Associations between neuromelanin depletion and cortical rhythmic activity in Parkinson’s disease

Abstract

Parkinson’s disease (PD) is marked by the death of neuromelanin-rich dopaminergic and noradrenergic cells in the substantia nigra (SN) and the locus coeruleus (LC), respectively, resulting in motor and cognitive impairments. Although SN dopamine dysfunction has clear neurophysiological effects, the association of reduced LC norepinephrine signalling with brain activity in PD remains to be established.

We used neuromelanin-sensitive T₁-weighted MRI (PD, n = 58; healthy control, n = 27) and task-free magnetoencephalography (PD, n = 58; healthy control, n = 65) to identify neuropathophysiological factors related to the degeneration of the LC and SN in patients with PD. We found pathological increases in rhythmic alpha-band (8–12 Hz) activity in patients with decreased LC neuromelanin, which were more strongly associated in patients with worse attentional impairments. This negative alpha-band–LC neuromelanin relationship is strongest in fronto-motor cortices, where alpha-band activity is inversely related to attention scores. Using neurochemical co-localization analyses with normative atlases of neurotransmitter transporters, we also show that this effect is more pronounced in regions with high densities of norepinephrine transporters. These observations support a noradrenergic association between LC integrity and alpha-band activity. Our data also show that rhythmic beta-band (15–29 Hz) activity in the left somatomotor cortex decreases with lower levels of SN neuromelanin; the same regions where beta activity reflects axial motor symptoms.

Together, our findings clarify the association of well-documented alterations of rhythmic neurophysiology in PD with cortical and subcortical neurochemical systems. Specifically, attention-related alpha-band activity is related to dysfunction of the noradrenergic system, and beta activity with relevance to motor impairments reflects dopaminergic dysfunction.

Using multimodal structural and functional neuroimaging approaches, Wiesman et al. show that in Parkinson's disease, alterations in alpha- and beta-band brain rhythms are related to degeneration of the noradrenergic and dopaminergic systems, respectively.

Introduction

One of the defining histopathological features of Parkinson’s disease (PD) is the loss of neuromelanin-rich cells in brainstem nuclei. Degeneration of the dopaminergic neurons in the substantia nigra (SN) leads to dysfunctional signalling along the subcortical nigrostriatal pathway, and contributes to motor impairments.¹ Loss of neuromelanin-rich neurons in the locus coeruleus (LC) can precede SN deterioration² and disrupts the production of norepinephrine (i.e. noradrenaline) in the CNS. The resulting changes in noradrenergic signalling projecting to the cortex are thought to induce cognitive symptoms,³ primarily impairment of attention functions.^4,5

A knowledge gap remains between these neurochemical observations and the well-documented alterations of rhythmic and arrhythmic neurophysiological activity in PD,^6-8 and how such associations relate to the cognitive and motor features of PD. Quantification of SN and LC integrity in vivo in PD has been difficult historically, but novel neuromelanin-sensitive MRI measurements hold promise for the non-invasive monitoring of the specific degeneration of dopaminergic and noradrenergic systems in these patients^9,10 in relation to the respective hallmark motor and cognitive features of the disease.^3,10 Typically based on a fast spin echo (FSE) sequence, this approach is preferentially sensitive to neuromelanin-rich structures via T₁-shortening associated with the melanin–iron complex,¹¹ in addition to lower macromolecular fraction in the LC.¹² Macro-scale neurophysiological activity can be measured non-invasively using magnetoencephalography (MEG) and EEG, with MEG offering superior spatial resolution.¹³ From the neurophysiological frequency spectra obtained with MEG, the rhythmic and arrhythmic components of cortical activity can be parameterized,^14,15 which enhances interpretational clarity.

Cortical alpha rhythms (8–12 Hz) are altered in PD and, similar to the deterioration of the LC, are related to mild cognitive decline and PD dementia.^16-20 However, whether PD alterations of alpha activity are connected to parallel LC degeneration remains an open question. Extensive research suggests that activation of the LC-norepinephrine system is associated with reduced cortical alpha synchronization in healthy humans and animal models, with relevance for attention.²¹ Animal models show that LC firing is related to peri-alpha cortical rhythmic activity.²² In healthy humans, the amplitude of cortical alpha activity is sensitive to noradrenergic modulation,²¹ and pharmacological increases in monoaminergic signalling, including norepinephrine, decrease alpha activity in fronto-central cortices.²³

The mechanism behind this inverse LC-alpha relationship is relatively well studied: tonic noradrenergic input from the LC depolarizes the membrane potential of thalamocortical neurons via α₁-adrenoceptors and suppresses their periodic activity,²¹ which is essential for the generation of the cortical alpha rhythms observed by M/EEG.²⁴ We therefore hypothesized that LC degeneration in PD is co-expressed in patients with region-specific increases in alpha activity and pronounced attention impairments.

Rhythmic beta activity (15–29 Hz) is also perturbed in PD and is linked to motor deficits,^1,25,26 but can be normalized by dopaminergic medications^19,27,28 and by deep brain stimulation of the subthalamic nucleus.^28-30 In the primary motor cortices of early-stage patients with PD, beta activity is increased compared to healthy levels,³¹ but this effect reverses as patients reach the moderate stages of the disease.^27,32 This signifies that in PD, altered dopamine-mediated signalling in the cortico-basal ganglia loop affects frequency-specific rhythmic cortical activity. Therefore, we asked whether SN degeneration is related to altered expressions of cortical beta-band activity in patients with PD. We hypothesized that loss of neuromelanin in the SN would be related to reduced beta activity and worse motor functions.

