Aging Associated Changes of Movement-Related Functional Connectivity in The Human Brain
Aging Associated Changes of Movement-Related Functional Connectivity in The Human Brain
Aging Associated Changes of Movement-Related Functional Connectivity in The Human Brain
PII: S0028-3932(18)30306-3
DOI: https://doi.org/10.1016/j.neuropsychologia.2018.07.006
Reference: NSY6842
To appear in: Neuropsychologia
Received date: 20 October 2017
Revised date: 15 May 2018
Accepted date: 6 July 2018
Cite this article as: Nils Rosjat, Liqing Liu, Bin A. Wang, Svitlana Popovych,
Tibor Tóth, Shivakumar Viswanathan, Christian Grefkes, Gereon R. Fink and
Silvia Daun, Aging-associated changes of movement-related functional
connectivity in the human brain, Neuropsychologia,
https://doi.org/10.1016/j.neuropsychologia.2018.07.006
This is a PDF file of an unedited manuscript that has been accepted for
publication. As a service to our customers we are providing this early version of
the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting galley proof before it is published in its final citable form.
Please note that during the production process errors may be discovered which
could affect the content, and all legal disclaimers that apply to the journal pertain.
Aging-associated changes of movement-related functional
connectivity in the human brain
Nils Rosjata,b, Liqing Liua,b, Bin A. Wanga, Svitlana Popovycha,b, Tibor Tóthb, Shivakumar
Viswanathana,c, Christian Grefkesa,c, Gereon R. Finka,c, Silvia Dauna,b*
a
Cognitive Neuroscience, Institute of Neuroscience and Medicine (INM-3), Research Centre Jülich, D-52425
Jülich
b
Heisenberg Research Group of Computational Neuroscience - Modeling Neural Network Function, Institute
of Zoology, University of Cologne, D-50674 Köln
c
Department of Neurology, University Hospital Cologne, D-50937 Köln
*
Corresponding author. Prof. Dr. Silvia Daun Cognitive Neuroscience, Institute of
Neuroscience and Medicine (INM-3), Research Centre Jülich, 52425 Jülich, Germany Tel.:
+49-2461-61 8638; fax: +49-2461-61 1518. [email protected]
Abstract
Motor performance declines with normal aging. Previous neuroimaging work revealed aging-
related general increases in neural activity, especially in the prefrontal and pre-motor areas,
underlying these changes and their relation to aging-associated motor decline to date remain
elusive.
To further elucidate the neural processes underlying aging-related motor decline, we recorded
EEG from younger and older subjects while they performed a finger-tapping task. As a measure
Behavioral data revealed increased movement times in older subjects. EEG data showed that
phase locking in the δ-θ frequencies is a general, age-independent phenomenon underlying the
execution of simple finger movements. In stark contrast, the extent of synchronization between
motor areas significantly differed dependent upon age of subjects: multiple additional intra- and
inter-hemispheric connections were observed in older subjects. Our data shed light upon the
1
results of previous neuroimaging studies showing aging-related increases in neural activation.
In particular, data suggest that the observed aging-dependent substantial intra- and inter-
the previously reported loss of lateralization in older subjects. The changes observed are likely
2
1. Introduction
Aging leads to increased difficulties in the coordination of movements, reflected by, e.g., an
increased variability in movement execution (e.g., Wu and Hallett, 2005; Seidler et al., 2010).
The wealth of data showing that human motor performance is affected by normal aging is
contrasted by the dearth of data providing information about the neural processes underlying
these changes.
Several fMRI studies suggest that the neural activity associated with self-initiated movements
changes during normal aging. An enhanced general activation, especially in the pre-frontal
and pre-motor areas, has been reported (Sailer et al., 2000; Heuninckx et al., 2005; Ward et
al., 2008), often referred to as the PASA (posterior-anterior shift in aging) phenomenon
(Dennis and Cabeza, 2008). Furthermore, consistent with the HAROLD (Hemispheric
Asymmetry Reduction in Older Adults) model, which proposes that reduced hemispheric
phenomenon (for a summary see Cabeza, 2002), reduced hemispheric asymmetry during
unilateral movements (Sailer et al., 2000; Rowe et al., 2006; Langan et al., 2010) has been
reported. The less lateralized activity patterns during cognitive or motor performance in older
relative to younger adults are assumed to reflect either compensation, i.e. the recruitment of
additional supporting brain regions to offset the effects of aging-related cognitive (or motor)
decline (e.g. Cabeza et al., 1997; Nolde et al., 1998; Reuter-Lorenz et al., 2000) or,
To date, the neural mechanisms underlying these reported aging-related changes in neural
activation remain elusive. They may result, for instance, from changes in inter-regional neural
cognitive tasks (e.g. Fell et al., 2004; Popovych et al., 2016). Neural synchronization refers to
the precise and coordinated timing of the activity of distinct neural populations (for a review,
3
see Palva and Palva, 2012), which is fundamental for motor and cognitive function (e.g.
different motor tasks have found that oscillations of remote neural populations exhibit short
coherence hypothesis posits that this transient synchronization reflects increased functional
patterns (Fries, 2005, 2015). This synchronization is believed to underlie the communication
Here, we investigated whether and to what extent the age-related changes in neural activation,
relationship.
