behaviour disorder

See Bohnen and Postuma (doi:10.1093/brain/awaa293) for a scientific commentary on this article.

Introduction

A pathological hallmark of Parkinson’s disease is the presence of intraneuronal α-synuclein inclusions termed Lewy pathology. Accumulating evidence suggests that misfolded α-synuclein protein behaves in a prion-like manner leading to cell-to-cell propagation of pathology (Uchihara and Giasson, 2016). However, it remains unknown from where the initial α-synuclein aggregates originate.

It has been hypothesized that α-synuclein inclusions initially form in nerve terminals of the enteric nervous system, and then subsequently spread via autonomic connections to the dorsal motor nucleus of the vagus and intermediolateral cell columns of the sympathetic system (Braak et al., 2003a, b; Borghammer, 2018). This proposal is supported by animal evidence showing gut-to-brain spreading of α-synuclein aggregates through autonomic nerves (Kim et al., 2019; Van Den Berge et al., 2019). In addition, Lewy pathology has been detected in gastrointestinal nerve fibres years prior to Parkinson’s disease diagnosis (Stokholm et al., 2016). There is also epidemiological evidence that complete but not partial vagotomy may protect against later Parkinson’s disease (Svensson et al., 2015; Liu et al., 2017).

However, other lines of evidence suggest that Parkinson’s disease probably does not start in the enteric nervous system in all cases. Autopsy studies have shown that a minority of cases with Lewy pathology do not have pathological inclusions in the dorsal motor nucleus of the vagus (Parkkinen et al., 2008), and that a fraction of cases display a limbic-predominant distribution of α-synuclein inclusions with less pathology in the brainstem (Kosaka et al., 1984; Beach et al., 2009; Raunio et al., 2019). Also, up to 50% of de novo Parkinson’s disease Hoehn and Yahr stage I patients have been reported to have preserved cardiac sympathetic innervation as measured by ¹²³I-metaiodobenzylguanidine (MIBG) scintigraphy (Kashihara et al., 2010; Kim et al., 2017).

Based on these considerations, we have hypothesized that Parkinson’s disease comprises two subtypes (Fig. 1): (i) a body-first (bottom-up) subtype, where the pathology originates in the enteric or peripheral autonomic nervous system, and then ascends via the vagus nerve and sympathetic connectome to the CNS (Borghammer and Van Den Berge, 2019; Van Den Berge et al., 2019); and (ii) a brain-first (top-down) subtype, in which the α-synuclein pathology initially arises in the brain itself or sometimes enters via the olfactory bulb, and subsequently descends to the peripheral autonomic nervous system.

Figure 1

Hypothetical body-first and brain-first subtypes of Parkinson’s disease. The figure depicts two contrasting spreading routes. The boxes summarize predicted symptoms and imaging findings in the two Parkinson’s disease (PD) subtypes, when motor symptoms first emerge. Numbered circles depict the starting point and subsequent propagation of pathology in the two subtypes. (A) In the body-first type, the initial α-synuclein pathology appears in the enteric or peripheral autonomic nervous system. It then propagates via the sympathetic connectome to the heart, and via the vagus nerve to the dorsal motor nucleus of the vagus. Ascending pathology affects pontine structures giving rise to RBD before the substantia nigra shows substantial involvement. When parkinsonism appears, signifying a loss of >50% of nigrostriatal dopamine terminals, all lower Braak stage structures show marked damage on relevant imaging markers. (B) In the brain-first type, the initial α-synuclein pathology appears in the CNS. The most likely site of origin seems to be the amygdala or connected structures, or the pathology may in some cases enter via the olfactory bulb. Rarely, the pathology arises in the upper brainstem (substantia nigra or locus coeruleus). The pathology spreads from the site of origin to the brainstem and cortex. When parkinsonism appears, the brainstem shows a rostro-caudal gradient of pathology with marked involvement of the substantia nigra, moderate involvement of the neighbouring pons, but little involvement of the medulla and autonomic nervous system. RBD is rarely present when parkinsonism emerges, but a top-down RBD can emerge later. Neuromelanin = MRI locus coeruleus/pons ratios.

Importantly, we hypothesize that premotor REM sleep behaviour disorder (RBD), i.e. the appearance of isolated REM sleep behaviour disorder (iRBD) well before parkinsonism is a strong marker of the body-first subtype, as propagating bottom-up pathology will affect the pons before reaching the substantia nigra (McKenna and Peever, 2017). In contrast, premotor RBD is not seen in the brain-first type, but a top-down RBD can emerge after the onset of parkinsonism caused by descending pathology to the pons (Dauvilliers et al., 2018) (Fig. 1B). Thus, body-first and brain-first Parkinson’s disease is not synonymous with RBD-positive and RBD-negative Parkinson’s disease.

To test this hypothesis, we recruited de novo patients with Parkinson’s disease and performed video-polysomnography to divide patients into de novo Parkinson’s disease without RBD (PD^RBD−) and de novo Parkinson’s disease with premotor RBD (PD^RBD+). The latter group was defined as patients with Parkinson’s disease whose subjective RBD sleep symptoms had appeared at least 1 year prior to the onset of motor symptoms. The patients were then studied with a multimodal imaging battery designed to characterize damage to Braak stages I, II and III structures (Knudsen et al., 2018). We used ¹¹C-donepezil PET/CT to assess cholinergic (including parasympathetic) innervation of the colon (Gjerloff et al., 2015; Fedorova et al., 2017), ¹²³I-MIBG scintigraphy to measure sympathetic cardiac innervation (Miyamoto et al., 2006), neuromelanin-sensitive MRI to measure the integrity of pigmented cell bodies of the locus coeruleus (Ehrminger et al., 2016; Sommerauer et al., 2018), and ¹⁸F-dihydroxyphenylalanine (FDOPA) PET to assess nigrostriatal dopamine storage capacity. The de novo Parkinson’s disease data were compared to previously published data from iRBD patients (Knudsen et al., 2018).

