Transcutaneous vagus nerve stimulation at nonspecific timings during training can compromise motor adaptation in healthy humans

Abstract

Adding afferent vagus nerve stimulation to motor training via implanted electrodes can modify neuromotor adaptation depending on the stimulation timing. This study aimed to understand neuromotor adaptations when transcutaneous vagus nerve stimulation (tVNS) is applied at nonspecific timings during motor skill training in healthy humans. Twenty-four healthy young adults performed visuomotor training to match a complex force trajectory pattern with the index and little finger abduction forces concurrently. Participants were assigned to the tVNS group receiving tVNS at the tragus or the sham group receiving sham stimulation to the earlobe. The corresponding stimulations were applied at nonspecific timings throughout the training trials. Visuomotor tests were performed without tVNS or sham stimulation before and after training sessions across days. The reduction in the root mean square error (RMSE) against the trained force trajectory was attenuated in the tVNS group compared with the sham group, while its in-session reduction was not different between groups. The reduction of RMSE against an untrained trajectory pattern was not different between groups. No training effect was observed in corticospinal excitability or GABA-mediated intracortical inhibition. These findings suggest that adding tVNS at nonspecific timings during motor skill training can compromise motor adaptation but not transfer in healthy humans.

NEW & NOTEWORTHY Adding vagus nerve stimulation via implanted electrodes during motor training can facilitate motor recovery in disabled animals and humans. No study examined the effect of transcutaneous vagus nerve stimulation (tVNS) during training on neuromotor adaptation in healthy humans. We have found that adding tVNS at nonspecific timings during motor skill training can compromise motor adaptation but not transfer in healthy humans.

INTRODUCTION

Developing and understanding noninvasive interventions for neuromotor adaptation are valuable for motor recovery and skill improvement in disabled and nondisabled individuals. Applying afferent vagus nerve stimulation (VNS, i.e., electrical stimulation of the afferent vagus nerve) during motor training is an emerging intervention that can modulate neural and functional motor adaptations substantially. VNS can induce central neuromodulations through the termination of afferent vagus nerves in the nucleus tractus solitarius (NTS) and its projection to neuromodulatory nuclei in the subcortical and cortical regions (1–3). Protocols and efficacies of VNS have been studied systematically in rats and mice via an implanted VNS device. For neuromodulation, brief VNS transiently enhanced the activity of adrenergic and cholinergic axons in the cortex, which were associated with enhanced motor activity (3, 4). In addition, motor training focusing on the distal or proximal forelimb in rats enlarged the corresponding representation in the motor cortex by pairing VNS with the training (5). For motor function, adding VNS to forelimb training led to a greater motor improvement compared with the training without VNS in nondisabled mice (4) and rats disabled due to an acute ischemic lesion (6–9), subcortical intracerebral hemorrhage (10), and traumatic brain injury (11). Although the number of human studies is small and the tested populations are limited to poststroke individuals, applying VNS during upper-limb motor rehabilitation enhanced functional motor recovery in humans as well (12–14). Nonetheless, all these studies applied VNS invasively via an implanted device, and all human studies applying VNS during training tested disabled individuals only. The fundamental effects of VNS on neuromotor adaptation in healthy humans are unknown because no study systematically examined the efficacy of applying VNS during motor training in nondisabled individuals.