Phase locking of neural activity is a key process of neural synchronization. A general statistic
to quantify phase locking in networks is the phase locking value (PLV) (Lachaux et al., 1999).
The PLV is a measure of variability in the phase relationship of neural activity at two distinct
locations. A large PLV implies low variability (nearly constant phase relation) and hence
strong connectivity between two distant neural populations. In contrast, a low PLV is a sign of
weak connectivity between two distinct neural populations. Task-related changes in the PLV
baseline).
To this end, here we analyzed EEG data recorded from younger and older participants, while
they performed a simple finger-tapping task. We specifically asked whether the functional
networks, defined using the relative phase locking values (rPLV), systematically vary between
4
younger and older age groups during movement preparation, execution, and outcome
evaluation.
Participants
We used EEG data of twenty-one younger healthy individuals (10F/11M, age: 22-35 years) as
well as a group of thirty-one older healthy subjects (15F/16M, age: 60-78 years) from two
earlier studies (Popovych et al., 2016 and Liu et al., 2017, respectively). In both groups, all
1971), had normal color vision, and no history of psychiatric or neurological disease (as
assessed by the Trail making test (Spreen and Strauss, 1998), the Mini-mental-status-test
(Folstein et al., 1975), the Clock-drawing test (Agrell and Dehlin, 1998), and the Beck
depression inventory (Beck et al., 1961). All participants gave their written informed consent
prior to the study. Both studies were approved by the local ethics committee of the Faculty of
Experimental design
We recorded EEG data of both groups of subjects while they performed a simple finger-
(Kornhuber and Deecke, 1965) where the subjects were asked to voluntarily press a button
with their left or right index finger every 4-8s. They were requested to perform the
movements in random order, i.e., without using a systematic pattern, with a roughly equal
number of left and right finger movements. The subjects were asked to avoid conscious time-
estimation and trial-balancing in order not to evoke brain activity unrelated to the motor task
itself. Subjects also separately performed a visually-cued task, where a left or right button
press was executed in response to a visual stimulus, and a vision-only task, which had the
5
same stimuli as in the visually-cued task but where no actions were required (for full details
see Popovych et al., 2016, Liu et al., 2017, Wang et al., 2017). The whole experiment lasted
about 70 min.
Our analysis was restricted to data acquired from the self-initiated condition, as this condition
was uncontaminated by the processing of visual stimuli. Each participant performed between
During the experiments, we continuously recorded EEG data from 64 active Ag/AgCl
electrodes (Brain Products GmbH, Munich, Germany), which were placed according to the
international 10-20 system. Following the generally accepted practice, the reference electrode
was placed at the left earlobe. Bipolar horizontal and left vertical electro-oculograms (EOG)
were recorded using three of the 64 scalp electrodes (FT9, FT10 and TP10 in the 10-20
system). These were placed at the bilateral outer canthi and under the left eye to monitor eye
movements and blinks. Before the experiment, we ensured that the impedance of the
electrodes was below 15 kΩ. The EEG signals were amplified, band-pass filtered in the
frequency range 0.87-500 Hz, and digitized at a sampling rate of 2.5 kHz. An acceleration
sensor was attached to the tip of each index finger to detect the movement onset. We used
these sensors instead of EMG electrodes, since the former are more sensitive to small
movements and are not restricted to specific muscle groups. We also used the information
from the acceleration signals to monitor the behavior of the subjects, for example, to exclude
errors, such as mirror movements. We defined the onset of the finger movement as the instant
of time at which the numerical time-derivative of the acceleration signal crossed a predefined
threshold.
6
The acquired raw data were band-pass filtered from 0.5-48 Hz to remove slow voltage drifts
from the data and down-sampled from 2.5 kHz to 200 Hz to reduce the file size and thereby
by means of independent component analysis (ICA) (Makeig et al., 1996). Finally, the data
were epoched to intervals centered around the movement onset. In the self-initiated tapping
had to be located before the start of the Bereitschaftspotential (readiness potential; Shibasaki
and Hallett, 2006). Thus, the epoch or trial was defined to be the interval [-2500, +1500] ms
For our analysis, we excluded trials containing EEG artifacts as well as trials that were
preceded by an interval of less than 4s from the previous button-press. Trials containing
movements during the baseline period, as detected using the accelerometer, were also
excluded. After excluding trials based on these criteria, we only included data from
participants who had at least 30 correct trials per hand remaining. Based on these quality
considerations, data from three younger subjects and seven older subjects were excluded from
further analysis. Thus, EEG data from 18 younger and 24 older subjects were used for further
processing.