We hypothesized that iRBD and de novo PD^RBD+ data would align with a continuous body-first trajectory, characterized by initial damage to the autonomic nervous system followed by dopaminergic damage. In contrast, we predicted that de novo PD^RBD− patients would align with the brain-first trajectory, characterized by marked dopaminergic damage but relatively little or no damage to lower brainstem structures and the peripheral autonomic nervous system.

Materials and methods

Study design and participants

The study was conducted between August 2018 and January 2020. We recruited 47 consecutive probable de novo patients with Parkinson’s disease from the Aarhus University Hospital clinic, private neurologists, and from Kiel University Hospital, Germany. Inclusion criteria: age 50–85 years with clinically probable Parkinson’s disease according to Movement Disorder Society (MDS) criteria (Postuma et al., 2015). For PD^RBD+ patients, subjective RBD sleep symptoms must have appeared at least 1 year before motor symptoms. Exclusion criteria: normal FDOPA PET or dopamine transporter scan (DaTscan) (this criterion only applied to patients with Parkinson’s disease, not iRBD patients), psychiatric disorders, cholinesterase inhibitors, diabetes, neuropathies, heart or kidney failure, current or previous cancer and/or major surgery on abdominal organs, inflammatory bowel disease. Five of the 47 recruited patients with Parkinson’s disease were immediately excluded because of unequivocally normal FDOPA PET or DaTscans. A further five patient datasets were excluded from analyses as these patients could not be assigned to the PD^RBD+ or PD^RBD− categories (explained below).

All de novo Parkinson’s disease and iRBD patients had video-polysomnography using the SOMNOmedics portable polysomnography equipment, as previously described (Knudsen et al., 2018). Each polysomnography dataset was analysed and diagnostic consensus obtained between two board-certified somnologists (M.O., M.S.), and RBD was diagnosed according to the International Classification of Sleep Disorders III criteria. RBD symptoms were assessed with the RBD screening questionnaire (Stiasny-Kolster et al., 2007), and all subjects and bed-partners were carefully interviewed about the time duration of RBD symptoms and of parkinsonism. Thus, the PD^RBD− group included RBD-negative patients, and the PD^RBD+ group included those patients with Parkinson’s disease, in whom the subjective RBD sleep symptoms with certainty had appeared at least 1 year prior to the onset of parkinsonism. This strategy was used to avoid including potential brain-first patients with Parkinson’s disease with top-down RBD into the study (Fig. 1).

Nine patients with Parkinson’s disease had just started medication at the time of inclusion (ropinirole/ropinirole+selegelin/pramipexole n = 8, levodopa n = 1). In those cases, motor assessment and radioisotope imaging were performed after >12 h of medication abstinence. The remaining patients had not yet started Parkinson’s disease medication. Four patients received drugs that can affect the clinical expression of RBD. One patient with RBD received serotonin reuptake inhibitor (SSRI) and benzodiazepine and one patient with RBD received melatonine. One excluded patient who experienced RBD symptoms after the Parkinson’s disease diagnosis took SSRI, benzodiazepine and melatonine. Finally, one patient without RBD received SSRI and benzodiazepine.

Previously published data from 22 iRBD patients were used for comparison (Knudsen et al., 2018). All comparator data were acquired with identical methodologies except for two deviations explained below. The study was approved by the Science Ethical Committees of the Central Denmark Region and Kiel University Hospital (case nr. 1-10-72-160-16). All subjects provided informed written consent according to the Declaration of Helsinki.

Imaging

The Parkinson’s disease subjects were scanned on a 3T Siemens SKYRA magnetic resonance system; protocols included T₁ MPRAGE used for co-alignment with FDOPA PET and a neuromelanin sensitive turbo spin echo T₁ for interrogation of the locus coeruleus neuromelanin signal (repetition time 600 ms, echo time 10 ms, flip 120°, nine slices, thickness 1.98 mm). Using a volume of interest approach, locus coeruleus/pons ratios were calculated, as previously described (Sommerauer et al., 2018). The magnetic resonance system used for our previously published studies was replaced with a new magnetic resonance scanner, so the current de novo Parkinson’s disease neuromelanin data were not comparable to our in-house iRBD data.

FDOPA PET was performed as previously described (Stokholm et al., 2017; Sommerauer et al., 2018). One hour before injection of 110 MBq FDOPA, 150 mg of carbidopa was administered orally. A 6-min transmission scan and a 20 min PET acquisition (70 to 90 min post-injection; four frames of 5 min) was obtained in list mode on an ECAT high-resolution research tomograph (Siemens/CTI). PET data were reconstructed using a 3D-OSEM algorithm yielding data volumes, which were motion corrected and normalized to MNI space with rigid matching of the subject’s PET to the anatomical MRI. Volumes of interest were defined in the putamen and occipital cortex using the built-in Hammers N30R83 atlas in PMOD and specific binding ratios were calculated (putamen/occipital − 1). Slight manual adjustments to the volumes of interest were made blindly to subject category. The lowest of left/right putamen specific binding ratios was always used for analyses in all groups.

