Sensory Tricks Modulate Corticocortical and Corticomuscular Connectivity in Cervical Dystonia

Abstract

Objective:

To examine interactions between cortical areas and between cortical areas and muscles during sensory tricks in cervical dystonia (CD).

Methods:

Thirteen CD patients and thirteen age-matched healthy controls performed forewarned reaction time tasks, sensory tricks and two tasks replicating aspects of the tricks (moving necks/arms). Control subjects mimicked sensory tricks. Corticocortical and corticomuscular coherence values were calculated from surface electrodes placed over motor, premotor, and sensory cortical areas and dystonic muscles.

Results:

During initial preparation (after the warning stimulus), the only between-task difference was found in the γ-band corticocortical coherence (higher during tricks than during voluntary neck movements). With movements (before/after the imperative stimulus), the γ-band coherence of CD patients significantly increased during tricks but decreased during voluntary movements, while opposite trends were observed in healthy subjects. Additionally, the α- and β-band coherence decreased in healthy subjects during movements. Between the two patient subgroups (typical vs. forcible tricks), only those with typical tricks showed significant decrease in corticomuscular coherence during tricks.

Conclusions:

Observed changes in the corticocortical coherence suggest that sensory tricks improve cortical function, which reduces corticomuscular connectivity and the dystonia.

Significance:

We demonstrated that sensory tricks fundamentally affect sensorimotor integration in CD, both in movement preparation and execution.

1. Introduction

Sensory tricks, also referred to as alleviating maneuvers or ‘geste antagoniste’, denote maneuvers that temporarily relieve dystonic postures (Patel et al., 2014a). Previous studies on the phenomenology of sensory tricks (Jahanshahi, 2000; Schramm et al., 2004; Martino et al., 2010) found that sensory tricks are effective in a majority of cervical dystonia (CD) patients, ranging from 75% to 90%. In an observational study with a larger cohort of 154 CD patients (Patel et al., 2014b), providing facial stimulus (e.g., touching face) instantly improved muscle spasms and involuntary head postures for more than 80% of them (40% marked improvement, 43% partial improvement).

Despite its effectiveness, the neurophysiological mechanisms of sensory tricks – how these ‘tricks’ alleviate the dystonia – have not been clearly demonstrated. While most studies on sensory tricks remain observational in nature, several physiological investigations have been performed using different measurement modalities, including electromyography (EMG) (Wissel et al., 1999), electroencephalography (EEG) (Tang et al., 2007), transcranial magnetic stimulation (TMS) (Amadio et al., 2014), and imaging techniques (e.g., positron emission topography) (Naumann et al., 2000). The sensory aspect of the maneuver was found certainly important, as its efficacy was related to the visuotactile discrimination ability of the patients (Kägi et al., 2013). However, the neurophysiological mechanisms underlying the sensory tricks are not confined to their sensory aspect, but also concerns the preparatory phase of the intended movement to perform the tricks. Some even suggested that the sensory input (from physical interaction) itself may not be an essential part of the sensory tricks; for instance, marked reduction of the EMG activity was found to occur during arm movements, even prior to contact between fingers and facial area, for many patients (Wissel et al., 1999). Moreover, execution of the sensory tricks involves different brain regions other than the sensory area, as increased activity of the superior/inferior parietal lobes were observed during sensory tricks (Naumann et al., 2000). Their efficacy also appears to be related to desynchronization of the oscillatory activity in basal ganglia (Tang et al., 2007).

Taken together, current evidence suggests that sensory tricks are not a simple phenomenon limited to a certain aspect of motor control (e.g., either sensory or motor), but rather a complex neurophysiological process of sensorimotor integration (Ramos et al., 2014). Proper understanding of the interactions between different cortical areas (Bullmore and Sporns, 2009; Bastos and Schoffelen, 2016) would thus be necessary to explain complex neurophysiological mechanisms underlying sensory tricks. Connectivity between cortical areas was indeed found significantly reduced among patients with dystonia (Jin et al., 2011; Kukke et al., 2015; Georgescu et al., 2018), suggesting that these measures are indicative of the neural changes that contribute to abnormal sensorimotor processing.

