OUP user menu

Dependence of subthalamic nucleus oscillations on movement and dopamine in Parkinson’s disease

Ron Levy, Peter Ashby, William D. Hutchison, Anthony E. Lang, Andres M. Lozano, Jonathan O. Dostrovsky
DOI: http://dx.doi.org/10.1093/brain/awf128 1196-1209 First published online: 1 June 2002


Local field potentials and pairs of neurones in the subthalamic nucleus (STN) of patients with Parkinson’s disease show high‐frequency oscillations (HFOs) at 15–30 Hz. This study explores how these HFOs are modulated by voluntary movements and by dopaminergic medication. We examined 15 patients undergoing implantation of bilateral deep brain stimulating electrodes using microelectrode recordings of pairs of STN neurones (eight patients) and macroelectrode recordings of local field potentials from the STN (14 patients). Synchronized HFOs between STN neurones were observed in 28 out of 37 pairs in five patients who had tremor in the operating room and none of 45 pairs in three patients who did not. In two of the three non‐tremulous patients, HFOs in the frequency spectra of local field potentials were detected but were weaker than in those patients with tremor. Active movement suppressed synchronized HFOs in three out of five pairs of neurones, independent of changes in firing rate. HFOs observed in the local field potentials in nine out of 14 patients were reduced with voluntary movement in six of the eight patients tested. Dopaminergic medication decreased the incidence of synchronized HFOs in STN neurone pairs, reduced HFO synchrony in a pair of tremor cells concurrent with a reduction in firing rate and limb tremor, and decreased HFOs of local field potentials in the STN. These results demonstrate that HFO synchronization in the STN is reduced by voluntary movements and by exogenous dopaminergic medication. A mechanism for neuronal oscillatory synchronization in basal ganglia is proposed. It is suggested that the firing of STN neurones can be synchronized by 15–30 Hz cortical beta oscillatory activity, particularly when dopamine deficiency results in a higher background firing rate of STN neurones, and that this synchronization contributes to parkinsonian pathophysiology.

  • Keywords: cortical beta oscillations; dopamine; Parkinson’s disease; subthalamic nucleus; synchronization
  • Abbreviations: APO = apomorphine; DBS = deep brain stimulation; EMG = electromyography; GPe = external segment of the globus pallidus; GPi = internal segment of the globus pallidus; HFOs = high‐frequency oscillations; MPTP = 1‐methyl‐4‐phenyl‐1,2,3,6‐tetrahydropyridine; STN = subthalamic nucleus; TC = tremor cell


Several lines of evidence indicate that the subthalamic nucleus (STN) is involved in the pathogenesis of Parkinson’s disease. Degeneration of the substantia nigra pars compacta and subsequent depletion of striatal dopamine and extrastriatal dopamine in 1‐methyl‐4‐phenyl‐1,2,3,6‐tetrahydropyridine (MPTP)‐treated monkeys leads to the emergence of STN neurones with increased spontaneous activity and periodic bursting or ‘oscillatory’ activity, such as tremor‐related neurones (tremor cells, TCs) (Bergman et al., 1994). Lesions or deep brain stimulation (DBS) of the STN dramatically reduce the symptoms of parkinsonism in MPTP‐treated monkeys (Bergman et al., 1990; Aziz et al., 1991; Benazzouz et al., 1993; Wichmann et al., 1994a; Guridi et al., 1996) and in patients with Parkinson’s disease (Limousin et al., 1995; Gill and Heywood, 1997; Obeso et al., 1997; Krack et al., 1998; Kumar et al., 1998; Rodriguez et al., 1998; Andres and Gerloff, 1999).

Synchronous neuronal discharge oscillations in the cortex play an important role in normal motor processing and are associated with such functions as binding of neuronal activity in disparate motor areas, recruitment of motor‐unit discharge, reduction of computational effort or processing load, and modifying motor states (Conway et al., 1995; Baker et al., 1997, 1999; Salenius et al., 1997; Donoghue et al., 1998; Volkmann, 1998; Kilner et al., 2000). The STN receives a substantial excitatory glutamatergic input from the cortex (Afsharpour, 1985; Rouzaire‐Dubois and Scarnati, 1985) and motor‐related cortical areas can modulate the activity of subthalamic efferent targets via the cortico‐subthalamic pathway (Nambu et al., 2000). Furthermore, synchronous oscillatory activity in the STN‐globus pallidus externus (GPe) network is intimately related to rhythmic cortical activity (Magill et al., 2000). Since the STN receives direct excitatory input from the primary motor cortex and the supplementary motor area (Monakow et al., 1978; Nambu et al., 2000), it is likely that voluntary movements which modulate oscillatory phenomena in the cortex (Crone et al., 1998; Pfurtscheller and Lopes da Silva, 1999) might influence oscillatory synchronization in the STN.

