Cortico-cortical coupling in Parkinson`s disease

doi:10.1093/brain/awh480
Brain (2005), 128, 1277–1291
Cortico-cortical coupling in Parkinson’s disease
and its modulation by therapy
Paul Silberstein,1,2 Alek Pogosyan,1 Andrea A. Kühn,1,4 Gary Hotton,3 Stephen Tisch,1,2
Andreas Kupsch,4 Patricia Dowsey-Limousin,1,2 Marwan I. Hariz1,2 and Peter Brown1
1
Sobell Department of Motor Neuroscience and Movement
Disorders and 2Unit of Functional Neurosurgery, Institute
of Neurology, 3Department of Clinical Neurosciences,
Institute of Psychiatry, London, UK and 4Department of
Neurology, Charité Campus Virchow, Humboldt University,
Berlin, Germany
Correspondence to: Professor Peter Brown, Sobell
Department of Motor Neuroscience and Movement
Disorders, Institute of Neurology and National Hospital for
Neurology and Neurosurgery, 2nd floor 8–11 Queen Square,
London WCIN 3BG, UK
E-mail: [email protected]
Summary
The role of changes in inter-regional cortical synchronization in the pathophysiology of Parkinson’s disease and
the mechanism of action of dopaminergic therapy and
high frequency subthalamic nucleus (STN) stimulation
is unclear. We hypothesized that synchronization between
distributed cortical areas would correlate with parkinsonism and that changes in synchronization with treatment
would correlate with improvements in parkinsonism.
To this end, we recorded scalp EEG in parkinsonian
patients off treatment (16 patients, 31 sides) and then
separately during high frequency stimulation (HFS) of
the STN (16 patients, 31 sides) and following drug treatment (12 patients, 24 sides). All recordings were made at
rest to avoid the confounding effects of differences in task
performance. The motor Unified Parkinson’s Disease
Rating Scale (UPDRS) score was determined in each
state. We found that EEG–EEG coherence over 10–
35 Hz correlated with the severity of parkinsonism, and
reductions in cortical coupling over this frequency range
with both L-dopa and STN stimulation correlated with
clinical improvement. These results suggest that both
dopaminergic therapy and STN stimulation may support
the restoration of normal cortico-cortical interactions in
the frequency domain. This mechanistic similarity may
underscore the strong clinical correlation between the
therapeutic effects of these treatment modalities.
Keywords: cortical coupling; Parkinson’s disease; dopamine; STN stimulation
Abbreviations: DBS = deep brain stimulation; HFS = high frequency stimulation; LFP = local field potential;
STN = subthalamic nucleus; tCoh = transformed coherence; UPDRS = Unified Parkinson’s Disease Rating Scale
Received August 17, 2004. Revised February 9, 2005. Accepted February 10, 2005. Advance Access publication
March 17, 2005
Introduction
Recent theories regarding the pathophysiology of Parkinson’s
disease have moved away from the neuronal firing rate-based
explanations encompassed in the model of Albin and DeLong
(Albin et al., 1989; DeLong, 1990) to focus on the importance
of alterations in the temporal patterning of neuronal discharge
in the development of Parkinsonian symptoms (Marsden and
Obeso, 1994; Obeso et al., 1997; Brown and Marsden, 1998).
In particular, recordings in patients with Parkinson’s disease
undergoing functional neurosurgery suggest excessive synchronization of neurons in the subthalamic nucleus (STN) and
globus pallidus. Evidence for this comes from microelectrode
recordings of pairs of units (Hurtado et al., 1999; Levy et al.,
2000, 2002b) and macroelectrode recordings of local field
potentials (LFPs), a surrogate marker of local synchronization
(Brown, 2003). Synchronization is particularly evident in the
beta band from 13 to 30 Hz. This is reduced by treatment with
levodopa (Levy et al., 2000, 2001, 2002a, b; Marsden et al.,
2000; Brown et al., 2001; Cassidy et al., 2002; Silberstein
et al., 2003; Priori et al., 2002, 2004; Williams et al., 2002).
Treatment may in turn be associated with synchronization in
the gamma band or even higher frequencies (Brown et al.,
2001; Cassidy et al., 2002; Williams et al., 2002; Foffani et al.,
2003). These spectral changes in oscillatory activity appear
at least in part to be network phenomena, as evidenced by
the finding of frequency- and dopaminergic state-dependent
coherence between population activity in STN, globus
# The Author (2005). Published by Oxford University Press on behalf of the Guarantors of Brain. All rights reserved. For Permissions, please email: [email protected]
1278
P. Silberstein et al.
pallidus internus and cerebral cortex (Brown et al., 2001;
Williams et al., 2002; Cassidy et al., 2002). Thus abnormal
synchronized neuronal activity in the basal ganglia is linearly
coupled to activity in the cortex. The issue of synchronization
within the cerebral cortex in Parkinson’s disease is a critical
one, as basal ganglia disease can only lead to motor dysfunction through effects on its executive motor projection
sites, the motor areas of the cerebral cortex and brainstem.
Moreover, oscillatory synchronization within and between
cortical areas is increasingly recognized as a key mechanism
in motor organization (Leocani et al., 1997; Farmer, 1998;
Gerloff et al., 1998; Andres et al., 1999; Marsden et al., 2001;
Ohara et al., 2001; Serrien and Brown, 2002, 2003; Serrien
et al., 2003).
Changes in synchronization within and between neuronal
populations in the cerebral cortex are readily determined
through recordings of EEG, and the oscillatory structure of
the synchronization evident in the EEG can be characterized
through spectral analysis (Pfurtscheller and Lopes da Silva,
1999). Increases in oscillatory synchronization within local
cortical neuronal populations are evident as increases in EEG
power, while increases in the synchronization between distributed cortical regions are manifest as cortico-cortical
coherence. That local cortical oscillatory activity is abnormal
in relation to movement in Parkinson’s disease is well
established. The normal suppression (desynchronization) of
mu activity prior to voluntary movement is delayed over
the contralateral sensorimotor cortex in Parkinson’s disease,
a delay that can be partially reversed by acute (Magnani et al.,
2002) or chronic (Defebvre et al., 1998; Devos et al., 2004)
treatment with levodopa or deep brain stimulation (DBS)
(Devos et al., 2004). Similarly, the normal increase (synchronization) of beta activity following voluntary movement
is impaired in Parkinson’s disease (Pfurtscheller et al., 1998)
and restored by levodopa or DBS (Devos et al., 2003b, 2004).
Impairments in the degree of suppression of mu power during
movement have been shown to correlate with bradykinesia
(Brown and Marsden, 1999; Wang et al., 1999). In addition,
levodopa-dependent changes in task-related cortico-cortical
coherence have been reported in Parkinson’s disease (Cassidy
and Brown, 2001).
However, the interpretation of the above studies is
complicated by differences in task performance during movement between treatment states, so that changes in cortical
power and cortico-cortical coherence might relate directly
to treatment or indirectly reflect the change in task performance with treatment. Changes in cortical oscillatory activity at
rest are less ambiguous, and EEG studies over the last few
decades have shown an increased incidence of background
and focal intermittent EEG slowing in Parkinson’s disease
(England et al., 1959; Enge et al., 1966; Yeager et al., 1966;
McPherson, 1970; Wiederholt, 1974; Yaar, 1977). The significant correlation between motor disability and slowing of
the background EEG suggests that at least part of this effect
is related to failure of normal nigrostriatal modulation of
basal ganglia input to cortex (Neufeld et al., 1988). Excessive
synchronization between neurons in the motor cortex has also
been reported at very low frequencies in the 1-methyl-4phenyl-1,2,3,6-tetrahydropyridine (MPTP) primate model
of Parkinson’s disease (Goldberg et al., 2002).
The above studies in resting subjects concentrated on local
synchronization of cortical activities evident in EEG power
and upon synchronization at low frequency. However,
important changes may also occur in the pattern of synchronization (cortico-cortical coherence) across distributed areas
of cortex. This is likely to be particularly true in the beta
and gamma bands, given the importance of cortico-cortical
coherence at these frequencies in motor organization and
the evidence that basal ganglia activity is preferentially synchronized in these bands. The role of changes in cortical
synchronization at higher frequencies in the pathophysiology
of Parkinson’s disease and the effect of DBS and dopaminergic therapy upon any such changes is hitherto unexplored.
We hypothesized that coherent activity between distributed
populations of cortical neurons would correlate with parkinsonism and that changes in cortico-cortical coherence with
treatment correlate with improvements in parkinsonism.
To this end, we recorded scalp EEG in resting patients
with Parkinson’s disease with chronically implanted STN
electrodes while without treatment, during therapeutic high
frequency stimulation (HFS) of the STN and following treatment with their usual oral antiparkinsonian medication.
We determined whether oscillatory cortico-cortical coupling
correlated with the severity of Parkinsonism in the untreated
state and whether STN HFS and dopaminergic medicationinduced reductions in cortical coupling correlated with reduction in motor impairment.
Material and methods
Patients and surgery
All patients (n = 16, mean age 56 6 7 years, range 42–66) participated with informed consent and the permission of the local ethics
committees. Their clinical details are summarized in Table 1. None
of the patients was demented as determined by pre-surgical neuropsychological assessment. Implantation of bilateral STN macroelectrodes was performed simultaneously in 12 subjects and
sequentially (separated by 6 months) in four subjects (cases 2, 3,
5 and 8) for treatment of severe Parkinson’s disease. The macroelectrode used was model 3389 (Medtronic Neurological Division,
MN) with four platinum–iridium cylindrical surfaces (1.27 mm diameter and 1.5 mm length) and a centre-to-centre separation of 2 mm.