Alterations in theta band (5–7 Hz) neurophysiological activity have also been shown in PD,³³ although less consistently and rarely in relationship to clinical features. More recently, broadband arrhythmic neurophysiological activity has also been shown to shift towards slower frequencies in PD,^8,34-36 signalling greater relative inhibition³⁷ and potentially confounding previously reported effects reported in low-frequency bands.¹⁴ We therefore tested for possible associations between these neurophysiological features, LC and SN neuromelanin, and clinical symptoms.

Ultimately, we used neuromelanin-sensitive MRI and MEG to test how associations between neuromelanin depletion, neurophysiological activity and clinical symptoms map onto the human cortex. We also relate these effects to the topography of relevant neurochemical systems using normative atlases of neurotransmitter system densities^38,39 and spatial autocorrelation-preserving permutation tests.^8,40

Materials and methods

Participants

The Research Ethics Board at the Montreal Neurological Institute reviewed and approved this study. Written informed consent was obtained from every participant following a detailed description of the study, and all research protocols complied with the Declaration of Helsinki. Exclusionary criteria for all participants included current neurological (other than PD) or major psychiatric disorder; MEG/MRI contra-indications; and unusable neuroimaging or demographic data. Data from patients with mild to moderate (Hoehn and Yahr scale: 1–3) idiopathic PD, as diagnosed by a treating neurologist, were included from the Quebec Parkinson Network database (QPN; https://rpq-qpn.ca/).⁴¹ All patients with PD were prescribed a stable dosage of antiparkinsonian medication, with satisfactory clinical response prior to study enrolment. Patients were instructed to take their medication as prescribed before research visits, and thus all data were collected in the practically defined ON state.

MEG¹³ data were included for 58 patients with PD who fulfilled the inclusion criteria. Detailed information regarding levodopa medication regimens was available for a subset (n = 23) of these participants and used to calculate the levodopa equivalent daily dose.⁴² Of the 46 patients who also provided information regarding their use of other medications for symptoms of PD, 13 reported taking dopamine agonists. These MEG data were compared with similar measurements from a sample of 65 healthy older adults (HC), collated from the Quebec Parkinson Network (n = 10),⁴¹ PREVENT-AD (n = 40)⁴³ and Open MEG Archive (OMEGA; n = 15)⁴⁴ data repositories. These control participants were selected such that their demographic characteristics, including age (Mann–Whitney U-test; W = 1774.00, P = 0.575), self-reported sex (χ² test; χ² = 0.01, P = 0.924), handedness (χ² test; χ² = 0.22, P = 0.894) and highest level of education (Mann–Whitney U-test; W = 1929.00, P = 0.217) did not differ statistically from those of the patient group. All participants underwent the same MEG data collection procedure using the same instrument.

Neuromelanin-sensitive MRI data were available for the same 58 patients with PD who fulfilled the inclusionary criteria and had usable MEG data. Neuromelanin-sensitive MRI data were not available for the healthy controls from the PREVENT-AD and OMEGA studies; therefore, neuromelanin-sensitive MRI data from a matched group of 27 healthy older adults from the Quebec Parkinson Network (QPN) were used. These participants underwent the same data collection protocol using the same MRI instrument as the patient group and were selected such that their demographic characteristics, including age (Mann–Whitney U-test; W = 892.50, P = 0.303), handedness (χ² test; χ² = 0.852, P = 0.653) and highest level of education (Mann–Whitney U-test; W = 726.50, P = 0.323) did not differ statistically from those of the patient group. The control sample of neuromelanin-sensitive MRI data from the QPN database could not be matched to the patient group on self-reported sex (χ² test; χ² = 5.34, P = 0.021), hence all group-wise neuromelanin-MRI comparisons included sex (in addition to age) as a nuisance covariate.

Group demographic summary statistics and comparisons, in addition to clinical summary statistics for the patient group, are provided in Supplementary Tables 1 and 2.

Neuromelanin-sensitive MRI acquisition and processing

Neuromelanin-sensitive MRI data were collected using a 3 T Siemens Prisma scanner with a 32-channel head coil at the Montreal Neurological Institute. Neuromelanin-sensitive sequences were collected using the following parameters: two-dimensional T₁ FSE; echo trains per slice, 60; repetition time (RT), 600 ms; echo time (ET), 10 ms; flip angle, 120°; field of view: 220 mm; slice thickness, 1.8 mm; and resolution, 0.7 mm × 0.7 mm × 1.8 mm. T₁-weighted sequences were collected in the same session using the following parameters: three-dimensional T₁ magnetization prepared-rapid gradient echo (MPRAGE); TR, 2300 ms; ET, 2.98 ms; flip angle, 9°; field of view, 256 mm; slice thickness, 1 mm; resolution, 1 mm isotropic. T₁-weighted images underwent denoising, non-uniformity correction, intensity normalization and registration to MNI stereotaxic space using the NIST Longitudinal Pipeline⁴⁵ and PD126 template⁴⁶ as the registration target.

The neuromelanin-sensitive acquisitions were registered into MNI stereotaxic space using the corresponding linear and non-linear transformations from T₁-weighted images processed using the NIST pipeline.⁴⁵ Regions of interest, including the SN, LC, cerebral peduncles (CP), and pontine tegmentum (PT), were segmented from the processed neuromelanin-sensitive images. SN, CP and PT regions of interest were defined manually by an expert on a separate in-house neuromelanin template and were warped to fit the PD126 template. The LC region of interest was obtained using a conservative 40% threshold on the Brainstem Navigator probabilistic atlas⁴⁷ aligned on the PD126 template. Segmented regions of interest were used to derive the neuromelanin score, which represents the integrity of brainstem nuclei (SN and LC) as the ratio of the normalized average SN neuromelanin intensity to the averaged neuromelanin intensity from the CP, and the ratio of the normalized average LC neuromelanin intensity to the neuromelanin intensity from the PT, respectively.

