Chronic intracranial recordings in the globus pallidus reveal circadian rhythms in Parkinson's disease

Abstract

Circadian rhythms have been shown in the subthalamic nucleus (STN) in Parkinson's disease (PD), but only a few studies have focused on the globus pallidus internus (GPi). This retrospective study investigates GPi circadian rhythms in a large cohort of subjects with PD (130 recordings from 93 subjects) with GPi activity chronically recorded in their home environment. We found a significant change in GPi activity between daytime and nighttime in most subjects (82.4%), with a reduction in GPi activity at nighttime in 56.2% of recordings and an increase in activity in 26.2%. GPi activity in higher frequency bands ( > 20 Hz) was more likely to decrease at night and in patients taking extended-release levodopa medication. Our results suggest that circadian fluctuations in the GPi vary across individuals and that increased power at night might be due to the reemergence of pathological neural activity. These findings should be considered to ensure successful implementation of adaptive neurostimulation paradigms in the real-world.

Fig. 1. Circadian rhythms in the GPi.

A Example subject exhibiting GPi beta band (24.41 ± 2.5 Hz) power decreasing at night. Pink vertical lines indicate the time of events marked by the subject and are used here as a surrogate marker of the awake/sleep period. Circular polar plots show the beta power (left panel) and the number of events marked (middle panel) in a 24-h circadian clock with a 1-h averaged increment. The mean and standard error of the normalized spectral power averaged over a 24-h cycle is shown on the right panel. The green and blue shades indicate the nighttime and daytime periods, respectively, used for subsequent analysis (see methods). B Example subject exhibiting GPi beta band (19.53 ± 2.5 Hz) power increasing at night. Same convention as in panel A. C Circadian heatmap showing GPi power across all individual recordings (n = 130) normalized and plotted over a 24-h circadian clock. GPi power is sorted from the most decreased (blue) to the most increased (red) at night after grouping into canonical frequency bands. The number of events reported by all subjects over a 24-h circadian clock is shown below the circadian heatmap and used as a surrogate marker of the awake/sleep period. Hz Hertz, arbu arbitrary units; Source data is provided in the Source Data file.

Fig. 2. Factors influencing circadian rhythms in the GPi.

A Scatter plot of the changes in spectral power between daytime and nighttime period (day–night), for each GPi hemisphere recording (n = 130), in relation to the center frequency of the recordings. A positive change indicates a decrease in power at nighttime and a negative change indicates an increase in power at nighttime. B Violin plot quantifying the changes in power during day and night for each canonical frequency band (mean and range); Alpha (≤12 Hz), low beta ( > 12 and ≤20 Hz), and high beta ( > 20 and ≤30 Hz). Low beta (p_corrected = 0.004) and high beta (p < 0.001) power was more decreased power at nighttime (indicated by positive values and *** below violins). At the same time, alpha power was not statistically different from 0 (p_corrected = 1.0), indicating that power was either increased or decreased at night. The high beta band was statistically different from the alpha band (p_corrected < 0.001) and low beta band (p_corrected = 0.003), and the low beta band was statistically different from the alpha band (p = 0.004) using post-hoc Tukey’s tests following a one-way ANOVA. C Scatter plot of circadian beta power ( > 12 and ≤30 Hz) in subjects treated with levodopa (L-Dopa) extended-release (ER) (purple) and those who do not take levodopa ER (orange). In subjects taking levodopa ER at nighttime, beta power was decreased more at night (p_corrected = 0.002) than in subjects who were not treated with levodopa ER. D Comparison of in-clinic off-medication off-stimulation awake baseline beta band power in subjects with increased and decreased power at nighttime (mean and standard error). There was a statistical difference between the two groups (p_corrected = 0.028; non-parametric Mann–Whitney U test). (*) denotes p < 0.05, (**) denotes p < 0.01, and (***) denotes p < 0.001. All p values were corrected for multiple comparisons with Bonferroni correction. n number of recordings; Hz Hertz; L-Dopa ER levodopa extended-release, arbu arbitrary units. Source data is provided in the Source Data file.

Fig. 3. Lead and sensing localizations.

Leads were localized in normalized atlas space (MNI152) in axial (left panel) and sagittal view (right panel). The segmentation of the globus pallidus internus (GPi) and externus (GPe) from the DISTAL atlas are shown in green and gray, respectively. The blue dots indicate the center of bipolar sensing contact pairs for subjects with increased beta power at nighttime and red dots for subjects with decreased beta power at nighttime. No cluster differentiation was found on visual inspection. Source data is provided in the Source Data file.