Contact 0 was the lowermost and contact 3 was the uppermost. The
intended coordinates at the tip of contact 0 were the midpoint of
the STN as determined by direct visualization and reference to the
anterior border of the red nucleus on preoperative MRI (Bejjani et al.,
2000). These coordinates corresponded closely with stereotactic
coordinates determined from the Schaltenberg and Wahren atlas
(Schaltenbrand and Wahren, 1977): 10–12 mm from the midline,
0–2 mm behind the midcommissural point and 4–5 mm below the
AC–PC line. Intraoperative electrode localization was tested by
macrostimulation in all patients. No microelectrode recordings
were made. Macroelectrodes were connected to a battery-operated
programmable pulse generator [Itrel II (4 subjects) or Kinetra 7428
55 F
60 M
51 M
66 M
64 M
61 M
59 M
3
4
5
6
7
8
9
10
12
20
16
11
15
16
16
PD on–off
fluctuations;
dyskinesias
PD on–off
fluctuations;
dyskinesias
PD on–off
fluctuations;
tremor
PD on–off
fluctuations;
freezing
PD on–off
fluctuations,
dyskinesias
PD on–off
fluctuations;
dyskinesias
PD on–off
fluctuations;
dyskinesias
(R) 3–1+/3.0/60/185;
(L) 3-C+ 3.2/60/185
(R) 5-C+/3.5/60/185;
(L) 2-C+/3.0/60/185
(R) 0–1+/4.5/90/185;
(L) 2-C+/2.8/90/185
King’s
College
Kinetra
Berlin;
Kinetra
NHNN
Kinetra
(R) 1-C+ 4.3/60/130;
(L) 5-C+ 4.4/60/130
(R) 1–2+ 4.0/60/130;
(L) 5–6+ 4.0/60/130
(R) 5-C+/2.3/60/130;
(L) 2-C+/1.5/60/130
Sweden; (R) 2-C+/2.8/60/185;
bilateral
(L) 0–1+/3.6/60/145
Itrel II;
NHNN;
bilateral
Itrel II
NHNN;
Kinetra
NHNN;
bilateral
Itrel II
NHNN;
bilateral
Itrel II
(R) 3-C+/2.6/60/130;
(L) 3–2+/3.6/60/145
62 M
2
NHNN;
Kinetra
(R) 4-C+/3.6/60/130;
(L) 0-C+/1.3/60/130
PD on–off
fluctuations;
freezing;
dyskinesia
PD on–off
fluctuations;
freezing
42 F
1
6
Usual stimulation
parameters:
contact/voltage/pulse
width/frequency
Case Age
Disease Predominant Surgical
(years)/ duration symptoms
centre/
(years) preoperatively stimulator
sex
type
21/29
28/46
42/79
20/35
14/61
Motor
UPDRS
part III
on/off STN
stimulation
postoperatively
36/63
24/40
5/41
25/46
26/42
5/33
Not available 38/51
23/49
15/61
9/42
21/39
13/61
Motor
UPDRS
part III
on/off drugs
preoperatively
Table 1 Patient clinical details and stimulation parameters used for the study
(R) 1–3+ 6/60/130;
(L) 5–6+ 6/60/130
(R) 1–2+ 4.0/60/130;
(L) 5–6+ 4.0/60/130
(R) 5–6+/2.9/60/130;
(L) 2–1+/2.0/60/130
(R) 2–0+ 3.6/60/185;
(L) 1–3+ 4.6/60/145
(R) 3–1+/3.0/60/185;
(L) 3–2+ 4.0/60/185
(R) 5–6+/4.4/60/185;
(L) 2–1+/3.7/60/185
(R) 0–1+/4.5/90/185;
(L) 2–1+/3.6/90/185
(R) 3–2+/3.0/60/130;
(L) electrode
misplaced on MRI
(R) 4–5+/4.6/60/130;
(L) 0–1+/1.7/60/130
Stimulation
parameters
used for study:
contact/voltage/pulse
width/frequency
(L) 25.5;
(R) 20
(L) 20;
(R) 15
(L) 10;
(R) 14
(L) 18;
(R) 13
(L) 19;
(R) 17
(L) 23;
(R) 25
(L) 18.5;
(R) 23
(L) 19
(L) 29.5;
(R) 22.5
UPDRS
hemibody
score
off-stim
off-med:
study
date
(L) 5.5;
(R) 3
(L) 7;
(R) 2
(L) 2;
(R) 2
(L) 4;
(R) 4
(L) 6;
(R) 3
(L) 9;
(R) 9
(L) 5.5;
(R) 5.5
(L) 6
(L) 9.5;
(R) 5
UPDRS
hemibody
tremor
score
off-stim
off-med:
study date
(L) 17.5;
(R) 14
(L) 10.5;
(R) 9
(L) 5.5;
(R)10.5
(L) 15;
(R) 9
(L) 10;
(R) 10
(L) 14;
(R) 12.5
(L) 8;
(R) 9.5
(L) 13
(L) 7;
(R) 13
UPDRS
hemibody
score with
study
stimulation
parameters
Dose failure
(L) 8.5;
(R) 8.5
(L) 5.5;
(R) 10
(L) 7;
(R) 4
Dose failure
(L) 8;
(R) 7
(L) 12;
(R) 12.5
(L) 12
L-dopa
800 mg
L-dopa 250 mg;
pergolide 750 mg;
epilim chrono
500 mg BD
L-dopa 400 mg;
cabergoline 2 mg BD;
amitryptiline
35 mg nocte
L-dopa 250 mg;
apomorphine
2.5–7.5 mg prn
L-dopa 400 mg
L-dopa 400 mg;
entacapone 1600 mg;
amantadine 100 mg;
benzhexol 4 mg;
doxazosin 4 mg
L-dopa 350 mg;
ropinirole 15 mg;
amantadine 200 mg;
amitryptiline 50 mg
L-dopa 200 mg
No medication No medication
given
UPDRS
Medications
hemibody
daily dose (mg)
scores on
medication off
stimulation
on study date
Cortico-cortical coupling in Parkinson’s disease
1279
53 F
63 M
59 F
46 M
56 M
12
13
14
15
16
PD on–off
fluctuations
PD on–off
fluctuations;
dyskinesias
PD on–off
fluctuations;
dyskinesias
PD on–off
fluctuations;
dyskinesias
NHNN
Kinetra
NHNN;
Kinetra
Berlin;
Kinetra
NHNN;
Kinetra
27/58
14/47
(R) 6-C+ 3.0/60/130;
(L) 1-C+ 2.8/60/130
(R) 5-C+ 3.8/60/130;
(L) 1-C+ 3.9/60/130
2/45
8/63
Stimulation
parameters
used for study:
contact/voltage/pulse
width/frequency
(L) 18.5;
(R) 10.5
(L) 11.5;
(R) 13.5
(L) 22;
(R) 17
UPDRS
hemibody
score
off-stim
off-med:
study
date
Not available (R) 6–4+ 4.3/60/130;
(L) 1–3+ 4.0/60/130
Not available (R) 5–7+ 5.0/60/130;
(L) 1–0+ 5.0/60/130
(L) 15.5;
(R) 22
(L) 24.5;
(R) 19
(R) 0–1+ 3.2/90/130; (L) 17;
(L) 5–6+ 2.7/90/130 (R) 13.5
Not available (R) 7–5+ 4.2/60/130; (L) 16;
(L) 3–1+ 5.3/60/130 (R) 20
17/26
(R) 1–3+ 4.0/60/180;
(L) 5–4+ 4.5/60/180
15/30
(R) 1–3+ 3.6/60/130;
(L) 5–7+ 4.0/60/130
Not available (R) 1–0+ 3.0/60/130;
(L) 6–5+ 4.1/60/130
25/51
Motor
UPDRS
part III
on/off STN
stimulation
postoperatively
(L) 5;
(R) 10.5
(L) 3;
(R) 3
(L) 1;
(R) 0.5
(L) 5.5;
(R) 8
(L) 5;
(R) 3
(L) 1;
(R) 1
(L) 2.5;
(R) 1.5
UPDRS
hemibody
tremor
score
off-stim
off-med:
study date
(L) 6.5;
(R) 10
(L) 17;
(R) 10.5
(L) 12;
(R) 8
(L) 9.5;
(R) 7
(L) 12;
(R) 4.5
(L) 8;
(R) 9
(L) 15.5;
(R) 9.5
UPDRS
hemibody
score with
study
stimulation
parameters
NHNN = National Hospital for Neurology and Neurosurgery, London; PD = Parkinson’s disease. The same surgeon as NHNN patients.