¹²³I-MIBG scintigraphy was performed using a dual-head gamma camera (Siemens Symbia SPECT/CT) with a low-energy high-resolution collimator. Fifteen minute images of the thorax were obtained 15 min (early) and 3.5 h (late) after injection of 111 MBq ¹²³I-MIBG. Regions of interest were defined on the heart myocardium and mediastinum. Mean heart-uptake/mean mediastinum-uptake (H/M) ratios were calculated on early and late images, and washout rates were calculated as H/M_late – H/M_early.

For ¹¹C-donepezil PET/CT, the subjects fasted for at least 6 h and abstained from drinking 4 h before PET. Forty-three minutes after injection of 440 MBq ¹¹C-donepezil, a 10-min static PET acquisition was performed in 3D mode with the abdominal organs in field of view. A CT with contrast enhancement was performed immediately prior to PET. The first 20 consecutive de novo patients with Parkinson’s disease had donepezil PET on a Siemens Biograph PET/CT (image reconstruction previously described, Fedorova et al., 2017), and the following 21 on a Siemens Biograph Vision PET/CT (Siemens Healthcare). On the latter, images were iteratively reconstructed using an ordinary Poisson ordered‐subset expectation maximization (OP‐OSEM) 3D‐iterative algorithm (four iterations, five subsets, time of flight, 440 × 440 matrix, 2-mm Gaussian filter). Body-weight corrected standard uptake values (SUVs) were calculated: SUV = concentration (kBq/ml)/[injected dose (kBq)/body weight (g)]. Using the CT, volumes of interest were defined on the colon, intra-luminal gas content excluded, and colon SUVs were extracted, as previously described (Fedorova et al., 2017). Colon SUVs were adjusted for volume to avoid underestimating the PET signal in Parkinson’s disease and iRBD patients (Fedorova et al., 2017). Colon volumes were corrected for sex, due to generally smaller volumes in females (Knudsen et al., 2017). Colonic transit time was assessed using the radiopaque marker (ROM) method. One capsule containing 10 ROM was ingested each morning (8 am) for 6 days, PET/CT performed on Day 7, and the number of ROM counted on CT. Colon transit time (days) was then calculated: [ROM(total number)+5]/10 (Knudsen et al., 2017).

All MRI and PET analyses were performed in PMOD 3.8. All image analyses were performed blinded to clinical patient category. There were a few missing imaging data-points in the iRBD and Parkinson’s disease groups due to technical failures of the scans. The exact number of patient scans are evident from the scatter plots in Fig. 1 and are listed in Supplementary Table 1.

Other assessments

Autonomic and non-motor symptoms were assessed with the Non-Motor Symptoms Scale (NMSS) (Chaudhuri et al., 2006), scales for outcomes in Parkinson’s disease-autonomic (SCOPA-AUT) (Rodriguez-Blazquez et al., 2010), constipation with ROME III diagnostic criteria (Rome, 2006), and motor symptoms with the MDS Unified Parkinson’s Disease Rating Scale part III (MDS-UPDRS III) (Goetz et al., 2008). Cognitive status was evaluated with the Montreal Cognitive Assessment (MoCA) battery, sleepiness with Epworth Sleepiness Scale (EPSS), and depression with Beck’s Depression Inventory-II (BDI-2). After 15 min of supine rest, blood pressure was measured for three consecutive minutes after standing up. Orthostatic hypotension was defined as systolic pressure drop of 20 mmHg or diastolic pressure drop of 10 mmHg. Olfaction was tested with the 16-item Sniffin’ Sticks identification battery (Hummel et al., 1997).

Statistical analyses

Statistical analyses were performed with Stata 13.1 and Graphpad PRISM 8.3. 3D plots were created with Plotly Chart Studio™. Normality of data was assessed with Shapiro-Wilk and Kolmogorov-Smirnov tests and Q-Q plots. Data are presented as mean [standard deviation (SD)] or median [interquartile range (IQR)]. Statistical outliers were identified with the robust regression and outlier removal test (ROUT) (Motulsky and Brown, 2006). Demographic and clinical data were interrogated using ANOVA, Kruskal-Wallis, and Fisher’s exact test as appropriate. Group comparisons of imaging data were performed with one-way ANOVA with Tukey’s multiple comparisons test; Brown-Forsythe ANOVA with Dunnett’s multiple comparisons test in case of unequal variance in the groups; and Kruskal-Wallis with Dunn’s correction for multiple comparisons for non-Gaussian distributed data. Effects of age and sex on the imaging parameters were tested with multiple linear regression models after inspection of diagnostic plots of residuals. All statistical tests were only performed on the three patient groups. Healthy control data from previous studies (Fedorova et al., 2017; Knudsen et al., 2017, 2018) are shown for illustrative purposes only and were not included in the statistical analyses. Their demographic data are listed in Supplementary Table 1. Associations between imaging data and clinical data were investigated with Pearson correlation or Spearman Rank correlation. Significance level was set at P < 0.05 corrected.

Data availability

The data that support the findings of this study are available from the corresponding author, upon reasonable request.

Results

Clinical and demographic data

Clinical and demographic data are summarized in Table 1. The final Parkinson’s disease dataset included 24 de novo patients with Parkinson’s disease without RBD (PD^RBD−) and 13 de novo Parkinson’s disease with premotor RBD (PD^RBD+). The median RBD symptom duration in the PD^RBD+ patients was 12.8 years (IQR 4.5–17.5), and median duration of parkinsonism was 1.4 years (IQR 1.0–2.0). Five de novo patients with Parkinson’s disease did not conform to the PD^RBD− and PD^RBD+ categories and these data were therefore not included in the overall analysis. Four of these patients were RBD-positive on the polysomnography but had no subjective experience of sleep disturbances by interview so could not be assigned to the PD^RBD+ category. Finally, one RBD-positive Parkinson’s disease patient stated that the RBD symptoms had begun ∼1 year after the onset of motor symptoms (i.e. possible top-down RBD).