The critical challenge of studying the effect of VNS intervention on basic human adaptation is the invasive procedure. The implantation of a VNS device involves various risks and higher costs. Exploring a noninvasive alternative to implanted VNS is crucial for studying basic human adaptation and providing more accessible procedures to clinical populations. Transcutaneous VNS (tVNS) is available that stimulates the afferent auricular branch of the vagus nerve at the outer ear (15–18). tVNS does not directly stimulate the vagus nerve as implanted VNS does and therefore may not have the same effect. Nevertheless, the feasibility, safety, and potential effects of isolated tVNS applications (i.e., not combined with motor training) have been explored in humans, including those with neurological symptoms, such as epilepsy, depression, and tinnitus (15, 17–19). Acute neuromodulations with isolated tVNS were confirmed in resting corticospinal excitability and GABA-mediated intracortical inhibition via transcranial magnetic stimulation (TMS) and encephalogram (EEG) in humans (20–23). The effect of pairing tVNS with motor training on TMS measures in healthy humans is unclear, either acute or chronic. Only one study applied tVNS during motor training in humans, which aimed to test the feasibility in poststroke individuals (24). This pilot study without a control group was not designed to determine the specific effect of adding tVNS to training. Hence, the effect of adding tVNS to motor training in healthy humans is unknown. Examining the effect of tVNS in healthy individuals is fundamental for future understandings of its potential utility and neurophysiological mechanisms for modulating the achievement of motor skills in humans. It also helps to obtain baseline knowledge in healthy humans for future examinations of injury- or patient-specific efficacies and involved mechanisms. Thus, the current study aimed to understand the effect of adding tVNS during motor training on adaptations in motor skills and corticospinal and intracortical excitability in healthy humans.

A number of animal studies have shown that the timing of VNS application during training is critical to the efficacy of VNS for neuromotor adaptation (4–10, 25, 26). Facilitatory effects on motor adaptations were observed only in the animals receiving VNS at a specific timing, i.e., immediately after “successful” motor trials specifically but not those receiving VNS at nonspecific timings in disabled and nondisabled rats and mice (5–10, 25). Interestingly, applying VNS at nonspecific timings during visuomotor skill training compromised early motor skill adaptations in healthy mice (4). These differences suggest that altering neuromodulators (e.g., noradrenaline, acetylcholine, GABA) can cause distinct effects depending on their temporal association with specific neuromotor activities. Without a previous study on timing specificity in humans, it was not feasible for us to design an appropriate experimental protocol for defining “successful” trials and yielding an unknown optimal frequency of “success” for adaptations in healthy humans. Therefore, we chose to apply tVNS at nonspecific timings during visuomotor skill training in healthy young humans in this first exploration into the emerging intervention. Furthermore, the study included examining the transfer of motor adaptation since improving motor skills beyond trained tasks is important in training and rehabilitation.

METHODS

Subjects

Twenty-four healthy young adults (18–30 yr old, means ± SD age: 21.4 ± 3.5 yr, 11 females) participated in the study. Participants provided written informed consent upon enrollment. They reported no history of any neuromuscular injury or disorder. Subjects were right-handed, and hand dominance was confirmed through the Edinburgh Handedness Inventory (27). Individuals who regularly perform activities that involve skilled use of the hands (e.g., professional musicians) were excluded. A screening questionnaire confirmed eligibility for transcranial magnetic stimulation (TMS) in each subject. The experimental procedure was approved by the Institutional Review Board of the Georgia Institute of Technology (H18151). This project is registered on ClinicalTrials.gov (NCT03628976).

Overall Design

Subjects were pseudorandomly assigned to the tVNS or sham group in equal numbers so that the difference in initial motor skill between groups becomes small. We employed the visuomotor task of abducting the index and little fingers since the primary muscles for each action are both innervated by the same nerve (i.e., the ulnar nerve). The experiment consisted of five sessions of testing and practice (Fig. 1). In each session, the first testing preceded the practice, followed by the second testing. Subjects performed visuomotor skill practice (pattern A) for five sessions with concurrent electrical stimulation to the auricular afferent vagus nerve at the tragus (tVNS group) or the earlobe (sham group) as outer ear stimulation. The visuomotor skill of the trained task (pattern A) was tested without outer ear stimulation before and after practice in each session. To examine the transfer of motor adaptation, the visuomotor skill of an untrained motor task (pattern B) was tested in the first and last sessions without outer ear stimulation. TMS tests were performed without outer ear stimulation before and after practice on the first and last sessions during a comparable time of day. Experimental sessions were scheduled for up to 48 h between consecutive sessions.

Figure 1.

Summary of overall research design. A, pattern A; B, pattern B; MVC, maximal voluntary contraction; TMS, transcranial magnetic stimulation.

