Download The role of the subthalamic nucleus in response inhibition: Evidence

NeuroImage 60 (2012) 271–278
Contents lists available at SciVerse ScienceDirect
NeuroImage
journal homepage: www.elsevier.com/locate/ynimg
The role of the subthalamic nucleus in response inhibition: Evidence from local field
potential recordings in the human subthalamic nucleus
Nicola J. Ray a,⁎, John-Stuart Brittain a, b, Peter Holland c, Raed A. Joundi a, John F. Stein a,
Tipu Z. Aziz c, Ned Jenkinson a, c
a
Department of Physiology, Anatomy and Genetics, University of Oxford, Oxford, OX1 3PT, UK
Centre of Excellence in Personalised Healthcare, Institute of Biomedical Engineering, Department of Engineering Science, University of Oxford, Old Road Campus Research Building,
Headington, Oxford OX3 7DQ, UK
c
Nuffield Department of Surgical Sciences, University of Oxford, John Radcliffe Hospital, Oxford, OX3 9DU, UK
b
a r t i c l e
i n f o
Article history:
Received 22 September 2011
Revised 16 November 2011
Accepted 13 December 2011
Available online 22 December 2011
Keywords:
Stop-signal reaction time task
Local field potentials
Beta oscillations
Gamma oscillations
Subthalamic nucleus
a b s t r a c t
Response inhibition as measured during a stop-signal task refers to the ability to halt an action that has already been set in motion. Cortical and sub-cortical structures, such as the subthalamic nucleus (STN), that
are active during attempts to inhibit action are thought to contribute to a ‘stop-process’ that must gain dominance over a ‘go-process’ if inhibition is to be successful. We recorded local field potential activity from the
STN of Parkinson's disease patients with implanted deep brain stimulation electrodes during a stop-signal
task. In particular we measured activity in the STN that has traditionally been associated with motor action
(gamma-band, 60–100 Hz) and inhibition (beta-band, 10–30 Hz). Our data support the idea that beta activity
in the STN is related to the inhibition of motor action. Further, we report that gamma oscillatory activity responds robustly to stop-signals as well as go-signals. This unexpected finding might suggest that gamma activity supports a go-process that not only responds to go-signals, but is also sensitive to stimuli that signal
stopping.
© 2011 Elsevier Inc. All rights reserved.
Introduction
The unpredictability of every-day life means that changing circumstances can render planned motor actions suddenly inappropriate. The ability to inhibit pre-planned or on-going motor action,
known as response inhibition, is therefore essential for the normal
control of movement. Response inhibition is frequently tested in
the laboratory using countermanding tasks that require participants
to ‘stop’ an on-going ‘go’ response. Functional imaging studies show
that response inhibition activates areas of frontal cortex and the
subthalamic nucleus (STN) of the basal ganglia (Aron and
Poldrack, 2006; Aron et al., 2007; Li et al., 2008; Sharp et al.,
2010). The importance of the STN in response inhibition has since
been demonstrated behaviourally in humans (Ray et al., 2009; van
den Wildenberg et al., 2006).
The opportunity to record electrical activity in the form of local
field potentials (LFP) directly from the STN arises in Parkinson's
⁎ Corresponding author at: Toronto Western Hospital and Institute, CAMH-PET Imaging Centre, University of Toronto, 250 College St, Toronto, ON, Canada.
E-mail addresses: [email protected] (N.J. Ray),
[email protected] (J.-S. Brittain), [email protected]
(P. Holland), [email protected] (R.A. Joundi), [email protected]
(J.F. Stein), [email protected] (T.Z. Aziz), [email protected]
(N. Jenkinson).
1053-8119/$ – see front matter © 2011 Elsevier Inc. All rights reserved.
doi:10.1016/j.neuroimage.2011.12.035
disease (PD) patients undergoing surgery to implant deep brain stimulation (DBS) electrodes. Such recordings have revealed that go responses during a go/no-go task are preceded by a decrease in beta
power, representing de-synchronised oscillatory activity, while inhibition of go responses during no-go trials are associated with the
early termination of beta desynchronisation (Kühn et al., 2004). Voluntary movements are also preceded by an increase in power or synchronisation of higher frequency gamma activity (Androulidakis et
al., 2007; Cassidy et al., 2002), but inhibition is not thought to be related to a modulation in gamma. These findings suggest that changes
in beta and gamma activity within the STN are associated with the
preparation of externally triggered movements, and that synchronised beta activity is involved in the inhibition of voluntary movements in a go/no-go paradigm.
During go/no-go paradigms, no-go trials require movements to be
inhibited post preparation but prior to execution. How oscillatory activity in the STN responds to the cancellation of an on-going movement – as required in stop-signal tasks – has not been investigated.
We report LFP data acquired from the STN of human participants performing a manual stop-signal task. We were primarily interested in
differences in these activities during go- and stop-trials. Previous research on beta and gamma activity in the STN during ‘going’ and
‘stopping’ (Androulidakis et al., 2007; Cassidy et al., 2002; Kühn et
al., 2004) leads us to expect beta synchrony to decrease after gosignals in both go- and stop-trials, but to re-emerge more quickly
272
N.J. Ray et al. / NeuroImage 60 (2012) 271–278
during stop-trials (i.e. following presentation of the stop-signal). We
also expect, based on previous data in the internal globus pallidus
(GPi), in which gamma-band activity is coherent with the STN
(Cassidy et al., 2002), that gamma activity is decreased during stoptrials compared to go-trials (Brücke et al., 2008).