In our recent study (Shin et al., 2021), we compared brain potentials of CD patients during sensory tricks and voluntary neck movements using a forewarned task paradigm with two sequential cues. A significant between-task difference in the brain anticipatory activities was found during motor preparation of CD patients, indicated by a significantly larger amplitude of contingent negative variation (CNV) during sensory tricks than voluntary neck movements. While this study showed that the anticipatory activities in premotor/motor areas are modulated before executing the trick, it is yet to be clarified how other cortical areas are involved in integrating sensory aspects of the tricks into the motor planning/execution (i.e., sensorimotor integration) when the tricks are actually performed.

In this study, therefore, we aimed to further elucidate the mechanism underlying the sensory tricks by examining rhythmic neuronal interactions between different cortical areas before and during sensory tricks of CD patients. Specifically, we examined the time course of the rhythmic interactions by employing a time-frequency analysis of the corticocortical coherence between different brain regions (premotor, motor, sensory), as well as corticomuscular coherence between those regions and dystonic muscles, in patients and healthy volunteers. We first hypothesized that the corticocortical connectivity during these tasks would be significantly different between healthy subjects and dystonia patients, similar to observations made during other motor tasks (Kukke et al., 2015; Georgescu et al., 2018). More importantly, we hypothesized that, during neck movements of CD patients, significant changes of brain connectivity would be induced by employing sensory tricks, which may not be present in healthy subjects mimicking the tricks. We also expected to detect significant between-condition differences (i.e., sensory tricks vs. voluntary neck movement) in the interaction between the cortical and muscle activities, quantified by the corticomuscular coherence values.

2. Methods

2.1. Subjects

Thirteen CD patients (age: 31–78, average 59.4) and 13 age-matched healthy volunteers (age: 43–82, average 58.8) participated, the same population as previously reported¹⁷. All patients demonstrated effective sensory tricks to improve their symptoms (i.e., move their heads to the neutral position) by touching their face, and at least 11 weeks from their last botulinum toxin injection at the time of their participation. The effectiveness of the sensory tricks was subjectively determined by the rater as either complete relief or partial relief, according to item C on Toronto Western Spasmodic Torticollis Scale (TWSTRS) severity scale. Participants were excluded if they had any abnormal neurological signs other than dystonia, or any history of brain tumor, stroke, or head trauma, or if they had taken benzodiazepine, anticholinergics, or selective serotonin reuptake inhibitor within the two weeks prior to the EEG recording.

The experimental protocol was approved by the National Institutes of Health Institutional Review Board. All participants gave written informed consent before participation.

2.2. Recording

EEG signals were recorded from 32 channel surface electrodes mounted on a cap (Braincap, Brain products, Germany) using the international 10–20 system referenced to Pz. EEG data, converted to the digitally linked earlobe reference, were acquired using BrainAmp (Brain products, Germany) at a sampling rate of 5 kHz and bandpass filtered (0.05Hz–100Hz).

Electromyography (EMG) signals were obtained from the right and left sternocleidomastoid (RSCM/LSCM) and the biceps brachii (BB) muscles of the arm that performed sensory tricks. The EMG signals were sampled at 1 kHz and bandpass filtered (5Hz–500Hz).

2.3. Experimental protocol

Details of the protocol of the experiment were previously reported¹⁷. Briefly, subjects were seated in a reclining armchair in a dimly lit, quiet room. Before task performance (rest/neutral condition), CD patients were instructed to keep their heads relaxed and not to correct their dystonic head position, while healthy subjects were asked to mimic dystonic posture. Subjects performed three types of forewarned reaction time tasks: ‘sensory trick (touch face)’, ‘touch shoulder’ and ‘move neck’. As both neck and arm movement are involved in performing tricks, the ‘touch shoulder’ and ‘move neck’ conditions were used to examine the effects of arm and neck movement alone, respectively. Subjects performed each task following the instructions presented on the screen (Fig. 1).

Fig. 1. Experimental paradigm of the forewarned task performance.

Stimulus 1 was seen on the screen in white letters for 2 seconds and stimulus 2 was given in green letters for 10 seconds.