Spatial filtering
A crucial pre-processing step was spatial filtering. A major challenge in estimating network
relationships with EEG is the phenomenon of volume conduction, whereby signals from a
particular neural generator can influence voltages at multiple distinct electrodes on the scalp
filtering using surface Laplacians has been shown to substantially reduce this problem,
thereby enhancing spatial resolution and enabling analyses of electrodes close to the region of
interest (Lachaux et al., 1999, Tenke et al., 2012). Therefore, our preprocessed data were re-
7
referenced to small Laplacian reference in order to improve the spatial resolution of the
signals and thus enhance their suitability for the subsequent connectivity analysis (Hjorth,
1975).
Phase-locking analysis
using Morlet wavelets (Lachaux et al., 1999) in a range of 2-48 Hz with a step size of 1 Hz (5
cycles). This time-frequency decomposition was performed using the Statistical Parametric
Mapping toolbox (SPM12, Wellcome Trust Centre for Neuroimaging, London, UK)
we analyzed the phase information gained from the time-frequency decomposition using
The complex Morlet wavelet transformation was used to compute the time evolution of the
amplitude A(f, t) and phase ϕ(f, t) for each frequency separately. In order to quantify
synchronization between the activities of the different brain regions, we used the single-
frequency phase-locking value (sPLV; adapting the phase-locking value defined in Lachaux et
8
1
!"#$,% (&, ') = -. exp /0 23$4 (&, ') − 3%4 (&, ')67-
+
9:;
Here 3$4 denotes the phase of the EEG at channel m in the k-th trial. N is the total number of
trials, and i is the imaginary unit. sPLV can take values in the interval [0,1]. While sPLV=0
represents a random distribution of phase differences over all trials without any coherence,
8
sPLV=1 occurs only in the case of perfect inter-trial phase locking of the phase differences
between the EEG signals at the two channels m and n over all trials.
Since we were interested in the synchronization effects during movement preparation and
movement execution, and not in persistent phase synchronization, we normalized the sPLV of
EEG at each pair of channels with respect to its baseline value and calculated its relative
change over the whole epoch. We refer to these normalized sPLVs as ‘relative phase-locking
values’ (rPLV):
Here, >>>>>>>>>>>>>>
!"#$,% (&) denotes sPLV of the baseline for each frequency, i.e., the mean sPLV at
frequency f in the baseline interval. The baseline interval used to compute >>>>>>>>>>>>>>
!"#$,% (&) was
from [-2300, -1500] ms after excluding the first 200ms of the original [-2500,-1500] ms
regions, we considered four major frequency bands: δ (2-3 Hz), θ (4-7 Hz), α (8-12 Hz), and β
(13-30 Hz). A widely used procedure to evaluate phase locking in these frequency bands is to
filter the signal to the frequency band of interest and calculate a single phase-locking value for
the whole frequency band. Here, we instead calculated the rPLV for each single frequency
separately with a resolution of 1 Hz before averaging these values over the frequencies of the
frequency band. This was done to maximize the contribution of each individual frequency to
the observed phase locking. Our primary focus was the average rPLV over the δ-θ frequency
9
band (2-7 Hz) where we previously found evidence of considerably local phase locking
Network of interest
We chose to perform our connectivity analysis on the level of electrodes rather than on
observed oscillations. As our focus was on the dynamics of the motor network, our analysis
was restricted to network relationships among ten electrodes clustered bilaterally over the
In our previous studies, the spatial distribution of local phase-locking in the δ-θ frequency
band (2-7 Hz) for unilateral movements was consistent with the lateralization of the motor
system, namely, being higher at electrodes contralateral to the moved hand than at ipsilateral
electrodes (Popovych et al., 2016; Liu et al., 2017). Based on this observation, the rPLV
values of electrode pairs during left and right finger movements were collapsed together based
on whether the electrodes were contralateral or ipsilateral to the moved hand. For
convenience, electrodes over the right hemisphere were assumed to be contralateral to the
The resulting network consisted of a total of 45 possible pairwise connections (Figure 1). Of
the 45 possible connections, there were 14 intra-hemispheric ones per hemisphere (6 between
non-medial electrodes and 8 between lateral and medial ones). One connection (FCz – Cz)
was strictly medial. The remaining 16 connections were strictly inter-hemispheric, i.e.,
We additionally included an occipital electrode Oz as a control electrode. As the task did not
involve any task-relevant visual stimuli, the rPLV of connections from the electrodes of
10
Statistical analysis.