Materials and methods
Participants
Participants were nine right-handed PD patients undergoing surgery to implant DBS electrodes into the STN. Seven of the patients received bilateral electrodes, while the remaining two patients received
electrodes implanted unilaterally into the left STN. Recordings were
made in the peri-operative period, prior to implantation of the DBS
pacemaker. During this time the electrodes can be used to record
electrical activity in the STN while the patients are awake and performing behavioural tasks. All patients were on their usual medication for PD. A summary of patient demographics and clinical
characteristics is presented in Table 1.
Behavioural task
For the stop-signal task patients were seated in front of a computer screen. The task started with a fixation cross (present for a variable
duration between 1000 and 2000 ms), followed by a white arrow
(the go-signal) pointing either left or right. Subjects were instructed
to respond by pressing the left or right button corresponding to the
direction of the arrow on a button box as fast as possible, using the
index and middle fingers of the left or right-hand. On 25% of the trials
the white arrow was followed by an auditory stop-signal, indicating
that the patients should withhold their response. The stop-signal is
given pseudo-randomly (appearing once during a set of 4 trials, but
in a random position within those trials) and is extinguished 1 s
later, or upon an uninhibited response from the subject. Trials were
shuffled at the beginning of the task. Subjects were instructed that
going is as important as stopping, that they must respond as quickly
as possible to all signals, and that they will be unable to inhibit their
responses on some of the stop-trials. Each task constituted 100 trials.
The delay between the go- and stop-signal was adaptive, increasing
by 50 ms after successful inhibition in the previous stop-trial and decreasing by 50 ms after failed inhibition.
Behavioural analysis
All subjects completed 2 runs of the stop-signal task with each
hand, contralateral to DBS leads (4 runs total in subjects with bilateral
electrodes). During the task, 3 different behavioural outcomes are calculated. Go reaction times (GORTs) are calculated as the median time
Table 1
Demographics and clinical characteristics of the patients.
Patient
Age
UPDRS
on meds
Medications
1
2
3
4
5
6
48
66
37
54
62
62
56
9
24
36
47
22
7
5
9
65
43
60
7
3
33
Sinemet, ropinirole
Madopar, tolcapone
Madopar, mirapexin, domperidone
Madopar, selegiline, mirapexin, entacapone
Madopar, tolcapone
Sinemet CR, Sinemet Plus, Rotigotine, Entacapone,
Rasagiline
Stalero, Madopar, Amantadine
Madopar Disp, Madopar, Entacapone
Madopar, Ropinirole
UPDRS = Unified Parkinson's disease rating scale III (Motor examination), performed
on medication and pre-surgery by 3 months (maximum). Total possible UPDRS
score = 104.
to respond during go-trials in which no stop-signal is present. Stopsignal delay (SSD), which is the adaptive delay, described above, between the go- and stop-signals, is calculated as the mean delay between go- and stop-signals. However, since this delay was set to
250 ms at the start of the task for all patients, we did not include
the first 12 delays in the calculation of the SSD in order to allow convergence from this initial value. Finally, stop-signal reaction time
(SSRT), following Logan et al. (1984), is estimated by subtracting
the SSD from GORT. This procedure is necessary for stop-signal
tasks because we are attempting to measure a response that is absent
(inhibited). It assumes that the reaction time to stop-signals occurs at
some point following the stop-signal and before the GORT at the 50th
percentile of the distribution of go reaction times for an individual
subject, based on the fact ~50% of stop-signals were successful and
~50% unsuccessful due to the adaptive staircase procedure. Thus,
the staircase procedure that controls the length of the SSD must
have become stable for estimates of the SSRT to be accurate i.e. by
the end of the task the participant should be approximately alternating between failure and success. This is because failure causes SSDs to
decrease and success causes SSDs to increase. Thus, when participants
failed to reach ~ 50% accuracy (±15%), both the behavioural and the
depth recording data was excluded from further analysis. This
resulted in the removal of 4 runs of the stop-signal task. At least
one run remained for each hand in all subjects.
LFP acquisition
Simultaneous recording of LFPs were made with three adjacent
pairs of electrode contacts in a bipolar configuration. Signals were
amplified (×10,000) and filtered (0.2–1000 Hz) with isolated CED
1902 amplifiers, sampled at 2500 Hz using a CED 1401 Mark II A-D
converter (Cambridge Electronic Design, Cambridge, UK) and Spike
2 (Cambridge Electronic Design) and stored on a computer hard
disk for off-line analysis.
LFP analysis
There are inherent difficulties in reconstructing the positions of
electrode contacts of DBS electrodes. Although the borders of the
STN can be resolved on thin-slice pre-operative T2 weighted MRI,
the borders of the STN are not always clearly defined on conventional
post-operative images due to artifact arising from the presence of the
macroelectrode. With this difficulty in mind several studies have
demonstrated that increased levels of beta oscillations recorded
from individual contacts on the macroelectrode itself are due to the
positioning of those electrode contacts within the STN (see Kühn et
al., 2005; Weinberger et al., 2006). Indeed, beta activity in the LFP is
a hallmark of activity within the STN in Parkinson's disease and can
be used during surgery to locate the STN (Chen et al., 2006). Therefore electrode contact pairs from each STN giving the strongest signal
within the beta range during rest periods in between tasks were selected for further analysis.
Only data from the STN contralateral to the hand used for the
behavioural task were analysed for each run of the tasks. Following
previous work, we treat each STN side as a separate sample (see
Kühn et al., 2004, 2009). This approach is used when attempting
to correlate performance of one hand with activity in the contralateral basal ganglia, but it must be acknowledged that it artificially increases our N from 9 to 16. That said, this approach may be
particularly appropriate for studies on response inhibition given
suggestions that neural control of this ability may be lateralised
(Aron and Poldrack, 2006).