A white cross was first displayed for five seconds before the warning stimulus (stimulus 1; S₁). The S₁ was then presented in white letters (‘face’, ‘neck’, or ‘shoulder’), denoting the task to be performed. After 2-second, the imperative stimulus (stimulus 2; S₂) was presented in green letters (‘touch face’, ‘move neck’, or ‘touch shoulder’), then subjects performed the instructed task. During ‘touch face’ condition, healthy subjects mimicked the patients by touching their face while moving the neck to the neutral position (sensory trick). During ‘move neck’ condition, they moved their necks to the neutral position simply by relaxing their neck muscles (i.e., stopping mimicking dystonic postures).

Each subject completed three experimental blocks of tasks, in which each of the three tasks were randomly presented 15 times; a total of 45 trials/task were included per subject.

2.4. Data processing

The following EEG electrodes were selected: electrodes over the motor cortex (C_c; either C₃ or C₄) and sensory cortex (P_c; either P₃ or P₄) contralateral to the dystonic muscle (SCM_d: RSCM or LSCM) and those ipsilateral to the dystonic side (C_i; C₄ or C₃, P_i; P₄ or P₃). Also, the two electrodes on the contralateral premotor (PM) areas (FC_c: FC₅ or FC₆) and the supplementary motor area (SMA; FC_z) were analyzed.

A notch-filter (60Hz) was first applied to remove power-line noise in the EEG signals. The filtered signals were then segmented from 2-second before S₁ (t = − 4 sec) until the end of the movement (t = 10 sec). Artifact rejection was initially performed using MARA (“Multiple Artifact Rejection Algorithm”; Winkler et al., 2011, 2014) removing components above a probability threshold of 90%. Visual inspection of the identified components was followed to further remove any remaining irrelevant components (e.g., eye blinks and/or muscle artifacts).

A notch-filter (60Hz) was also applied to the EMG data. The EMG signals were not rectified as previous studies suggested that the rectification could interfere with the identification of oscillatory input to the muscles (Neto et al., 2010; McClelland et al., 2012).

2.4.1. Corticocortical coherence

The corticocortical connectivity was assessed by coherence between the EEG channels contralateral to the dystonic muscles. The following pairs were examined: C_c-P_c (motor-sensory), FC_z-C_c, FC_c-C_c, (PM/SMA-motor), and FC_z-P_c, FC_c-P_c (PM/SMA-sensory).

For each pair, the time-dependent EEG-EEG coherence values were estimated using a sliding window of duration 250ms with an increment of 100ms, within the MATLAB environment (MathWorks, Inc., Natick, MA, US) employing a script by Neurospec (www.neurospec.org) (Halliday et al., 1995). Briefly, the magnitude-squared coherence at frequency f was computed as:

$$C_{xy}(f)=\frac{{\left|{\Phi }_{xy}(f)\right|}^2}{{\Phi }_{xx}(f){\Phi }_{yy}(f)}$$

(Φ_xx (f), Φ_yy(f): power spectra, Φ_xy(f): cross-spectrum).

The coherence values above the 95% confidence level were then z-transformed to yield values normally distributed with a standard deviation of approximately 1 (Rosenberg et al., 1989). The average of z-transformed coherence was estimated for each of the three time periods (T₁: after S₁, −2s < t < −1s; T₂: before S₂, −1s < t < 0s; and T₃: after S_2, 0s < t < 1s) for the four frequency bands (α-band: 8–12Hz, β-band: 13–35Hz, low γ-band, γ_L: 36–55Hz and high γ-band, γ_H: 66–85Hz).

2.4.2. Corticomuscular coherence

The connectivity between cortical areas and dystonic muscles was assessed by coherence between the EEG and EMG signals, which were estimated only from the movement time period (T₃: 0s < t < 1s, after S₂), during which willful contraction and/or relaxation of the muscles was performed. The coherence values above the 95% confidence level were z-transformed then integrated for the β- and γ_L-bands, where a majority of significant coherence values were observed. The following pairs were examined: C_c-SCM_d, C_i-SCM_d, P_c-SCM_d (c – contralateral; i – ipsilateral; d – dystonic).