Subject Setup

Subjects were seated in an upright chair with a computer monitor placed ∼50 cm away. The monitor displayed the real-time visual feedback of the abduction forces produced by the index and little fingers during motor tasks and a cross symbol for subjects to focus on during other tasks. Subjects extended their right arm along with the adjacent table and placed their right hand in a hand task device. The shoulder was slightly abducted in a naturally comfortable position, the elbow was flexed, and the pronated forearm was rested on a table. The hand task device was assembled with a base and brackets to house two force transducers (model 31 or model 34, depending on the task, Honeywell, Charlotte, NC) fitted to their hand. Each force transducer was horizontally placed against the proximal interphalangeal joints of the index finger and little finger to measure their abduction forces. A block designed to fit the contour of a hand was secured in the center of the board. The middle finger, ring finger, and thumb were constrained with Velcro straps. The thumb was constrained to a bracket. The force transducers were powered, and the signals were amplified with a powered amplifier (Transbridge TBM4M, World Precision Instruments, Sarasota, FL). The amplified force signal was routed to an analog-to-digital conversion board (BNC-2111 M Series, National Instruments, Austin, TX) and a computer and sampled at 1,024 samples/s.

Electromyogram

Electromygram (EMG) was recorded from the first dorsal interosseous (FDI) and the abductor digiti minimi (ADM) with pairs of surface electrodes attached to the overlaying skin. After cleaning the skin surface with abrasive gel, a pair of surface electrodes was placed in a belly-tendon configuration and taped down for stability. The electrodes were plugged into battery-powered amplifiers (Y03, Motion Lab Systems, Baton Rouge, LA). A grounding wristband was placed around the wrist. EMG signals were band-pass filtered (15–2,000 Hz) and amplified. The signals were digitally converted at 5,000 samples/s and collected (Power1401 and Signal, Cambridge Electronic Design, Cambridge, UK) for monitoring and analysis.

Maximal Voluntary Contraction

Forces during maximal voluntary contraction (MVC) of the index and little finger abduction were determined at the beginning of session 1 to normalize the intensity of the motor tasks in each subject. MVC was also performed at the beginning of subsequent sessions to provide a comparable contraction history before the visuomotor skill task and to compare MVC force across sessions. In the posture described earlier, subjects were instructed to simultaneously use their index and little fingers to exert forces against the transducers (model 34, Honeywell, Charlotte, NC) while receiving visual feedback on the average finger forces. They were reminded to avoid engaging forces other than these two fingers and simultaneously give maximal effort for both fingers. They were also reminded to produce similar force profiles between fingers. Subjects attempted to make the force line on the screen reach as high as possible while an experimenter verbally encouraged the subject. After reaching the maximal effort, subjects maintained this level for 3 s, then relaxed. Peak force and corresponding EMG were recorded for each finger after each trial. This procedure was repeated for three or more trials until less than a 10% difference in peak force was observed between trials. Subjects were allowed to rest for at least 1 min between trials. MVC force was selected from the trial in which the sum of peak forces from the index and little fingers was greatest.

Visuomotor Skill Tasks

As the first step to examine neuromotor adaptations in healthy young adults in this basic scientific study, we prioritized internal validity in motor tasks while compromising external validity (i.e., functional applicability). We thus employed visuomotor skill tasks in which the performance would improve substantially due to training, and the involved muscles are those often tested with TMS. Accordingly, we employed a complex low-intensity visuomotor hand task not experienced in everyday life. The novel visuomotor task was expected to progressively improve motor skills with training in healthy individuals, and the low-intensity task allowed many practice repetitions without neuromuscular fatigue. Subjects were instructed to produce abduction forces with their index and little fingers to push against the transducers (model 31, Honeywell, Charlotte, NC) to match each force to a visually provided target pattern as accurately as possible. Two visualized target patterns were employed, pattern A and pattern B. Subjects were familiarized with each target pattern before the test.