Table 1

Demographic and clinical information

	PDRBD−	PDRBD+	iRBD	P	PDRBD− versus PDRBD+	PDRBD− versus iRBD	PDRBD+ versus iRBD
Sample size, n	24	13	22	–
Age	62.3 (7.8)	72.6 (5.3)	68.6 (8.6)	0.002	0.002	0.05	0.29
Sex, male/female	18/6	10/3	18/4	0.85	–	–	–
ON PD medication, yes/no	7/17	2/11	–	0.45	–	–	–
MDS-UPDRS III	21 (15–24)	19 (15–20)	1 (0–2)	0.36	–	–	–
Hoehn and Yahr (1/2/3)	7/15/2	4/8/1	–	0.92	–	–	–
Disease duration, months	2.5 (1–8.5)	7(2–7)	–	0.43	–	–	–
Symptom duration, months	24 (8.5–29.3)	17 (12–24)	–	0.83	–	–	–
RBD symptom duration, years	–	12.8 (9.8)	7.1 (5.2)	0.03	–	–	–
RBDSQ	3.5 (3–5)	8 (7–10)	10 (9–11)	<10−7	0.001	<10−7	0.65
NMSS	23 (14–34)	28 (17–50)	6.5 (5–10)	<10−6	0.99	<10−5	<10−4
NMSS GI section	0 (0–1)	3 (1–5)	0 (0–1)	0.002	0.007	0.99	0.002
SCOPA-AUT	11 (7–15)	13 (9–14)	12.5 (10–17)	0.68	0.84	0.78	0.84
ROME III - constipation	4.5 (2.8–10)	7 (5–8)	4 (1.5–7)	0.47	0.99	0.99	0.73
ROME III 9-15	3 (1.8–4.3)	4 (1–7)	3 (0–6.25)	0.83	0.99	0.99	0.99
ROME III nausea	0.5 (0–4)	0 (0–1)	0 (0–0)	0.02	0.85	0.01	0.54
Sniffin’ sticks	8.3 (3.2)	5.7 (2.6)	6.2 (2.6)	0.03	0.05	0.08	0.86
MoCA	27.5 (25–29)	29 (26–29)	27 (26–28)	0.96	0.99	0.99	0.99
EPSS	7 (3.8–9.3)	5 (4–7)	5.5 (3–7.8)	0.49	0.63	0.63	0.76
BDI-2	7 (4.8–9.5)	4 (3–9)	–	0.49	–	–	–
OH (yes/no)	4/20	6/7	7/15	0.15	–	–	–

Data from de novo PD without RBD (PD^RBD−) and with RBD (PD^RBD+), and isolated RBD (iRBD) presented as mean (SD) or median (25%–75% quartile). The three groups were compared with chi-squared, ANOVA, or Kruskal-Wallis tests with post-tests corrected for multiple comparisons. When data from only two groups were available, t-tests, Mann Whitney U or Fischer’s Exact test were used. P designates the overall group test P-value. Significant group post-test P-values are listed in the three rightmost columns. UPDRS-III scores were only compared between the Parkinson’s disease groups. OH = orthostatic hypertension.

No significant differences were seen between the PD^RBD+ and PD^RBD− on UPDRS-III scores, Hoehn and Yahr stage, symptom duration of parkinsonism, and BDI-2 (P > 0.05). No significant differences were seen between PD^RBD+, PD^RBD−, and iRBD patients on sex-distribution, SCOPA-AUT, ROME-III constipation module scores, MoCA, EPSS, and frequency of orthostatic hypotension (P > 0.05).

The PD^RBD+ patients had significantly longer duration of RBD symptoms compared to the iRBD group (P = 0.03), and were significantly older than the PD^RBD− (P = 0.002) and iRBD patients (P = 0.05). The PD^RBD+ patients also had lower olfaction scores than PD^RBD− (P = 0.05). The iRBD group had lower NMSS total scores and NMSS gastrointestinal (GI) scores compared to the two Parkinson’s disease groups (P < 10⁻⁴).

Imaging data

Figure 2 depicts the six different imaging parameters from the three patient groups and healthy control subjects. Table 2 lists the imaging data of the patient groups. The median duration from the first to the last imaging session was 3.5 months (IQR 0.3–5.2).

Figure 2

Plots of imaging measures. (A) FDOPA PET of the nigrostriatal dopaminergic innervation [minimum putamen specific binding ratios (SBR)]. Four additional patients with Parkinson’s disease (two PD^RBD+ and two PD^RBD−) had abnormal DaTscans (data not shown). (B) Neuromelanin (NM)-sensitive MRI of the locus coeruleus (LC). One PD^RBD+ was a significant outlier (C) ¹²³I-MIBG late H/M ratio of the cardiac sympathetic innervation. Two iRBD MIBG values were significant outliers (P < 0.05, ROUT test). (D) ¹¹C-donepezil PET of the colonic cholinergic (parasympathetic) innervation (SUVs). (E) Colonic transit time measured by radiopaque markers (# of markers retained at Day 7). The PD^RBD+ patients had significantly delayed transit compared to the two other groups. (F) Sex-corrected colon volume measured on abdominal CT. The PD^RBD+ patients had significantly enlarged colon volumes compared to the two other groups. Healthy control data (HC) acquired with identical methodology are shown for visual comparison.