To test whether the difference between the mean rPLV (averaged over frequency band, see
Figure 2A) and the baseline was statistically significant, we compared the rPLV obtained for
each pair of electrodes at each time point in the interval [-1500, 1500] ms with its baseline
value using a pointwise t-test with a significance level of p < 0.05, corrected for multiple
comparisons (false discovery rate, FDR, q = 0.05) (Benjamini and Hochberg, 1995) (Figure
2B, middle column). The correction was performed with regard to the number of conditions,
time points, age groups, and electrodes. The baseline was constructed by generating normally
distributed random values that had the same mean value and standard deviation as the EEG at
For the construction of the rPLV networks, we defined a 'connection', i.e., an edge of the
network, between two given nodes to exist if rPLV between these two nodes was significant.
Our analysis was restricted to the time interval [-300, 500] ms, which was divided up into a
preparatory phase [-300, 0] ms, an execution phase of the movement [0, 200] ms, and a post-
movement phase [200, 500] ms. Moreover, we split the whole interval [-300, 500] ms into
bins of 100 ms length in order to assess the dynamics of the transient synchrony between the
electrode pairs of interest. Two electrodes were regarded to be connected in such a subinterval
if rPLV between them was significant over the whole duration (100 ms) of this interval (solid
line), or at least for 75 ms in this time interval (dashed line in Figure 2B, right column).
3. Results
Behavioral results.
We first tested whether motor performance differed between groups. To this end, we measured
the movement time, i.e., the time from movement onset (determined using the accelerometer)
to the actual button press. We tested the movement times of younger (M = 81.9 ms, Std = 8.6)
11
vs. older (M = 88.3 ms, Std = 8.3) subjects for significant differences and found that the finger
movements of older subjects were significantly (on average 6 ms) slower than those of the
Prior to any test of age-related differences, we first sought to evaluate whether the expected
task-induced changes in phase-locking were statistically detectable at all over the electrode
pairs in the motor network (Figure 1). This was indeed the case:
For both populations, the relative phase-locking value (rPLV) in the δ−θ frequency band
showed a significant increase over the baseline pre-, during, and post-movement (t=0 denotes
movement onset) for many electrode pairs (increase in rPLV up to 50%)(Figure 3). Significant
phase locking occurred mainly between movement-related electrodes (see Figure 3, rPLV
and older subjects. Importantly, phase-locking between electrodes in the motor network and
the occipital control electrode Oz was comparatively negligible (bottom row, Figure 3A,B).
Since our analysis revealed that inter-regional phase locking in the δ−θ frequency band was
indeed movement-related, occurring in both younger and older subjects, we next turned to
we first analyzed the time evolution of synchronization in both age groups separately. The
rPLV was used to construct network snapshots of the dynamic connectivity process during the
different stages of the action (i.e., preparation, execution, and post-movement) defined by
each of eight 100 ms subintervals from -300 ms before movement onset to +500 ms after
12
movement onset (Figure 4). If the rPLV was statistically significant between two nodes of the
For each age-group, the networks defined by connections that survived continuously over the
entire 100 ms of each subinterval (solid lines) as well as those that only survived for 75 ms of
that subinterval (dashed lines) are shown in Figure 4, first row (younger participants) and
Rather than a uniform loss or gain of connections, the network for older participants exhibited
a complex relationship to that of the younger participants showing both a gain in connections
(Figure 4, third row) while accompanied by a loss of other connections (Figure 4, fourth row).
Only a small fraction of connections were shared across both groups over all subintervals
When lowering the threshold for connectivity (see Methods) in our analysis from 100ms to
75ms, we observed that the resulting networks underwent only minor changes, i.e., a maximal
increase by five edges per time-interval was observed (see dashed lines in Figure 4).
Importantly, these minor changes had virtually no bearing on the network properties reported
here. The network properties in question can thus be regarded as robust against small-to-
We next tested whether the strength of phase locking differed between the two groups. Our
test was restricted to connections that survived for at least one 100 ms interval in either age
group, namely, all connections displayed in the first and second rows of Figure 4 (solid lines).
Age-related modulations of the mean rPLV of the above connections over each 100 ms
interval was assessed with a three-way ANOVA with factors Age Group {Young, Old} x Time
{8-subintervals} x Connection {23-pairs}. The 3-way interaction between these factors was
not statistically significant [F(154,15242) = 0.44, p > 0.05]. However, we found a significant
interaction between age group and connection (F(22,15242) = 11.75, p < 0.0001). The main
13
effects of all factors were also statistically significant [age group: (F(1,15242) = 64.52, p <
0.0001); time: F(7, 15242) = 64.4, p < 0.0001; connection: (F(22,15242) = 15.5, p < 0.0001)].