It has been suggested that the STN might play a role in synchronizing oscillatory activity in the GPe (Plenz and Kital, 1999) and the globus pallidus internus (GPi) (Wichmann et al., 1994a), and that an increase in 3–8 Hz oscillatory synchronization underlies the development of parkinsonian limb tremor (Nini et al., 1995; Bergman et al., 1998a; Hurtado et al., 1999; Raz et al., 2000). Recent reports have demonstrated that STN shows synchronized oscillatory activity at 15–30 Hz (Levy et al., 2000; Marsden et al., 2001; Brown et al., 2001). These synchronized ‘high‐frequency’ oscillations are closely associated with tremor‐related neuronal activity in the basal ganglia of tremulous parkinsonian patients and MPTP‐treated monkeys (Bergman et al., 1994; Levy et al., 2000; Raz et al., 2000). Dopaminergic medication has also been shown to modulate oscillatory activity in the STN and thus may play a role in the pathology of akinesia and rigidity, in addition to limb tremor, by affecting oscillatory synchronization in the basal ganglia (Allers et al., 2000; Brown et al., 2001; Levy et al., 2001; Marsden et al., 2001).

In the present study we used simultaneous microelectrode recordings of pairs of STN neurones and macroelectrode recordings of local field potentials in the STN to demonstrate that synchronous 15–30 Hz or ‘high‐frequency’ oscillations in the STN (Levy et al., 2000) are reduced by voluntary activity and dopaminergic medication. It is proposed that synchronous high‐frequency oscillations in the STN are the result of rhythmic cortical input to the STN, namely beta frequency activity, which is present when patients are at rest but not during active movement, and that this mechanism promotes synchronous oscillations in the basal ganglia‐thalamo‐cortical loop in the dopamine‐deficient state.


Patient group

Fifteen patients with Parkinson’s disease participated in this study. The group consisted of four females and 11 males who, at the time of operation, had a mean age of 55 years (range 34–67). The average disease duration was 12 years (range 7–19). There were five who had a prior pallidotomy and one who had a prior thalamotomy. The mean time since these previous procedures was 44 months (range 24–74). All patients gave free and informed consent, and procedures were approved by the University Health Network Research Ethics Board.

Microelectrode and macroelectrode recording procedures

All 15 patients underwent microelectrode‐guided placement of bilateral DBS electrodes in the STN for the treatment of their parkinsonian symptoms. The localization procedure of the STN using microelectrode recording is described elsewhere (Hutchison et al., 1998). There were eight patients in whom pairs of STN neurones were recorded during the stereotaxic procedure using a dual microelectrode assembly. Microelectrode procedures for recording from neurone pairs have been described elsewhere (Levy et al., 2000). Recordings were performed in the ‘OFF’ drug state (i.e. following a 12 h holiday from parkinsonian medications). In one of these patients, apomorphine (APO), a non‐selective D1/D2 dopamine receptor agonist, was administered at a dose determined during pre‐operative assessment to provide significant motor benefit and relief of limb tremor (Levy et al., 2001).

Macroelectrode recordings of field potentials in the STN were obtained in 13 patients post‐operatively and one patient intra‐operatively. Bipolar macroelectrode recordings were performed post‐operatively through adjacent contacts of the DBS leads between 2 and 7 days after implantation. Recordings were performed in the morning at ∼1–3 h following administration of dopaminergic and other anti‐parkinsonian medication (unless otherwise indicated). In most cases, recordings were performed bilaterally. The DBS leads consisted of four concentric cylindrical contacts (length 1.52 mm, diameter 1.27 mm) separated by 1.5 mm. Contacts were labelled 0, 1, 2 and 3 from the most ventral to the most dorsal contact, respectively. Differential recordings between pairs of contacts were amplified, band‐pass filtered (2–500 Hz) and digitally recorded (CED1401; Cambridge Electronic Design, Cambridge, UK) at a sampling rate of 1200–2000 Hz. Patients had the DBS leads internalized (i.e. placed under the scalp and down the side of the neck to be connected to the internal pulse generator in the chest) 7 days following the implantation surgery, after which time field potential recordings were no longer possible. Intra‐operative recording of STN field potentials in the fourteenth patient was performed simultaneously with microelectrode recording of single neurone activity. This was accomplished by recording single neurones with a microelectrode (25 µm tip, platinum plated) that was inserted in parallel with a 30 gauge stainless steel tube (Small Parts, Miami Lakes, FL, USA) insulated to 0.5 mm of the tip (28 Kapton tubing; Micro ML, College Point, NY, USA) at a centre to centre distance of 600 µm. Monopolar recording was performed through the exposed tip of the macroelectrode.

Task conditions

Neuronal recordings during microelectrode surgery were performed with the patients at rest. Movement‐related modulation of synchronous neuronal oscillations was examined by having four patients perform voluntary chest to target reaching movements. Following a ‘go’ cue, indicated with a light, patients made three movements back and forth between their chest and a button board located ∼50 cm in front of their chest. Patients kept their hand at their chest until one of five buttons was lit, then reached out and pressed the button indicated and brought their hand back to their chest. There was a 2‐s delay between buttons being lit. The maximum distance between the furthest two buttons was 20 cm and the buttons were oriented in a horizontal row. Multiple trials were performed.