Table 2

Imaging data

	PDRBD−	PDRBD+	iRBD	P	PDRBD− versus PDRBD+	PDRBD− versus iRBD	PDRBD+ versus iRBD
FDOPA min (SBR)	0.71 (0.64–0.86)	0.89 (0.72–1.13)	1.54 (1.39–1.80)	<10−13	0.44	<10−15	<10−8
MIBG early (H/M)	1.42 (1.28–1.58)	1.23 (1.18–1.27)	1.33 (1.26–1.38)	<10−3	<10−3	0.006	0.38
MIBG late (H/M)	1.39 (1.21–1.61)	1.12 (1.08–1.20)	1.19 (1.11–1.28)	<10−5	<10−4	<10−3	0.58
MIBG wash-out rate	−0.04 (−0.08 to −0.01)	−0.08 (0.12 to −0.06)	−0.09 (−0.13 to −0.06)	<10−3	0.02	0.001	0.85
Donepezil colon (SUV)	1.48 (1.31–1.60)	1.12 (1.02–1.26)	1.32 (1.16–1.43)	0.008	0.005	0.33	0.14
MRI LC/pons ratio	1.278 (0.055)	1.239 (0.065)	–	–	0.07	–	–
Colon volume, cm3	918 (756–1072)	1323 (1230–1385)	1007 (895–1167)	<10−3	<10−3	0.42	0.006
Radiopaque marker, n	22 (13)	37 (10)	24 (16)	0.01	0.01	0.99	0.03
Colon transit, days	2.7 (1.3)	4.1 (1.2)	2.9 (1.6)	0.02	0.02	0.99	0.05

Data are presented as mean (SD) or median (25–75% quartile). The three groups were compared with ANOVA or Kruskal-Wallis tests with post-tests corrected for multiple comparisons. MRI locus coeruleus (LC)/pons data were only available for the two Parkinson’s disease groups (t-test). P designates the overall group test P-value. Significant group post-test P-values are listed in the three rightmost columns.

FDOPA PET

A highly significant overall between-group difference was seen in FDOPA putamen specific binding ratios values (P < 10⁻¹³, ANOVA; Fig. 2A). No significant difference was seen between the PD^RBD+ and PD^RBD− groups, but both Parkinson’s disease groups had significantly lower specific binding ratios compared to the iRBD group (P < 10⁻⁸).

Neuromelanin MRI

The mean locus coeruleus/pons ratio was decreased in the PD^RBD+ compared to the PD^RBD− group, but the finding was not significant (P = 0.07, t-test, Fig. 2B). No comparable neuromelanin MRI data were available for the iRBD group due to a change of MRI scanner between this and previous studies (Knudsen et al., 2018).

¹²³I-MIBG scintigraphy

Highly significant between-group differences were seen in late image MIBG H/M ratios (P < 10⁻⁵, ANOVA; Fig. 2C), in the early image H/M ratios (P = 0.0002, ANOVA, Table 2), and washout rates (P = 0.0009, ANOVA). No significant difference was seen between the PD^RBD+ and iRBD groups, but the PD^RBD− group had significantly more normal late H/M ratios compared to the PD^RBD+ (P < 10⁻⁴) and the iRBD groups (P < 10⁻³).

¹¹C-donepezil PET

A significant overall between-group difference was seen in the colonic ¹¹C-donepezil SUV (P = 0.008, ANOVA; Fig. 2D). The PD^RBD+ group had significantly lower uptake compared to the PD^RBD− group (P = 0.005), but not compared to the iRBD group. Of note, the patients with Parkinson’s disease were imaged on two different Siemens Biograph PET cameras. However, the distribution of PD^RBD+ and PD^RBD− patients scanned on the two cameras were similar (Fisher’s exact test, P = 0.73). Adjusting for scanner type did not change the difference of colon ¹¹C-donepezil SUV between the two groups (adjusted difference = 0.33, P = 0.0001).

Colonic transit and volume

Significant overall differences were seen in colonic transit times (P = 0.01, Kruskal-Wallis test, Fig. 2E) and colon volumes between the groups (P = 0.0002, ANOVA; Fig. 2F). The PD^RBD+ had significantly more radiopaque markers compared with the PD^RBD− (P = 0.01) and the iRBD patients (P = 0.03). The PD^RBD+ group also had significantly enlarged colon volumes compared to the PD^RBD− (P = 0.0002) and the iRBD patients (P = 0.006).

Correlations and corrections

In the combined iRBD and PD^RBD+ group, no correlations were seen between the RBD symptom duration or RBD Screening Questionnaire (RBDSQ) scores and other parameters, including MIBG, donepezil, colon volume, or transit data (P > 0.05). As previously described, a significant correlation was seen between colonic volume and transit (P < 0.0001, r = 0.72). No associations were seen between colon volume or transit and ROME, NMSS, or SCOPA questionnaire constipation items (P > 0.05). No correlation was found between putaminal FDOPA uptake and UPDRS-III in any of the three groups alone or in the Parkinson’s disease group as a whole (P > 0.22). Only when all three groups were combined (PD^RBD−, PD^RBD+, and iRBD) was a significant correlation found (Spearman's r = –0.73, P < 0.001). The between-group differences in imaging measures (FDOPA, locus coeruleus/pons neuromelanine, MIBG, donepezil) were unchanged after correction for age- and sex-effects by multiple linear regression analyses.