2.4.3. Muscle activation

The activation profile of the SCM muscles was obtained by rectifying and low-pass filtering (at 5Hz) the EMG signals. For each patient, the average muscle activity during the movement (0 < t < 10sec) was compared to those before the movement (−4 < t < 0sec) to compute the relative change in the muscle activity (change in mean activation level; ΔMAL), which were compared between the two conditions (sensory trick vs. move neck) to evaluate the efficacy of the trick for each patient. Previous studies reported different patterns of sensory tricks across dystonia patients, such as decrease in dystonic muscles by the touch (‘typical’ sensory tricks) or ‘forcible’ tricks that require voluntary activation of antagonistic muscles (Ramos et al., 2014).

2.4.4. Statistical analysis

Changes in the corticocortical coherence values (Δ_LC) during each of the three time periods (T₁ – T₃), in comparison to the baseline, were compared between the groups (healthy vs. patient) and the tasks (sensory trick, move neck, touch shoulder) by multivariate analyses of variance (IBM SPSS Statistics; IBM Corp., Armonk, NY, USA). The log change in percent (L%; Δ_L) was used to address asymmetry and non-additivity of the relative measure (Törnqvist, 1985; Cole, 2000). For pairwise comparisons, the p-values were corrected for multiple comparisons (Bonferroni adjustment). A post hoc univariate analysis of variance (ANOVA) was implemented to the coherence values of each pair to avoid type-I error (Wetcher-Hendricks, 2011). Significance level was set to 0.05.

Changes in the muscle activation during movements, quantified by the log change in percent (ΔMAL), were compared between the two tasks involving neck muscles (sensory tricks and move neck) by a paired t-test, which were used to distinguish two types of tricks (typical vs. forcible tricks). Differences in the EEG-EMG coherence during movement period (0 < t < 10sec) of patients between the sensory tricks and the move neck conditions were tested by a paired t-test.

3. Results

3.1. Corticocortical connectivity/coherence

Significant between-task or between-group differences were found between the move neck and sensory tricks conditions, while no between-task or group differences were found between the move neck and touch shoulder conditions. The observed between-task (move neck vs. sensory tricks) and between-group differences became larger as the time progressed. During the early preparation period (T₁), only a between-task difference in the γ_H-band was found (p = 0.003; Table 1a), while no significant group-effect or task×group interaction was observed. During the late preparation period (T₂), in contrast, a between-group difference and task×group interaction were observed in the β- and γ_L-bands (Table 1b). During the movement execution (T₃: after S₂), significant effects of the task, group, and their interaction were found across all frequency bands (Table 1c), indicating the corticocortical connectivity significantly differed between the tasks (sensory tricks vs. move neck), as well as between the groups (CD patients vs. healthy).

Table 1.

p-values from multivariate analysis of variance (MANOVA) and post-hoc analysis of variance (ANOVA) for change in the corticocortical coherence (Δ_LC)