During the week following DBS insertion, macroelectrode recordings of field potentials were performed in all patients. Patients also performed maintained isometric contractions of the wrist extensors or self‐paced hand tapping movements contralateral to the side of the macroelectrode recordings. Tasks were always alternated with trials in which macroelectrode activity was recorded with the patients at rest. The length of each trial was 60 s. Patients participated in as many trials of each voluntary activity as they wished, but tests were terminated if the patient became tired. Therefore not all tasks were tested in all patients. In most cases, spontaneous limb tremor was absent due to the acute subthalamotomy effects of the DBS electrode in the week following the stereotaxic procedure.

Data analysis

Macroelectrode signals recorded post‐operatively were low‐pass filtered (<100 Hz) and then down‐sampled to a common sampling rate of 200 Hz to facilitate comparisons between the power spectra of different patients and different trials. Standard spectral techniques resulted in 256 spectral estimates between 0 and 100 Hz, thereby yielding a frequency resolution of 0.39 Hz (unless otherwise indicated). Statistical significance of spectral peaks was assessed by comparing each macroelectrode signal to a white noise signal with the same mean power in the range 0–55 Hz (i.e. mean spectral ‘noise’). The upper limit of 55 Hz was used to avoid any contamination due to 60 Hz line noise. Spectral estimates were deemed significant if they were equal to or greater than the upper bound of a 100 × (1–α)% confidence interval about this white noise signal. The upper bound was given by 2N × f(ω)/χ22N,1–α/2, where f is the power, ω is the frequency, 2N is the effective degrees of freedom, and N is the number of sampling windows (Chatfield, 1996). In addition, the size of significant peaks in the frequency spectra of the macroelectrode recordings was calculated by averaging the signal‐to‐noise ratio over the range of frequencies given by the width of the peak at half the maximum peak value. The mean signal‐to‐noise ratio of the high‐frequency peaks was estimated using the macroelectrode recording from DBS contacts or intra‐operative cannula with the greatest high‐frequency signal, and multiple trials were averaged.

Coherence analysis was used to compare the linear correlation between two waveforms over a range of frequencies. Coherence is a function of frequency and is calculated from the cross‐spectral density between the two waveforms, normalized by the power spectral density of each waveform. Coherence values can range from 0 if the signals are not linearly related to 1 if the signals have a perfectly linear relationship. Since coherence is a measure of linear similarity, the phase shift must be constant and the amplitudes of the two waveforms must have a constant ratio to be completely coherent at a particular frequency over a given time range. A 100 × β% confidence level was determined by calculating a coherence value given by the equation: coherence = 1 – (1 – β)1/(L–1), where β = 0.99 and L is number of windows used (Rosenberg et al., 1989). This value or greater was considered to indicate a significant probability (P < 0.01) of a linear relationship between the two waveforms and the presence of oscillatory synchronization. Phase relations were also assessed for two waveforms that had a significant coherence at a frequency fi, using the equation: phase (fi) = arctan (–Q(fi)/L(fi)), where fi is the ith spectral estimate, Q is the real part and L is the imaginary part of the cross‐spectra between a pair of neurones (Glaser and Ruchkin, 1976).

Spectral analysis of single‐neurone discharge activity has previously been described in detail (Levy et al., 2000). Briefly, the event times of neuronal discharges were converted to waveforms (with a final sampling rate of 1000 Hz), representing the discharge density over time (10 ms bins) using standard software (Spike2; Cambridge Electronic Design). Subsequent Fourier analysis was used to determine the oscillatory modulation of this discharge density due to voluntary movement or APO. In addition, frequency–time plots were constructed by analysing data in consecutive, non‐overlapping, 10‐s windows. Spectral data in each window were normalized by the spectral noise and significance was calculated in a similar manner to the macroelectrode recordings discussed above. Analysis of the movement‐related modulation of oscillatory neuronal discharge involved segmenting the discharge density waveform into 1.024‐s non‐overlapping windows. These windows had a fixed temporal relationship to the beginning of the voluntary movement, as indicated with an accelerometer placed on the dorsum of the index finger or wrist extensor electromyography (EMG). Application of a 1024‐point fast Fourier transform then gave a frequency resolution of 0.98 Hz. Spectral estimates and coherence plots were derived from windows sharing the same relative time delay to the start of movement.


Neurone pair and field potential ‘high‐frequency’ oscillations (15–30 Hz) recorded during rest

Synchronized 15–30 Hz oscillatory activity was observed in 28 out of 82 pairs of neurones that were sampled in eight patients. All 28 pairs were found in those patients that had limb tremor in the operating room. Synchronized 15–30 Hz activity was present during time periods with limb tremor, but also during some periods without any noticeable limb tremor. There was no effect of previous pallidotomy on synchronized activity in the STN. An example of intra‐operative simultaneous microelectrode recordings of two STN TCs with high‐frequency oscillations is shown in Fig. 1A. The corresponding frequency spectra of the neuronal firing and the coherence between these two signals are displayed below the two traces. Coherence between the oscillatory discharge of the two neurones is significant only for the high‐frequency components (P < 0.01).