Pattern A was the primary target pattern for training the skills and testing the improvement. It was composed of three low-frequency sinusoids centered around 5%MVC force, with each sinusoid at different frequencies (0.15, 0.50, and 0.95 Hz) and amplitudes (2.75, 1.50, and 0.75%MVC force). The resulting pattern had a range of 0–10%MVC forces of the subject. This pattern spanned 20 s as the target (Fig. 2A). The frequencies that compose the target pattern were selected so that no harmonics appeared across the trial. The complexity of the task was intentional by design so that any ceiling effect in skill improvement could be mitigated. Subjects received instantaneous visual feedback for finger forces as two separate colored traces alongside the target trace over the full 20 s. Subjects were instructed to match the target as closely as possible. To motivate the subjects to improve their skills, they were further provided the root-mean-square error (RMSE) from the target pattern after each trial in both practice and test.

Figure 2.

The target pattern and representative force traces during a motor task trial (session 1). The target pattern (thick black line), index finger force (thin black line), and little finger force (thin gray line) are shown. A: pattern A. B: pattern B. MVC, maximal voluntary contraction.

Pattern B was used only for testing the transfer of motor adaptation in the first and last sessions. This untrained pattern was three series of square waves with varying target intensities (Fig. 2B): the first target intensity at 2.5%MVC for 7 s, the second target at 7.5%MVC for 7 s, and the third target returning to 2.5%MVC for 7 s. The full target trace was visually provided on the monitor, and subjects were asked to reach the target as quickly as possible and maintain the force as steadily as possible.

Visuomotor Skill Tests

The visuomotor skill tests were administered without stimulation to the outer ear. Pattern A was tested before and after practice for each session. Pattern B was tested before practice on the first and last sessions (sessions 1 and 5). Only one trial was performed per test. Force trajectory against each target pattern was used for subsequent analysis. After each test, subjects were asked to rate the perceived difficulty by marking a line on a visual analog scale (VAS). The scale was a 10 cm horizontal line with “not difficult at all” at 0 cm and “extremely difficult” at 10 cm. The perceived difficulty of the test was measured as the length between 0 cm and the marked line on the scale.

TMS Tests

The TMS tests were performed to examine corticospinal excitability in the first and last sessions (sessions 1 and 5). In addition, we examined short-interval intracortical inhibition (SICI) as a measure of GABA-mediated inhibition. The TMS tests were administered without outer ear stimulation immediately before and after a block of motor practice (2 × Magstim 200 with a Bistim module, Magstim, Whitland, UK). Baseline EMG and motor evoked potential (MEP) were obtained from the FDI and ADM. A figure-of-eight TMS coil (diameter: 7 cm, D70, Magstim, Whitland, UK) was placed over the primary motor cortex in the left hemisphere. The coil was held with the handle pointing posteriorly at ∼45° to the sagittal plane yielding an E-field perpendicular to the central sulcus (28). A TMS coil navigation system (TMS navigator, NDI, Waterloo, ON, Canada) was used to track the three-dimensional coordinates of the coil on the head. For identifying a hotspot and resting motor threshold (RMT), MEPs should have a peak-to-peak amplitude greater than 50 µV in FDI in more than half of the trials for a given TMS intensity (29, 30). In this study, three out of five MEPs were required to be greater than 50 µV in the resting FDI (31, 32). Note, both FDI and ADM are innervated by the ulnar nerve. While the hotspot of ADM could be slightly different from that of FDI, the hotspot of FDI was used across the experiment to accurately maintain the TMS coil at one spot throughout the experiment and complete the experiment within an acceptable duration and burden to subjects. For determining the active motor threshold of FDI, subjects maintained ∼5% of their maximal EMG amplitude in FDI with visual feedback during the procedure. AMT was the lowest TMS intensity at which at least three out of five stimulations produced > 200 µV MEPs in actively contracting FDI (33). In session 5, the same hotspot was used as in session 1 except when a lower-intensity hotspot was found in an adjacent location due to interday variability in head shape registration with the navigation system. RMT and AMT were determined independently in each session, and there was no difference in RMT or AMT between sessions (Table 1).

Table 1.