Clinical and neuropsychological testing

Standard clinical assessments were available for most of the patients with PD, including gross motor impairment [Unified Parkinson’s Disease Rating Scale—part III (UPDRS-III); n = 44]⁴⁸ and general cognitive function [Montreal Cognitive Assessment (MoCA); n = 51].⁴⁹ In those patients with UPDRS sub-score data available (n = 42), we summed these scores using established criteria⁵⁰ to represent two sets of motor features. Bradykinesia and rigidity, but not tremor, are the motor impairments most commonly linked to cortical beta oscillations in PD.²⁶ We summed the rigidity, finger tapping, pronation/supination of hands and leg agility sub-scores to represent these features. Axial motor symptoms were also considered in isolation, because they are resistant to administration of levodopa.⁵¹ The following sub-scores of the UPDRS-III were summed to represent axial symptoms: speech, facial expression, arising from chair, posture, gait, postural stability and body bradykinesia. We summed the resting tremor sub-scores to consider tremor symptoms as a potential confound.

Detailed neuropsychological data were available for 49 patients with PD and, as described and validated previously,^8,52 were used to derive composite scores across five domains: attention (Digit Span: Forward, Backward and Sequencing; Trail Making Test Part A), executive function (Trail Making Test Part B; Stroop Test: Colors, Words and Interference; Brixton Spatial Anticipation Test), memory [Hopkins Verbal Learning Test-Revised (HVLT-R) Learning Trials 1–3, Immediate and Delayed Recall; Rey Complex Figure Test (RCFT): Immediate and Delayed Recall], language (Semantic Verbal Fluency: Animals & Actions; Phonemic Verbal Fluency: F, A & S; Boston Naming Test) and visuospatial function (Clock Drawing Test: Verbal Command & Copy Command; RCFT: Copy). To use as much available data as possible, missing values were excluded pairwise from analysis per each test. Negatively scored test values were sign inverted; the data for each individual test were standardized to the mean and standard deviation (SD) of the available sample and these z-scores were then averaged within each domain listed earlier to derive domain-specific metrics of cognitive function. We focused our analyses on the attention domain a priori, owing to its established associations with alpha rhythms^53-55 and noradrenergic function.^56-58 Data regarding disease duration (i.e. years since diagnosis) were also available for n = 46 patients (mean = 6.17 years, SD = 4.94). Additional clinical data are provided in Supplementary Table 2.

Magnetoencephalography data collection and analyses

Eyes-open resting-state MEG data were collected using a 275-channel whole-head CTF system (Port Coquitlam, BC, Canada) at a sampling rate of 2400 Hz and with an antialiasing filter with a 600 Hz cut-off. Noise cancellation was applied using CTF’s software-based built-in third-order spatial gradient noise filters. Recordings lasted a minimum of 5 min⁵⁹ and were conducted with participants in the upright position as they fixated a centrally presented crosshair. The participants were monitored during data acquisition via real-time audio-video feeds from inside the MEG shielded room, and continuous head position was recorded during all sessions.

MEG preprocessing was performed with Brainstorm⁶⁰ unless otherwise specified, with default parameters and following good-practice guidelines.⁶¹ The data were bandpass filtered between 1 and 200 Hz to reduce slow-wave drift and high-frequency noise, and notch filters were applied at the line-in frequency and harmonics (i.e. 60, 120 and 180 Hz). Signal space projectors were derived around cardiac and eye-blink events detected from ECG and electrooculogram channels using the automated procedure available in Brainstorm,⁶² reviewed and corrected manually where necessary, and applied to the data. Additional signal space projectors were also used to attenuate stereotyped artefacts on an individual basis. Artefact-reduced MEG data were then epoched into non-overlapping 6 s blocks and downsampled to 600 Hz. Data segments containing major artefacts (e.g. superconducting quantum interference device jumps) were excluded from each session based on the union of two standardized thresholds of ±3 median absolute deviations from the median: one for signal amplitude and one for the numerical gradient. An average of 78.72 (SD = 14.55; 83.50%) epochs were used for further analysis [patients, 83.14 (SD = 7.65); controls, 74.77 (SD = 17.82)], and the percentage of epochs rejected did not differ between the groups (Mann–Whitney U-test; W = 1670.00, P = 0.276). Empty-room recordings lasting ≥2 min were collected on or near the same day as the participants’ visits and were processed using the same pipeline, with the exception of the artefact signal space projectors, to model environmental noise statistics for source mapping.

MEG data were co-registered to each individual’s segmented T₁-weighted MRI (Freesurfer recon-all) using approximately 100 digitized head points. For participants with useable MEG but not MRI data (HC, n = 3; PD, n = 14), we produced an individualized template with Brainstorm, by warping the default Freesurfer anatomy to the participant’s head digitization points and anatomical landmarks.⁶³ We produced source maps of the MEG sensor data with overlapping-spheres head models (15 000 cortical vertices, with current flows of unconstrained orientation) and the dynamic statistical parametric mapping approach, informed by estimates of sensor noise covariance derived from the empty-room MEG recordings.