20
13
21
18
9
8
13/30
Motor
UPDRS
part III
on/off drugs
preoperatively
(R) 0-C+ 2.5/90/130; 9/25
(L) 5-C+ 2.0/90/130
(R) 7-C+ 3.2V/60/130; 15/52
(L) 3-C+ 4.0V/60/130
51 M
11
Berlin;
Kinetra
Berlin;
Kinetra
Berlin;
Kinetra
(R) 1-C+ 3.2/60/180;
(L) 5-C+ 3.2/60/180
(R) 1-C+ 2.8/60/130;
(L) 5-C+ 3.0/60/130
(R) 1-C+ 1.7/60/130;
(L) 6-C+ 2.2/60/130
PD on–off
fluctuations
PD on–off
fluctuations
PD on–off
fluctuations
51 M
10
18
Usual stimulation
parameters:
contact/voltage/pulse
width/frequency
Case Age
Disease Predominant Surgical
(years)/ duration symptoms
centre/
sex
(years) preoperatively stimulator
type
Table 1 Continued
(L) 7.5;
(R) 8.5
(L) 3.5;
(R) 1.5
(L) 7.5
(R) 5.5
Dose Failure
(L) 6.5;
(R) 4
(L) 8;
(R) 6
(L) 12.5;
(R) 9.5
L-dopa 600 mg;
cabergoline 4 mg;
propranolol 80 mg;
disopyramide 200 mg;
oxybutinin 7.5 mg;
simvastatin 10 mg
L-dopa 700 mg;
entacapone 1200 mg;
pergolide 9 mg;
amantadine 300 mg
L-dopa 1650 mg;
entacapone 1600 mg
L-dopa 125 mg;
cabergoline 3 mg;
amantadine 200 mg
L-dopa 500 mg;
cabergoline 3 mg
L-dopa 800 mg;
pergolide
L-dopa 300 mg
UPDRS
Medications
hemibody
daily dose (mg)
scores on
medication off
stimulation
on study date
1280
P. Silberstein et al.
Cortico-cortical coupling in Parkinson’s disease
(12 subjects), Medtronic]. All patients received postoperative
imaging that consisted of MRI (n = 15) or CT (n = 1, case 3).
Postoperative imaging was consistent with the placement of at
least one macroelectrode contact in the STN, except for the left
side in case 2 which was excluded from further analysis. Patients
had an overall improvement of 47 6 5% in Unified Parkinson’s
Disease Rating Scale (UPDRS) motor score during continuous
HFS off medication rated at least 6 months after surgery, which
further supports a satisfactory placement of the macroelectrodes
(subjects 1–11 and 13). Comparable postoperative UPDRS scores
were not yet available in subjects 12, 14, 15 and 16. In these subjects,
however, the mean improvement in hemibody score with contralateral stimulation using the study stimulation parameters (see Table 1)
was 46 6 6%, also supporting satisfactory electrode placement.
Study protocol
Patients were studied after overnight withdrawal of medications at
least 2 months (range 2–56 months) postoperatively. Subjects were
seated and recorded at rest. They were asked to maintain visual
fixation on a coloured dot on a computer monitor. Scalp EEG at
rest and hemibody UPDRS motor assessments were recorded under
the following conditions: (i) off medication off stimulation (off-med
off-stim); (ii) off medication on left STN stimulation; (iii) off
medication on right STN stimulation; and (iv) on medication off
stimulation (on-med off-stim). Half points were used to increase
the sensitivity of the UPDRS score. This method has been used
previously when assessing the efficacy of STN DBS (Limousin
et al., 1995).
The off-med off-stim condition was always recorded first. Both
STN stimulators were turned off for a minimum of 10 min prior to
this recording. Subsequent recordings in the off medication state
were performed in randomized order. Recordings lasted 120 s and
were performed twice in each condition, separated by 5 min rest.
We waited a minimum of 5 min between each stimulator change.
STN stimulation was always performed bipolarly as monopolar STN
stimulation led to significant artefact in the scalp EEG. In patients
who were usually stimulated bipolarly, usual stimulation parameters
were used. In patients who were stimulated monopolarly, the
maximally effective bipolar pair was determined on clinical
grounds, and this contact pair was used for subsequent EEG and
UPDRS recordings. In the latter case, stimulation amplitude had
to be increased by 25–30% in order to achieve a similar clinical
effect with bipolar stimulation. Usual stimulation pulse width and
frequency were not altered in any subjects for the recordings (see
Table 1).
Patients were then instructed to take their usual morning dopaminergic medications (L-dopa 6 dopamine agonists). Patients were
examined at intervals and asked to report when they felt the medications were acting. After 45–60 min, EEG was recorded (120 s) in
the on medication off stimulation state on two occasions separated
by at least 10 min. Hemibody UPDRS part III scores were also
determined in this state. Three out of the 16 subjects (cases 5, 9
and 14) experienced medication dose failure during the study. One
subject (case 1) had stopped all medications postoperatively and was
therefore not given L-dopa during the study. Consequently, on medication recordings could only be performed in 12 subjects.
EEG recordings and data analysis
Scalp EEG was recorded according to the 10 : 20 international
system and referenced to linked ears (Fig. 1A). Signals were
1281
amplified, pass band filtered between 0.25 and 90 Hz and sampled
at 184 Hz (Biopotential Analyzer Diana, St Petersburg, Russia). An
example of the raw EEG at rest and during left sided HFS is shown
for case 8 in Fig. 1B. Note that no stimulus artefact is seen during
stimulation.
EEG was examined off-line, and eye movement-, blink- and
EMG-contaminated sections removed as far as possible by visual
inspection before frequency analysis was performed. The two
records performed in any given state were concatenated. After artefact rejection, the average total length of EEG records was 185 6 8 s.
Nineteen electrodes were chosen for further analysis to facilitate a
comparison of the topography of spectral changes with stimulation
and L-dopa therapy (Fp1, Fp2, F7, F3, Fz, F4, F8, T3, C3, CZ, C4,
T4, T5, P3, PZ, P4, T6, O1 and O2; see Fig. 1A). Spectral analysis
was performed using the MATLAB function ‘fft’. This function
employs a high speed radix-2 fast Fourier transform algorithm if
the data length is a power of two, otherwise a slower mixed radix
algorithm if it is not. Coherence is a measure of the linear association
(correlation) between two signals across frequencies. It is a bounded
measure taking values from 0 to 1, where 0 indicates that there is no
linear association (i.e. that one process is of no use in linearly predicting another process) and 1 indicates a perfect linear association.
The coherence |Rab(l)|2 was calculated as described previously
(Halliday et al., 1995) by using the formula: |Rab(l)|2 = | fab(l)|2/
aa(l) fbb(l). In this equation, f characterizes the spectral estimate of
two EEG signals a and b for a given frequency (l). The numerator
includes the cross-spectrum for a and b ( fab), whereas the denominator includes the autospectra for a ( faa) and b ( fbb). Spectra were
estimated by dividing the data epochs into a number of disjoint
sections of 1 s duration. Frequency resolution was 1 Hz. Data
were Hanning-windowed to control spectral leakage. Figure 1C
shows the coherence between C3–C4 at rest and during HFS in
case 8. The square root of the coherence was normalized using a
Fisher transform, and the variance of spectral power estimates
stabilized by logarithmic transformation (Halliday et al., 1995).
For each subject, data were analysed in four conditions: (i) off
medication off stimulation; (ii) off medication left STN stimulation;
(iii) off medication right STN stimulation; and (iv) on medication off
stimulation. Transformed coherence (tCoh) and log power changes
were examined across five frequency bands: 3–7, 8–12, 13–24,
25–45 and 60–80 Hz. Synchronization in the first four frequency
bands is evident in basal ganglia LFP recordings made in the
untreated state in patients with Parkinson’s disease (Brown et al.,
2001; Cassidy et al., 2002; Levy et al., 2002a; Priori et al., 2002;
Williams et al., 2002). The 60–80 Hz band was chosen to match the
frequency of oscillatory synchronization sometimes seen in the basal
ganglia following L-dopa (Brown et al., 2001; Cassidy et al., 2002;
Williams et al., 2002).
For each subject, the stimulation or drug-related changes in tCoh
or log power in each frequency band for each electrode pair or single
electrode were calculated by subtracting the tCoh or log power in the
off-med off-stim condition from the tCoh or log power in the stimulation or drug condition. To determine the clinical significance of
the frequency- and topography-dependent changes in scalp EEG
coupling, three correlations were made with the UPDRS hemibody
scores. (i) Correlation between off-med off-stim tCoh and off-medoff stim UPDRS scores: for each electrode coupling, the tCoh in the
off-med off-stim state was correlated with the contralateral hemibody UPDRS scores (UPDRS part III items 18–26) in this condition.
(ii) Correlation of change in tCoh off-med on-stim and difference in
hemibody UPDRS: for each channel pair, the change in tCoh with
1282
P. Silberstein et al.
A
C
1
0.9
0.8
Coherence
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0
10
20
30
40
50
60
50
60
70
80
Frequency
B
D
Off Med off Stim
250
C4
200
Power (a.u.)
C3
Left HFS STN
C3
150
100
50
C4
50µV
0
1 second
0
10
20
30
40
70
80
Frequency
Fig. 1 (A) EEG channels used for analysis. Electrodes within the dotted line are ‘central’ channels and electrodes outside the dotted line are
‘peripheral’ channels. (B) Examples of EEG data: raw EEG data recorded from C3 and C4 at rest and during high frequency stimulation
(HFS) (referenced to linked ears). Note that there is no stimulus artefact during HFS of the STN region. (C) Coherence spectrum (C3-C4)
at rest (grey line) and during HFS (black line). Coherence between 45 and 55 Hz is unreliable due to mains artefact and has been omitted.