Brain-first versus body-first Parkinson’s disease

Figure 3A depicts the FDOPA and MIBG data plotted for the iRBD, PD^RBD+, and PD^RBD− groups. The iRBD and PD^RBD+ patient data are compatible with a continuous body-first trajectory, characterized by initial loss of cardiac MIBG signal followed by loss of putaminal FDOPA signal. In contrast, the PD^RBD− data seem to follow a brain-first trajectory, characterized by primary loss of putaminal FDOPA signal followed by a secondary loss of cardiac MIBG signal. Two iRBD cases had normal MIBG scans (statistically significant outliers, P < 0.05, ROUT test), which could be compatible with prodromal multiple system atrophy. Similar brain-first and body-first trajectories were suggested when plotting FDOPA versus ¹¹C-donepezil PET data, despite the somewhat higher variance of ¹¹C-donepezil PET data within each group (Fig. 3B).

Figure 3

XY-plots illustrating brain-first versus body-first Parkinson’s disease. (A) The figure depicts FDOPA specific binding ratio (SBR) values on the x-axis and late image ¹²³I-MIBG H/M ratios on the y-axis. Healthy aged control data (HC) are shown in the top right corner. Prodromal patients with Parkinson’s disease with iRBD initially lose the cardiac MIBG signal, while their dopamine system is fairly intact. As the dopamine system is progressively damaged, iRBD patients convert to de novo Parkinson’s disease with RBD (PD^RBD+). In contrast, de novo Parkinson’s disease without RBD (PD^RBD−) initially show damage to the nigrostriatal dopamine system and only later damage to the cardiac sympathetic system. The two iRBD outliers may represent multiple system atrophy cases. (B) FDOPA specific binding ratios values on the x-axis and colon ¹¹C-donepezil SUVs on the y-axis. Despite the higher variance in ¹¹C-donepezil data, body-first and brain-first trajectories are clearly suggested. Two extreme PD^RBD− donepezil values and one extreme iRBD value are excluded from this plot (Fig. 2D).

Figure 4 depicts a combined 3D plot of FDOPA-MIBG-donepezil data from those patients, who had all three imaging markers. In-house healthy control data from the three imaging modalities are included as a visual reference of the normal ranges. The plot supports that the RBD-negative and RBD-positive patients follow two very different trajectories through this parameter space. Additional 3D projections and a video are shown in the Supplementary material.

Figure 4

3D plot illustrating brain-first versus body-first Parkinson’s disease. FDOPA PET specific binding ratios, ¹²³I-MIBG H/M late image ratios, and ¹¹C-donepezil colon SUVs in de novo patients with Parkinson’s disease with RBD (PD^RBD+), de novo patients with Parkinson’s disease without RBD (PD^RBD−), and isolated RBD patients. Healthy control (HC) in the top right corner are plotted through the healthy control mean of the three imaging modalities. Two PD^RBD− and one iRBD with very high donepezil values are omitted from this plot (shown in Fig. 2D). The hypothesized brain- and body-first trajectories are illustrated by transparent red and blue arrows, respectively.

Discussion

Our PD^RBD− and PD^RBD+ patients were very similar on a range of standard clinical tests. They showed a comparable burden of motor and most non-motor symptoms, and a similar degree of nigrostriatal dopaminergic deficit on FDOPA PET. Nevertheless, this study revealed that these two groups of patients display strikingly different profiles on our multimodal imaging battery. These profiles support the existence of a brain-first and body-first subtype of Parkinson’s disease, and furthermore, that premotor RBD is a highly predictive marker of the body-first subtype.

Neuropathological autopsy studies of patients with Parkinson’s disease, dementia with Lewy bodies (DLB), or incidental Lewy body disease have shown discrepant results. Braak’s staging system was derived from a cohort of patients, who were selected based on the presence of Lewy pathology in the dorsal motor nucleus of the vagus, and all these patients conformed to a brainstem predominant type (Braak et al., 2003a). Other studies reported that some post-mortem cases do not harbour Lewy pathology in the dorsal motor nucleus of the vagus, and a minority of patients do not follow the Braak staging scheme (Parkkinen et al., 2008; Jellinger, 2019).

Alternative staging systems of Lewy pathology have been proposed (Kosaka et al., 1984; McKeith et al., 2005; Beach et al., 2009). These systems differentiate between the clear brainstem predominant cases described by Braak, and other cases where the limbic system is preferentially involved. A recent population-based study investigated 124 non-selected Lewy-body positive cases aged 85 years or older and demonstrated two common theoretical progression patterns (Raunio et al., 2019). Approximately two-thirds of these cases displayed a brainstem predominant pattern with most severe involvement of the lower brainstem and sympathetic nuclei in the spinal cord. The remaining cases showed an amygdala-predominant profile with significantly less Lewy pathology in the lower brainstem.

We hypothesize that our body-first patients reflect the brainstem predominant type documented in neuropathology studies. In such patients, the α-synuclein pathology could have originated in the enteric and autonomic nervous system, and propagated to form progressive Lewy pathology predominantly in the lower brainstem and sympathetic structures. Ascending pathology will then, in most cases, lead to RBD prior to parkinsonism. In contrast, our brain-first patients, many of whom had normal MIBG and donepezil PET scans, represent cases in whom the α-synuclein pathology arose in the CNS itself. It seems probable that the first α-synuclein pathology arises in the amygdala or sometimes enters via the olfactory route, since post-mortem studies show that the amygdala and olfactory bulb are the two most common sites of ‘single-location’ α-synuclein pathology in incidental Lewy body disease (Parkkinen et al., 2008; Beach et al., 2009). Brain-first patients with Parkinson’s disease will therefore in general develop motor symptoms before RBD.