Fig. 1 Examples of an intra‐operative recording from a pair of microelectrodes (A), a post‐operative recording from a DBS electrode (B), and the spectral analysis of an intra‐operative simultaneous recording from a microelectrode and a macroelectrode (C). The dashed line in the plots represents the significance level for the power spectra estimates (P < 0.01) and for the coherence function (P < 0.01). The units of the power spectra are arbitrary. The number beside the peak in the coherence spectrums is the phase difference in degrees. (A) Raw traces of extracellular microelectrode recordings of two simultaneously recorded STN TCs with high‐frequency oscillations. The tips of the two microelectrodes were separated by 670 µm. The frequency spectra of the neuronal oscillatory discharges and the corresponding coherence function are shown below (1.95 Hz resolution). Spectra and coherence were calculated over a 10‐s interval that included the period shown in the recordings. (B) A segment of a bipolar recording of STN field potentials from a pair of DBS contacts (low pass filtered below 55 Hz). The frequency spectrum of this signal is shown to the right (0.78 Hz resolution) and was calculated from 10 s of data containing this segment. (C) Simultaneous microelectrode and macroelectrode intra‐operative recordings in Patient N. The tips of the electrodes were positioned 2 mm within the dorsal border of the STN. The raw traces shown are an example of a period of highly significant coherence (note that the field potential data was low pass filtered below 55 Hz). The lower left and middle plots show the power spectrums (0.98 Hz resolution) for the single STN neurone recorded with the microelectrode and a monopolar field potential recorded through a 30 gauge cannula, respectively (see Methods). The lower right plot shows the coherence between the two signals. Spectra and coherence were calculated over a 70‐s period. Coherence between the macroelectrode recording and the single unit was limited to the high frequency component of each signal.

Local field potential recordings were obtained from 14 patients. An example of a bipolar field potential recording between adjacent contacts of the DBS electrode is shown in Fig. 1B. In this example there was significant high‐frequency activity (P < 0.01). Field potential recordings performed when the patients were at rest revealed three characteristic spectra that are displayed in Fig. 2. Data from multiple trials are shown in the representative examples. The spectra of each trial recorded when the patients were at rest is indicated by a solid black line. Three patients had broad bandwidth oscillations in the macroelectrode recordings centred between ∼20 and 35 Hz (Fig. 2A). There were 11 patients who displayed significant low‐frequency oscillations (lower than ∼12 Hz) (Fig. 2B and C) and six of these patients also displayed a significant high‐frequency component (Fig. 2C). The spectral profiles were always similar on both sides in the eight patients in whom bilateral recordings were performed. There were no obvious differences in the frequency spectra due to previous pallidal lesions when recordings from lesioned and non‐lesioned sides were compared. Since field potentials were recorded from three pairs of contacts per side, the signal‐to‐noise ratio of the low‐ and high‐frequency oscillations at the three sites were compared. As shown in the two plots in Fig. 2D, only the signal‐to‐noise ratio of the high‐frequency oscillations was maximal for one pair of DBS contacts (P < 0.01), indicating that the spatial extent of these oscillations was more focused than the low‐frequency oscillations.

Fig. 2 Representative examples of STN field potential spectra recorded from two patients displaying only high‐frequency oscillations (A), only low‐frequency oscillations (<12 Hz) (B), and significant low‐ and high‐frequency field potential oscillations (recorded during rest) (C). Black lines in these plots show the spectra of individual trials recorded while the patients were at rest; grey lines show the spectra of individual trials recorded during sustained isometric contractions of the wrist extensors. High‐frequency field potential activity was reduced by voluntary movements only in patients displaying spectra with both low‐ and high‐frequency activity. Labels in each graph such as ‘left 01’ indicate that bipolar recordings of the field potentials were performed between contact 0 and contact 1 of the left STN. The dashed horizontal line in each graph represents the upper bound for a 99.8% confidence interval with respect to the mean spectral noise (calculated between 0 and 55 Hz from the recordings performed with the patients at rest). Spectral estimates above this line indicate statistically significant frequencies. PAL = previous pallidotomy; THAL = previous thalamotomy. (D) The signal‐to‐noise ratio of the high‐frequency (left) and low‐frequency (right) spectral peaks recorded from the three pairs of DBS contacts when the patients were at rest (all recorded sides shown). Values from the same DBS electrode are joined by a line and multiple trials were averaged. Data were aligned with respect to the location of the greatest signal‐to‐noise ratio from each side (dashed vertical line). The mean signal‐to‐noise ratio is indicated by a filled circle. *P < 0.01, one‐way ANOVA (analysis of variance). The relative distance between the bipolar recordings from adjacent DBS contacts was 3 mm.

