Age-related changes in the motor planning strategy slow down motor initiation in elderly adults

Age-related changes in the human brain functioning crucially affect the motor system, causing increased reaction time, low ability to control and execute movements, difficulties in learning new motor skills. The lifestyle and lowered daily activity of elderly adults, along with the deficit of motor and cognitive brain functions, might lead to the developed ambidexterity, i.e. the loss of dominant limb advances. Despite the broad knowledge about the changes in cortical activity directly related to the motor execution, less is known about age-related differences in the motor initiation phase. We hypothesize that the latter strongly influences the behavioral characteristics, such as reaction time, the accuracy of motor performance, etc. Here, we compare the neuronal processes underlying the motor planning of fine motor tasks between elderly and young subjects. We demonstrate that aging significantly reduces the speed of motor initiation in the dominant hand task due to the different motor planning strategies employed by elderly and young adults. Based on the results of the whole-scalp electroencephalography (EEG) analysis, we suggest that young adults tend to use the efficient and fast mechanism of motor working memory. In contrast, elderly adults involve a more demanding sensorimotor integration process similar to the non-dominant hand task.

Healthy aging affects neural processes by changing the neurochemical and structural 2 properties of the brain [1]. It determines the cognitive and motor performance decline 3 during a daily activity of elderly adults and negatively influences the quality of their life.
While the age-related differences cortical activation directly related to motor 23 execution and control is extensively studied, less is known about the effect of healthy 24 aging on the motor planning phase and its influence on RT. Exploring these mechanisms 25 is crucial to deeper understand motor control in humans. Motor planning is also 26 subjected to the age-related changes due to the following: (i) motor initiation process 27 involves many higher cognitive functions such as sensory processing, working memory, 28 motor embodiment, and sensorimotor integration [18][19][20][21], which are known to decline 29 strongly with age; (ii) the theta activity underlying the majority of these processes 30 exhibits significant age-related changes -abnormally increased theta activity in elderly 31 people indicates subjective cognitive dysfunction and suspected dementia [22,23]. 32 Based on the above, we hypothesize that the age-related changes in the motor 33 planning mechanism also affect the slowing of the motor initiation phase in elderly 34 adults. To address the issue, we considered the differences in cortical activity during the 35 controlled execution of fine motor tasks between elderly adults and young adults using 36 electroencephalography (EEG). Consistent with the dedifferentiation theory [8,17], we 37 found that the motor cortex of younger adults activated much faster during the 38 dominant hand task, while in elderly adults, the time required for motor activation was 39 equal for both hands and approached the level of the non-dominant hand of younger 40 adults. Further, as expected, we found significant differences in cortical activation 41 during the time interval preceding the motor action. In elderly adults, as well as in 42 young adults performing the non-dominant hand task, we observed the increased 43 theta-band power in sensorimotor and frontal areas, whereas theta-activation was 44 insignificant in young adults during the dominant hand task. Finally, based on the 45 results of between-subject functional connectivity analysis, we revealed that motor 46 planning involves different types of cortical interactions in young adults and elderly 47 adults, which allows concluding about age-related changes in motor planning 48 mechanisms. 49 Materials and methods 50 Participants 51 Two groups of healthy volunteers, including 10 elderly adult subjects (EA group; age: 52 65±5.69 (MEAN±SD); range: 55-72; 4 males, 6 females) and 10 young adult subjects 53 (YA group; age: 26.1±5.15 (MEAN±SD); range: 19-33; 7 males, 3 females), 54 participated in this study. All subjects were right-handed and had no history of brain 55 tumors, trauma or stroke-related medical conditions. The experimental protocol was 56 approved by the local research Ethics Committee of Innopolis University. The 57 experimental study was performed in accordance with the Declaration of Helsinki. All 58 participants were pre-informed about the goals and design of the experiment and signed 59 a written informed consent. Timelines of the experimental session (A) and a single motor task (B). Here, t b is the duration of the beep, which is 0.3 s for the LH movement command, and 0.75 s for the RH movement command.

61
All participants were instructed to sit on the chair with their hands lying comfortably 62 on the table desk in front of them, palms up. The timeline of the experimental session is 63 presented in Fig. 1A. First, the background EEG was recorded at the beginning of the 64 experiment when the participants were instructed to sit relaxed, eyes open, and not to 65 think about anything particular (5 minutes). The active phase of the experiment 66 included sequential execution of 60 fine motor tasks. Each task required squeezing one 67 of the hands into a wrist after the audio signal and holding it until the second signal (30 68 tasks per hand). The duration of the signal determined the type of movement: short 69 beep (0.3 s) was given to perform a non-dominant hand (left hand, LH) movement and 70 long beep (0.75 s) was given to perform a dominant hand (right hand, RH) movement.