Data were imported into EEGLAB (Delorme and Makeig, 2004) filtered (between 6 Hz and 100 Hz) and aligned to triggers representing
the presentation of task stimuli and subject responses. Epochs were
then extracted from each record beginning 1 s before until 2 s after
N.J. Ray et al. / NeuroImage 60 (2012) 271–278
each trigger. To determine event related beta and gamma desynchronisation and synchronisation a time-frequency analysis was carried
out on each epoch using a Hermite functions analysis (Baraniuk and
Bayram, 2000) with time-frequency area A/2 = 5 (following Brittain
et al., 2007), in which frequency spectra in bins of 0.33 Hz were calculated from 6 Hz to 100 Hz in 10 ms steps. The extracted timefrequency data were then smoothed using a 333 ms sliding window.
For each patient, the peak beta and gamma signal was identified visually, and a mean time evolving beta/gamma signal was calculated as
the average power in the peak and ±10 frequency bins (6.6 Hz).
This trace was then Z normalised, and was used to determine event
related change (see later).
Go-trials
Change point analysis was carried out on the LFP data, triggered
from the go-signal, using scripts written by Gallistel, described and
available here (http://ruccs.rutgers.edu/distribution/change_point/).
Using this method, the time point at which beta synchronisation occurred post go-signal could be determined for 14 of our 16 STN during
go-trials (2 of the patients had slowly changing beta synchronous activity during go-trials, making determination of the time point of
change unreliable), and for all 16 STN during stop-trials. Consistent
with previous data (Trottenberg et al., 2006), not all of the datasets
revealed a distinctive gamma increase after go-signals that could be
subjected to change-point analysis. We therefore restricted the procedure to beta activity. For gamma signals, we instead used the
more objective measure of peak latency.
Stop-trials
Failed versus successfully inhibited stop-trials
In the stop-signal paradigm, differences in response latency (due
to the adaptive staircase procedure) incur bias when comparing failed
with successfully inhibited stop-trials. Failed stop-trials tend to occur
at longer delays after the go-signal than in successfully inhibited trials. Beta and gamma activity related to the preceding go-signal will
therefore have undergone prolonged desynchronisation and synchrony respectively, in failed stop-trials compared with successful stoptrials. To account for this we align stop-trial data to the preceding
go stimulus, and subtract latency matched go-signal activity. Note
that the SSRT in stop-signal tasks is calculated by assuming that the
faster 50% (median split) of go-responses will, on average, fail to be
inhibited while the slower 50% will be successfully inhibited (for detailed description see Logan et al., 1984). Thus, LFP activity during the
faster half of go-responses was subtracted from activity present during failed stop-trials, while LFP activity during the slower half of goresponses was subtracted from the LFP activity during successfully
inhibited stop-trials. The residual represents the (latency adjusted)
deviation in stop-signal response compared with go-signal activity.
Again, change point analysis was used to determine the time-point
at which beta synchronous activity during the failed and successfully
inhibited stop-trials deviated from the faster and slower go-trials, respectively, but the low signal-to-noise in individual datasets prevented us from doing this for gamma synchrony.
Removing the influence of SSD
We sought to determine whether the timing of the beta ERS following stop-signals was related to SSRT. However, the onset of the
stop-signal (SSD) was different for each patient. Since beta ERS occurs following the stop-signal, this means that beta ERS will unsurprisingly correlate with SSD. Given that SSRT and SSD are closely
related (SSRT = GORT − SSD), this poses a problem when attempting to determine a relationship between beta ERS and SSRT; any
correlation between SSRT and beta ERS may be an artifact of the relationship each of these have with SSD. To statistically remove the
273
influence of SSD on the timing of the beta ERS, we regressed the
time point at which beta synchrony became greater during successful stop-trials than slow go-trials, on to individual mean SSDs. The
residuals of this regression reflect variation across individuals in
their tendency to increase beta synchrony following stop-signals,
relative to the SSD. These residuals can therefore be thought of as
the timing of the beta response to stop-signals with the activity relating to the preceding go-signal and differences in the SSD between
patients are controlled for. Thus, negative numbers indicate quicker
changes in beta synchronous activity following the stop-signal.
Finally, to determine the event related beta and gamma changes
that followed the stop-signal itself, we triggered the data from the
stop-signal and analysed the increase in beta or gamma power by
determining the percentage change in synchrony post versus pre
stop-signal via the following calculation: (post stop-signal mean
power (at 200 ms to 400 ms) − pre stop-signal mean power (at
−200 ms to 0 ms))/post stop-signal mean power (at 200 ms to
400 ms). We also calculated the time point at which peak beta
and gamma synchrony was reached.
Statistical analysis
Matlab (The Mathworks, Natick, MA, USA) was used to perform
statistical analyses. Correlations between reaction times and beta
and gamma metrics were analysed with a Pearson's r. In general
GORTs and SSRTs were predicted to shorten for earlier changes in
gamma and/or beta synchrony. Due to lowered signal-to-noise in
the gamma signals during the infrequent stop-trials, we could only
reliably compute statistics on beta activity in these trials. Comparisons between trial types were evaluated with paired t-tests. One
tailed t-test predicted that beta activity would desynchronise quicker
for stop-trials compared with go-trials, and for successfully inhibited
stop-trials versus failed stop-trials. We mainly performed planned
comparisons, for which no correction for multiple comparisons was
applied. For unplanned exploratory comparisons, correction was applied as indicated in the text.