Note that coherence is generally reduced during HFS. (D) Power spectrum at C3 off-med off-stim (grey line) and off-med on left-stim
(black line). Power between 45 and 55 Hz is unreliable due to mains noise and has been omitted. Power spectra at C4 were similar (data not
shown). Note that power is also reduced at these electrodes, and therefore reductions in coherence are related to an absolute reduction in
coupling related to stimulation between these electrodes.
STN stimulation (off-med on-stim–off-med off-stim) was correlated
with the difference in contralateral hemibody UPDRS score with
STN stimulation (off-med off-stim–off-med on-stim). (iii) Correlation of change in tCoh on med off-stim and difference in UPDRS: for
each channel pair, the change in tCoh with medication (on-med
off-stim–off-med off-stim) was correlated with the difference in
hemibody scores on medication (off-med off-stim–on-med off-stim).
To increase the statistical power of our study to detect correlations, EEG coherence values were correlated with both hemibody
scores, but only after inverting right intra-hemispheric coherences
about the midline so that F4–P4 became F3–P3 etc., while inverting
the corresponding left hemibody score. Thus left and inverted right
intra-hemispheric coherences were correlated with right and inverted
left hemibody scores to give EEG correlations with ‘contralateral’
scores. The remaining midline and interhemispheric coherence values were also correlated with both hemibody scores. Thus for a
significant correlation between an interhemispheric coherence,
such as F4–F3, and hemibody score to occur, the coherence values
were likely to have a similar proportional relationship with the
left and right hemibody scores across patients. Equally, for a
significant correlation between an intrahemispheric coherence,
such as F4/F3–C4/C3 (e.g. contralateral fronto-central coherence),
and hemibody score to occur, then F3–C3 would tend to share the
same proportional relationship with right hemibody scores across
patients as (inverted) F4–C4 would have with (inverted) left
hemibody scores. Note that this approach neglects any effects of
hemispheric dominance.
To determine whether there was a topographic distribution of
correlations by frequency band, we performed similar correlations
to those outlined above, but only for nine ‘central’ EEG channels
lying over frontal and parietal regions (F3, FZ, F4, C3, CZ, C4, P3,
PZ and P4), and compared these results with the frequencydependent changes in tCoh and log power in 10 ‘peripheral channels’ (Fp1, Fp2, F8, T4, T6, O1, O2, T3, T5 and F7) (see Fig. 1A). In
those frequency bands with at least five significant correlations,
central and peripheral correlations were Fisher transformed and
separately averaged in each subject. These were then analysed
using two-tailed Student’s t tests for populations of unequal variance. Significance levels of t tests were Bonferroni corrected for
multiple comparisons.
Cortico-cortical coupling in Parkinson’s disease
All correlations were performed by frequency band using
Spearman’s correlation test (SPSS for Windows version 11,
SSPS Inc, Chicago, IL) and Fisher transformed prior to statistical
evaluation. A P value <0.005 was considered significant for bandwise correlations and is the threshold used in the topographic maps
of coherence. An asymmetry index was also estimated in those
frequency bands with at least five significant correlations to determine whether correlations were lateralized. The mean Fishertransformed correlation coefficients for each connection involving
F3, C3 and P3 with themselves or with Fz, Cz and Pz (total possible
12) were compared with the mean transformed correlation coefficients for all connections involving F4, C4 and P4 with themselves
or with Fz, Cz and Pz (total possible 12) and an unpaired t test
performed across subjects in each frequency band. Significance
levels were Bonferroni corrected for multiple comparisons.
The division of coherent frequencies into specific frequency
bands, although generally supported in the literature, inevitably
uses somewhat arbitrary definitions. Accordingly, we also examined
the correlations detailed above per 1 Hz frequency bin, rather than
within frequency bands. To this end, we plotted significantly coherent electrode connections as a percentage of the total number of
possible connections, with significance defined here as correlations
with P < 0.005. Thus only 0.25% positive and 0.25% negative correlations were to be expected to arise by chance at a given frequency.
Correlation between spectral effects of STN
stimulation and dopaminergic medication
In addition, we determined if there was any correlation between the
effects of STN stimulation and dopaminergic medications on scalp
EEG coherence in terms of the topography in the different frequency
bands. Only subjects with satisfactory bilateral STN macroelectrodes who also experienced an L-dopa response on the date
of the study were included for this analysis (n = 11; patients 3, 4,
6, 7, 8, 10, 11, 12, 13, 15 and 16). Here, we determined the average
change in coherence with left and right STN stimulation in the offmed off-stim condition at each EEG channel pair in each frequency
band: (left STN stim – off-med off-stim) + (right STN stim – off-med
off-stim)/2. This value was then averaged across subjects and
correlated with the average change in coherence with medication
at each contact pair (on-med off-stim–off-med off-stim) for the
corresponding frequency band.
Results
Correlations between transformed coherence
and motor state off medication off stimulation
In each frequency band, the tCoh off-med off-stim for each
electrode pair was correlated with the off treatment contralateral hemibody UPDRS scores across subjects. Here we found
multiple positive correlations, which almost exclusively
occurred within the 13–24 Hz band (Fig. 2A). Significant
positive correlations were seen within and between hemispheres and tended to cluster around the central region,
with the exception of some correlations that bridged central
(including frontal electrodes) and occipital electrodes. Positive correlations indicated that motor difficulties were greater
the higher the cortico-cortical coherence at rest. The average
Fisher-transformed r was significantly higher in correlations
1283
involving the central than peripheral electrodes in the
13–24 Hz band (P < 0.005). No significant lateralization
was present as determined by the asymmetry index.
The percentages of significant correlations between tCoh
off-med off-stim at each individual frequency and off treatment contralateral hemibody UPDRS scores are shown in
Fig. 2B. This confirmed that virtually all correlations were
positive and most occurred over 10–36 Hz. In addition, note
that the percentage of significant correlations within the nine
central electrodes exceeded that within the 10 peripheral
electrodes at almost all individual frequencies (Fig. 2B).
The central predominance of the positive correlations with
UPDRS motor scores argues against the spurious correlation
of EEG parameters and motor state through tremor- (and its
harmonics) or rigidity-related EMG contamination of EEG.
Such contamination would be expected to have been much
more prominent in the more peripheral EEG electrodes. The
number of significant correlations within each hemisphere at
each individual frequency did not show any asymmetry (data
not shown).
Note that in the above, tCoh off-med off-stim for each
electrode pair was correlated with the off treatment contralateral hemibody UPDRS. As tCoh was absolute and not the
result of a subtraction between two states (see below), it will
include the influence of both physiological cortico-cortical
coupling and volume conduction. There is, however, no
plausible reason for expecting volume conduction to differ
in proportion to the contralateral hemibody UPDRS. Nevertheless, we repeated the above analysis for the tCoh off-med
off-stim between all possible combinations of central bipolar
electrode pairs rather than between all possible combinations
of central monopolar electrodes. Correlating the tCohbipolar at
each individual frequency with off treatment contralateral
hemibody UPDRS, we again found that virtually all correlations (thresholded at P < 0.005, as before) were positive and
occurred over 10–30 Hz (95 positive correlations and 18
negative correlations over this band, x2 test P < 0.0001).
Correlations between coherence and
motor state during STN HFS
In each frequency band, the change in tCoh related to STN
stimulation (off-med on-stim–off-med off-stim) in each electrode pair was correlated with the difference in hemibody
UPDRS score (off-med off-stim–off-med on-stim) across
subjects. Here we found only significant negative correlations
so that the greater the reduction in tCoh, the greater was the
difference in UPDRS hemibody score and hence improvement in parkinsonism. Multiple correlations were found in all
but the highest frequency band (Fig. 3A). In the 3–7 Hz band,
significant correlations involved predominantly peripheral or
peripheral to central EEG channel pairs, so that it is possible
that the loss of rest tremor contributed to the correlation
through volume conduction of EMG to the EEG electrodes
or the induction of movement artefact. In contrast, significant
1284
P. Silberstein et al.
A
3-7Hz
8-12Hz
25-45Hz
13-24Hz
60-80Hz
Significant positive correlations
Significant negative correlations
Percent significant correlations
B
80
70
60
50
40
30
20
10
0
0
3
6
9
12
15
18
21
24
27
30
33
36
39
42
45
48
51
54
57
60
63
66
69
72
75
78
Frequency
All positive correlations
Central positive correlations
Peripheral positive correlations
Fig. 2 Off medication off stimulation correlations between EEG and motor state. (A) Maps denote tCoh in EEG channel pairs that
correlated significantly (P < 0.005) with contralateral hemibody UPDRS motor scores at rest. Most correlations are positive (red in A) and
occur in the 13–24 Hz band. (B) Each line graph refers to the percentage of significant correlations between tCoh and contralateral
hemibody UPDRS motor scores at rest (P < 0.005) across all (total possible 171), central (total possible 36) or peripheral channel pairs
(total possible 45 electrodes) at 1 Hz frequency intervals. Most correlations are preferentially centrally distributed and occur
from 10 to 32 Hz.
correlations at higher frequencies were preferentially distributed over central EEG channel pairs, and are therefore
unlikely to reflect the effects of tremor or its harmonics.