A posteriori t-tests on these mean rPLVs revealed that the EEG of younger subjects exhibited
significantly higher phase locking in 30 pairs of electrodes in various time intervals than the
EEG of older subjects (Figure 5 top row, solid lines) (all ts(82) > 1.7622, all ps < 0.05, FDR
corrected q = 0.05). The EEG of older subjects showed an increased phase locking in only 7
pairs of electrodes (all ts(82) < -1.7131, all ps < 0.05 FDR corrected q = 0.05) relative to that
The strength of mean rPLV differed between the age groups for several connections as evident
in Figure 5 (top row). To further assess the age-dependence of mean rPLV strength at each of
these connections, we pooled the two populations together and tested the correlation between
the numerical age and mean rPLV (averaged in each of the 100 ms time intervals) for these
connections. There was a significant negative correlation between age and inter-regional
phase-locking (i.e., rPLV) for several electrode pairs, namely, F3 - FCz, F3 - C4, FC3 – FCz,
FC3 – C4, C4 –CP4, C3 – CP4 and FCz – C4 (all ρs < -0.2467, all ps < 0.05, FDR corrected q
= 0.05) (Figure 5, bottom row). That is, the older the subjects, the smaller was the rPLV for
those connections.
Since movement time was significantly higher for the older group relative to the younger
group (see Behavioral results above), we also tested whether rPLV was correlated with
movement time. However, this correlation was not statistically significant suggesting that the
age-dependence of the rPLV did not directly arise due to movement differences.
Betweenness centrality.
14
Applying the search procedure for hub nodes in networks (i.e., nodes through which the most
paths with the shortest length pass (Bullmore and Sporns, 2009)), which are usually the ones
with the largest degree, to the networks in Figure 4 suggests that the brain region at both
nodes Cz and FCz revealed hub-like properties during the preparatory, the executional, and
Network summary.
some simplifying anatomical assumptions: We assumed that the channels C4 and CP4
represent activity of the contralateral sensorimotor cortex (cSM), C3 and CP3 that of the
ipsilateral sensorimotor cortex (iSM). The frontal and fronto-central electrodes F4 and FC4
were assumed to represent activity from the contralateral prefrontal and pre-motor cortex
(cPFC/cPM), and similarly, electrodes F3/FC3 represent ipsilateral iPFC/iPM. Finally, the
central electrodes Cz and FCz were deemed to reflect activity related to the medial pre-frontal
cortex and the supplementary motor area (mPFC/SMA). Based on these simplifying
assumptions, Figure 6 summarizes the results obtained for the network of ten electrodes in
Figure 4 in a compact form. As in Figure 4, edges between two brain regions denote
significant, temporary inter-regional phase locking that existed in a given phase of the
networks in the case of separately analyzed left and right hand tappings can be found in the
In this simplified form, we see that the motor network for younger subjects (top row, Figure
mPFC/SMA - cSM (light green) and iPFC/iPM - cSM (orange), which persisted in all phases
of the movement. Only in the post-movement phase, an additional edge between cPFC/cPM
15
While some of the connections representing phase locking and being present in younger
subjects were significantly reduced in older ones (Figure 6, second row, black solid lines),
certain other connections emerged (Figure 6, second row, dashed lines). For instance, an
additional connection between iSM and mPFC/SMA was established in the preparatory phase
of the movement. During movement execution three connections showed an increased phase
inter-hemispheric edge connecting iSM and cSM, and an inter-hemispheric edge connecting
iPFC/iPM and cPFC/cPM, the latter being also present in the post-movement phase.
In addition to the significantly decreased phase locking observed in older subjects for the
connections that form the sub-network in younger subjects, we found significant correlations
of the mean rPLV of these connections with the age of the subjects in all phases of the
movement. That is, the older the subject the weaker was the phase locking between iPFC/iPM
and mPFC/SMA and between iPFC/iPM and cSM in all movement phases. Furthermore, the
phase locking between mPFC/SMA and cSM was negatively correlated with age in the post-
movement phase.
between different brain regions. The majority of the edges lasted over several phases of the
4. Discussion
We carried out inter-regional phase-locking analyses on EEG data obtained from younger and
older healthy subjects, while they performed a simple self-initiated finger-tapping task. We
compared the time evolution of inter-regional phase locking between the participating brain
synchronization between the involved neural populations. Phase locking was assessed, since it
constitutes a key process underlying the communication between neural populations (Moore
16
et al., 2006; Schmorrow and Fidopiastis, 2011). According to the communication through
between neural populations expressed by coherently coordinated firing patterns (Fries, 2005,
2015).
To gain an overall picture of the functional connectivity during motor task performance, we
calculated the rPLV, which measures the extent of instantaneous synchronization between two
distinct brain regions. We thereby identified a network of areas associated with the motor task.
The great advantage of the rPLV is that it is capable of extracting information from fast
transient signals. Other methods, like dynamic causal modeling (DCM) or Granger causality,
require statistically stationary signals. To start with, the nodes were assigned to electrodes of
interest and the edges represented significant inter-regional rPLVs computed in the δ−θ
frequency band from the EEGs at these electrodes in all movement phases (Figure 4). We thus
obtained dynamic networks that describe the temporal changes in functional connectivity in
the different movement phases. Special nodes, so-called hubs, essentially nodes with largest
degrees, were identified within these networks. During almost the whole length of the
movement, FCz and Cz turned out to constitute such hub nodes, in both age groups.