(a) Period 1 (T1: after S1; −2s < t < −1s)
Band*	α (pt=.11; pg=.30; pi=.70)					β (pt=.66; pg=.42; pi=.42)					γL (pt=.24; pg=.87; pi=.67)					γH (pt=.01; pg=.60; pi=.94)
Region**	M-S	P-M		P-S		M-S	P-M		P-S		M-S	P-M		P-S		M-S	P-M		P-S
Pair	Cc-Pc	Cc-FCz	Cc-FCc	Pc-FCz	Pc-FCc	Cc-Pc	Cc-FCz	Cc-FCc	Pc-FCz	Pc-FCc	Cc-Pc	Cc-FCz	Cc-FCc	Pc-FCz	Pc-FCc	Cc-Pc	Cc-FCz	Cc-FCc	Pc-FCz	Pc-FCc
Task	.99	.66	.02	.56	.43	.23	.62	.40	.60	.37	.02	.18	.45	.16	.16	.01	.02	.09	.05	.00
Group	.40	.96	.58	.39	.94	.94	.78	.98	.17	.56	.25	.69	.51	.23	.46	.62	.82	.93	.35	.45
Interaction	.46	.85	.22	.40	.82	.28	.39	.11	.93	.39	.52	.26	.65	.84	.70	.98	.78	.453	.49	.90
(b) Period 2 (T2: before S2; −1s < t < 0s)
Band*	α (pt=.48; pg=.13; pi=.59)					β (pt=77; pg=.04; pi=.78)					γL (pt=.43; pg=.84; pi=.09)					γH (pt=.42; pg=.93; pi=.63)
Region**	M-S	P-M		P-S		M-S	P-M		P-S		M-S	P-M		P-S		M-S	P-M		P-S
Pair	Cc-Pc	Cc-FCz	Cc-FCc	Pc-FCz	Pc-FCc	Cc-Pc	Cc-FCz	Cc-FCc	Pc-FCz	Pc-FCc	Cc-Pc	Cc-FCz	Cc-FCc	Pc-FCz	Pc-FCc	Cc-Pc	Cc-FCz	Cc-FCc	Pc-FCz	Pc-FCc
Task	.92	.56	.21	.70	.60	.34	.66	.44	.91	.83	.04	.67	.25	.38	.25	.10	.47	.10	.31	.03
Group	.82	.40	.71	.04	.07	.30	.73	.20	.02	.19	.84	.71	.37	.93	.78	.33	.57	.99	.63	.63
Interaction	.54	.43	.65	.50	.87	.79	.44	.32	.92	.37	.39	.02	.17	.66	.07	.15	.15	.19	.23	.10
(c) Period 3 (T3: after S2; 0s < t < 1s)
Band*	α (pt=.08; pg=.01; pi=.99)					β (pt=.49; pg=.00; pi=.33)					γL (pt=.20; pg=.60; pi=.01)					γH (pt=.24; pg=.83; pi=.48)
Region**	M-S	P-M		P-S		M-S	P-M		P-S		M-S	P-M		P-S		M-S	P-M		P-S
Pair	Cc-Pc	Cc-FCz	Cc-FCc	Pc-FCz	Pc-FCc	Cc-Pc	Cc-FCz	Cc-FCc	Pc-FCz	Pc-FCc	Cc-Pc	Cc-FCz	Cc-FCc	Pc-FCz	Pc-FCc	Cc-Pc	Cc-FCz	Cc-FCc	Pc-FCz	Pc-FCc
Task	.67	.79	.05	.84	.91	.93	.18	.35	.11	.80	.31	.62	.43	.64	.22	.66	.95	.21	.42	.05
Group	.70	.22	.79	.00	.07	.12	.39	.29	.00	.07	.68	.28	.12	.29	.94	.72	.27	.96	.28	.96
Interaction	.95	.91	.82	.78	.93	.43	.12	.67	.05	.06	.07	.00	.72	.27	.08	.08	.21	.62	.18	.31

^*p_t: p-value for task-effect; p_g: p-value for group-effect; p_i: p-value for task-group interaction

^**M-S: Motor-Sensory; P-M: Premotor/supplementary-Motor; P-S: Premotor/supplementary-Sensory

^***Shaded area indicates factors identified as statistically significant by MANOVA (p < 0.05) or trend towards significance (p < 0.1).

Similar patterns of the corticocortical coherence changes were observed across the periods (Fig. 2). When significant group effects were found, the coherence values of the healthy controls decrease while those of the CD patients remained unchanged (or increased) with movement. Between tasks, the coherence values generally increased during the sensory trick, but remained unchanged (or decreased) during the move neck task. The coherence decreased during sensory tricks and increased during move neck for healthy subjects, while the opposite trend was observed for CD patients (task×group interaction).

Fig. 2. Change in the corticocortical coherence values of electrode pairs, grouped by the cortical regions, in different phases.

A. Period 1 - early preparation phase (−2s < t < −1s); B. Period 2 - late preparatory phase (−1s < t < 0s); C. Period 3 - movement phase (0s < t < 1s) (Electrodes - M: motor; P: premotor; S: sensory). During Period 1, the increase in the γ-band coherence was greater for sensory trick for both groups. During period 2, the degree of increase in the β-band coherence was greater for cervical dystonia (CD) patients. The γ-band coherence increased during sensory trick but not during move neck for CD patients, while an opposite patterns was observed for healthy. During Period 3, some degree of group- and task-effects, as well as their interactions were found in different bands (α-, β-, and γ-bands).