This distribution was confirmed by a comparison of the
average Fisher-transformed Spearman’s r for all connections
between the nine central electrodes with that for all connections involving the 10 peripheral electrodes. The average
Fisher-transformed r was higher in correlations involving
the central electrodes in 8–12 (P < 0.001), 13–24 (P < 0.001)
and 25–45 Hz (P < 0.001) frequency bands. The one exception to this general rule was the presence of some correlations
involving connections bridging fronto-central and occipital
electrodes.
We also examined the percentage of significant correlations between tCoh related to STN stimulation (off-med
on-stim–off-med off-stim) at each individual frequency
with the difference in hemibody UPDRS (off-med offstim–off-med on-stim) (Fig. 3B). This confirmed that virtually all correlations were negative and most occurred over
10–32 Hz. The percentage of significant correlations
within the nine central electrodes exceeded that within the
10 peripheral electrodes except at tremor-related frequencies
(Fig. 3B). Lateralization was only found in the upper beta
band (P = 0.02). Here the asymmetry index indicated that
there was a greater (negative) correlation between tCoh over
the hemisphere ipsilateral to stimulation with the contralateral hemibody UPDRS than correlation between tCoh over
the hemisphere contralateral to STN stimulation and contralateral (to STN stimulation) hemibody UPDRS.
Correlations between transformed coherence
and motor state after levodopa
In each frequency band, the change in coherence with medication (on-med off-stim–off-med off-stim) in each electrode
pair was correlated with the averaged right and left hemibody
UPDRS scores (off-med off-stim–on-med off-stim) across
subjects. Again we found only significant negative correlations so that the greater the reduction in tCoh, the greater was
the difference in UPDRS hemibody score and hence improvement in parkinsonism. Multiple correlations were found, and,
unlike STN HFS or off-med off-stim correlations, included
the highest frequency bands (Fig. 4A). In the 3–7 Hz band,
significant correlations involved predominantly peripheral
Cortico-cortical coupling in Parkinson’s disease
1285
A
3-7Hz
8-12Hz
25-45Hz
13-24Hz
60-80Hz
Significant negative correlations
Percent significant correlations
B
80
70
60
50
40
30
20
10
0
0
3
6
9
12
15
18
21
24
27
All negative correlations
30
33
36
39
42
45
48
Frequency
Central negative correlations
51
54
57
60
63
66
69
72
75
78
Peripheral negative correlations
Fig. 3 Off medication on stimulation correlations between EEG coherence and contralateral summed UPDRS part III scores normalized to
left stimulation and right limbs. (A) Maps denote tCoh changes in EEG channel pairs that correlated significantly (P < 0.005) with
contralateral hemibody UPDRS motor scores at rest. Most correlations are negative (blue in A) and are concentrated in the 13–24 Hz band.
(B) Each line graph refers to the percentage of significant correlations between tCoh changes and contralateral hemibody UPDRS
motor scores at rest (P < 0.005) across all (total possible 171), central (total possible 36) or peripheral channel pairs (total possible
45 electrodes) at 1 Hz frequency intervals. Correlations are preferentially centrally distributed and concentrated from 10 to 32 Hz.
or peripheral to central EEG channel pairs, so, as for STN
stimulation, it is possible that the loss of rest tremor contributed to the correlation through volume conduction of EMG to
the EEG electrodes or the induction of movement artefact. In
contrast, significant correlations at higher frequencies were
preferentially distributed over central EEG channel pairs
(with the exception of some correlations involving connections between occipital and fronto-central electrodes). This
was confirmed by a comparison of the average Fishertransformed Spearman’s r for all connections between the
nine central electrodes with that for all connections involving
the 10 peripheral electrodes. The average Fisher-transformed
r was higher in correlations involving the central electrodes
in the 8–12 (P < 0.001), 13–24 (P < 0.001), 25–45 (P < 0.001)
and 60–80 Hz (P < 0.001) frequency bands. No significant
lateralization was determined by the asymmetry index.
We also examined the percentage of significant correlations with medication (on-med off-stim–off-med off-stim) at
each individual frequency with the difference in hemibody
UPDRS (off-med off-stim–on-med off-stim) (Fig. 4B). This
confirmed that virtually all correlations were negative.
Although correlations predominated over a similar band
as with correlations in the untreated and HFS states
(i.e. 10–35 Hz), less frequent correlations were seen
additionally at higher frequencies following levodopa.
The percentage of significant correlations within the nine
central electrodes exceeded those within the 10 peripheral
electrodes except at tremor-related frequencies (Fig. 4B).
The presence of correlations at higher frequencies than
with STN HFS raises the question of whether these correlations might be spurious and arise through dyskinesia-related
EMG contamination of EEG following drug treatment. This
seems unlikely as this would have caused positive rather than
negative correlations and would have failed to be centrally
predominating.
Correlation of tCoh and power changes
Power was assessed primarily so as to ensure that changes in
coherence were not the result of modulations of non-linearly
related frequency components (Florian et al., 1998). Coherence denotes that proportion of a pair of signals that covaries
with respect to phase and amplitude at a given frequency.
Thus, increases in activities that do not covary may lead to
1286
P. Silberstein et al.
A
3-7Hz
8-12Hz
25-45Hz
13-24Hz
60-80Hz
Significant negative correlations
Percent significant correlations
B
80
70
60
50
40
30
20
10
0
0
3
6
9
12
15
18
21
24
27
30
33
36
39
42
45
48
51
54
57
60
63
66
69
72
75
78
Frequency
All negative correlations
Central negative correlations
Peripheral negative correlations
Fig. 4 On medication off stimulation correlations between EEG coherence and contralateral summed UPDRS part III scores. (A) Maps
denote tCoh changes in EEG channel pairs that correlated significantly (P < 0.005) with contralateral hemibody UPDRS motor scores
at rest. Most correlations are negative (blue) and are concentrated in the lower and upper beta frequency bands. (B) Each line graph
refers to the percentage of significant correlations between tCoh changes and contralateral hemibody UPDRS motor scores (P < 0.005)
across all (total possible 171), central (total possible 36) or peripheral channel pairs (total possible 45 electrodes) at 1 Hz frequency
intervals. Most correlations are preferentially centrally distributed but involve more frequencies than STN HFS (see B).
reductions in coherence, even though the absolute degree
of coupling between areas may not have changed. Such a
relationship can be suspected when coherence and power
changes occur in opposite directions, i.e. are negatively correlated. Figure 1C and D shows that this was not the case in
example raw spectra: power and coherence both dropped with
stimulation. In addition, across patients, therapy-induced
changes in the log power averaged over all electrodes failed
to correlate or showed a positive correlation with therapyinduced changes in tCoh averaged over all connections
(Table 2).
Table 2 Correlation of therapy-induced changes in
log power versus therapy induced changes in tCoh
by frequency band
Spearman’s r
P-value
Correlation of stimulation- versus
medication-related change in tCoh
STN stimulation correlations (n = 30)
3–7 Hz
8–12 Hz
13–24 Hz
25–45 Hz
60–80 Hz
Medication correlations (n = 12)
3–7 Hz
8–12 Hz
13–24 Hz
25–45 Hz
60–80 Hz
In each frequency band, we correlated the change in tCoh
by EEG electrode pairs in the STN stimulation condition
with the change in tCoh in the medication condition by electrode pair (see Material and methods for details). Here we
found positive correlations in all five frequency bands: 3–7
(r = 0.629, P < 0.001), 8–12 (r = 0.694, P < 0.001), 13–24
(r = 0.714, P < 0.001), 24–45 (r = 0.729, P < 0.001) and 60–
80 Hz (r = 0.552, P < 0.001). This confirmed a strong correlation in the frequency band-dependent topographic effects of
both therapeutic modalities on cortical coupling (Fig. 5).
0.650
0.732
0.530
0.121
0.269
<0.005
<0.005
0.015
NS
NS
0.734
0.704
0.685
0.692
0.210
0.035
NS
NS
NS
NS
Cortico-cortical coupling in Parkinson’s disease
1287
tCoh off med on stim – tCoh off med off stim
.1
0.0
-.1
-.2
-.2
-.1
0.0
.1
tCoh on med off stim – tCoh off med off stim
Fig. 5 Correlation of change in tCoh by electrode pair in the off medication on stimulation condition with change in tCoh by electrode pair
in the on medication off stimulation condition in the 13–24 Hz band (r 2 = 0.714, P < 0.001).
Discussion
We have demonstrated that EEG–EEG coherence over 10–
35 Hz correlates with severity of parkinsonism in untreated
Parkinson’s disease. The STN stimulation-induced reductions
in EEG–EEG coherence correlated with clinical improvement at similar frequencies. Collectively these observations
suggest that elevated cortico-cortical coupling in this band
may be an important feature of parkinsonism, that reverses
with STN HFS. Dopaminergic therapy also induced reductions in EEG–EEG coherence correlated with clinical
improvement, again predominantly in the 10–35 Hz band,
although less marked correlations extended to higher frequencies. This suggests that the two treatments may achieve some
of their beneficial effects through common actions at the
cortical level. However, before examining our findings in
detail, we should bear in mind some of the potential limitations of our study.