Results
Behavioural data
In 4 of the 32 runs of the stop-signal task there was a failure to
achieve ~50% accuracy, due either to impulsivity (failure to stop) or
delayed response (withholding). These were removed from further
analysis. The data for the remaining 28 runs are summarised in
Table 2. The reaction times for the failed (uninhibited) stop-trials
were faster than GORTs (t = 5.60, P b 0.001, 1-tailed). GORTs and
SSRTs were not correlated (r = 0.07, P = 0.78, 2-tailed).
LFP data — go-trials
Beta synchrony and desynchronisation
Figs. 1a and g, show a frequency spectrogram and trace, respectively, of the beta range for the time period 750 ms before and
Table 2
Behavioural data.
GORT
GORT during failed stop-trials
SSRT
SSD
Mean
SD
777 ms
683 ms
410 ms
366 ms
100 ms
70 ms
86 ms
128 ms
GORT = go reaction time, SSD = stop-signal delay, SSRT = stop-signal reaction time,
Stop-respond RT = reaction time for failed stop-signal trials, Accuracy = percentage
successfully inhibited stop-trials.
274
N.J. Ray et al. / NeuroImage 60 (2012) 271–278
Fig. 1. a, b and c are spectrograms of the beta frequency range recorded from the contralateral STN. 1a is an average of all go-trials, 1b is an average of all failed stop-trials, and 1c is
an average of all successful stop-trials. 1 d, e and f are as above, but for the gamma range. Plot 1g shows the average z-normalized beta signal (centered on the beta peak for each
patient and including +/− 10 bins (6.6 Hz)) for each of the trials types, and the equivalent is shown for 1h but for the gamma range. In all plots the go-signal is at t = 0, and the
onset of the stop-signal (mean SSD) is indicated by the vertical dotted line.
1750 ms after go-signals. Consistent with previous studies using go/
no-go tasks, go-signals were followed by a reduction in beta synchrony and a subsequent re-emergence of beta after a delay.
We used change-point analysis to determine the point at which
the mean beta signal for each patient began to desynchronise after
go-signals. We found that the time point of the onset of beta desynchronisation was positively correlated with GORT (r = 0.63, P b 0.05,
1-tailed; Fig. 2a). This is consistent with previous data showing a correlation between beta desynchronisation and GORTs (Kühn et al.,
2004). There was no relationship between SSRTs and onset of beta
desynchronisation (r = 0.43, P = 0.12).
Gamma synchrony and desynchronisation
Figs. 1d and h show frequency spectrogram and plots, respectively, of the gamma range for the time period 750 ms before to 1750 ms
after go-signals. Increases in gamma synchrony were seen during gosignals, consistent with previous data (Androulidakis et al., 2007;
Cassidy et al., 2002).
Changes in gamma synchrony following go-signals did not easily
lend themselves to change point analysis as increases in power occurred gradually, or not at all, for some patients. As the most objective
measure of gamma activity, we determined the time point at which
gamma power after the go-signal was at its peak. Gamma peak latency was found to be significantly correlated with GORTs (r = 0.59,
P = 0.007, 1-tailed; Fig. 2b), but not with SSRTs (r = 0.16, P = 0.55,
2-tailed).
LFP data — stop-trials
Beta synchrony and desynchronisation
Figs. 1b, c and g show spectrograms and plots of the beta range for
successful (Figs. 1c and g) and failed (Figs. 1b and g) stop-trials, starting 750 ms before to 1750 ms after the go-signal. Fig. 3a shows the
mean beta power for failed stop-trials and the latency matched (faster) go-trials. Fig. 3b shows the mean beta power for successfully
inhibited stop-trials and the latency matched (slower) go-trials.
Fig. 3c shows beta activity during fast and slow go-trials together,
demonstrating decreases in beta power occurring slightly earlier but
to a similar degree for fast as compared to slow go-trials, with reemergence of beta occurring slightly earlier for the faster go trials.
Fig. 3a suggests that unsuccessful stopping is associated with an initial increase in beta synchrony following the stop-signal (i.e. post
the desynchronisation that occurred due to the go-signal), followed
by subsequent desynchronisation. Fig. 3b suggests that successful
stopping is associated with less beta desynchronisation following
the go-signal, and an increase in beta synchrony. For successfully
inhibited and failed stop-trials, we determined (using change-point
analysis) the time point at which beta synchrony became greater
than that occurring during the slower and faster go trials, respectively. A repeated measures t-test revealed that beta synchronous
activity increases during successfully inhibited and failed stop-trials,
relative to the latency matched go-signals, occurs at similar time
points after the go-signal (t = 0.97, P = 0.18; Fig. 3g).
N.J. Ray et al. / NeuroImage 60 (2012) 271–278
275
that which had occurred during the latency matched go trials (i.e.
the time point at which there was a difference in the degree of synchrony during these trials) was positively correlated with the time
at which the stop-signal occurred (SSD) (r = 0.58, P = 0.03). Importantly however, we then removed the influence of the SSD during
these trials from the time point at which beta synchrony increases occurred using linear regression (see Methods section Removing the
influence of SSD). Having removed the influence of the SSD, we
found that the time point at which beta synchrony became greater
during the successfully inhibited stop trials compared with the slower
go-trials was significantly correlated with SSRTs (r = 0.55, P = 0.02;
Fig. 2c). Thus, faster synchronisation of beta activity relative to what
would be expected given the onset of the stop-signal for individual
patients was associated with faster response inhibition.
Finally, we looked at the beta synchronous activity that was triggered by the stop-signal. Calculating the percentage change in synchrony post versus pre stop-signal revealed that event related beta
synchrony was not different for failed and successful stop-trials
(t = 1.09, P = 0.28). Further the time point at which beta began to
synchronise post stop-signal was not different between the failed
and successful stop-trials (t = 0.51, P = 0.62; Fig. 4a).