71
Thus, we conducted a mixed-design experimental study with the Movement Type (LH 72 and RH conditions) as within-subject factor and the Age (EA and YA groups) as 73 between-subject factor.

74
The timeline of a single task movement is presented in Fig. 1B. The time interval 75 between the signals during the task and the pause between the tasks were chosen 76 randomly in the range 4-5 s and 6-8 s, respectively. The types of tasks were mixed in 77 the course of the session and given randomly to exclude possible training or 78 motor-preparation effects caused by the sequential execution of the same tasks. The 79 overall experimental session lasted approximately 16 minutes, including the background 80 cortical activity recording and series of motor task execution.

81
EEG data acquisition and preprocessing 82 We acquired EEG signals using the monopolar registration method (a 10-10 system 83 proposed by the American Electroencephalographic Society [24]). According to this, we 84 recorded EEG signals with 31 sensors (O2, O1, P4, P3, C4, C3, F4, F3, Fp2, Fp1, P8,   85   P7, T8, T7, F8, F7, Oz, Pz, Cz, Fz, Fpz, FT7, FC3, FCz, FC4, FT8, TP7, CP3, CPz, 86 CP4, TP8) and two reference electrodes A1 and A2 on the earlobes and a ground 87 electrode N just above the forehead. We used the cup adhesive Ag/AgCl electrodes 88 placed on the "Tien-20" paste (Weaver and Company, Colorado, USA). Immediately 89 before the experiments started, we performed all necessary procedures to increase skin 90 conductivity and reduce its resistance using the abrasive "NuPrep" gel (Weaver and  The raw EEG and EMG signals were sampled at 250 Hz and filtered by a 50-Hz 100 notch filter by embedded hardware-software data acquisition complex. Additionally, raw 101 EEG signals were filtered by the 5th-order Butterworth filter with cut-off points at 1 Hz 102 and 100 Hz. Eyes blinking and heartbeat artifact removal was performed by the 103 Independent Component Analysis (ICA) [25] motor-related activity. Data was then inspected manually and corrected for remaining 108 artifacts. Epochs which we failed to correct manually mostly due to the strong muscle 109 artifacts were rejected. Finally, each set contained 15 corrected epochs, which was equal 110 to the minimal number of the artifact-free epochs over all participants.

111
All preprocessing steps including filtering, artifact removal and epoching were 112 performed using MNE package (ver. 0.20.0) for Python 3.7 [26]. The analyzed EEG 113 data is available online [27]. were averaged over epochs for each subject.

123
Estimation of the motor brain response time. A priory knowledge about the 124 cortical activation during movements execution implies that motor brain response is 125 determined as a pronounced event-related desynchronization (ERD) of mu-oscillations 126 in the contralateral area of the motor cortex. Notably, a wide body of EEG studies 127 reports that symmetrical C4 and C3 sensors evidence brain motor response in case of 128 the left-and right-hand movements, respectively [28][29][30][31]. Here, we used mu-band 129 event-related spectral power (ERSP µ ) at C4 and C3 sensors to estimate motor brain 130 response time (MBRT) associated with LH and RH conditions for each subject of both 131 groups. We manually inspected each time-series and defined MBRT as the first 132 minimum of the mu-band spectral power below the 2.5th baseline level (Fig. 2, A). 133 Thus, we collected four sets of MBRT corresponding to each (group, condition)-set.

134
Statistical comparison of the MBRT was performed using a two-way mixed-design 135 ANOVA test implemented in JASP open-source statistical software [32]. 136 Within-subject time-frequency analyses. We performed within-subject 137 spatio-temporal clustering analyses to reveal arrays of sensors associated with the 138 motor-related brain activity for each age group and experimental condition. We 139 considered baseline-corrected topo-maps averaged in the frequency bands of interest in 140 the non-overlapping 0.2 s windows. Pairwise comparison of (time,sensor)-pairs was 141 performed via one-sampled t-test (DF = 9, p pairwise = 0.01, t critical = ±2.821) and 142 spatio-temporal clustering was assessed using non-parametric permutation test with 143 r = 2000 random permutations (p cluster = 0.01) following Maris and Oostenveld [33].

144
Between-subject time-frequency analyses. During between-subject analyses, 145 we compared brain activity of the age groups in the same experimental conditions. (p cluster = 0.05) [33].

152
Mixed-design analyses. Based on the results of within-and between-subject 153 spatio-temporal clustering analysis, we localized the effect of significant spectral power 154 change in the spatio-temporal domain. Further, for each (group, condition)-set we averaged spectral power over the corresponding spatio-temporal clusters and compared 156 it using mixed-design ANOVA.