Moreover, one node occasionally took over this role from the other one. This hub-property
emphasizes the functional significance of the central brain region, where the medial PFC and
Although a reconstruction of source activity would probably have provided more detailed
information on the origin of the oscillations, we performed our connectivity analysis on the
level of electrodes. We did so in order to avoid the ill-posed inverse problem, which is only an
assumption of the source activity and could thus lead to flawed interpretations (Bradley et al.,
2016; Grech et al., 2008; Wendel et al., 2009). We used the small Laplacian reference to
substantially reduce the effect of volume conduction on our data (Lachaux et al., 1999).
Though not an explicit source localization method, the small Laplacian reference renders the
17
electrodes maximally sensitive to radial sources that directly lie beneath them (Rigoni et al.,
2013; Steinmetz et al., 1989; Tenke et al., 2012). Another way of dealing with the problem of
common sources could have been to compute measures that are not sensitive to 0-phase-lag
interactions, e.g. weighted phase-lag index or imaginary PLV (Vinck et al., 2011; Nolte et al.,
2004). Even so, our analysis revealed only a small number of neighboring electrodes that
were connected (as opposed to all 45 possible connections cf. Figs. 1 and 3). Furthermore,
neighboring connections that were present in younger subjects were absent in older ones. It is
therefore unlikely that the reported connections would solely result from volume conduction.
They are much more likely to arise from distinct brain sources.
At the second stage, we identified invariant sub-networks that existed in all or almost all
phases of the movement. These sub-networks occurred in both age groups, albeit that they
fMRI connectivity data in healthy younger participants who performed a left- or right-handed
fist closure task. Here, a connectivity increase from SMA to cSM was observed (Grefkes et
al., 2008; Pool et al., 2013). However, given the poor time resolution of fMRI, such studies
could not identify the connectivity networks that were active during the different phases of a
simple and rapid motor task (i.e., preparation, execution, and post-movement evaluation).
In older subjects, in contrast to younger ones, the strength of these rPLVs, i.e., the strength of
the connections forming this strongly lateralized sub-network, was significantly reduced.
Conversely, phase locking between iM1 and the central areas as well as between contralateral
frontal and the central areas was significantly higher than in younger subjects. Additionally,
older subjects showed a significant increase in inter-hemispheric phase locking, i.e., between
(i.e., a loss of lateralization) in the motor and prefrontal cortex of older subjects. The data are
compatible with a DCM analysis of fMRI data reported by Boudrias and colleagues (Boudrias
18
et al., 2012), which revealed that older subjects display stronger connections onto iSM from
cSM.
Further fMRI studies showed a strong general activation, especially in the PFC and the (pre)
motor areas (Sailer et al., 2000; Heuninckx et al., 2005; Ward et al., 2008), as well as a loss of
lateralization in older subjects (Sailer et al., 2000; Rowe et al., 2006; Langan et al. 2010).
Based on our results, we hypothesize that the changes observed in neural activation result, at
least in part, from changes in phase locking, i.e., from altered patterns of transient
synchronization between the participating brain regions. Specifically, our data suggest that i)
cM1), and ii) the increase in intra- and inter-hemispheric connectivity (i.e., iSM-mPFC/SMA,
lateralization in older subjects. Thus, there is not only an increase in overall activity, which is
commonly interpreted as a loss of lateralization in the power spectral density (Sailer et al.,
In an earlier study (Popovych et al., 2016), we found an increase in local phase locking in the
δ−θ frequency band of EEG at the electrodes that lie above the motor areas of the cortex
contralateral to the moving hand. We hypothesized that the preparation and execution of
motor actions, irrespective of whether the movement was initiated internally or by an external
cue, rely on local phase locking in the motor cortex prior to the actual execution, i.e., during
preparation. In our follow-up study (Liu et al., 2016), we found that this local phase locking is
age independent. The fact that local synchronization is age independent while the inter-
regional one changes quite drastically with age is compatible with the 'network view' of the
brain region is not independent of the effects of aging in other brain regions involved in a
certain task, i.e., that it rather results from a global network change.
19
Summing up, we found that inter-regional phase locking in the δ−θ frequency band between
and within the frontal and motor regions underlies the execution of a finger tapping task.
Since enhanced activity of, and also increased connectivity between, brain regions is often
Grady, 2012), we suggest that the observed increase in the inter-hemispheric and cPFC/cPM-
mechanism that aim at counteracting the loss of the strongly lateralized sub-network
motor act.
Older subjects showed a significant increase (about 6 ms) in movement time, i.e., the time
from movement onset to the button press. One reason for increased movement times might be
the reduced conduction velocity of nerve fibers in older subjects (Nishihara et al., 2013).