Gamma synchrony and desynchronisation
Figs. 1e, f and h show that in successful stop-trials there is an increase in gamma synchrony following the go-signal that is terminated
earlier than gamma synchrony occurring during the failed stop-trials.
Following the protocol for beta synchrony analysis during these trials,
we plotted both trials with their latency matched go-trials (see
Figs. 3d and e). It was expected that stop-trials, particularly successfully inhibited stop-trials in which no movement occurs, would be associated with lowered gamma activity as compared to the latency
matched go-trials. Figs. 3d and e suggest that this is not the case,
but the poor signal-to-noise ratio present in for the gamma signals
prevented us from confirming this statistically (i.e. see Fig. 3h for
the difference plots).
In fact, failed stop-trials in particular seemed to be associated with
a more distinctive gamma signal which was sustained beyond the signal present in the latency matched go-trials (see Fig. 3d). To test
whether stop-signals were modulating gamma synchrony, we
aligned the stop-trial data from the stop-signal instead of the gosignal. Fig. 4b shows the results of this, and it is clear that gamma synchronous activity is evoked by the stop-signal. To determine if any
differences in event related gamma synchrony following stopsignals existed between the failed and successfully inhibited stoptrials we determined the size of the power change by calculating
the percentage change in power post versus pre stop-signal, as we
had for beta synchrony. This revealed that there was a significantly
greater change in gamma synchrony after stop-signals in which inhibition failed (t = 3.03, P = 0.01, corrected). However the time point at
which the peak in gamma synchrony following the stop-signal was
reached was not different between failed and successful stop-trials
(t = 1.29, P = 0.21).
Fig. 2. 2a Scatterplot showing the positive relationship between GORTs and the time
point at which beta activity in the STN began to desynchronise following the gosignal (r = 0.63, P b 0.05). b Relationship between GORTs and the time point at which
gamma synchronous activity was at a peak (r = 0.59, P = 0.007). Fig. 2.c relationship
between SSRTs and the time point at which beta ERS during successful stopping
exceeded that during failed stopping, relative to the SSD (see Methods section Removing the influence of SSD) (r = 0.55, P = 0.02). Negative numbers indicate quicker
changes in beta synchronous activity following the stop-signal. For all plots 95% confidence intervals of the regression line are indicated. Patient number and hand used during the task are also listed for each data point.
We then determined the mean SSD for the successfully inhibited
stop-trials for each subject individually. Unsurprisingly, during successful stop-trials, the time point at which beta became greater than
Discussion
The impact of Parkinson's disease on our results
Excessive beta activity in the basal ganglia is a hallmark of basal
ganglia activity in PD (Brown et al., 2001; Kühn et al., 2006b, 2009;
Levy et al., 2000; Ray et al., 2008; Weinberger et al., 2006, 2009). Patients were studied in the medicated state, which minimises pathological beta activity (Kühn et al., 2006b; Ray et al., 2008). However,
as with all studies undertaken during a pathological state, our results
may not transfer directly to the normal population (Kühn et al.,
2006b; Ray et al., 2008).
276
N.J. Ray et al. / NeuroImage 60 (2012) 271–278
Fig. 3. 3a is a plot of z-normalised mean beta synchronous activity during the fastest 50% of go-signals for each patient and failed stop-trials. 3b is the slower 50% of go trials and the
successful stop-trials. 3c shows the faster and slower go-trials together. Plots 3d, e and f are as above but for z-normalised mean gamma synchronous activity. 3 g shows the difference between the traces in plot 3a (solid line) against the difference between the traces in 3b (dotted line) i.e. the time evolving differences between beta synchronous activity
during the failed and successful stop-trials and their latency matched go-trials. Plot 3 h is equivalent to 3 g but for gamma synchronous activity.
Going: beta ERD
Fig. 4. Plot 4a shows the data during the successful and failed stop-trials triggered from
the stop-signal. The onset of the stop-signal is at 0 and the approximate onset of the
go-signal is indicated by the horizontal dotted line. Plot 4b is equivalent to 4a but for
gamma synchronous activity.
Consistent with previous work, go-signals were associated with
beta de-synchronisation (Cassidy et al., 2002; Kühn et al., 2004,
2006a). Also consistent with previous work (Kühn et al., 2004), we
report a correlation between the time of onset of beta desynchronisation and GORTs. The task used in the study by Kühn et
al. (2004) allowed movements to be entirely pre-prepared before
the onset of the go-signal, as a pre-cue revealed the appropriate upcoming movement. In our stop-signal task, participants did not
know which of the two movements (index finger or middle finger
push) would be required for the upcoming trial. Thus unlike go/nogo tasks, our task does not allow for pre-programing of the movement
before the action is executed. This difference may explain the weaker
correlation between event related beta desynchronisation and GORTs
in the present study (the correlation coefficient between beta desynchronisation and reaction time is greater than 0.9 in the Kuhn et al.
study, but only 0.63 in the present study (see Fig. 2a, and section
Beta synchrony and desynchronisation of the results). Thus, these
findings suggest that beta activity may remain high in the STN
while the correct movement is selected, blurring, but not obscuring,
the relationship between beta desynchronisation and reaction
times. In this regard it is interesting to note that DBS of the STN induces reduced beta synchronous activity (Kühn et al., 2009) and impairs patients' normal ability to slow decision making in the presence
of conflict (Frank et al., 2007). Also, the degree of beta synchrony reduction is found to correlate with DBS-induced treatment of the more
N.J. Ray et al. / NeuroImage 60 (2012) 271–278
akinetic symptoms of PD (Kühn et al., 2009; Ray et al., 2008). In other
words, beta activity in the STN promotes slowness, benefiting decision making by affording extra time while the appropriate action is
selected, but also contributes to the paucity of movement in PD.