We obtained vertex-wise estimates of power spectrum density from the source-imaged MEG data using Welch’s method (3 s time windows with 50% overlap), which we normalized to the total power of the frequency spectrum at each cortical location. These power spectrum density data were next averaged over all artefact-free 6 s epochs for each participant, and the power spectrum density root-mean-squares across the three orthogonal current flow orientations at each cortical vertex location were projected onto a template cortical surface (FSAverage) for comparison across participants. Note that it was not necessary to register the MEG cortical maps into the same stereotaxic space as the neuromelanin-sensitive MRI data. The neuromelanin scores were derived as single, cross-sectional values for each brainstem nucleus prior to any statistical comparisons (Supplementary Fig. 1).

To consider rhythmic versus arrhythmic cortical activity separately, we parameterized the power spectrum densities with specparam (Brainstorm MATLAB version; frequency range = 2–40 Hz; Gaussian peak model; peak width limits = 0.5–12 Hz; maximum number of peaks = 3; minimum peak height = 3 dB; proximity threshold = 2 SD of the largest peak; fixed aperiodic; no guess weight)¹⁴ and extracted the vertex-wise exponent of arrhythmic spectral components. The rhythmic (i.e. aperiodic-corrected) spectra were derived by subtracting these arrhythmic components from the original power spectrum densities. Rhythmic components of interest were then extracted by averaging over canonical frequency bands (delta, 2–4 Hz; theta, 5–7 Hz; alpha, 8–12 Hz; beta, 15–29 Hz).⁶² This procedure produced five maps of neurophysiological brain activity per participant: one for rhythmic activity in each of the four canonical frequency bands and one for broadband arrhythmic activity. These maps were regressed on SN and LC neuromelanin scores in whole-cortex analyses (see the ‘Statistical analyses’ section).

The parameters of the specparam rhythmic model fits were also extracted for each vertex location to investigate which accounted for significant relationships. The extracted parameters for each Gaussian-modelled rhythmic peak included the maximum amplitude, frequency-at-maximum amplitude and bandwidth (full-width at half-maximum). For the alpha-frequency relationships, these features were extracted within the 4–15 Hz frequency range (bandwidth extended to account for potential alpha-frequency slowing in PD⁶⁴), and for beta-frequency relationships within the 15–29 Hz range. These values were averaged within each participant across the vertices of each significant cluster, excluding any vertices where no rhythmic component was identified. The number of rhythmic peaks detected within the frequency range was summed across all vertices of the relevant cortical cluster and divided by the total number of possible peaks (i.e. three maximum peaks × number of cluster vertices) to derive a percentage peak detection probability.

To investigate the possible confounding effects of head motion, eye movements and heart-rate variability on our MEG data, we extracted the root sum square of the reference signals from the MEG head position indicators, electrooculogram and ECG channels, respectively. None of these metrics differed significantly between patients with PD and controls (Mann–Whitney U-tests; head motion, W = 1818.00, P = 0.848; electrooculogram, W = 1590.00, P = 0.220; ECG, W = 1898.00, P = 0.715).

Normative atlases of catecholamine transporter densities

We used neuromaps³⁸ to obtain cortical maps of norepinephrine transporter (NET; data from 77 participants using ¹¹C-MRB PET) and dopamine transporter (DAT; 174 participants; ¹²³I-FP-CIT), following previously established procedures,^8,39,52,65 and parcellated the resulting maps based on both the Desikan–Killiany⁶⁶ and Brainnetome⁶⁷ atlases, to ensure that none of our findings was biased by the choice of atlas.

Statistical analyses

Participants with missing data were excluded across analyses pairwise. We used a threshold of P < 0.05 to indicate statistical significance and ran two-tailed tests unless otherwise specified. Significance testing relied on non-parametric models for all cases, to account for any effects of outliers and potential non-normality of the data. All relationships concerning neuromelanin scores in either the LC or SN included the other region as a covariate of no interest to control for any non-focal effects (e.g. of absolute signal quality or general neurodegeneration).

We derived statistical comparisons across the cortical maps produced, covarying out the effect of age, using SPM12. We tested for group differences in neurophysiological features, beyond the effects of age, using independent-samples t-tests. We formulated the relationships between MEG derivatives and neuromelanin scores as multiple regressions for each neurophysiological feature:

$$\mathrm{neurophysiological}\mathrm{feature}\sim \mathrm{SN}\mathrm{neuromelanin}+\mathrm{LC}\mathrm{neuromelanin}+\mathrm{nuisance}\mathrm{covariate}(\mathrm{s}).$$

For models where LC neuromelanin was the covariate of interest, both age and SN neuromelanin were included as nuisance covariates. Disease duration (i.e. time since diagnosis) was not included in these models, because it did not covary significantly with LC neuromelanin. Disease duration and SN neuromelanin were found to be significantly associated (r = −0.39, P_PERM = 0.007). As such, for models where SN neuromelanin was the covariate of interest, disease duration was also included as a nuisance covariate, alongside age and LC neuromelanin. This limited the participant sample for these models to those patients whose disease duration was known (n = 46). Initial whole-cortex tests used parametric general linear models, with secondary corrections of the resulting F-contrasts for multiple comparisons across cortical locations with non-parametric threshold-free cluster enhancement (extent threshold = 1.0, height threshold = 2.0; 5000 permutations).⁶⁸ We applied a final cluster-wise threshold of P_FWE < 0.05 to determine statistical significance and used the threshold-free cluster enhancement clusters at this threshold to mask the original statistical values (i.e. vertex-wise F-values) for visualization. We extracted data from the cortical location exhibiting the strongest statistical relationship in each cluster (i.e. the ‘peak vertex’) for subsequent analysis (e.g. to test for potential confounds) and visualization.