Another reason could be, since we did not find any correlation between rPLV and movement
time, that the brain of older subjects compensated for the loss in iPFC/iPM-mPFC/SMA-cSM
which subjects would have performed worse (i.e., when relying solely on the sub-network
present in younger subjects only). Our results indicate that not a single connection but an
the naturally occurring decline of motor functions. Our results therefore support the
compensation hypothesis of the HAROLD model, which proposes that age-related asymmetry
reductions may help to counteract neurodegenerative decline. Finally, our results did not
reveal evidence for the realization of the PASA phenomenon through phase locking in the δ-θ
frequency range.
20
Role of the funding source
This work was funded by the University of Cologne Emerging Groups Initiative (CONNECT
group) within the framework of the Institutional Strategy of the University of Cologne and the
Conflict of interest
References
Agrell B, Dehlin O (1998) The clock-drawing test. Age and Ageing 27:399-403.
Babcock RL, Laguna KD, Roesch SC (1997) A comparison of the factor structure of
processing speed for younger and older adults: Testing the assumption of measurement
Baker SN, Olivier E, Lemon RN (1997) Coherent oscillations in monkey motor cortex and
hand muscle EMG show task-dependent modulation. The Journal of physiology 501(1):225-
241.
Baker SN, Spinks R, Jackson A, Lemon RN (2001) Synchronization in monkey motor cortex
of Neurophysiology 85(2):869-885.
Beck AT, Ward CH, Mendelson M, Mock J, Erbaugh J (1961) An inventory for measuring
21
Benjamini Y, Hochberg Y (1995) Controlling the false discovery rate: a practical and powerful
(Methodological):289-300.
Boudrias MH, Gonçalves CS, Penny WD, Park CH, Rossiter HE, Talelli P, Ward NS (2012)
Age-related changes in causal interactions between cortical motor regions during hand grip.
Neuroimage 59(4):3398-3405.
Bullmore E, Sporns O (2009) Complex brain networks: graph theoretical analysis of structural
Cabeza R (2002) Hemispheric Asymmetry Reduction in Older Adults: The HAROLD Model.
Cabeza R, Grady CL, Nyberg L, McIntosh AR, Tulving E, Kapur S, Jennings JM, Houle S,
Craik FIM (1997) Age-related differences in neural activity during memory encoding and
Dennis NA, Cabeza R (2008) Neuroimaging of healthy cognitive aging. In The Handbook of
Aging and Cognition (Craik FIM, Salthouse TA, eds.), pp1-54. New York: Psychology Press,
3rd Edn.
Fernandez G (2004) Neural bases of cognitive ERPs: more than phase reset. Journal of
Folstein MF, Folstein SE, McHugh PR (1975) “Mini-mental state”. A practical method for
grading the cognitive state of patients for the clinician. Journal of Psychiatric Research
12(3):189–198.
22
Fries P (2015) Rhythms for cognition: communication through coherence. Neuron 88(1):220-
235.
13(7):491-505.
Grech R, Cassar T, Muscat J, Camilleri KP, Fabri SG, Zervakis M, Xanthopoulos P, Sakkalis
V, Vanrumste B (2008) Review on solving the inverse problem in EEG source analysis.
Grefkes C, Nowak DA, Eickhoff SB, Dafotakis M, Küst J, Karbe H, Fink GR (2008) Cortical
connectivity after subcortical stroke assessed with functional magnetic resonance imaging.
Hjorth, B. (1975). An on-line transformation of EEG scalp potentials into orthogonal source
Lachaux JP, Rodriguez E, Martinerie J, Varela FJ (1999) Measuring phase synchrony in brain
Langan J, Peltier SJ, Bo J, Fling BW, Welsh RC, Seidler RD (2010) Functional implications
Liu L, Rosjat N, Popovych S, Wang BA, Yeldesbay A, Tóth TI, Viswanathan S, Grefkes CB,
Fink GR, Daun S (2017) Age-related changes in oscillatory power affect motor action. PLoS
Makeig S, Bell AJ, Jung TP, Sejnowski TJ (1996) Independent component analysis of
23
Mitrushina M, Satz P (1991) Analysis of longitudinal covariance structures in assessment of
Moore RA, Gale A, Morris PH, Forrester D (2006) Theta phase locking across the neocortex
Nishihara K, Kawai H, Kanemura N, Hara M, Naruse H, Gomi T (2013) A novel approach for
Nolde SF, Johnson MK, Raye CL (1998) The role of prefrontal cortex during tests of episodic
Nolte G, Bai O, Wheaton L, Mari Z, Vorbach S, Hallett M (2004). Identifying true brain
interaction from EEG data using the imaginary part of coherency. Clinical Neurophysiology
115:2292-2307.
Oldfield R (1971) The assessment and analysis of handedness: The Edinburgh inventory.
Neuropsychologia 9:97–113.
Pool E-M, Rehme AK, Fink GR, Eickhoff SB, Grefkes C (2013) Network dynamics engaged
Popovych S, Rosjat N, Tóth TI, Wang BA, Liu L, Abdollahi RO, Viswanathan S, Grefkes C,
Fink GR, Daun S (2016) Movement-related phase locking in the delta–theta frequency band.