To examine the relationship between high‐frequency oscillatory discharge recorded from single neurones and high‐frequency oscillations in the field potentials, simultaneous microelectrode and macroelectrode recordings were performed in one patient (Patient N). The tip of the microelectrode was located ∼350 µm from the surface of the macroelectrode. As shown in Fig. 1C, the STN neurone recorded with the microelectrode had oscillations at tremor frequency (∼5 Hz) and at ∼20 Hz (lower left plot). This patient was at rest, but during the period of these recordings was experiencing bilateral limb tremor. The macroelectrode recording of the field potential had a high‐frequency peak centred at ∼20 Hz (lower middle plot). There was a striking coherence at ∼20 Hz between the firing activity of the neurone and the macroelectrode recording, and both these oscillatory signals were in‐phase throughout the recording period (lower right plot).

Comparisons of intra‐operative microelectrode data and intra‐/post‐operative field potential activity (recorded with the patients at rest) were made in seven patients and are shown in Table 1. In the patients in whom pairs of single neurones showed a significant high‐frequency coherence (M, A, K, N), the signal‐to‐noise ratio of the high‐frequency peak recorded with the macroelectrode was greater than in those patients who did not display high‐frequency coherence between pairs of neurones (J, G, B). High‐frequency neurone pair synchronization was consistently in‐phase (maximum absolute phase difference was 30°) and occurred at frequencies similar to those from the macroelectrode recordings in Patients A, K and N. In Patient M, the neurone pair oscillatory synchronization frequency was half of that recorded using the macroelectrode.

View this table:
Table 1

The relationship between high‐frequency activity from microelectrode and macroelectrode recordings

PatientNeurone pairs with high frequency synchronization/total pairs sampledMean synchronization frequency of STN neurone pairs (±SE)Signal‐to‐noise ratio of high‐frequency peak in macroelectrode recordings*Mean frequency of the high‐frequency field potential oscillation from the macroelectrode recordings**
J0/28 (right side)1.7 (right12)23 Hz (18–26)
G0/5 (left side)non‐significant
B0/8 (right side) 0/4 (left side)2.1 (right 12) 2.3 (left 12)32 Hz (25–40) 29 Hz (23–35)
M5/6 (right side)15.7±0.3 Hz2.5 (right 23)27 Hz (24–30)
A5/10 (right side)23.6±0.4 Hz2.9 (right 12)26 Hz (23–29)
K6/8 (right side)21.3±0.5 Hz3.4 (left 12)18 Hz (15–21)
N9/10 (left side)19.6±0.1 Hz5.8 (left)19 Hz (18–20)

*Values in parentheses indicate the side and DBS contacts used for the macroelectrode recordings (see Methods).

**Values in parentheses indicate the frequency range given by width of peak at half the maximum peak value.

Macroelectrode recordings were not available on the side where single unit pairs were recorded, therefore peak frequencies from the other side are reported.

Monopolar macroelectrode recording performed intra‐operatively in this patient only and when patient was OFF medication.

Voluntary movement suppresses neurone pair and field potential high‐frequency oscillations

Simultaneous microelectrode recordings were made from five pairs of STN neurones that displayed a modulation of their firing rates with voluntary movements in addition to displaying synchronized high‐frequency oscillations. For three pairs, the synchronized high‐frequency oscillatory activity between the pair of neurones was reduced during the period of voluntary movements. Two examples from separate patients are shown in Fig. 3 (right column versus left column). Both of these patients performed chest‐to‐target reaching movements (see Methods). In Fig. 3A and C, it can be observed that the spontaneous firing rate was decreased in three neurones and increased in the fourth (*P < 0.05). Corresponding power spectra plots are displayed at the top and middle of Fig. 3B and D, and coherence plots are displayed at the bottom. High‐frequency oscillations in all four neurones were reduced during the movement, and high‐frequency coherence between neurone pairs was only significant before and after the task.

Fig. 3 The suppression of synchronized high‐frequency oscillations of pairs of STN neurones during voluntary pointing movements in two patients (right column versus left column). Data were calculated with respect to the onset of movement (0 s in plots, see Methods). The top trace of both columns is the rectified and averaged accelerometer signal (Acc). (A and C) Plots of the mean firing rate of each neurone (1.024 s bins, ±SE). The patient whose data are displayed in the left column performed 15 trials of the task and the patient whose data are displayed in the right column performed eight trials of the task. The firing rates of all four neurones were modulated by the movement [*P < 0.05 (ANOVA) versus discharge from 5.12 to 0 s before the onset of movement]. (B and D) Time–frequency plots of the spectra of each neurone (top and middle) and the coherence between them (bottom, 1.024 s bins). Both pairs of neurones displayed a movement‐related desynchronization and post‐movement synchronization of their high‐frequency oscillations (i.e. significant coherence before and after the movement). The legends for these plots are located at the bottom of each column.