When different pairs were compared (spatial distribution), connections to the premotor areas (FC_z/FC_c) were found affected largely in the early periods, while the connections to the sensory areas (P_c) were mainly affected in the later periods.

1. Early preparation period (T₁; Fig. 2a), only significant task-effects were observed. The γ_H-band coherence increased during sensory tricks for the pairs including premotor area (FC_z/FC_c), but remained same or decreased during move neck for both groups.

2. Late preparation period (T₂; Fig. 2b), significant effects of both group and task×group interaction were found: the β-band coherence values between premotor and sensory areas (FC_z/FC_c-P_c) generally decreased for healthy subjects, while they did not significantly change for CD patients (group-effect). The γ_L-band coherence values of healthy subjects were not different between the tasks, while those of CD patients between premotor and motor areas (FC_z/FC_c-C_c) were higher during sensory tricks than move neck (task×group interaction).

3. Movement period (T₃; Fig. 2c), all three factors (task, group, task×group) were found significant. In the α- and β-bands, similar to Period 2, the coherence values decreased for healthy subjects, but not for CD patients (group-effect). Significant task×group interaction was found in three bands (β-, γ_L-, γ_H-bands), all of which exhibited similar patterns; for patients, corticocortical coherence increased during sensory tricks but decreased during move neck, while opposite trends were observed in healthy. For most bands, this trend was observed in the pairs including the sensory area (P_c).

A typical pattern of the between-task difference in CD patients is described in a time-frequency plot of the corticocortical coherence of a patient (patient 4; Fig. 3), for whom significant changes were observed during move neck, but not during sensory tricks.

Fig. 3. Between-task difference: Representative time-frequency plot of the corticocortical coherence of the three pairs (C_c-P_c, FC_z-C_c, FC_z-P_c) (Patient 4).

A. Move neck; B. Sensory tricks. For this patient, during move neck (T₂), the coherence values in the low and high γ-bands (γ_L: 35 – 55Hz, γ_H: 65 – 85Hz) started to decrease about 1-second before the imperative stimulus (S₂), then remained lower during the movement (white box), when compared to the pre-movement period (black box) (A). However, such patterns were not observed during sensory tricks (T₁) (B).

3.2. Muscle activities during sensory trick: two subgroups

For six patients (Patient 2, 4, 5, 6, 9, 12; subgroup 1), the increase in the SCM muscle activity during sensory tricks (ΔMAL: mean±SD = 2.3±31.5% for dystonic side; 22.9±70.6% for contralateral side) was found to be significantly smaller than the activity during move neck condition (mean±SD = 164.5±187.1% for dystonic side; 68.3±130.1% for contralateral side; ‘typical’ trick). For the others (Patients 1, 3, 7, 8, 10, 11, 13; subgroup 2), the change in the SCM muscle activity was similar between the sensory tricks (mean±SD = 57.4±117.3% for dystonic side; 18.0±25.5% for contralateral side) and the move neck (mean±SD = 61.8±135.4% for dystonic side; 22.6±29.9% for contralateral side) conditions, consistent with the description of ‘forcible tricks’ in the literature (Ramos et al., 2014).

3.3. Corticomuscular connectivity/coherence

When the two conditions were compared (sensory tricks vs. move neck), the two subgroups of patients showed contrasting trends in the change in their corticomuscular coherence. For subgroup 1 (typical tricks), the corticomuscular coherence values during sensory trick condition was significantly smaller, when compared to move neck condition, for all three pairs: mean±SD percent difference = −95%±61% for motor area-dystonic muscle (p = 0.01), −101%±78% for premotor area-dystonic muscle (p = 0.01), and −59%±51% for sensory area-dystonic muscle (p = 0.02), as shown in the representative case (subject 12; Fig. 4). However, for subgroup 2 (forcible tricks), the corticomuscular coherence values were similar or often greater during sensory trick condition: mean±SD percent difference = 10%±24% for motor area-dystonic muscle (p = 0.24), 11%±21% for premotor area-dystonic muscle (p = 0.01), and 3%±25% for sensory area-dystonic muscle (p = 0.02). We did not find significant differences between the two subgroups in disease duration and total TWSTRS score. TWSTRS severity score was higher in patients with typical sensory tricks (p = 0.01) (see Figure S1 in Supplementary Material).