NeuroImage 139:439-449.
Reuter-Lorenz PA, Jonides J, Smith ES, Hartley A, Miller A, Marshuetz C, Koeppe RA (2000)
Age differences in the frontal lateralization of verbal and spatial working memory revealed by
24
Rigoni D, Brass M, Roger C, Vidal F, Sartori G (2013) Top-down modulation of brain activity
underlying intentional action and its relationship with awareness of intention: An ERP/Laplacian
(2006) Aging is associated with contrasting changes in local and distant cortical connectivity
Sailer A, Dichgans J, Gerloff C (2000) The influence of normal aging on the cortical
Directing the Future of Adaptive Systems: 6th International Conference, FAC 2011, Held as
Part of HCI International 2011, Orlando, FL, USA, July 9-14, Proceedings (Vol. 6780).
Seidler RD, Bernard JA, Burutolu TB, Fling BW, Gordon MT, Gwin JT, Kwak Y, Lipps BB
(2010) Motor control and aging: links to age-related brain structural, functional and
117(11):2341-2356.
Tenke CE, Kayser J, Manna CG, Fekri S, Kroppmann CJ, Schaller JD, Alschuler DM, Stewart
JW, Mcgrath PJ, Bruder GE (2012) Current source density measures of EEG alpha predict
Uhlhaas PJ, Roux F, Rodriguez E, Rotarska-Jagiela A, Singer W (2010) Neural synchrony and
25
Vinck M, Oostenveld R, van Wingerden M, Battaglia F, Pennartz CMA (2011) An improved
Wang BA, Viswanathan S, Abdollahi RO, Rosjat N, Popovych S, Daun S, Grefkes C, Fink
Ward NS, Swayne OB, Newton JM (2008) Age-dependent changes in the neural correlates of
doi:10.1155/2009/656092
Physiol 562(2):605-615.
26
Fig 1: Summary of all possible connections in the network. 14 possible ipsilateral intra-
(green) are shown on the left and 16 possible inter-hemispheric ones on the right (orange).
corresponding phase locking networks from the rPLVs. See text for details.
Fig 3: Colormap showing the time evolution of the rPLV for all electrode pairs of interest (cf.
Methods, Figure 1). The upper part of each of the two subfigures displays the rPLV between
electrode pairs) and the bottom part of each subfigure the one between movement-related
electrodes and Oz (Oz) in younger (A) and older subjects (B) in the δ-θ frequency band,
averaged over 2-7 Hz. Only significant rPLV values are shown with their respective colors (p
< 0.05, FDR corrected, cf. Methods). Non-significant ones are displayed in green (0%
Fig 4: Time evolution of rPLV networks. The networks are computed from EEG signals at the
length. Significant inter-regional phase locking (rPLV) of EEG between electrode pairs for the
whole 100 (or at least 75) ms is represented by an edge drawn in solid (or dashed) line.
Results on the left and right index finger-tapping movements have been merged and are
shown in the two upper rows for younger and older subjects, as indicated. Connections are
classified as ipsilateral (blue), contralateral (green), and inter-hemispheric (orange). The 2nd
and 3rd row depict the connectivity changes in the older subjects by displaying additional
connections and lost connections, respectively. The last row shows the connections that are
Fig 5: Results of the statistical analysis. Top row: connections for which rPLV was
significantly different between age groups. Solid lines depict rPLVs that were significantly
27
decreased, dashed lines the ones that were significantly increased in older subjects. Bottom
row: Connections for which rPLVs correlated negatively with the age of the subjects.
Fig 6: Summary of the results presented as a network of five virtual electrodes: younger
subjects (first = upper row); significant differences between connections (rPLVs) in younger
and older subjects (second row); and connections that significantly correlated with age (third
row). The networks in each row appeared during the preparation [-300,0] ms, the execution
phase of the movement [0,200] ms, and in the post-movement phase [200,500] ms.
Significantly decreased connections (rPLVs) are shown as solid lines, significantly increased
connections as dashed lines, and correlations of rPLVs with age as dash-dotted lines.
The connections in the top row are drawn using the same color-codes as in Figure 4, with the
green) and those from contralateral PFC/PM (cPFC/cPM) to mPFC/SMA (dark green).
28
Figure 1
Figure 2
Intra-hemispheric Inter-hemispheric
F3 F4 F3 F4
C3 Cz C4 C3 Cz C4
Oz Oz
Figure 3
Figure 4
Significant phase locking
A rPLV (in %)
50
40
ipsilateral
30
20
10
contralateral
0
Younger
-10
-30
Oz
-40
-50
-2500 -2000 -1500 -1000 -500 0 500 1000
rPLV (in %)
50
B 40
ipsilateral
30
20
10
contralateral
0
Older
-10
-30
Oz
-40