We derived general linear models to test for univariate group differences and bivariate linear relationships, with non-parametric permutation testing to derive frequentist P-values using lmPerm in R. Where appropriate, we assessed Bayesian evidence for null versus alternative models using Bayes factors (BF) computed with the BayesFactor package in R. All reported linear relationships survived multiple comparisons corrections across related tests using a Benjamini–Hochberg false discovery rate (FDR) approach.

We conducted spatial co-localization analyses between group-level statistical maps and neurochemical atlas data from neuromaps³⁸ using the cor.test function in R and non-parametric spin tests with autocorrelation-preserving null models (5000 Hungarian spins; threshold, P_SPIN < 0.05).⁴⁰ We averaged all relevant neurophysiological and neurochemical data over the 68 parcels of the Desikan–Killiany atlas prior to analysis. We replicated our observation over the 210 regions of the Brainnetome atlas,⁶⁷ confirming that our results are not likely to be biased by the cortical parcellation. We obtained group-level statistical maps by regressing region-wise neurophysiological features on covariates of interest, controlling for age. We then extracted, for each parcel, the resulting unstandardized beta weights for the covariate of interest to represent the spatial topography of unthresholded group-level statistical effects.

Results

Pipelines for our MEG, neuromelanin-sensitive MRI and neurochemical co-localization processing and analyses are visualized in Supplementary Fig. 1.

Atypical cortical neurophysiology and brainstem neuromelanin depletion in Parkinson’s disease

We found expressions of atypical neurophysiological activity in patients with PD (n = 58) compared with healthy older adults (n = 65) in bilateral posterior [theta band; PD > HC; cluster P_FWE < 0.001; peak vertex (x, y, z) = 33, −45, −11; Fig. 1A, top], precentral [alpha band; PD > HC; cluster P_FWE = 0.011; peak vertex ( = x, −43, y, 14, z, 41; Fig. 1A, middle) and occipito-temporal [slope of the aperiodic spectrum; PD < HC; cluster P_FWE = 0.013; peak vertex (x, y, z) = 48, −64, 25; Fig. 1A, bottom] cortices. These findings align with previous literature.^33,36 As expected in patients ‘on’ dopaminergic medications that normalize beta-band activity in PD,^19,27,28 we did not find significant group differences in beta rhythmic activity. Therefore, it is unclear whether the variance in beta rhythmic activity observed in these patients is inherently pathological, because it does not differ statistically from healthy levels and should be interpreted with caution.

Figure 1.

Alterations of regional cortical neurophysiology and neuromelanin depletion of brainstem nuclei in Parkinson’s disease. (A) Cortical maps represent regional clusters of differences in rhythmic (i.e. theta and alpha oscillations) and arrhythmic (i.e. aperiodic 1/f slope) neurophysiological features between patients with Parkinson’s disease and healthy older adults. Colours indicate the strength of the statistical effect (in F-values), thresholded based on the cluster limits identified using threshold-free cluster enhancement (P_FWE < 0.05). No significant differences were observed in the delta and beta bands. (B) Top: Representative neuromelanin-sensitive MRI data focused on the substantia nigra (SN; left panel) and the locus coeruleus (LC; right panel) for one participant with Parkinson’s disease (PD) and one age-matched healthy control participant (HC). Bottom: Group differences in neuromelanin scores for substantia nigra (left panel) and locus coeruleus (right panel), with individual points representing participants. Boxes and whiskers indicate the median, upper and lower quartiles and minima/maxima for each group, and violin plots show the associated density distributions.

We also found reduced neuromelanin MRI signals in both the SN (t = 5.06, P_PERM < 0.001) and the LC (t = 2.74, P_PERM = 0.009; Fig. 1B) of patients (n = 58) compared with healthy older adults (n = 27). It is important to note that the majority of healthy older adult participants in this study did not have both MEG and neuromelanin MRI data available, hence all analyses that connected these two types of data were conducted exclusively within the PD patient cohort.

Neuromelanin depletion is associated with atypical cortical neurophysiology in Parkinson’s disease

We found that depletion of neuromelanin in the LC is associated with increased alpha activity in the bilateral fronto-motor cortices of patients (n = 58; cluster P_FWE = 0.006; peak P_PERM < 0.001; peak vertex (x, y, z) = −52, 7, 34; Fig. 2A], beyond the effects of neuromelanin depletion in the SN. This relationship is contributed by the presence (i.e. peak likelihood; r = −0.38, P_PERM < 0.001; post hoc tests) and amplitude (r = −0.30, P_PERM = 0.018) of the alpha peak in the frequency spectrum of regional activity (Supplementary Fig. 2), not by other spectral parameters (centre frequency, P_PERM = 0.583, BF₀₁ = 3.19; width, P_PERM = 0.961, BF₀₁ = 3.36; all-features model versus frequency-features model, BF₁₀ = 5.40). Note that this relationship remains significant (P_PERM < 0.001) even after excluding a subjective outlier data point (see orange data-point in Fig. 1B). This point did not exert undue influence on the model (Cook’s distance = 0.047; 1.47 SD from the group mean), hence it could not be excluded based on an empirical threshold. Furthermore, no significant hemispheric lateralization of this LC-alpha effect was observed (laterality index; t = −0.25, P_PERM = 0.706). This association between LC neuromelanin and alpha activity is more pronounced in patients with stronger attention impairments (n = 49; moderating effect, t = 4.01, P_PERM < 0.001; Fig. 2A, bottom right). Here too, the subjective outlier point did not exert undue influence on the model (Cook’s distance = 0.014; −0.18 SD from the group mean).