We want to stress that our body- and brain-first subtypes should be understood as two main phenotypes, which may capture most of the phenotypic variation among patients with Parkinson’s disease. However, variations in the site of initial Lewy pathology may produce aberrant cases, which do not follow the common sequence of symptom development. For instance, in a few cases of incidental LBD, isolated α-synuclein pathology has been reported in the substantia nigra or the locus coeruleus (Adler and Beach, 2016; Raunio et al., 2019). In such patients, parkinsonism, RBD, and dysautonomia could arise more or less simultaneously. Multi-focal Parkinson’s disease, where α-synuclein pathology arises synchronously in two or more anatomical regions, is also a theoretical possibility. Such cases would also not fit the body- and brain-first patterns described here. We do, however, believe that such cases must be rare, since the Lewy pathology seen in incidental LBD is most commonly found in neighbouring, interconnected regions—suggesting an onset in a single location followed by spreading to neighbouring regions.

It is known that iRBD patients convert to approximately equal numbers of Parkinson’s disease and DLB cases (Postuma et al., 2019). If our hypothesis is correct, i.e. that iRBD is a marker of a body-first subtype, it means that body-first aetiology can nevertheless progress to a ‘dementia-first’ clinical phenotype. However, this is not necessarily a contradiction. It is known that parkinsonism first ensues once ∼50% of putaminal dopaminergic terminals are lost, which represents several years of clinically silent nigrostriatal degeneration (Cheng et al., 2010). Also, Alzheimer-type pathology is very frequent in the elderly population, and is a common finding in DLB patients. Thus, in patients with concomitant Alzheimer pathology, dementia may develop quickly (ahead of parkinsonism) even when the severity of cortical Lewy pathology is relatively low (Chetelat et al., 2013). Also, cognitive decline may, at least in part, be caused by damage to several subcortical modulatory systems, including the locus coeruleus, raphe nuclei, and nucleus basalis, all of which are heavily involved in brainstem predominant Lewy body positive cases at post-mortem.

Interestingly, our PD^RBD− patients were less hyposmic than the PD^RBD+ patients, suggesting that in at least some cases, the pathology arose within the CNS in the PD^RBD− group. In addition, our finding that more PD^RBD+ cases were hyposmic actually fit well with the dual-hit hypothesis. Braak proposed that α-synuclein pathology can develop in the olfactory bulb and spread from there to closely related areas (tubercle, piriform cortex, entorhinal cortex), but the lesions do not then seem to advance further in a meaningful manner. He therefore concluded that, although Parkinson’s disease may have a dual-hit aetiology (gut and olfactory bulb), it seems to be the ascending propagation of pathology through the brainstem that drives the disease (Braak et al., 2003a). Furthermore, a recently developed prodromal Parkinson’s disease mouse model exhibited RBD-like behaviour after 5 months, hyposmia after 9 months and had not developed parkinsonism when they were sacrificed at 18 months (Taguchi et al., 2020). Within this framework, it is therefore not surprising that our PD^RBD+ and iRBD patients were more hyposmic than the PD^RBD− patients.

The present study was cross-sectional and comprised three distinct groups of patients. The brain- and body-first trajectories depicted in Figs 3 and 4 are therefore based on inference, as we do not yet possess longitudinal data. However, we would argue that the longitudinal trajectories are the most salient interpretation of the data. First, it can be assumed that all of our patients originally had normal FDOPA, MIBG, and donepezil imaging parameters and therefore started out in the healthy control region of Figs 3 and 4. Second, it is well documented that nearly all later stage patients with Parkinson’s disease (Hoehn and Yahr stage 3–5) display profoundly reduced putamen signal on dopamine imaging scans and severely reduced MIBG uptake on scintigraphies (Hamada et al., 2003; Kashihara et al., 2010). One follow-up MIBG study found that patients with Parkinson’s disease with low MIBG H/M values at baseline remained low at follow-up, but that >80% of patients with normal or near-normal baseline MIBG showed progressively declining values when re-scanned after 2–4 years (Tsujikawa et al., 2015). Thus, all patients with Parkinson’s disease will eventually develop severe cardiac sympathetic denervation and so end up in the lower left corner of Fig. 3A. The two subtypes will therefore be indistinguishable on our imaging battery at late disease stages, but when studied early, we have demonstrated that they comprise two different clustered profiles with little overlap between them. Put together, patients with Parkinson’s disease show two distinct types of sequential neuronal degeneration—either the dopamine system degenerates before the peripheral autonomic systems, or vice versa. This interpretation also agrees with the known distribution of Lewy pathology discussed above.

Moreover, our data suggest a significant lag time between degeneration of one system and the other. This observation has important consequences for imaging-based strategies to monitor the effect of neuroprotective treatment. MIBG scintigraphies may be the most sensitive measure for monitoring de novo PD^RBD− patients, whereas iRBD patients should be monitored with dopamine imaging. However, if prodromal patients with Parkinson’s disease could be identified and enrolled in trials before any neuronal damage to any system has occurred, the opposite strategy would be needed. Here, PD^RBD− should be monitored with dopamine imaging, since that is the first system to undergo degeneration, but PD^RBD+ patients should be monitored with MIBG. Ideally, it might be advantageous to employ both MIBG and dopamine imaging markers in future clinical trials.

In our study, five RBD-positive patients with Parkinson’s disease were excluded, since they could not be assigned to the PD^RBD+ category, defined by the presence of clear RBD symptoms before appearance of parkinsonism. Some or all of these patients could have developed top-down RBD as part of a brain-first Parkinson’s disease. Data from the five patients are provided in Supplementary Fig. 2.