MVC force, motor skill for the untrained pattern, and the resting and active motor thresholds before practice in the first and last sessions in the tVNS and sham groups

	tVNS		Sham
	Session 1	Session 5	Session 1	Session 5
MVC (N)
Index finger	20.62 ± 4.51	20.35 ± 3.75	22.18 ± 3.37	22.44 ± 2.64
Little finger	13.68 ± 2.01	12.91 ± 2.29	17.66 ± 2.00	17.24 ± 2.22
RMSE for pattern B (%MVC)
Index finger	1.091 ± 0.073	0.900 ± 0.048	1.362 ± 0.121	1.144 ± 0.075
Little finger	1.253 ± 0.139	1.014 ± 0.049*	1.225 ± 0.071	1.103 ± 0.041*
Motor threshold (%MSO)
RMT	48.57 ± 2.41	46.57 ± 1.95	46.13 ± 3.15	46.45 ± 3.12
AMT	40.86 ± 2.34	40.57 ± 2.06	40.96 ± 2.93	39.13 ± 3.49

AMT, active motor threshold; MSO, maximal stimulator output; MVC, maximal voluntary contraction force; RMSE, root mean square error of force; RMT, resting motor threshold; tVNS, transcutaneous vagus nerve stimulation. *P < 0.05, main effect of session. Means ± SD; n = 12 in each group.

Corticospinal excitability of the resting hand muscles was assessed with single-pulse stimulation at five different intensities, ranging from 100–160%RMT in 15% intervals. The intensities were ordered pseudorandomly, balanced between groups, to end with 130% RMT as a reference for the subsequent paired-pulse responses. Twelve single pulses were delivered for each intensity with 5–9 s randomized intervals between pulses. To assess GABA-mediated short-interval intracortical inhibition (SICI), paired-pulse TMS was applied with an interstimulus interval at 2 ms 12 times (34, 35). The conditioning and test pulses for SICI were 80% AMT and 130% RMT, respectively. The interstimulus interval and stimulus intensities follow those in the previous studies in our laboratory (36) and others (37, 38). Baseline EMG was monitored throughout the TMS tests, and trials with background EMG activity were discarded and recollected. In addition, the peak-to-peak amplitude of maximal M-wave (Mmax) was obtained from each muscle by applying supramaximal (130% max.) electrical stimulation to the ulnar nerve percutaneously 5 cm proximal to the wrist joint (Digitimer, Digitimer North America, Fort Lauderdale, FL). The stimulation electrodes were positioned in parallel with the ulnar nerve with an interelectrode distance of ∼2 cm (39).

Visuomotor Training

Subjects were trained to follow pattern A with concurrent outer ear stimulation. Training is defined here as the total practice across all five experimental sessions. Each of the five sessions in the experiment had a designated practice period. Practice is defined as the block(s) of trials with concurrent outer ear stimulation between each motor test within an experimental session. Sessions 1 and 5 had one block of ten trials, and sessions 2–4 had five blocks of 10 trials (50 trials total). Subjects were given 1 min to rest between trials and 3 min between blocks. Subjects received the skill score (i.e., RMSE) after each trial and were instructed to make an effort to improve the score. The highest personal score of each practice session was also displayed for each subject throughout the session. When a new personal record was achieved, the highest score was updated with a sound. Each practice session was ∼1 h long and separated by 1–2 days.

tVNS/Sham Stimulation

Subjects in the tVNS and sham groups received tVNS and sham stimulations, respectively, while practicing the visuomotor task. In the tVNS group, stimulation was administered to the afferent auricular branch of the vagus nerve located medial of the tragus at the entry of the left ear canal. Alternatively, stimulation was administered to the earlobe in the Sham group. A recent comprehensive review concluded that the tragus is one of the suitable sites for vagal modulation (40). In addition, the presence and absence of brainstem activation (including locus coeruleus) with the stimulation of the tragus and earlobe, respectively, have been confirmed with functional magnetic resonance imaging (41, 42). On the left outer ear, the surfaces of the tragus and earlobe were cleaned with alcohol wipes. On the tragus, one electrode was placed on the inner medial surface and another on the outer surface. Surface electrodes (Patients Choice, Balego, Saint Paul, MN) were trimmed into a diameter of ∼5 mm to fit the space. Both electrodes were secured in place with putty and tape. Electrodes on the earlobe were placed on the outer and inner surfaces and secured with tape.