Figure 2.

Regional cortical rhythmic neurophysiological features related to brainstem nuclei neuromelanin. (A) The maps show the cortical regions where locus coeruleus (LC) neuromelanin scales with rhythmic alpha power, beyond the effects of substantia nigra (SN) neuromelanin and age. The nature of this relationship is displayed in scatter plots of peak vertex alpha power values on the bottom left, with the partial correlation coefficient and permuted P-value (P_PERM) overlaid. The line plots on the bottom right reflect the significant moderation of this relationship by attention abilities [indicated by colours of lines, separated into the top (n = 12) and bottom (n = 13) quartiles for visualization], such that individuals with worse attentional impairments exhibited a stronger relationship between locus coeruleus degeneration and rhythmic alpha power. Note that the quartile separation was only for visualization, and attention was modelled as a continuous moderating variable in this analysis (n = 49). (B) The map shows the cortical regions where substantia nigra neuromelanin scores are associated with rhythmic beta power, beyond the effects of locus coeruleus neuromelanin, disease duration and age. The nature of this relationship is displayed in the scatter plot below using peak vertex beta power values, with the partial correlation coefficient and permuted P-value overlaid. Colours on the cortical maps indicate the strength of the statistical effect (in F-values), thresholded based on the cluster limits identified using threshold-free cluster enhancement (P_FWE < 0.05).We used non-parametric permutation testing to account for the influence of outliers in these associations. Also note that the subjective outlier in the top left of the scatterplot in A did not exert undue influence on the model (defined as a Cook’s distance > 3 standard deviations from the group mean).

We found that the association of alpha activity with LC neuromelanin is stronger in the bilateral fronto-motor cortex, where alpha activity also scales negatively with attention scores (r = 0.69, P_SPIN < 0.001; Fig. 3A) and which features the highest concentrations of noradrenaline transporters (NET; r = −0.65, P_SPIN < 0.001; Fig. 3B). Both these alignments were replicated using a different parcellation (Brainnetome), with results for attention (r = 0.64, P_SPIN < 0.001) and NET (r = −0.65, P_SPIN < 0.001).

Figure 3.

Co-localization of cortical associations between rhythmic neurophysiological features, clinical symptoms and brainstem nuclei neuromelanin. (A) Cortical maps indicate the region-wise linear relationships, in unstandardized regression (i.e. beta) weights, between alpha power and locus coeruleus (LC) neuromelanin (top) and attention (bottom). The scatter plot to the left indicates the alignment between these maps (each dot is from a parcel of the Desikan–Killiany cortical atlas), with r-values and P-values overlaid. (B) Co-localization of the region-wise alpha–LC relationships shown in (A) with the density of cortical norepinephrine transporter (NET). (C) Similar to the effects shown in (A), but concerning relationships between rhythmic beta power, substantia nigra (SN) neuromelanin and axial motor symptoms. (D) Similar to the effects shown in (B), but concerning co-localization of the region-wise beta–SN relationships with the density of cortical dopamine transporter (DAT). Note that null distributions for estimating P-values were generated using 5000 autocorrelation-preserving spatial permutations of the data.

Our data also show that depletion of neuromelanin in the SN is associated with decreased beta activity in the left lateral cortices of patients [n = 46; cluster P_FWE = 0.010; peak P_PERM = 0.001; peak vertex (x, y, z) = −64, −7, 28; Fig. 2B], with the strongest peak in the postcentral gyrus. This relationship remains significant even when controlling for the effects of disease duration and neuromelanin depletion in the LC. This association is specific to the amplitude of beta activity (r = 0.41, P_PERM = 0.006; Supplementary Fig. 2) and is not related to other peak parameters (peak likelihood, P_PERM = 0.150, BF₀₁ = 0.96; centre frequency, P_PERM = 0.804, BF₀₁ = 2.77; width, P_PERM = 0.725, BF₀₁ = 2.97; all-features model versus frequency-features and detection probability model, BF₁₀ = 6.61). This effect is not significantly moderated by motor symptoms (summed sub-scores from UPDRS-III; n = 40; axial, P_PERM = 0.667; bradykinesia and rigidity, P_PERM = 0.128).

The cortical topography of the association between beta activity and SN neuromelanin is aligned with that of the association of regional beta activity with axial (r = 0.69, P_SPIN < 0.001; Fig. 3C) but not bradykinesia and rigidity (r = −0.01, P_SPIN = 0.543) motor symptoms. This alignment was also replicated using a different parcellation (i.e. Brainnetome; r = 0.52, P_SPIN < 0.001). There was no significant alignment of the beta–SN relationship with the cortical atlas of dopamine transporters (DAT; r = −0.04, P_SPIN = 0.471; Fig. 3D).

These effects were not affected substantively by additional nuisance covariates (i.e. disease duration, eye movements, heart rate variability, head motion and number of accepted trials; all P-values remained <0.01). Neither the alpha–LC relationship nor the beta–SN relationship was significantly moderated by the use of dopamine agonists (alpha–LC: n = 53, P_PERM = 0.941; beta–SN: n = 42, P_PERM = 0.444) or the levodopa equivalent daily dose (alpha–LC: n = 31, P_PERM = 0.922; beta–SN: n = 25, P_PERM = 0.583), and both these relationships remained significant when the severity of resting-tremor symptoms was included as a nuisance covariate (alpha–LC: n = 42, P_PERM < 0.001; beta–SN: n = 40, P_PERM = 0.007). We also found evidence that these relationships were not attributable to the use of either dopaminergic or non-dopaminergic medication (see the Supplementary material, Testing for Medication Confounds).