157
Functional connectivity analysis 158 Functional connectivity measures the similarity of activation in the different brain 159 regions based on the recorded signals of brain activity. According to the review 160 papers [34,35], there exists a variety of functional connectivity metrics that evaluate 161 this similarity in the different aspects. Moreover, functional connectivity analysis based 162 on EEG or MEG recordings suffers from such problems as volume conduction/field 163 spread effect, signal-to-noise ratio, common input, etc. due to the nature of these 164 neuroimaging techniques [36]. Thus, the choice of the particular functional connectivity 165 measure requires both a prior knowledge about the analyzed neuronal processes and an 166 understanding of possible problems that may potentially interfere with the adequate 167 interpretation of functional connectivity results. 168 Functional connectivity measure. In accordance with the prior knowledge that 169 motor-related activity is associated with certain frequency bands, first of all we expect 170 the similarity of oscillatory behavior in remote brain regions in terms of phase-locking. 171 Among the variety of FC measures based on the phase-synchronization, phase lag index 172 (PLI) seems to be an appropriate metric [37]. PLI is robust to the common source 173 problem as it ignores simultaneous phase similarity, less sensitive to the intrinsic EEG 174 noise and allows reasonable interpretation of the obtained results. PLI is traditionally 175 defined as: where φ i,j (t) are phases of signals at i th and j th EEG sensors introduced via Hilbert 177 transform and operator • averaging over time points k. It clearly follows from Eq. (2), 178 that PLI lies between 0 and 1, where PLI = 1 corresponds to perfect phase-locking and 179 PLI = 0 implies a complete lack of synchrony.

180
P LI is also formulated in the frequency domain. In this case, the definition of PLI 181 given in Eq. (2) is rewritten as: where S i,j is a complex-valued Fourier-based cross-spectrum of i th and j th time-series  interval required for the brain to activate a corresponding motor area for both groups. 205 We estimated MBRT for each subject in both experimental conditions (Fig. 2B) and  Based on the above MBRT analysis, we assumed that age-related changes affecting the 224 speed of brain motor activation should be found in the pre-motor period. With this aim, 225 we performed within-subject spatio-temporal clustering analysis of the spectral power in 226 the theta and alpha/mu frequency bands for each (group, condition)-set in the premotor 227  involved sensors in the motor, frontal, and bilateral temporal areas. In the EA group, 234 strong theta-band synchronization spanned widely across the frontal and motor areas. 235 Thus, in LH condition, both groups shared the same activation mechanism and timing 236 of the motor initiation process.

237
On the contrary, the way of cortical activation during the pre-motor period in the 238 RH condition (dominant hand movement) was different in considered age groups 239 (Fig. 4). In the YA group, the theta-band spectral power did not change significantly To address the age-related changes of the pre-motor theta-band activation in detail, we 247 provided a between-subject analysis of spectral power topo-maps. in the theta-band activation was not observed. On the contrary, the between-subject 251 differences were found in the spatial cluster, which included Cp3, Cpz, and Cp4 sensors 252 (dorsal stream region of the sensorimotor area) in a 0.4-0.6s window before the RH 253 movement execution. Thus, we localized the effect in the spatio-temporal domain.

254
To estimate age-related differences of theta-band activation taking into account both 255 Age and Movement Type factors, we compared mean theta-band spectral power over 256 the evaluated spatio-temporal cluster via mixed-designed ANOVA (Fig. 5B). The  Table 4. 270 Table 3. Pre-motor theta-band spectral power (Two-way mixed-design ANOVA Summary) To support and extend our observations of the pre-motor cortical activation, we 272 explored age-related changes in terms of the underlying functional interactions between 273 remote brain regions. As we demonstrated above that theta-band activity preceded the 274 motor execution phase, we provided a between-subject comparison of the sensor-level  Here, ∆PLI defines the difference between group-level mean functional connectivity (EA versus YA). Element-wise comparison of mean connectivity matrices between Age groups was performed via unpaired t-test with p pairwise = 0.05 (dF = 9, t critical = ±2.101).
midline connections, was strongly coupled in YA group compared to EA subjects. At 278 the same time, we found the significant bilateral coupling increase between the motor 279 cortex, temporal, and frontal lobes in EA participants (Fig. 6 B). Here, the large-scale 280 neuronal communication was provided through the strong hub located in the primary 281 motor cortex (Cz-sensor), which aggregated sensory information via coupling with the 282 bilateral cortical sensorimotor circuits (C3-TP7 and C4-TP8), temporal area (Cz-FT7, 283 Cz-T7) and transferred it for further processing to the frontal area (Cz-F3, Cz-F4,

286
We considered the effect of healthy aging on the cortical activation in the motor 287 initiation phase during the controlled execution of fine motor tasks -squeezing a hand 288 into a fist paced by the audio command. We found that the time required for 289 motor-related mu-band desynchronization, which we referred to as a motor brain mechanism of use-dependent plasticity [16], causing the degradation of well-trained 309 motor functions due to the reduced activity and sedentary lifestyle of elderly individuals. 310 Also, J. Langan et al. [11] supported these results and showed less-lateralized 311 task-related motor activity in elderly adults compared to the younger control group.