The effects of voluntary movement on field potential oscillations were examined in seven patients. Figure 2 displays examples of movement‐related suppression of high‐frequency field potential activity. Data from multiple trials of each task are shown for these patients. It was consistently observed that high‐frequency peaks were largest during the rest condition in all patients (black lines). A grey line indicates the spectra of each trial recorded when the patients performed sustained isometric contractions. Only in those patients displaying both low‐ and high‐frequency oscillations (see Fig. 2C) did sustained isometric contractions (Patients M, I) or tapping (L, J, K) decrease high‐frequency activity (five patients tested). Voluntary movement did not modulate the field potential spectra of those patients with only high‐frequency oscillations (two patients tested; see Fig. 2A). In those patients displaying only low‐frequency oscillations, there was no noticeable effect of voluntary movement on frequencies <12 Hz with one exception. In Patient G, isometric contraction of the wrist extensors suppressed a discrete peak at 10 Hz (not shown). Although two of the patients with only low‐frequency field potential oscillations (Patients G, H) did not display statistically significant high‐frequency activity, small high‐frequency peaks can be observed, and sustained contraction of the wrist extensor muscles did cause a clear and reproducible reduction of high‐frequency activity (see Fig. 2B, alternate trials of rest versus contraction for Patient H).

Reduction of neurone pair and field potential high‐frequency oscillations with dopaminergic medication

The effects of dopaminergic medication were examined on neurone pairs in one patient and on field potential activity in another. Single neurone discharge oscillations were examined by administering APO to a patient while performing simultaneous microelectrode recordings of neurones with synchronized high‐frequency oscillations. The changes in firing rate and oscillatory activity of a pair of STN TCs with high‐frequency oscillations following a subcutaneous injection of 6.5 mg of APO are displayed in detail in Fig. 4. In‐phase coherent oscillations between the two neurones at their high‐frequency components occurred throughout the OFF period (bottom two plots of Fig. 4B, P < 0.01). In contrast, several periods without any coherence in the tremor frequency range were observed during this time (i.e. before ∼500 s). Following APO administration, there was a decrease in the firing rates of both neurones, concurrent with a loss of limb tremor, and the patient reported that he felt the effects of medication (Fig. 4A). It can be observed that a decrease in high‐frequency oscillations and a loss of synchronization precede the effect of APO on firing rates and were maintained during the period of recording when the patient was ON. High‐frequency synchronization was not present during the ON period. Figure 5 shows coherence plots of other neurone pairs in this patient recorded before and after APO administration. Before the administration of APO (non‐medicated or OFF state), three pairs of TCs displayed high‐frequency synchronization (Fig. 5A). In four pairs of neurones recorded 15–29 min after APO administration, only one pair showed high‐frequency coherence (Fig. 5B). All pairs recorded before and after APO were located <2 mm from one another.

Fig. 4 The effect of APO administration on neuronal activity of a pair of STN TCs with high‐frequency oscillations. (A) Histograms of the firing rate of the pair of STN neurones (10 s bins) demonstrate that APO reduced the spontaneous discharge of the two neurones. (B) Abolition of high‐frequency synchronization of the same pair of STN TCs following the administration of APO. The top two panels show the power spectra (1.95 Hz resolution) over the same time period and bin size as the firing rate histograms in (A). Both neurones displayed tremor‐related activity and high‐frequency oscillations before the administration of APO. The third plot shows the coherence between the oscillatory activity of the two neurones. The bottom panel is the phase of significantly coherent oscillations at tremor frequency (∼5 Hz) and at high frequency (∼15 Hz).

Fig. 5 Dopaminergic modulation of high‐frequency synchronized oscillations in neurone pairs recorded before (A) and after (B) APO administration and in field potential activity in another patient (C). (A) Coherence plots (0.98 Hz resolution) of three pairs of STN neurones recorded before APO administration. All neurones displayed tremor‐related activity and two pairs (right and left plots) were synchronized in the tremor frequency range (∼3–8 Hz). The dashed line in the plots is the significance level for the coherence function (P < 0.01). The number beside the peaks is the phase difference in degrees. All neurone pairs in (A) and (B) were sampled for 3 min. (B) Coherence plots of four pairs of STN neurones recorded after APO administration when the patient was ON (the time of recording after APO dosing is indicated above each plot). (C) Effect of levodopa on high‐frequency field potential signals recorded from Patient H. The high‐frequency oscillations recorded during the OFF levodopa state were localized to the dorsal contacts of the DBS electrodes. Note the reduction of high‐frequency activity during the levodopa‐induced ON period or with contraction in the OFF state (two alternate trials of 60 s each).

Field potential recordings were made in one patient (Patient H, see Fig. 2B) before and after levodopa administration and are shown in Fig. 5C. High‐frequency activity was observed in the OFF state but was not present in the levodopa‐induced ON state. Isometric wrist extensor contraction reduced high‐frequency activity in the OFF state (two alternate trials with rest condition shown).