Experimental limitations
Surgical placement of stimulating
macroelectrode
To determine accurately the physiological effects of STN
stimulation in humans, it is first important to be sure of correct electrode placement. Examination of the postoperative
images in our patients was consistent with placement of at
least one of the electrode contacts in the STN; however, it is
important to keep in mind the limitations of image interpretation in reaching this conclusion. Whilst the borders of the
STN may be depicted on preoperative, thin slice T2-weighted
MRI (Hariz et al., 2003), the borders of the STN are not
always clearly defined on conventional postoperative images,
due to artefact arising from the macroelectrode, making any
estimation as to contact position presumptive, based on relationships to clearly defined surrounding anatomic structures
(Bejjani et al., 2000). Further, tissue compression, inevitable
in even thin MRI slices, may overestimate the proximity of
electrode contacts to the STN. Support for correct electrode
placement may, however, also be provided by the clinical
effects of stimulation, and the reduction in the postoperative
UPDRS score of almost 50% in our patients is consistent with
satisfactory electrode placement. We also limited the effects of
variance in electrode positioning on STN stimulation-induced
changes in cortical coupling by recording as many patients
and sides as possible. To achieve this, we collaborated across
several surgical centres (see Table 1), but this itself may have
introduced systemic bias, albeit non-intended, in target localization between centres. Here again, MRI findings in individual cases along with UPDRS improvement with stimulation
would argue against this as a significant confound.
Experimental protocol
Our protocol required assessment of patients under the different treatment conditions in a single sitting. Randomization
of recordings between subjects and separate recordings of left
and right sides necessitated switching on and off each stimulator side on a number of occasions. In order that the study
was not excessively long and uncomfortable for our subjects,
we limited the off medication–no stimulation period to 10 min
1288
P. Silberstein et al.
and the period between stimulation changes to 5 min. The
recurrence of Parkinsonian signs after stimulation is switched
off increases with time (Temperli et al., 2003), making it
possible that the limited time interval between recordings
resulted in less clinical difference in each stimulation state
than might be seen with longer intervals. On the other hand,
clinical improvement related to stimulation, as determined by
changes in UPDRS hemibody score of at least 30%, was determined in all cases, so any underestimate of effects was
probably limited. Furthermore, the intervals used were also
sufficient to allow appreciable changes in EEG dynamics to be
detected, and both changes in EEG coupling and improvements in clinical score correlated. In conclusion, timing limitations could have only served to underestimate the relationships
demonstrated. The use of standard dosages of medications after
overnight withdrawal may have also contributed to a possible
underestimation of the effects of L-dopa on cortical coupling.
EEG recordings and analysis
We recorded scalp EEG referenced to linked ears. This introduced the possibility of volume conduction between electrode
sites, leading to overestimates of coherence and blurring of
any topographic differences. We limited these effects by
using a subtractive approach, with respect to the consequences of stimulation and L-dopa on power and coherence
change. Furthermore, correlations with clinical state were
unlikely to have highlighted coupling that related to volume
conduction or common reference, rather than that associated
with functional change. The central predominance of the
correlations at individual frequencies and in frequency
bands higher than those associated with tremor, although
expected from proximity to mesial and lateral motor areas,
raises the possibility of another confound. Given that all of
our patients had frontal burr holes, our EEG recordings necessarily included a partial breach in the skull and associated
tissues. Whilst the position of the burr hole varied between
subjects, it tended to be located between F3, FZ, C3 and CZ
on the left and corresponding electrodes on the right hand
side. The burr holes may have improved the signal to noise
ratio of the cortically derived EEG signal, but could not have
caused the correlations between coherence changes and
improvements in parkinsonism. On the other hand, burr
holes may have contributed to the central predominance of
these correlations, although any effect was insufficient to
obscure the peripheral predominance of correlations at
those frequencies under 10 Hz associated with parkinsonian
rest and action tremor. The latter finding means that correlations at such frequencies may have been spurious and related
to decreases in tremor contamination of peripheral scalp
recordings with treatment.
Statistical considerations
Our primary approach was correlation to demonstrate an association between changes in spectral differences in the EEG and
changes in UPDRS motor scores upon treatment across
patients, and thereafter to show a similarity in effects between
treatment types. Correlation does not necessarily imply causation, and, moreover, there were a number of confounding
variables, such as the potential variability in precise electrode
positioning between patients, disparate stimulation parameters
and the variable timing of recordings following surgery. However, factors such as these would have acted to bias against
the finding of significant correlations and led to an underestimation of effects rather than generate spurious effects.
We assumed an empirical significance level of 0.005 in
correlations by frequency band. Thus one in 200 of the correlations tested may have been spuriously significant. The
contribution of these spurious results to the overall picture
seems likely to have been insignificant, given that correlations occurred with a much higher incidence than expected by
chance, were relatively frequency selective and were of uniform sign. Spurious correlations would have been equally
represented across all frequency bands and would have
been just as likely to be positive as negative. Similar arguments apply to the correlations at individual frequencies,
where we applied the same significance level of 0.005.
The proportion of coherent connections in the 10–35 Hz
band far exceeded the 0.5% expected by chance and correlations were almost entirely of consistent sign. Nevertheless,
the qualitative similarity between correlations of STN stimulation and levodopa-induced EEG–EEG coherence changes
with clinical improvement should be stressed rather than any
quantative differences or similarities. The most obvious factor militating against a direct comparison of these correlations
was the lower number of observations and hence reduced
statistical power of the levodopa correlations.
In order to investigate similarities in the frequencydependent topography of STN stimulation and L-dopa effects,
we correlated frequency-dependent changes in coupling by
electrode pair between these therapeutic modalities, finding
strong correlations in each frequency band. As we did not
record subjects during simultaneous left and right STN stimulation, we determined a surrogate measure of bilateral
stimulation effects by averaging the coupling change from
separate left and right stimulation recordings in each subject.
These were then compared with the frequency-dependent
changes in coupling after L-dopa. Data sets were not paired
given that one of the subjects that achieved a satisfactory
medication response was excluded from the off-med onstim group due to unilateral macroelectrode misplacement,
and three subjects in the off-med on-stim group were not
included in the average calculations in the on-med off-stim
group due to medication dose failure. One could argue that
since only a surrogate measure of the off-med on-stim state
and non-paired data was used, our findings may be unreliable.
Against this, correlation analysis depends on similarity of
effects, and both of these factors would be likely to introduce
greater variance to the difference between the treatment
states. This suggests that a direct comparison of bilateral
STN stimulation with levodopa effects may have yielded
even stronger correlations.
Cortico-cortical coupling in Parkinson’s disease
Cortico-cortical coherence varies with
clinical state
There are three possible, non-mutually exclusive, explanations for the correlation between untreated cortico-cortical
coupling and clinical state and between treatment-induced
changes in coupling and improvements in clinical state.
First, correlations may have been related, in part, to decreased
movement artefact or volume conduction related to tremor
upon treatment. This may have been a confound in the 4–7 Hz
band, where correlations were not central in their distribution,
but is unlikely to have been a significant factor at frequencies
higher than those represented in tremor, where correlations
were maximal centrally.
Secondly, the relationship between cortico-cortical coherence and clinical state might represent a direct effect of the
impairment of dopaminergic tone. In this regard, it is notable
that the dopaminergic system has considerable and widespread modulatory cortical influences (Steiner and Kitai,
2001). Furthermore, LFP recordings in STN and its major
target, globus pallidus internus, are characterized by a preponderance of activity in the 10–30 Hz range in the untreated
parkinsonian state (Brown et al., 2001; Marsden et al., 2002;
Priori et al., 2002; Silberstein et al., 2003). This activity is
coupled between these nuclei and cerebral cortex (Marsden
et al., 2001; Williams et al., 2002). The direction of this
coupling at rest, however, suggests net cortical driving of
synchronization in the basal ganglia, rather than the converse
(Marsden et al., 2001; Williams et al., 2002).
Thirdly, the relationship between cortico-cortical coherence and clinical state might reflect the existence of compensatory cortical mechanisms in the off state. The possibility
of compensatory mechanisms in Parkinson’s disease has been
raised by imaging, transcranial magnetic stimulation (TMS)
and previous EEG studies. Imaging studies demonstrate taskrelated hyperactivity in untreated Parkinson’s disease in the
lateral motor system and caudal supplementary motor area,
although hypoactivity is present in the rostral supplementary
motor area and prefrontal cortex (Samuel et al., 1997). This
apparent upregulation of the lateral motor system has been
linked to a greater utilization of and attention to visual cues as
a compensatory motor strategy, and it is interesting to note
that we found a number of connections between the occipital
electrodes and more frontal ones that correlated with severity
of parkinsonism in untreated Parkinson’s disease and negatively correlated with clinical improvement. TMS studies have
shown some inconsistencies, but in general suggest a reduction in intracortical inhibition in Parkinsonian patients
(Ridding et al., 1995; Cantello et al., 2002), ostensibly inconsistent with the Albin and DeLong model of basal ganglia
function (Alexander et al., 1986; Albin et al., 1989). Consequently, Cunic et al. (2002) have reasoned that the reduction in intracortical inhibition may be a compensation for
akinesia/bradykinesia. In EEG studies, movement-induced
desynchronization of the mu rhythm in untreated de novo
patients or subjects withdrawn from medication was delayed
1289
over the contralateral central area, but occurred earlier over
frontocentral regions in comparison with treated subjects
or normal controls performing the same self-paced task
(Defebvre et al., 1996; Devos et al., 2004). These alterations
in the timing of the mu desynchronization have been interpreted as showing that other cortical areas, especially the
ipsilateral primary sensorimotor cortex, are activated to compensate for deficiencies in cortical motor preparation (Devos
et al., 2004).