Discussion

Neurophysiological changes in patients with PD affect several aspects of the electrophysiological spectrum and relate to the cognitive and motor features of the disease.^{8,17,33,34,52,69} Reductions in beta-band activity are the most well studied in PD and have associations with both the severity of motor impairments^1,25,26 and their therapeutic amelioration.^19,27-30 Increased alpha-band cortical rhythms have received less attention but are thought to be related to cognitive impairments and progression to PD dementia.^16-20

Here, using a new multimodal approach, we report that alpha- and beta-band cortical neurophysiology are differentially associated with the degeneration of neuromelanin-rich cells in brainstem nuclei in PD. Our data show that beta activity is related to loss of dopaminergic neurons in the SN and alpha activity to the depletion of noradrenergic cells in the LC. The cortical regions where alpha activity is the most strongly related to LC neuromelanin also have higher concentrations of norepinephrine transporter. We found that cortical regions showing a strong association between alpha activity and LC neuromelanin are also those where alpha activity scales with the severity of attentional impairments. We also found that in regions where regional beta activity is related to SN neuromelanin, stronger beta activity signals more severe axial motor impairments.

These findings suggest dissociable norepinephrine–alpha–cognitive and dopamine–beta–motor pathophysiological pathways in PD. The relationship between LC degeneration and alpha activity is significant from a pathophysiological standpoint, as it establishes a new link. This link unites previous lines of research showing associations between noradrenergic dysfunction,³ atypical alpha activity,^16-20 and cognition in PD, and between norepinephrine and alpha activity in the healthy brain.^21-23 We also report associations between beta activity, SN degeneration and motor impairments, thereby confirming previous, albeit scattered, observations.¹ Our results provide converging evidence for the association between dopaminergic nigrostriatal dysfunction and motor-related beta changes in PD. Furthermore, these findings broaden the existing literature by highlighting associations between motor impairments and beta rhythmic activity in lateral peri-central cortices in patients on effective levodopa regimens. It should be noted, however, that these beta–SN relationships were weaker than their alpha–LC counterparts, and we did not find significant differences in the amplitude of cortical beta activity between patients with PD and control participants. Furthermore, the beta–SN effects we observed did not map to the primary motor cortices. This might be attributable to the patients being under their normal regimen of dopaminergic medications when they participated in the study. These medications ameliorate motor symptoms⁷⁰ and normalize beta activity,^19,27 which likely reduces the meaningful variability available for modelling in our study, and therefore the size of the observed effects. Additionally, only the axial motor symptoms that are resistant to levodopa treatment⁵¹ were found to be associated with the same beta oscillations that corresponded with SN neuromelanin depletion. The relationship between SN neuromelanin and beta activity was not stronger in cortical regions of high DAT density compared to other regions. However, dopaminergic projections from the SN synapse more prominently with the striatum than directly with the neocortex.⁷¹ Given that reduced dopaminergic signalling from the SN affects cortical activity indirectly, it is less probable that cortical dopamine systems significantly modulate the relationship between beta activity and SN degeneration. Nonetheless, the absence of data from patients in their ‘off’ medication state represents a significant limitation of our study, particularly regarding the relationships observed with beta rhythmic activity. In contrast, our findings suggest relatively strong evidence that the association between LC neuromelanin and alpha rhythmic activity is mediated by noradrenergic mechanisms, which are unlikely to be influenced by levodopa administration.

We also emphasize that few studies so far^{8,35-37,72,73} have considered the respective associations of rhythmic and arrhythmic neurophysiological activity with PD. Yet, pathological changes in rhythmic versus arrhythmic electrophysiology call for distinct mechanistic interpretations. For instance, the arrhythmic slope of the neurophysiological power spectrum is steeper in patients with PD, potentially indicating a broadband reduction in the balance of excitatory-versus-inhibitory firing.^37,74 This arrhythmic shift was present in our participants but did not relate significantly to degeneration of the SN or LC. Instead, we found that only rhythmic spectral activity in the alpha and beta bands was related to the integrity of brainstem nuclei. This reinforces previous conceptualizations of PD as affecting primarily rhythmic neurophysiological activity.⁶⁹ We also found that these relationships were replicated only with the presence (i.e. detection probability) and strength (i.e. peak amplitude) of band-limited rhythmic activity in fronto-motor cortices. They were not related to frequency characteristics such as peak frequency or bandwidth.

Theta band activity also differed between patients with PD and healthy older adults, but did not significantly relate to neuromelanin levels in either the LC or SN. These neurophysiological alterations may be associated with other neurochemical depletions in PD, such as acetylcholine.⁷⁵ Alternatively, theta band effects might be secondary to the primary neuropathological hallmarks of PD (e.g. Lewy bodies). Future studies investigating theta band activity in PD in more detail are therefore necessary.

A key limitation of our present work is the lack of neuroimaging data collected while patients with PD were ‘off’ their normal antiparkinsonian medication regimens. This particularly complicates the interpretation of the SN–beta relationship, given the effect of levodopa on beta rhythmic activity^19,27 and motor symptoms.⁷⁰ Future studies examining this relationship in both ‘on’ and ‘off’ medication states will help determine whether the observed association results from imperfect dopamine therapies or levodopa-resistant beta rhythms in lateral cortices.