312
They found that longer reaction time in elderly adults was correlated with greater 313 activation of the ipsilateral primary motor cortex during the motor task performance 314 and weaker resting-state interhemispheric coupling, which was also observed in 315 Refs. [10,38,39]. Described changes provided the compensatory mechanism to maintain 316 the level of motor performance consisting in the reorganization of functional networks 317 aimed at overcoming the age-related chemical and structural changes [14,15]. Our 318 results also evidence the motor-related over-activation of the brain areas in elderly 319 adults as a large cluster of mu-band desynchronization covering additional areas of 320 frontal, motor, parietal, and temporal regions (see Fig. 3 and Fig. 4).

321
However, the aforementioned mechanisms are not the only ones that support the 322 brain's motor response slowdown. Our results showed that the prevalent theta-band  [19,42]. In their studies, they 331 concluded that while mu-band suppression (a traditional hallmark of the motor-related 332 brain activity) reflected cortical activation directly during the motor task execution, the 333 increased theta power between stimuli presentation and motor execution was associated 334 with sensorimotor integration similarly to rodents. Along with this, several EEG-studies 335 reported the increase of theta-band power during the planning phase in the 336 choice-related, catching, and imagery motor tasks [43][44][45]. Specifically, M. Tambini et   337 al. [44] demonstrated a positive correlation between theta-power and task performance. 338 On the contrary, we found that increased theta-band power was associated with 339 prolonged motor initiation. It should be noted that the significant increase of the 340 theta-band power related to the dominant hand decline in elderly adults was observed 341 in the dorsal stream region of the sensorimotor cortex and associated with the motor 342 planning but not with audio command processing. Following the recent study by J. 343 Dushanova et al. [46], such a result should be explained by the different strategies of the 344 motor task initiation between age groups. While the degraded plasticity in elderly 345 adults requires higher cortical activation for motor planning, younger subjects optimize 346 their cognitive resources for the familiar and well-trained motor task accomplishment.

347
The latter was represented as a lower theta-band activation. Therefore, less effective use 348 of cognitive resources prolonged the motor planning phase in elderly adults compared to 349 the younger control group during the dominant hand tasks.

350
These conclusions about the age-related differences in the motor initiation strategies 351 during the dominant hand activity were supported and extended by the pre-motor 352 functional connectivity analysis. The differences in theta-band functional connectivity 353 pattern between two groups could be interpreted as a meaning of the different 354 mechanisms of motor planning in elderly adults and young subjects. First, we showed 355 the stronger neuronal interaction within the frontoparietal cognitive network involving 356 midline connections in younger adults compared to the elderly group. It reflects the 357 increased perceptual-motor facility and motor working memory [47][48][49][50]. We suppose 358 that in young adults, initiation of the familiar motor activity emphasizes motor working 359 memory and enables the formation and processing of the motor memories, i.e., the 360 stored information about the motor action obtained from prior experience, for accurate 361 motor performance [51]. The process of memory representation is fast and efficient in 362 terms of cognitive demands. On the contrary, in elderly adults, functional connectivity 363 inferred stronger coupling between the frontal, motor, and bilateral temporal areas with 364 a central hub in the primary motor cortex. As the working memory decline with age is 365 well-documented [52][53][54], we suggest that memory representation of motor actions is less 366 accessible in elderly adults. Thus, the involvement of sensorimotor integration 367 mechanisms relative to the Bland's Type 1 motor-related theta activation [19], which 368 requires higher cognitive resources, is prevalent in the elderly group. Based on the 369 above functional connectivity results, we conclude that the difference of the motor 370 planning mechanisms is more demanding and could not be optimized as well as in 371 younger subjects causing the significantly delayed motor initiation.

373
Elderly adults exhibited the approach to ambidexterity in term of the slowdown in 374 cortical activation related to the execution of the dominant hand task. We 375 demonstrated that the observed age-related loss of the dominant hand advance was due 376 to the difference in motor initiation strategy between elderly adults and young subjects. 377 Our results suggest that while young participants tend to activate motor working