If the relationship between cortico-cortical coherence and
clinical state were to be partly due to cortical compensation
for the motor dysfunction accrued through the basal ganglia
defect in Parkinson’s disease, then one might predict corresponding changes in patients with motor dysfunction due to
different pathological mechanisms. This seems to be the case
in patients recovering from motor stroke. In such patients,
there is increased coherence in the 13–24 Hz band over mesial
areas during gripping with the affected relative to the unaffected hand compared with the intermanual difference in
healthy controls. This increased hand-related asymmetry is
negatively correlated with recovery, consistent with a compensatory role for these cortical changes (Strens et al., 2004).
In the above, we have considered the possibility of compensatory changes in cortical function in Parkinson’s disease
and stroke as adaptive and beneficial, making up for the core
dysfunction related to the underlying pathology. However,
we cannot discount the alternative hypothesis that any secondary increase in cortical coupling is maladaptive and actually compounds the pathological dysfunction. For example, a
net cortical drive to the basal ganglia in the beta band might
increase subcortical synchronization in this frequency range,
thereby exacerbating an oscillatory activity that has been
considered essentially antikinetic (Brown, 2003).
STN HFS and dopaminergic drugs have
similar effects on cortical oscillatory activity
We found strong frequency-dependent correlations in the
topographic effects of HFS STN and dopaminergic medications on cortical coupling, suggesting that, at rest, dopaminergic medications and L-dopa affect cortical oscillatory
coupling in a similar manner. This finding is in keeping
with other lines of evidence suggesting that the cortical
effects of these therapies, both at rest and with movement,
are similar, as determined by TMS (Ridding et al., 1995;
Pierantozzi et al., 2001, 2002; Cunic et al., 2002), imaging
(Jenkins et al., 1992; Limousin et al., 1997) and EEG studies.
In the latter case, Devos et al. (2004) showed that a similar
improvement in contralateral pre-movement- and bilateral
movement-related mu desynchronization was achieved by
STN stimulation as with L-dopa. In an earlier study, the
same group (Devos et al., 2003a) demonstrated that attenuated beta Event related synchronisation (ERS) over the central region was significantly increased with STN stimulation
and L-dopa.
1290
P. Silberstein et al.
Overall, these findings are consistent with the notion that
STN stimulation and L-dopa alter the pattern and extent of
cortical activity in a similar manner. Nevertheless, our data
revealed one aspect in which correlations between EEG
coherence and clinical state differed between STN HFS
and oral dopaminergic therapy. Correlations following
drug therapy, although concentrated over 10–35 Hz, did
extend over higher frequencies and, as indicated in the Results, this could not be explained by dyskinesia-related EMG
contamination of EEG following drug treatment. This additional effect of dopaminergic therapy may relate, in part, to a
more complex effect of dopamine on the patterning of striatal
output, or have involved extra-striatal effects, such as direct
dopaminergic effects on the cerebral cortex acting to suppress
the synchronization between cortical regions at high frequency. Note that the correlation between changes in coherence at frequencies >35 Hz and difference in hemibody scores
on medication was also negative. Specifically, inter-regional
synchronization in the 60–80 Hz band dropped as the
clinical state improved. This result was unexpected, given
that dopaminergic medication induces increases in corticosubcortical coupling in this frequency band (Williams et al.,
2003). There may be several explanations for this. Corticosubcortical coupling in the 60–80 Hz band may only promote
local (intra-regional) rather than the distributed inter-regional
cortical synchronization examined here. In addition, not
all treated Parkinson’s disease patients show an increase in
60–80 Hz cortico-subcortical coupling (Brown, 2003) and,
when present, the effect is most marked during task performance (Cassidy et al., 2003), whereas our recordings were
made at rest.
In conclusion, our results have shown that cortico-cortical
coupling in Parkinson’s disease relates to clinical state, and
decrements in cortico-cortical coupling with STN stimulation
or levodopa positively correlate with clinical improvement.
The changes in cortical dynamics responsible for these correlations may be partially compensatory or maladaptive,
while the similarity between the effects of STN stimulation
and levodopa suggests that the two treatments may achieve
some of their beneficial effects through common effects on
the pattern of interaction between cortical areas.
Acknowledgements
P.B. is supported by the Medical Research Council of Great
Britain, P.S. by a fellowship from the Parkinson’s disease
Society UK, A.K. by a fellowship from the German
Academic Exchange Service (DAAD), S.T. by the Brain
Research Trust, UK, and P.D.-L. by the Medical Research
Council of Great Britain and Parkinson’s appeal.
References
Albin RL, Young AB, Penney JB. The functional anatomy of basal ganglia
disorders. Trends Neurosci 1989; 12: 366–75.
Alexander GE, DeLong MR, Strick PL. Parallel organization of functionally
segregated circuits linking basal ganglia and cortex. Annu Rev Neurosci
1986; 9: 357–81.
Andres FG, Mima T, Schulman AE, Dichgans J, Hallett M, Gerloff C. Functional coupling of human cortical sensorimotor areas during bimanual skill
acquisition. Brain 1999; 122: 855–70.
Bejjani BP, Dormont D, Pidoux B, Yelnik J, Damier P, Arnulf I, et al.
Bilateral subthalamic stimulation for Parkinson’s disease by using
three-dimensional stereotactic magnetic resonance imaging and electrophysiological guidance. J Neurosurg 2000; 92: 615–25.
Brown P. Oscillatory nature of human basal ganglia activity: relationship to
the pathophysiology of Parkinson’s disease. Mov Disord 2003; 18: 357–63.
Brown P, Marsden CD. What do the basal ganglia do? Lancet 1998; 351:
1801–4.
Brown P, Marsden CD. Bradykinesia and impairment of EEG desynchronization in Parkinson’s disease. Mov Disord 1999; 14: 423–9.
Brown P, Oliviero A, Mazzone P, Insola A, Tonali P, Di Lazarro V. Dopamine
dependency of oscillations between subthalamic nucleus and pallidum in
Parkinson’s disease. J Neurosci 2001; 21: 1033–8.
Cantello R, Tarletti R, Civardi C. Transcranial magnetic stimulation and
Parkinson’s disease. Brain Res Brain Res Rev 2002; 38: 309–27.
Cassidy M, Brown P. Task-related EEG–EEG coherence depends on
dopaminergic activity in Parkinson’s disease. Neuroreport 2001; 12:
703–7.
Cassidy M, Mazzone P, Oliviero A, Insola A, Tonali P, Di Lazarro V, et al.
Movement-related changes in synchronization in the human basal ganglia.
Brain 2002; 125: 1235–46.
Cunic D, Roshan L, Khan FI, Lozano AM, Lang AE, Chen R. Effects of
subthalamic nucleus stimulation on motor cortex excitability in Parkinson’s
disease. Neurology 2002; 58: 1665–72.
Defebvre L, Bourriez JL, Destee A, Guieu JD. Movement related desynchronisation pattern preceding voluntary movement in untreated Parkinson’s
disease. J Neurol Neurosurg Psychiatry 1996; 60: 307–12.
Defebvre L, Bourriez JL, Derambure P, Duhamel A, Guieu JD, Destee A.
Influence of chronic administration of L-DOPA on event-related desynchronization of mu rhythm preceding voluntary movement in Parkinson’s
disease. Electroencephalogr Clin Neurophysiol 1998; 109: 161–7.
DeLong MR. Primate models of movement disorders of basal ganglia origin.
Trends Neurosci 1990; 13: 281–5.
Devos D, Labyt E, Cassim F, Bourriez JL, Reyns N, Touzet G, et al.
Subthalamic stimulation influences postmovement cortical somatosensory
processing in Parkinson’s disease. Eur J Neurosci 2003a; 18: 1884–8.
Devos D, Labyt E, Derambure P, Bourriez J, Cassim F, Guieu JD, et al.
Effect of L-Dopa on the pattern of movement-related (de)synchronisation
in advanced Parkinson’s disease. Neurophysiol Clin 2003b; 33: 203–12.
Devos D, Labyt E, Derambure P, Bourriez JL, Cassim F, Reyns N, et al.
Subthalamic nucleus stimulation modulates motor cortex oscillatory activity in Parkinson’s disease. Brain 2004; 127: 408–19.
Enge S, Heppner F, Diemath HE, Lechner H. Effects of L-dopa on the EEG
before, during and after stereotactic operations. Electroencephalogr Clin
Neurophysiol 1966; 20: 268.
England AC, Schwab RS, Peterson E. The electroencephalogram in
Parkinson’s syndrome. Electroencephalogr Clin Neurophysiol 1959; 11:
723–31.
Farmer SF. Rhythmicity, synchronization and binding in human and primate
motor systems. J Physiol 1998; 509: 3–14.
Florian G, Andrew C, Pfurtscheller G. Do changes in coherence always reflect
changes in functional coupling? Electroencephalogr Clin Neurophysiol
1998; 106: 87–91.
Foffani G, Priori A, Egidi M, Rampini P, Tamma F, Caputo E, et al.
300-Hz subthalamic oscillations in Parkinson’s disease. Brain 2003;
126: 2153–63.
Gerloff C, Richard J, Hadley J, Schulman AE, Honda M, Hallett M. Functional coupling and regional activation of human cortical motor areas
during simple, internally paced and externally paced finger movements.