Going: gamma ERS
We also found that the time point at which gamma synchronous
activity peaked was positively correlated with GORTs (Fig. 2b) in concordance with previous data showing increases in gamma synchrony
in the GPi during voluntary movement (Brücke et al., 2008). Since it
was not possible to determine the time point at which gamma activity began to synchronise, the time point at which gamma reached
peak synchrony was selected as an objective metric for the gamma
traces. Thus, the significant correlation between peak gamma synchrony and GORTs may be secondary to a relationship between
peak gamma synchrony and the onset of gamma synchrony. Nonetheless, this data supports the notion that gamma ERS may mediate
actions that follow cues.
Stopping: beta ERS
The plots in Figs. 3a, b and g reveal that stop-trials are associated
with an increase in beta synchrony following the stop-signal after
go-signal induced beta de-synchronisation. These findings corroborate previous reports that beta synchronous activity in the STN supports a stop-process. The relationship has previously been
demonstrated using go/no-go tasks in which the movement to be executed is prepared, but has not yet been initiated at the onset of the
stop-signal. Our data suggest that beta activity in the STN is also associated with the inhibition of movements that have already been set in
motion. Further, the correlation between SSRTs and the time point at
which beta activity during stop-trials becomes greater than that observed during go-trials (Fig. 2c), with the influence of SSD removed
(see Methods section Removing the influence of SSD), suggests that
individuals who are better able to inhibit responses are also those in
whom beta synchronises early following the stop-signal.
Surprisingly, we did not find that synchronous beta activity during
stopping was different during successfully inhibited and failed stoptrials when the synchronisation present in latency matched go-trials
was controlled for (Fig. 3g). This suggests that an increase in beta synchrony following a stop-signal alone does not determine success or
failure of response inhibition. Alternatively, this finding might suggest that a minimum threshold of synchronous beta activity must
be reached if actions are to be successfully inhibited. Clearly, this is
easier to do for the successfully inhibited stop trials because the
stop-signal during these trials occurs earlier following the go-signal.
Thus it may be that beta activity has not had time to desynchronise
and the threshold of synchronisation needed for inhibition is therefore easier to reach. This idea fits with our finding that the size and
the timing of event related beta synchrony triggered by the stopsignal itself is not different during failed and successful stop-trials
(Fig. 4a). Thus, beta activity synchronises similarly following stopsignals of varying delays, and it is the degree of synchrony already
present before the stop-signal occurs that determines whether response inhibition will be successful or unsuccessful. Taken together
with our finding of an association between the onset of beta synchrony and SSRTs (see Fig. 2c), this suggests that while quicker beta reactivity across subjects to stop-signals means better response
inhibition, within an individual the timing and size of the beta response does not differ during failed and successful inhibition.
Stopping: gamma ERS
Since movement occurs during failed stop-trials and none occurs
during successfully inhibited stop-trials, it was expected that
277
gamma synchronous activity would be greater in the former. Fig. 1h
certainly seems to suggest that this is the case, since gamma synchrony during failed stop-trials is elevated beyond that present during
successfully inhibited stop-trials from around the time that the
stop-signal occurs. However, we also see differences in gamma synchronous activity in the fast compared with the slow go-trials
(Fig. 3f). We therefore examined synchronous gamma activity during
failed and successfully inhibited stop-trials against the latency
matched go-trials (Figs. 3d and e respectively). It was not possible
to confirm any differences in the gamma signals statistically due to
the low signal to noise in individual datasets. However, during successfully inhibited stop-trials there was an increase in gamma synchrony that occurred prior to the increase that was seen for slower
go-trials. Conversely, during failed stop trials there was an increase
in gamma synchrony that occurred after the increase that was seen
during faster go-trials. A probable, but unexpected, explanation for
this is that stop-signals induce an increase in gamma synchrony i.e.
stop-signals during successfully inhibited stop-trials occur early in relation to the go-response, while stop-signals that occur during failed
stop-trials occur later. To test this, we triggered the data from the
stop-signal and examined the data in the gamma range. We found a
distinctive event related gamma synchrony, which unlike the beta
signal, was different for the failed compared with the successful
stop-trials. Thus, synchronous gamma activity evoked by the stopsignal occurred at similar time points following the SSD, but was significantly greater for failed compared with successful stop-trials
(Fig. 4b).
A gamma response to stop-signals was not expected. Gamma synchrony is traditionally associated with voluntary movements and responses to cued movements in the basal ganglia and cortex (Brücke et
al., 2008; Crone et al., 1998; Kempf et al., 2009; Lalo et al., 2008;
Pfurtscheller et al., 2003). However, we find increases in gamma synchrony are induced by stop-trials, and in particular by stop-signals in
which inhibition fails (Fig. 4b). It is possible that this gamma synchrony may support the stop-process, but this would be inconsistent with
previous data cited above. It is also possible that gamma simply reflects a startle response (Kempf et al., 2009) i.e. detection of an infrequent and important stimulus for the task at hand may have
increased arousal. However, a startle response would have the opposite pattern i.e. it would be greater for successful stop-trials. A more
parsimonious explanation is that the go-process, mediated by
gamma activity, is sensitive to the stop-signal, suggesting that stopsignals can influence go-process activity. This outcome supports recent work suggesting the go- and stop- processes are interactive
(Boucher et al., 2007; Lo et al., 2009) rather than independent
(Logan et al., 1984).