Fig. 4. Between-task difference: Representative time-frequency plot of the corticomuscular coherence of the patients in the subgroup 1 (typical sensory tricks).

Under move neck condition (A), the corticomuscular coherence values of most pairs (motor-sensory, premotor-motor, premotor-sensory) significantly increased after the warning stimulus S₁ (t = −2 sec) and remained at the level during the movement (0 sec < t < 4 sec). In contrast, the corticomuscular coherence values remained low throughout the movement under sensory tricks condition (B). Increase in the activation level of both sternocleidomastoid (SCM) muscles was significantly lower during sensory tricks condition.

Among the thirteen CD patients, the peak frequency of the corticomuscular coherence during the voluntary move neck movements was found mostly in the low γ-band (36Hz – 55Hz; 6 patients), as shown in a representative case (Fig. 4), or in the high γ-band (56Hz – 75Hz; 3 patients). For four CD patients, no clear peak was observed in the coherence profile.

4. Discussion

In this study, we examined the time course of the rhythmic interactions between cortical areas of CD patients and healthy volunteers during sensory tricks (mimicked by the healthy volunteers) and other voluntary movements. Significant differences in the corticocortical coherence were observed between voluntary neck movements and sensory tricks in CD patients, but not in healthy controls, supporting our hypothesis that sensory tricks affect sensorimotor integration process of CD patients. Two subgroups of patients were identified (typical vs. forcible trick), for whom different patterns of the corticomuscular coherence were observed.

4.1. Effects of sensory tricks on corticocortical coherence

The observed differences between tasks and groups in brain connectivity indicate that sensory tricks fundamentally affect the sensorimotor integration of CD patients, both in movement preparation and execution.

The main effects of sensory tricks on cortical connectivity appeared to emerge during the late preparation phase, which persisted through the movement phase. During the early preparation phase (T₁), no between-group difference was found, and the observed between-task differences may simply represent the differences in motoneuron excitability due to task condition. In contrast, during the late preparation phase (T₂), different trends emerged between CD patients and healthy subjects. In the β-band, corticocortical coherence generally decreased in healthy subjects, while no significant changes were observed in CD patients. More importantly, the γ_L-band coherence of CD patients significantly increased during sensory tricks more than during voluntary neck movements, which was not the case in healthy subjects, indicating that sensory tricks enhanced the brain connectivity of CD patients in this band during the preparatory phase (T₂). Such between-group and between-task differences appear to be further amplified during the movement phase (T₃) (Fig. 2).

Our observation supports the notion that dystonia is a disorder of movement preparation (Hallett, 2000) as well as execution. During the course of sensory trick, we identified a spatial shift in the changes in the corticocortical connectivity, from the premotor area to the sensory area. Connectivity to the premotor areas was mostly affected during the preparation period, while the between-task difference was mainly observed in the pairs including sensory areas during the movement period. This indicates that the effects of sensory tricks on the sensorimotor integration are initiated at the motor preparation stage. Note that deficits in movement preparation were demonstrated in focal dystonia patients (abnormal premovement gating of somatosensory input) (Murase et al., 2000). Other studies also suggested that actual sensory inputs during sensory tricks may not be required to relieve dystonic symptoms, as the reduction in dystonic muscle activity was found to be initiated before ‘touch’ (Wissel et al., 1999). Our recent study (Shin et al., 2021) also demonstrated that late CNV of CD patients was significantly greater during sensory trick, indicating that sensory tricks may normalize impaired motor preparation in dystonia. The current study further demonstrated that significant between-task differences in brain connectivity of patients emerge during late preparation phase, which were further amplified through the execution periods while shifting (spatially) from premotor to sensory areas. These findings further support the notion that the sensory tricks fundamentally affect the sensorimotor integration process, initiated at the motor preparation stage.