Brain 1998; 121: 1513–31.
Goldberg JA, Boraud T, Maraton S, Haber SN, Vaadia E, Bergman H.
Enhanced synchrony among primary motor cortex neurons in the
1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine
primate
model
of
Parkinson’s disease. J Neurosci 2002; 22: 4639–53.
Cortico-cortical coupling in Parkinson’s disease
Halliday DM, Rosenberg JR, Amjad AM, Breeze P, Conway BA, Farmer SF.
A framework for the analysis of mixed time series/point process data—
theory and application to the study of physiological tremor, single motor
unit discharges and electromyograms. Prog Biophys Mol Biol 1995; 64:
237–78.
Hariz MI, Krack P, Melvill R, Jorgensen JV, Hamel W, Hirabayashi H, et al.
A quick and universal method for stereotactic visualization of the subthalamic nucleus before and after implantation of deep brain stimulation
electrodes. Stereotact Funct Neurosurg 2003; 80: 96–101.
Hurtado JM, Gray CM, Tamas LB, Sigvardt KA. Dynamics of tremor-related
oscillations in the human globus pallidus: a single case study. Proc Natl
Acad Sci USA 1999; 96: 1674–9.
Jenkins IH, Fernandez W, Playford ED, et al. Impaired activation of the
supplementary motor area in Parkinson’s disease is reversed when akinesia
is treated with apomorphine. Ann Neurol 1992; 32: 749–57.
Leocani L, Toro C, Manganotti P, Zhuang P, Hallett M. Event-related coherence and event-related desynchronization/synchronization in the 10 Hz
and 20 Hz EEG during self-paced movements. Electroencephalogr Clin
Neurophysiol 1997; 104: 199–206.
Levy R, Hutchison WD, Lozano AM, Dostrovsky JO. High-frequency
synchronization of neuronal activity in the subthalamic nucleus of parkinsonian patients with limb tremor. J Neurosci 2000; 20: 7766–75.
Levy R, Dostrovsky JO, Lang AE, Sime E, Hutchison WD, Lozano AM.
Effects of apomorphine on subthalamic nucleus and globus pallidus
internus neurons in patients with Parkinson’s disease. J Neurophysiol
2001; 86: 249–60.
Levy R, Ashby P, Hutchison WD, Lang AE, Lozano AM, Dostrovsky JO.
Dependence of subthalamic nucleus oscillations on movement and dopamine in Parkinson’s disease. Brain 2002a; 125: 1196–209.
Levy R, Hutchison WD, Lozano AM, Dostrovsky JO. Synchronized neuronal
discharge in the basal ganglia of parkinsonian patients is limited to oscillatory activity. J Neurosci 2002b; 22: 2855–61.
Limousin P, Greene J, Pollak P, Rothwell J, Benabid AL, Frackowiak R.
Changes in cerebral activity pattern due to subthalamic nucleus or internal
pallidum stimulation in Parkinson’s disease. Ann Neurol 1997; 42: 283–91.
Limousin P, Pollak P, Benazzouz A, Hoffman D, Le Bas JF, Brousolle E, et al.
Effect of parkinsonian signs and symptoms of bilateral subthalamic nucleus
stimulation. Lancet 1995; 345: 91–5.
Magnani G, Cursi M, Leocani L, Volonte MA, Comi G. Acute effects of
L-dopa on event-related desynchronization in Parkinson’s disease. Neurol
Sci 2002; 23: 91–7.
Marsden CD, Obeso JA. The functions of the basal ganglia and the paradox of
stereotaxic surgery in Parkinson’s disease. Brain 1994; 117: 877–97.
Marsden JF, Werhahn KJ, Ashby P, Rothwell J, Noachtar S, Brown P. Organization of cortical activities related to movement in humans. J Neurosci
2000; 20: 2307–14.
Marsden JF, Limousin-Dowsey P, Ashby P, Pollak P, Brown P. Subthalamic
nucleus, sensorimotor cortex and muscle interrelationships in Parkinson’s
disease. Brain 2001; 124: 378–88.
McPherson A. Convulsive seizures and electroencephalogram changes in
three patients during levodopa therapy. Electroencephalogr Clin Neurophysiol 1970; 20: 41–5.
Neufeld MY, Inzelberg R, Korczyn AD. EEG in demented and non-demented
parkinsonian patients. Acta Neurol Scand 1988; 78: 1–5.
Obeso JA, Rodriguez MC, DeLong MR. Basal ganglia pathophysiology.
A critical review. Adv Neurol 1997; 74: 3–18.
Ohara S, Mima T, Baba K, Ikeda A, Kunieda T, Matsumoto R, et al. Increased
synchronization of cortical oscillatory activities between human supplementary motor and primary sensorimotor areas during voluntary movements. J Neurosci 2001; 21: 9377–86.
Pfurtscheller G, Lopes da Silva FH. Event-related EEG/MEG synchronization
and desynchronization: basic principles. Clin Neurophysiol 1999; 110:
1842–57.
1291
Pfurtscheller G, Pichler-Zalaudek K, Ortmayr B, Diez J, Reisecker F.
Postmovement beta synchronization in patients with Parkinson’s disease.
J Clin Neurophysiol 1998; 15: 243–50.
Pierantozzi M, Palmieri MG, Marciani MG, Bernardi G, Giacomini P,
Stanzione P. Effect of apomorphine on cortical inhibition in Parkinson’s
disease patients: a transcranial magnetic stimulation study. Exp Brain Res
2001; 141: 52–62.
Pierantozzi M, Palmieri MG, Mazzone P, Marciani MG, Rossini PM,
Stefani A, et al. Deep brain stimulation of both subthalamic nucleus
and internal globus pallidus restores intracortical inhibition in Parkinson’s
disease paralleling apomorphine effects: a paired magnetic stimulation
study. Clin Neurophysiol. 2002; 113: 108–13.
Priori A, Foffani G, Pesenti A, Bianchi A, Chiesa V, Baselli G, et al.
Movement-related modulation of neural activity in human basal ganglia
and its L-DOPA dependency: recordings from deep brain stimulation electrodes in patients with Parkinson’s disease. Neurol Sci 2002; 23 Suppl 2:
S101–2.
Priori A, Foffani G, Pessenti A, Tamma F, Bianchi AM, Pellegrini M, et al.
Rhythm specific pharmacologic modulation of subthalamic activity in
Parkinson’s disease. Exp Neurol. 2004; 189: 369–79.
Ridding MC, Inzelberg R, Rothwell JC. Changes in excitability of motor
cortical circuitry in patients with Parkinson’s disease. Ann Neurol 1995;
37: 181–8.
Samuel M, Ceballos-Baumann AO, Blin J, Uema T, Boecker H,
Passingham RE, et al. Evidence for lateral premotor and parietal overactivity in Parkinson’s disease during sequential and bimanual movements. A
PET study. Brain 1997; 120: 963–76.
Schaltenbrand G, Wahren W. Atlas of stereotaxy of the human brain.
Stuttgart: Thieme; 1977.
Serrien DJ, Brown P. The functional role of interhemispheric synchronization
in the control of bimanual timing tasks. Exp Brain Res 2002; 147: 268–72.
Serrien DJ, Brown P. The integration of cortical and behavioural dynamics
during initial learning of a motor task. Eur J Neurosci 2003; 17: 1098–104.
Serrien DJ, Fisher RJ, Brown P. Transient increases of synchronized neural
activity during movement preparation: influence of cognitive constraints.
Exp Brain Res 2003; 153: 27–34.
Silberstein P, Kuhn AA, Kupsch A, Trottenberg T, Krauss JK, Wohrle JC,
et al. Patterning of globus pallidus local field potentials differs between
Parkinson’s disease and dystonia. Brain 2003; 126: 2597–608.
Steiner H, Kitai ST. Unilateral striatal dopamine depletion: time-dependent
effects on cortical function and behavioural correlates. Eur J Neurosci
2001; 14: 1390–404.
Strens LHA, Asselman P, Pogosyan A, Loukas C, Thompson AJ, Brown P.
Cortico-cortical coupling in chronic stroke: its relevance to recovery.
Neurology 2004; 63: 475–84.
Temperli P, Ghika J, Villemure JG, Burkhard PR, Bogousslavsky J,
Vingerhoets FJ. How do parkinsonian signs return after discontinuation
of subthalamic DBS? Neurology 2003; 60: 78–81.
Wang HC, Lees AJ, Brown P. Impairment of EEG desynchronisation before
and during movement and its relation to bradykinesia in Parkinson’s disease. J Neurol Neurosurg Psychiatry 1999; 66: 442–6.
Wiederholt WC. EEG and reaction time in patients with parkinsonism before
and during L-dopa therapy. Agents Actions 1974; 4: 61–6.
Williams D, Tijssen M, Van Bruggen G, Bosch A, Insola A, Di Lazarro V,
et al. Dopamine-dependent changes in the functional connectivity
between basal ganglia and cerebral cortex in humans. Brain 2002; 125:
1558–69.
Yaar I. EEG power spectral changes secondary to L-dopa treatment in
Parkinsonian patients: a pilot study. Electroencephalogr Clin Neurophysiol
1977; 43: 111–8.
Yeager CL, Alberts W, Delattre LD. Effect of stereotaxic surgery upon
electroencephalographic status of parkinsonian patients. Neurology
1966; 16: 904–10.