Gamma ERS in context with previous data
Our study is the first to report gamma synchrony after stopsignals, and is therefore inconsistent with previous work on oscillations within the GPi during a go/no-go task in which no-go trials did
not influence gamma signals (Brücke et al., 2008). However, that
paper examined gamma responses to stop-signals with zero delay
after the go-signal (i.e. during a go/no-go task). In both independent
and interactive race-models however, gamma activity after stopsignals would not synchronise for stop-trials of zero delay, as the
go-process remains inhibited by elevated stop-process activity i.e.
it is only when beta activity has become sufficiently desynchronised
following go-signals that the go process (or gamma activity related
to it) can begin to increase in strength. This fits with our finding
of an increased gamma synchrony following failed stop-trials than
successfully inhibited stop-trials, i.e. the stop-signal occurs later
during failed stop-trials, and as such, beta activity will have desynchronised to a greater degree, resulting in a less inhibited go
(gamma) response.
278
N.J. Ray et al. / NeuroImage 60 (2012) 271–278
Conclusion
We confirm that beta ERD follows go-signals, and that beta ERS
follows stop-signals during a stop-signal task. The degree of beta
ERS following stop-signals was different between individuals (those
with quicker beta ERS responses had shorter SSRTs), but not within
an individual (beta ERS following stop-signals was not different in
timing or in power for failed and successfully inhibited stop-trials
when beta activity related to the preceding go-signal was controlled
for). Thus, successful inhibition during stop-trials may depend on
beta ERS following the stop-signal, but this beta ERS mediated stopping may be successful only during trials in which beta has not fully
desynchronised following the preceding go-trial. We also found that
gamma ERS followed go-signals, and, surprisingly, was also evoked
by stop-signals. In particular, gamma ERS was strongest for failed
stop-trials. A gamma ERS response to stop-signals is a new finding,
and we suggest that the go-process may react to stop-signals by increasing in strength, perhaps in order to overcome attempts to inhibit
it. This finding provides support for interaction between go- and stopprocesses at the level of the STN.
Acknowledgments
The authors acknowledge the financial support from the UK Medical Research Council, The Norman Collisson Foundation, Charles
Wolfson Charitable Trust and the Oxford Collaborative Biomedical Research Centre.
References
Androulidakis, A.G., Kühn, A.A., Chen, C.C., Blomstedt, P., Kempf, F., Kupsch, A., Schneider,
G.-H., Doyle, L., Dowsey-Limousin, P., Hariz, M.I., Brown, P., 2007. Dopaminergic therapy promotes lateralized motor activity in the subthalamic area in Parkinson's disease.
Brain 130, 457–468.
Aron, A., Poldrack, R., 2006. Cortical and subcortical contributions to Stop signal response inhibition: role of the subthalamic nucleus. J. Neurosci. 26, 2424–2433.
Aron, A., Behrens, T., Smith, S., Frank, M., Poldrack, R., 2007. Triangulating a cognitive
control network using diffusion-weighted magnetic resonance imaging (MRI)
and functional MRI. J. Neurosci. 27, 3743–3752.
Baraniuk, R.G., Bayram, M., 2000. Multiple Window Time Varying Spectrum Estimation.
Cambridge University Press.
Boucher, L., Palmeri, T., Logan, G., Schall, J., 2007. Inhibitory control in mind and brain:
an interactive race model of countermanding saccades. Psychol. Rev. 114, 376–397.
Brittain, J.-S., Halliday, D.M., Conway, B.A., Nielsen, J.B., 2007. Single-trial multiwavelet
coherence in application to neurophysiological time series. IEEE Trans. Biomed.
Eng. 54, 854–862.
Brown, P., Oliviero, A., Mazzone, P., Insola, A., Tonali, P., Di Lazzaro, V., 2001. Dopamine
dependency of oscillations between subthalamic nucleus and pallidum in Parkinson's disease. J. Neurosci.: Off. J. Soc. Neurosci. 21, 1033–1038.
Brücke, C., Kempf, F., Kupsch, A., Schneider, G.-H., Krauss, J.K., Aziz, T., Yarrow, K.,
Pogosyan, A., Brown, P., Kühn, A.A., 2008. Movement-related synchronization of
gamma activity is lateralized in patients with dystonia. Eur. J. Neurosci. 27,
2322–2329.
Cassidy, M., Mazzone, P., Oliviero, A., Insola, A., Tonali, P., Lazzaro, V.D., Brown, P., 2002.
Movement-related changes in synchronization in the human basal ganglia. Brain
125, 1235–1246.
Chen, C., Pogosyan, A., Zrinzo, L., Tisch, S., Limousin, P., Ashkan, K., Yousry, T., Hariz, M.,
Brown, P., 2006. Intra-operative recordings of local field potentials can help localize the subthalamic nucleus in Parkinson's disease surgery. Exp. Neurol. 198,
214–221.
Crone, N., Miglioretti, D., Gordon, B., Lesser, R., 1998. Functional mapping of human
sensorimotor cortex with electrocorticographic spectral analysis. II. Event-related
synchronization in the gamma band. Brain 121, 2301–2315.
Delorme, A., Makeig, S., 2004. EEGLAB: an open source toolbox for analysis of singletrial EEG dynamics including independent component analysis. J. Neurosci.
Methods 134, 9–21.
Frank, M., Samanta, J., Moustafa, A., Sherman, S., 2007. Hold your horses: impulsivity,
deep brain stimulation, and medication in parkinsonism. Science 318, 1309–1312.