There is limited research suggesting an effect of dopaminergic medications on alpha cortical rhythms.^27,76,77 However, increases in alpha activity have also been observed in patients with PD after withdrawal of dopaminergic medication⁷⁸ and in PD patients who were medication naive,⁶⁴ indicating that the alpha increases observed in our study are unlikely to be a result of levodopa administration. We also found evidence that dopaminergic medication effects were unlikely to mediate or moderate the observed alpha–LC relationship. This relationship, in addition to its co-localization with norepinephrine transporter densities, remained significant after controlling for the influence of these medications, supporting a noradrenergic interpretation.

We were also unable to test whether associations between rhythmic neurophysiological activity and brainstem neuromelanin levels exist in healthy individuals without PD, owing to the non-overlap in MEG and neuromelanin-sensitive MRI data between our healthy control groups. However, we found it prudent to first demonstrate that patients with PD in our sample exhibited aberrant increases in fronto-motor alpha activity and decreased neuromelanin levels in the LC and SN, compared to healthy controls. This was done before testing the association between these variables within the patient group.

Additionally, although the presence of such a relationship in healthy older adults would certainly be of interest, it might be more difficult to detect owing to reduced variability of neuromelanin scores in this group (Fig. 1B). Regardless, in our view, the absence of such a relationship in healthy individuals does not diminish the clinical relevance of the observed associations in the PD group.

Finally, owing to our use of large open datasets for this research, we lacked detailed information on medication use and clinical metrics for many patients in our sample. Although this approach imposes certain limitations on the analyses and interpretations that follow, we feel that it also provides several key benefits, such as a larger and more representative participant sample. We also emphasize that, even when considering only participants with detailed medication or clinical information available, the resulting sample sizes still match or exceed much of the extant MEG literature on PD.³³

Conclusion

In sum, the present study supports the possibility of a noradrenergic basis for atypical regional alpha rhythmic activity in patients with PD, with relevance for attention functions; it also strengthens the understanding of the role of dopaminergic dysfunction in changes to rhythmic beta activity related to motor functions in those patients. Future research is needed to substantiate the causal nature of these relationships through longitudinal studies, pharmaco-MEG studies targeting the noradrenergic system and non-invasive neuromodulation approaches. These findings are potentially translatable as a biomarker of treatment response in PD. Neurophysiological measures are expected to respond more dynamically to disease-modifying therapeutics than measures of brainstem nucleus integrity, making them valuable within the relatively short time frame of clinical trials. The present results also hint at the possibility that clinical interventions combining noradrenergic pharmacotherapy with frequency-targeted neurostimulation might be beneficial in PD. Overall, this study advances our understanding of the neurochemical bases of neurophysiological alterations in patients with PD, where non-dopaminergic mechanisms remain particularly underexplored.

Data availability

Magnetoencephalography and T₁-weighted MRI data used in the preparation of this work are available through the Clinical Biospecimen Imaging and Genetic (C-BIG) repository (https://www.mcgill.ca/neuro/research/c-big-repository/resources-researchers),⁴¹ the PREVENT-AD open resource (https://openpreventad.loris.ca/),⁴³ and the OMEGA repository (https://www.mcgill.ca/bic/resources/omega).⁴⁴ Normative neurotransmitter density data are available from neuromaps (https://github.com/netneurolab/neuromaps).⁶⁶ Neuromelanin-sensitive MRI data will be made openly available in the future, and until then are available upon reasonable request.

Funding

This work was supported by a Banting Postdoctoral Fellowship (BPF-186555) and the Canada Research Chair (Tier 2; CRC-2023-00300) in Neurophysiology of Aging and Neurodegeneration to A.I.W. from the Canadian Institutes of Health Research (CIHR) and grant F32-NS119375 to A.I.W. from the National Institutes of Health (NIH); by a Foundation Grant to E.A.F. from the Canadian Institutes of Health Research (CIHR; FDN-154301) and the CIHR Canada Research Chair (Tier 1) of Parkinson’s Disease to E.A.F.; from a Rapid Response Grant from the Weston Family Foundation (RR171117) to D.L.C.; and from a NSERC Discovery grant (RGPIN-2020-06889), a grant from the Healthy Brains for Healthy Lives initiative of McGill University under the Canada First Research Excellence Fund and the CIHR Canada Research Chair (Tier 1; CRC-2017-00311) for Neural Dynamics of Brain Systems, and a grant from the NIH (R01-EB026299) to S.B.. Data collection and sharing for this project was provided by the Quebec Parkinson Network (QPN), the Pre-symptomatic Evaluation of Novel or Experimental Treatments for Alzheimer’s Disease (PREVENT-AD; release 6.0) program, and the Open MEG Archives (OMEGA). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. The QPN is funded by a grant from Fonds de recherche du Québec—Santé (FRQS) and the Canada First Research Excellence Fund, awarded through the HBHL initiative at McGill University. PREVENT-AD was launched in 2011 as a $13.5 million, 7-year public–private partnership using funds provided by McGill University, the FRQS, an unrestricted research grant from Pfizer Canada, the Levesque Foundation, the Douglas Hospital Research Centre and Foundation, the Government of Canada, and the Canada Fund for Innovation. Private sector contributions are facilitated by the Development Office of the McGill University Faculty of Medicine and by the Douglas Hospital Research Centre Foundation (http://www.douglas.qc.ca/). OMEGA and the Brainstorm app are supported by funding to S.B. from the NIH (R01-EB026299), a Discovery grant from the Natural Science and Engineering Research Council of Canada (436355-13), and the CIHR Canada Research Chair in Neural Dynamics of Brain Systems (CRC-2017-00311).

Competing interests

The authors report no competing interests.

Supplementary material

Supplementary material is available at Brain online.