This study demonstrates that the 15–30 Hz local field potential activity that can be recorded from the contacts of DBS electrodes in STN (Brown et al., 2001; Marsden et al., 2001) is the result of in‐phase synchronous 15–30 Hz oscillatory discharge activity in STN neurones. Similar observations between neurones with oscillatory synchronization and local field potentials have been made in the sensorimotor cortex of monkeys (Murthy and Fetz, 1996b). In addition, our results indicate that these synchronous high‐frequency oscillations are reduced by voluntary movements. The suppression of STN single unit high‐frequency oscillatory discharge with movement occurred independently of increases or decreases in firing rate, suggesting that the modulation of high‐frequency activity reflects a change in the pattern of input to the STN. This also indicates that the temporal pattern, in addition to rate coding, conveys information that may be involved in the normal and pathological functioning of the STN. Our observations of the reduction of 15–30 Hz field potential oscillations in the STN during voluntary movement (six out of eight patients in total) support and complement the findings of Brown et al. (2001), who reported that the 15–30 Hz coherence between STN and GPi local field potentials is not as high during tonic contractions as during rest.

Cortical input to the STN (Monakow et al., 1978; Carpenter et al., 1981; Afsharpour, 1985; Rouzaire‐Dubois and Scarnati, 1985; Canteras et al., 1990) from the primary motor cortex and the supplementary motor area is somatotopically arranged (Nambu et al., 1996). Short latency cortical input is relayed by the STN to other basal ganglia nuclei (Wichmann et al., 1994b; Nambu et al., 1996; Levy et al., 1997; Magill et al., 2000; Nambu et al., 2000). The STN can strongly influence neuronal activity in the basal ganglia (Carpenter and Strominger, 1967; Smith et al., 1990; Hamada and DeLong, 1992; Hazrati and Parent, 1992; Parent and Hazrati, 1995; Ni et al., 2000) and plays a role in synchronizing oscillatory population behaviour (Wichmann et al., 1994a; Plenz and Kital, 1999; Brown et al., 2001). It has been shown in the rat that rhythmic cortical activity leads to low‐frequency oscillatory activity in the STN‐GPe network in the dopamine‐depleted state (Magill et al., 2001). Similar to our results in the STN, comparable changes in cortical rhythms during voluntary movements in humans are also observed. The 15–30 Hz cortical beta frequency sensorimotor oscillation is observed during periods of rest and is suppressed during tasks activating the sensorimotor cortex (Pfurtscheller, 1981; Arroyo et al., 1993; Salmelin and Hari, 1994; Pfurtscheller et al., 1996; Crone et al., 1998; Andres and Gerloff, 1999; Pfurtscheller and Lopes da Silva, 1999). It is possible that the synchronous 15–30 Hz STN oscillations observed in this study are due to cortical beta oscillation input transmitted via the cortico‐subthalamic pathway.

It has been hypothesized that the cortico‐subthalamic pathway synchronizes oscillatory activity in the basal ganglia (Bergman et al., 1994; Wichmann et al., 1994a; Hurtado et al., 1999; Deuschl et al., 2000; Magill et al., 2000) and that a high level of oscillatory synchronization of the whole basal ganglia in the parkinsonian state underlies the clinical features of Parkinson’s disease (Raz et al., 2001). An association between the cortical beta oscillations and basal ganglia function has previously been demonstrated by Brown and Marsden (1998, 1999), wherein they suggested that akinesia and bradykinesia were due to an inability of the parkinsonian basal ganglia to release cortical elements from alpha (∼10 Hz) and beta oscillatory activity during voluntary movement. Furthermore, it has recently been shown that 15–30 Hz field potential oscillations in the STN can be coherent with cortical electroencephalogram oscillations and that stimulation through DBS contacts with these cortically coherent oscillations produces the most effective motor benefit in patients with Parkinson’s disease (Marsden et al., 2001). These studies suggest that synchronous 15–30 Hz cortical oscillations play a role in parkinsonian pathophysiology by affecting basal ganglia function via the STN.

It should be noted that in the present study and in a previous study by our group (Levy et al., 2000), an inherent limitation of using single unit correlation techniques to examine synchronization between single neurones is that weaker neuronal synchronizations involving a larger population of neurones may be missed. For example, it was demonstrated in Table 1 that several patients without neurone pair synchronization (and without limb tremor) still displayed significant high‐frequency peaks in the local field potentials (albeit at a lower signal‐to‐noise ratio than in tremulous patients). Therefore, it is possible that high‐frequency oscillatory synchronization detected in the local field potentials but not seen between neurone pairs contributes to the non‐tremor symptoms of Parkinson’s disease. It is interesting to note that DBS of the STN that is performed at 15 Hz, which would be expected to synchronously drive STN neurones, worsens parkinsonian disability through an increase in akinesia (Demeret et al., 1999).