Kempf, F., Brücke, C., Salih, F., Trottenberg, T., Kupsch, A., Schneider, G.-H., Doyle Gaynor,
L.M.F., Hoffmann, K.-T., Vesper, J., Wöhrle, J., Altenmüller, D.-M., Krauss, J.K., Mazzone,
P., Di Lazzaro, V., Yelnik, J., Kühn, A.A., Brown, P., 2009. Gamma activity and reactivity
in human thalamic local field potentials. Eur. J. Neurosci. 29, 943–953.
Kühn, A.A., Williams, D., Kupsch, A., Limousin, P., Hariz, M., Schneider, G.-H., Yarrow, K.,
Brown, P., 2004. Event-related beta desynchronization in human subthalamic nucleus correlates with motor performance. Brain 127, 735–746.
Kühn, A.A., Trottenberg, T., Kivi, A., Kupsch, A., Schneider, G.-H., Brown, P., 2005. The relationship between local field potential and neuronal discharge in the subthalamic
nucleus of patients with Parkinson’s disease. Exp. Neurol. 194 (1), 212–220.
Kühn, A.A., Doyle, L., Pogosyan, A., Yarrow, K., Kupsch, A., Schneider, G.-H., Hariz, M.I.,
Trottenberg, T., Brown, P., 2006a. Modulation of beta oscillations in the subthalamic area during motor imagery in Parkinson's disease. Brain 129, 695–706.
Kühn, A.A., Kupsch, A., Schneider, G.-H., Brown, P., 2006b. Reduction in subthalamic
8–35 Hz oscillatory activity correlates with clinical improvement in Parkinson's
disease. Eur. J. Neurosci. 23, 1956–1960.
Kühn, A.A., Tsui, A., Aziz, T., Ray, N., Brücke, C., Kupsch, A., Schneider, G.-H., Brown, P.,
2009. Pathological synchronisation in the subthalamic nucleus of patients with
Parkinson's disease relates to both bradykinesia and rigidity. Exp. Neurol. 215,
380–387.
Lalo, E., Thobois, S., Sharott, A., Polo, G., Mertens, P., Pogosyan, A., Brown, P., 2008. Patterns of bidirectional communication between cortex and basal ganglia during
movement in patients with Parkinson disease. J. Neurosci. 28, 3008–3016.
Levy, R., Hutchison, W.D., Lozano, A.M., Dostrovsky, J.O., 2000. High-frequency synchronization of neuronal activity in the subthalamic nucleus of Parkinsonian patients with limb tremor. J. Neurosci. 20, 7766–7775.
Li, C.-S., Yan, P., Sinha, R., Lee, T.-W., 2008. Subcortical processes of motor response inhibition during a stop signal task. Neuroimage 41, 1352–1363.
Lo, C.-C., Boucher, L., Pare, M., Schall, J.D., Wang, X.-J., 2009. Proactive inhibitory control
and attractor dynamics in countermanding action: a spiking neural circuit model. J.
Neurosci. 29, 9059–9071.
Logan, G.D., Cowan, W.B., Davis, K.A., 1984. On the ability to inhibit simple and choice
reaction time responses: a model and a method. J. Exp. Psychol. Hum. Percept. Perform. 10, 276–291.
Pfurtscheller, G., Graimann, B., Huggins, J.E., Levine, S.P., Schuh, L.A., 2003. Spatiotemporal patterns of beta desynchronization and gamma synchronization in corticographic data during self-paced movement. Clin. Neurophysiol. 114, 1226–1236.
Ray, N., Jenkinson, N., Wang, S., Holland, P., Brittain, J., Joint, C., Stein, J., Aziz, T., 2008.
Local field potential beta activity in the subthalamic nucleus of patients with Parkinson's disease is associated with improvements in bradykinesia after dopamine
and deep brain stimulation. Exp. Neurol. 213, 108–113.
Ray, N.J., Jenkinson, N., Brittain, J., Holland, P., Joint, C., Nandi, D., Bain, P.G., Yousif, N.,
Green, A., Stein, J.S., Aziz, T.Z., 2009. The role of the subthalamic nucleus in response inhibition: evidence from deep brain stimulation for Parkinson's disease.
Neuropsychologia 47, 2828–2834.
Sharp, D.J., Bonnelle, V., De Boissezon, X., Beckmann, C.F., James, S.G., Patel, M.C., Mehta,
M.A., 2010. Distinct frontal systems for response inhibition, attentional capture,
and error processing. Proc. Natl. Acad. Sci. 107, 6106–6111.
Trottenberg, T., Fogelson, N., Kühn, A.A., Kivi, A., Kupsch, A., Schneider, G.-H., Brown, P.,
2006. Subthalamic gamma activity in patients with Parkinson's disease. Exp. Neurol. 200, 56–65.
van den Wildenberg, W., van Boxtel, G., van der Molen, M., Bosch, D., Speelman, J.,
Brunia, C., 2006. Stimulation of the subthalamic region facilitates the selection
and inhibition of motor responses in Parkinson's disease. J. Cogn. Neurosci. 18,
626–636.
Weinberger, M., Mahant, N., Hutchison, W., Lozano, A., Moro, E., Hodaie, M., Lang, A.,
Dostrovsky, J., 2006. Beta oscillatory activity in the subthalamic nucleus and its relation to dopaminergic response in Parkinson's disease. J. Neurophysiol. 96,
3248–3256.
Weinberger, M., Hutchison, W.D., Dostrovsky, J.O., 2009. Pathological subthalamic nucleus oscillations in PD: can they be the cause of bradykinesia and akinesia? Exp.
Neurol. 219, 58–61.