One of these RBD-positive cases reported that the RBD symptoms had emerged ∼1 year after the onset of motor symptoms, which is clearly suspect of top-down RBD. The imaging profile of this Parkinson’s disease case resembled that of the PD^RBD− patients with normal MIBG scintigraphy, normal colon volume and transit, and a completely normal blood pressure response on the orthostatic hypotension test. Of note, a published MIBG follow-up study reported that baseline MIBG scans of three patients with Parkinson’s disease, who only later developed RBD, were closer to normal than those of most iRBD patients and patients with Parkinson’s disease with pre-motor RBD of that same study (Miyamoto et al., 2011).

Of the other four excluded RBD-positive patients in our study, two had an imaging profile resembling body-first patients, and the other two resembled brain-first patients.

It is possible that the severity of dream enactment and degree of vigorous movements during sleep may vary widely among RBD-positive cases. Thus, interviewing patients and spouses about RBD symptoms is an imperfect method to reliably assess when the RBD symptoms appeared. However, more than half of all patients with Parkinson’s disease will eventually develop RBD, but only ∼30% do so during the prodromal phase of the disease (Zhang et al., 2017; Dauvilliers et al., 2018). This suggests that a fraction of our PD^RBD− patients will develop RBD later, and since their current baseline data are compatible with the brain-first subtype, such future RBD-converters will also be compatible with a top-down RBD category.

The PD^RBD+ group had lower locus coeruleus/pons ratios compared to the PD^RBD− group, in accordance with previous studies (Ehrminger et al., 2016; Sommerauer et al., 2018), but the difference was not statistically significant (P = 0.07). MRI-based neuromelanin measures of the locus coeruleus are noisy and prone to head movement artefacts, so our study may have been underpowered with respect to this analysis.

Interestingly, nearly all our PD^RBD+ patients had pathologically enlarged colon volumes and transit times. The iRBD group was less affected on these parameters, despite the fact that the iRBD and PD^RBD+ groups showed a similar degree of autonomic neuronal dysfunction assessed by MIBG and donepezil PET. Possibly, the aggravated gastrointestinal dysfunction in the PD^RBD+ group was caused by a synergistic effect of autonomic and nigrostriatal denervation, since animal studies have shown that nigrostriatal lesions in isolation can elicit some degree of constipation and gastroparesis (Zhu et al., 2012).

This study has several limitations. The cross-sectional design has already been discussed above. Our sample size was modest, but still sufficient to demonstrate highly significant differences in the most important imaging parameters. The average age in the PD^RBD+ group was higher than in the two other groups. However, MIBG H/M ratios are not known to show an age-related decline (Hamada et al., 2003). FDOPA uptake does show an age-related decline (Kumakura et al., 2010), but since the average FDOPA specific binding ratios were higher in the older PD^RBD+ group, the overall interpretation of our FDOPA and MIBG results was probably not affected by the differing age among the groups. This was confirmed by multiple linear regression analyses with age correction. Two different PET cameras were used for ¹¹C-donepezil PET/CT. However, any unknown bias from different resolution and reconstruction parameters would have been evenly distributed between the two Parkinson’s disease groups, since 13 PD^RBD− and six PD^RBD+ were scanned on the Biograph, and 11 PD^RBD− and seven PD^RBD+ on the Vision camera (Fisher’s exact test, P = 0.73). Of note, correction for scanner type in the analyses did not change the difference in colon donepezil SUV between the two groups. Moreover, the use of two different scanners would tend to increase data variance in the groups, and the significance level of our observed between-group difference would therefore, if anything, have been underestimated. It should also be noted that, although we applied multiple comparison correction to analyses within each imaging modality, the large number of comparisons performed in this study still produces some risk for type 1 errors. Finally, despite our use of the MDS diagnostic criteria, our dataset may have included misdiagnosed cases with atypical parkinsonism, such as multiple system atrophy or progressive supranuclear palsy. However, given the relative rarity of these disorders, we would expect to have included at most one or two such cases, which does not change the overall interpretation of the data.

In conclusion, we have shown that prodromal and de novo patients with Parkinson’s disease can be categorized by means of multimodal imaging into distinct clusters, which are compatible with a brain-first and body-first Parkinson’s disease subtype. The presence of RBD in the premotor phase is a marker of the body-first type, probably a reflection of ascending α-synuclein pathology reaching the pons before the substantia nigra. It is crucial that these findings be replicated by independent researchers. It will also be important to include other types of prodromal patients with Parkinson’s disease, including mutation carriers, and RBD-negative prodromal cases to study how closely such patient groups align with the brain- and body-first trajectories.

Finally, further research is now needed to uncover the aetiological factors underlying the subtypes. The gut-brain axis and microbiota are currently undergoing intensive study. However, it seems probable that such factors may primarily be causative for the body-first type, but are perhaps irrelevant for the brain-first subtype. Such discoveries will be pivotal for the development of subtype-specific therapies to treat and eventually prevent Parkinson’s disease.

Acknowledgements

We thank all study participants.

Funding

The study was funded by Lundbeck Foundation (R190-2014-4183). The funding source had no role in the study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all data in the study and had final responsibility for the decision to submit for publication.

Competing interests

The authors report no competing interests.

 
• iRBD = 

  isolated REM sleep behaviour disorder

 
• MIBG =

  ¹²³I-metaiodobenzylguanidine

 
• FDOPA =

  ¹⁸F-dihydroxyphenylalanine

 
• H/M ratio =

  mean heart/mediastinum uptake ratio

 
• SUV =

  standard uptake value