Alpha oscillations in the pedunculopontine nucleus correlate with gait performance in parkinsonism

Abstract

The pedunculopontine nucleus, a component of the reticular formation, is topographically organized in animal models and implicated in locomotor control. In Parkinson's disease, pedunculopontine nucleus stimulation is an emerging treatment for gait freezing. Local field potentials recorded from pedunculopontine nucleus electrodes in such patients have demonstrated oscillations in the alpha and beta frequency bands, reactive to self-paced movement. Whether these oscillations are topographically organized or relevant to locomotion is unknown. Here, we recorded local field potentials from the pedunculopontine nucleus in parkinsonian patients during rest and unconstrained walking. Relative gait speed was assessed with trunk accelerometry. Peaks of alpha power were present at rest and during gait, when they correlated with gait speed. Gait freezing was associated with attenuation of alpha activity. Beta peaks were less consistently observed across rest and gait, and did not correlate with gait speed. Alpha power was maximal in the caudal pedunculopontine nucleus region and beta power was maximal rostrally. These results indicate a topographic distribution of neuronal activity in the pedunculopontine nucleus region and concur with animal data suggesting that the caudal subregion has particular relevance to gait. Alpha synchronization, proposed to suppress 'task irrelevant' distraction, has previously been demonstrated to correlate with performance of cognitive tasks. Here, we demonstrate a correlation between alpha oscillations and improved gait performance. The results raise the possibility that stimulation of caudal and rostral pedunculopontine nucleus regions may differ in their clinical effects.

Figure 1

Localization of electrodes and contact locations represented in MNI space (sagittal view). The colouring of electrode tips identifies the electrodes of individual patients. PM = pontomesencephalic line connecting the pontomesencephalic junction to the caudal end of the inferior colliculi.

Figure 2

Example data from Patient 3, left PPN electrode at contact pair 01. (A) Raw local field potential (LFP) during rest. (B) Local field potential autospectrum averaged over the period (127 s) of rest. Note peak at 9 Hz (‘peak rest local field potential’). (C) Autospectrum of local field potential averaged over the period of unconstrained walking. Note peak at 8 Hz (‘peak gait local field potential’). (D) Spectra of coherence between the local field potential and EEG (FzCz) during the period of walking. Horizontal line is the 95% confidence limit. Note the peak in coherence at 8–9 Hz. (E) Peak alpha power (sum of normalized power over 6–10 Hz) correlation with relative gait speed (rGS). Linear regression line and its 95% confidence limit are shown. (F) Cross frequency local field potential (normalized) correlation with relative gait speed. Dotted line is the raw correlation and continuous line is the 6-point (backwards) moving average. Note the peak in correlation in the alpha band.

Figure 3

(A) Relationship between the peak rest local field potential (LFP) and peak gait local field potential. The strong correlation (r = 0.90, P < 0.001) suggests that the two local field potential peaks may be related. (B) Relationship between the peak rest local field potential and the frequency of the peak in the local field potential–relative gait speed (rGS) correlation spectra. The strong correlation (r = 0.779, P = 0.013) suggests that the peak rest local field potential may be relevant to the performance of gait. (C) Group average un-normalized local field potential and relative gait speed correlation spectra demonstrating a peak in correlation in the alpha band. Dotted line is the average correlation and continuous line is the 6-point (backwards) moving-average. (D) Group averaged normalized (across 5–40 Hz) local field potential and relative gait speed correlation spectra demonstrating persistence of the peak in correlation in the alpha band. Dotted line is the average correlation and continuous line is the 6-point (backwards) moving-average. Correlations were Fisher transformed prior to averaging in C and D.

Figure 4

(A) Normalized power (mean ± SEM) of alpha peaks during gait and beta peaks during rest or gait grouped according to caudal (deeper than 2 mm below the pontomesencephalic junction) or rostral recording site. Alpha peak power was greater caudally and beta power was greater rostrally. (B) Log alpha peak power at rest (mean ± SEM) divided according to recording site depth into 4-mm subregions (denoted on the x-axis in mm relative to the pontomesencephalic junction). Log alpha peak power is maximal in the −2 to −6 mm region, falling significantly at surrounding sites. *P < 0.05. In grey has been superimposed a plot of the variation in per cent improvement in the Gait and Falls questionnaire (GFQ) postoperatively according to stimulation depth at most recent follow-up [depth selected blinded to the local field potential (LFP) data]. One data point was available from each subject (Table 2), with the exception of the patient from London, where the Gait and Falls questionnaire was not assessed. Three of these data points fell −6 to −10 mm below the pontomesencephalic junction, so that here the SEM is also displayed. (C) Representation of the rostrocaudal location of peak un-normalized alpha power at rest, which correlated strongly in frequency and location with peak alpha power that correlated with gait (but which required normalization to remove movement artefact thereby also diminishing power gradients). Relative log alpha power is represented in grey-scale intensity whereby black is highest and white lowest alpha power. Regions relative to the pontomesencephalic line are numbered as follows: 1, +2 to +6 mm; 2, −2 to +2 mm; 3, −6 to −2 mm; 4, −10 to −6 mm; and 5, −14 to −10 mm. Alpha power was maximal at location 3 (−6 to −2 mm below the pontomesencephalic line). Beta power was highest in regions 1 and 2 combined. Note: no inference is made regarding location in ventrodorsal or mediolateral planes. DBS = deep brain stimulation; GFQ = Gait and Falls questionnaire; LFP = local field potential; PM = pontomesencephalic line connecting the pontomesencephalic junction to the caudal end of the inferior colliculi.

Figure 5

Change-point analysis of the time series of mean local field potential power over 7–9 Hz averaged to onset (0 s) of freezing episodes (n = 24) in Case 6. Horizontal dotted lines are the 95% confidence limit of the whole record and the grey blocks represent stable periods between changes in power as defined by change-point analysis. There is a significant drop in 7–9 Hz power ∼1 s before the onset of freezing, and 7–9 Hz activity continues to be attenuated for just over 2 s thereafter.