Our results also suggest that synchronized high‐frequency oscillations in the STN are more pronounced in the dopamine‐depleted state. We observed that APO decreased the spontaneous discharge of two STN TCs, concurrent with a reduction in limb tremor and a loss of synchronized high‐frequency oscillatory activity. Furthermore, pairs of neurones recorded before APO administration displayed high‐frequency synchronization, but only weak synchrony was observed in pairs of neurones recorded during the APO‐induced ON period. We also observed that levodopa administration reduced high‐frequency oscillations in the field potential of another patient. These results are consistent with the findings of Brown et al. (2001) who demonstrated that 15–30 Hz synchronization between STN and GPi local field potentials is reduced by levodopa. Our results further suggest that 15–30 Hz synchronization in the STN is not dependent on the GPi because these oscillations were detected in the STN in patients with pallidal lesions. The increase in the spontaneous activity of the STN following MPTP‐treatment in monkeys (Bergman et al., 1994) and presumably in parkinsonian patients (Hutchison et al., 1998) could also act to enhance the prevalence of 15–30 Hz oscillations in the STN. It is interesting to note that in the GPi and STN of both MPTP‐treated monkeys and patients with Parkinson’s disease, neurones with oscillatory activity have a higher spontaneous discharge rate than neurones without oscillatory activity (Bergman et al., 1994; Levy et al., 2001). A possible mechanism is that the increase in STN firing rates occurring in Parkinson’s disease has the effect of increasing the maximum oscillatory frequency that can be transmitted by a neuronal spike train. For example, a neurone discharging at a mean firing rate of 50 spikes/s could better carry a 25 Hz oscillation than a neurone discharging at 30 spikes/s. Smaller neuronal populations would be required to express synchronized activity and/or a greater proportion of the neurones in the STN would then present higher frequency oscillations. Dopaminergic therapy has been shown to decrease STN firing rates (Kreiss et al., 1997) and thus could lessen the influence of high‐frequency cortical oscillations on the basal ganglia. These ideas are supported by the demonstration that the incidence of STN neurones with high‐frequency oscillations is reduced following the administration of APO in parkinsonian patients (Levy et al., 2001).

Although oscillatory synchronization between groups of neurones is hypothesized to be necessary for limb tremor (Llinás and Paré, 1995; Bergman et al., 1998a, b), it is presently unclear what mechanisms might contribute to or promote TC synchrony in Parkinson’s disease. Since the GPe‐STN network can maintain low frequency (<2 Hz) oscillatory synchronization that is driven by the STN (Plenz and Kital, 1999), it is feasible that 15–30 Hz in‐phase synchronous activity in the STN is also present in the GPe and that this network behaviour may in turn promote tremor‐related synchronization in the basal ganglia. This idea is supported by the demonstration in tremulous MPTP‐treated monkeys and a patient with Parkinson’s disease that pairs of GPe cells exhibit high‐frequency oscillatory synchronization with phase differences that are centred around 0° (Raz et al., 2000; Levy et al., 2002), similar to our findings in the STN of parkinsonian patients (Levy et al., 2000). If rhythmic oscillatory activity in the STN‐GP network in disease states is driven by the cortex (Magill et al., 2000), the modulation of 15–30 Hz oscillations in the cortex, and hence the STN‐GPe, might underlie some of the clinical features of parkinsonian rest tremor. Event‐related desynchronization and rebound synchronization of 15–30 Hz cortical beta oscillations occurring during tasks activating the sensorimotor cortex (Pfurtscheller, 1981; Salmelin and Hari, 1994; Salmelin et al., 1995; Andrew and Pfurtscheller, 1996; Pfurtscheller et al., 1996; Crone et al., 1998) may account for the common clinical observation that parkinsonian rest tremor shows a decreasing amplitude when a limb is voluntarily activated (Deuschl et al., 1998). Furthermore, synchronous oscillations in the sensorimotor cortex are affected by arousal or attention (Murthy and Fetz, 1996a, b) and may play a role in ‘enhanced tremor’, whereby the amplitude of the parkinsonian rest tremor increases during mental stress (i.e. enhanced tremor can be readily observed by asking patients to perform mental arithmetic) (Deuschl et al., 1998). It is interesting to note that increased coupling between muscles in a tremulous limb is observed when parkinsonian patients perform a mental arithmetic task (Hurtado et al., 2000), suggesting that tremor‐related oscillations in the basal ganglia are more synchronized during this period (Bergman et al., 1998b). In addition, since the same cortical regions that are used for voluntary movement are involved in the pathogenesis of parkinsonian limb tremor (Alberts et al., 1969; Parker et al., 1992; Duffau et al., 1996; Volkmann et al., 1996; Hellwig et al., 2000), it is feasible that limb tremor itself could eventually lead to an instability in the regularity of sensorimotor beta oscillations (Makela et al., 1993) and this mechanism might account for the variability or intermittent nature of parkinsonian tremor observed clinically (Schwab and Cobb, 1939; Scholz and Bacher, 1995) or when recording from individual neurones (Bergman et al., 1998a; Hurtado et al., 1999; Levy et al., 2000; Raz et al., 2000).

In summary, this study proposes that 15–30 Hz cortical beta oscillations gain access to the basal ganglia through the cortico‐subthalamic pathway, possibly via hyperactive STN neurones that are present in the dopamine‐depleted state. These synchronized high‐frequency oscillations would then promote oscillatory synchronization in the basal ganglia and contribute to the symptoms of Parkinson’s disease.


We wish to thank the patients for their participation in this study. Funding was provided by the US National Institutes of Health (NS 40872), Canadian Institute of Health Research (MOP‐42505) and the Parkinson’s Foundation of Canada. A.M.L. is a Canadian Institute of Health Research clinician scientist.


View Abstract