All subjects had both sets of electrodes placed on the left ear, but only the set corresponding to the assigned group was stimulated. The stimulation was applied for 1.5-s burst sequences (0.5-s on and 1-s off) continuously, resulting in the bursts of stimulations at nonspecific timings during practice. The stimulation was 1.5-s bursts (0.5-s on, 1-s off) of the symmetrical biphasic rectangular waveform (frequency: 30 Hz; pulse width: 100 µs, Ultima Neo Advanced Multi-Mode Stimulator, Pain Management Technologies, Akron, OH). The stimulation intensity was at the midpoint between the perception and pain threshold (19, 21). These stimulation conditions are within the range employed to stimulate the afferent vagus nerve pathway (43). Outer ear stimulation was applied only during each trial of the visuomotor training. tVNS or sham stimulation started and ended 3 s before and 1 s after the execution of each trial (20 s). No outer ear stimulation was applied during the resting period between trials.

Data Reduction

Visuomotor skill was expressed as RMSE of the normalized finger force about the target trajectory. For Pattern A, RMSE was calculated for the middle 16 s of each trial. RMSE before practice on each session was normalized to the session 1 value (in the first trial) to assess the relative improvement of visuomotor skills across sessions. Developing synchronized force control between fingers would help simplify the motor control for simultaneous force matching with individual finger forces. Hence, a cross-correlation function was computed between the index and little finger forces to examine the temporal correlation of the concurrent force profiles between the fingers. To assess the improvement of the temporal correlation, the relative change in the peak value of the cross-correlation function from that of session 1 was determined. Relative changes from session 1 in prepractice RMSE, the peak of cross-correlation function, and VAS on the perceived difficulty for pattern A were determined to examine the adaptation with training. Note, the effect of VNS was most apparent in the early rapid improvement of motor skills in healthy mice (4). To assess the rate of early rapid change in this study, the slope of the linear regression analysis for these variables was determined against the initial three sessions. For pattern B, RMSE was calculated for the entire trial except for the initial 3 s. RMSE in the last session was normalized to the first session to determine the relative improvement for pattern B to assess the transfer of motor adaptation.

Corticospinal excitability in FDI and ADM was determined as the peak-to-peak MEP amplitude averaged across twelve responses. The first two pulses of each intensity were discarded. MEP amplitude was normalized to Mmax of each subject to reduce the potential effect of variability in the peripheral response. The peak-to-peak MEP amplitude was averaged across 115–160%RMT to yield a mean peak-to-peak MEP amplitude above RMT. MEP responses for SICI were normalized to those with single-pulse TMS at 130%RMT. Background EMG amplitude was determined before TMS for 100 ms for each TMS trial.

Statistics

MVC force was tested with a two-way mixed model analysis of variance (ANOVA) for the group (tVNS and sham) and session (sessions 1–5). For pattern A, to examine the in-session effect of practice on visuomotor skill in each session, RMSE in each finger was tested with a three-way mixed-model ANOVA for the group (tVNS and sham), time (Pre and Post practice), and session (sessions 1–5). To compare the rate of early rapid changes, the slope of prepractice RMSE, the peak of cross-correlation function, and VAS on the perceived difficulty for pattern A in sessions 1–3 were obtained with linear regression. The slope of prepractice RMSE was tested with a two-way mixed-model ANOVA for the group and finger. The slope for the cross-correlation peak and VAS on the perceived difficulty were tested with an unpaired t test between groups. To examine the overall adaptation, the relative changes in prepractice RMSE, the peak of cross-correlation function, and VAS on the perceived difficulty for pattern A from session 1 were tested with a two-way mixed-model ANOVA for the group (tVNS and sham) and session (sessions 2–5). To examine the transfer of motor adaptation, prepractice RMSE and VAS on the perceived difficulty for pattern B were tested with a two-way mixed model ANOVA for group and session (sessions 1 and 5). The effect of the group on the change in RMSE and VAS for pattern B from session 1 was further tested with an unpaired t test. For the TMS tests, EMG for FDI and ADM were analyzed. To assess the acute effect of the interventions, MEP amplitude for single-pulse TMS, normalized MEP amplitude for SICI, and background EMG amplitude were tested with a three-way mixed-model ANOVA for group, time, and session (sessions 1 and 5). The assumption of normality and the assumption of homogeneity of the variance were confirmed before conducting ANOVAs. The statistical test was performed with SPSS (version 26). Data are presented as means ± SD. An α value of 0.05 was used for statistical significance. P values less than 0.001 were presented as P < 0.001.