It should be noted that a significant task-group interaction in corticocortical coherence was observed mainly in the γ-band (Fig. 2), suggesting that the intracortical connectivity in the γ-band may represent the neural mechanisms of the sensory tricks in CD. In healthy subjects, the γ-band coherence has been linked to higher-level (cognitive) functions, such as sensory information processing (Miltner et al., 1999), skill acquisition (Serrien and Brown, 2003), and perceptual learning (Gruber et al., 2002). The γ-band corticocortical coherence was found to be reduced in dystonia patients performing simple tasks, particularly between the sensory and motor areas (Melgari et al., 2013), indicating that the decrease in the γ-band coherence reflects a reduced efficiency of sensorimotor integration in dystonia. A recent case study with a CD patient also demonstrated that the γ-band connectivity (measured by magnetoencephalography) increased by sensory tricks (Mahajan et al., 2018). Our study further provides a detailed description of the spatiotemporal changes in the intracortical connectivity during sensory trick, which span from motor preparation to execution of the trick (temporal), and from premotor to motor to sensory areas (spatial).

4.2. Corticomuscular coherence and muscle contraction: two subgroups of CD patients

In accordance with the literature, we identified two subgroups of CD patients during sensory trick. For about the half of subjects (6 patients), activity of the dystonic muscles significantly decreased during sensory trick. However, for the other seven patients, no significant difference in the muscle activity was found between the two conditions (sensory tricks vs. move neck), which is consistent with the description of ‘forcible tricks’ in the literature (Ramos et al., 2014; Ochudlo et al., 2007). The changes of corticomuscular coherence during sensory tricks were significantly different between these subgroups, suggesting that the connectivity between the central (brain) and the periphery (muscles) indicates different mechanisms of sensory tricks. Previous studies attempted to link several subject-specific neurophysiological features, including biomechanics (head position; Schramm et al., 2004), sensory impairments (visuotactile discrimination; Kägi et al., 2013), and severity of the symptom (Patel et al., 2014), to the efficacy of sensory tricks (forcible tricks vs. typical tricks). As of yet, however, differences in the neurophysiological mechanisms underlying these variants of sensory tricks remain unclear. Our study demonstrated that the degrees of the corticomuscular synchronization (i.e., coherence) could present a direct measure to differentiate typical and forcible tricks. Note that previous studies showed that the abnormality in sensorimotor processing in dystonia can be indicated by the changes in the corticomuscular/intermuscular coherence (McClelland et al., 2020).

It is possible that the between-group difference in the corticomuscular coherence could be, in part, attributed to the difference in the higher muscle activation level of the patients who used forcible tricks. For instance, a previous study found that, during a force-tracking task, a significant modulation in the corticomuscular coherence was observed as the force level increases (Chakarov et al., 2009). However, in our study, the increase in the corticomuscular coherence was found largely in the γ-band (9 out of 13 patients). Other studies (Andrykiewicz et al., 2007; Omlor et al., 2007) also showed that the force magnitude does not modulate the γ-band coherence during simple dynamic force modulations, while that the force level could mainly affect the β-band corticomuscular coherence. As the difference in the corticomuscular coherence was observed mainly in the γ-band (Fig. 4), the observed between-group difference in coherence is unlikely due to their difference in the muscle activation level.

4.3. Limitations of the study

Due to the exploratory nature of the study, the number of subjects was not large enough to further analyze the difference between the two subgroups of CD patients (typical tricks vs. forcible tricks). The observed between-group differences in the brain activity could be in part due to the difference in the nature of the tasks performed, as control subjects were asked to mimic the dystonic symptoms of the patients. Note that, however, our results do not depend entirely on group differences since a significant between-task differences were also found. Lastly, source localization was not performed due to the relatively small number of electrodes used in the experiment.

Highlights.

• Corticocortical coherence was significantly higher during sensory tricks of CD patients, compared to voluntary neck movements.

• The main effects of sensory tricks on coherence emerged during late preparation phase, then persisted through the movement phase.

• Corticomuscular coherence was found lower for CD patients who demonstrated typical tricks.