RESULTS

MVC Force

MVC force was 22.18 ± 1.65 N (means ± SD) for index finger abduction and 15.56 ± 1.09 N for little finger abduction when averaged across groups and sessions. MVC force was comparable across groups and sessions (no significant effect of group, session, or interaction). MVC forces in sessions 1 and 5 are presented in Table 1.

Motor Adaptation

As an in-session effect of practice on visuomotor skill, RMSE for pattern A decreased immediately after practice by 14%, on average, in each session in the index finger (Prepractice: 1.228 ± 0.078%MVC vs. Postpractice: 1.059 ± 0.086%MVC, F_1,22 = 95.591, P < 0.001, d = 0.401) and by 10%, on average, in the little finger (Pre: 1.288 ± 0.096%MVC vs. Post: 1.162 ± 0.104%MVC, F_1,22 = 10.967, P = 0.002, d = 0.346) when averaged across groups and sessions. There was no significant effect of group or interaction on RMSE in either finger, indicating comparable in-session improvement across groups.

The changes across sessions were examined using the prepractice RMSE normalized to the first session. When the effect of training was analyzed across sessions 2–5, the main effect of the session (index finger: F_3,66 = 10.324, P < 0.001; little finger: F_3,66 = 8.878, P < 0.001) indicated reductions of RMSE with sessions in both groups, reaching 38% and 40% reductions, on average, in the index (Fig. 3A) and little (Fig. 3B) finger forces, respectively, by session 5 when averaged across groups. RMSE for little finger force showed qualitatively similar trends, but there was no significant effect of group (F_1,22 = 0.864, P = 0.364) or group × session interaction (F_3,66 = 0.503, P = 0.682). An attenuated RMSE reduction in the tVNS than in sham group across sessions 2–5 was evident for the index finger by the main effect of group (F_1,22 = 6.984, P = 0.0096, d = 0.448) with a 16% difference, on average, between groups without a significant group × session interaction (F_3,66 = 0.629, P = 0.599). When focused on the early rapid adaptation as in the previous research on healthy mice (4), the negative slope for RMSE reductions from sessions 1 to 3 was less steep in the tVNS than in sham group across fingers (main effect of group, F_1,44 = 9.536, P = 0.004, d = 1.010) (Fig. 3C), indicating an attenuated improvement for the early rapid change in the tVNS group across fingers. There was no significant main effect of finger (F_1,44 = 0.008, P = 0.929) or interaction of finger and group (F_1,44 = 0.334, P = 0.566) on the slope, indicating that the attenuated early improvement is not different between fingers.

Figure 3.

Changes in RMSE of force during the prepractice tests without outer ear stimulation. Data in the tVNS (filled circles with solid lines) and sham (open squares with broken lines) groups are presented as a function of session. A: index finger force. B: little finger force. RMSE of force are normalized to the corresponding value in session 1. C: slope of initial rapid change in RMSE of force from session 1 to session 3. *P < 0.05, main effect of group across sessions 2–5. **P < 0.01, main effect of group across fingers. Data are means ± SD. n = 12 in each group. RMSE, root mean square error; tVNS, transcutaneous vagus nerve stimulation.

Note that the prepractice RMSE was measured without outer-ear stimulation because the study purpose was to examine the effect of pairing stimulation during training on the performance in a standard condition, i.e., without outer-ear stimulation. Nonetheless, to investigate a possible task specificity (i.e., stimulation status) in the improvement of motor performance, the RMSE of the first trial during practice was also examined in each session while the corresponding outer-ear stimulation was received in the group (Fig. 4). When the effect of training was tested across sessions 2–5, the main effect of the session (index finger: F_3,66= 7.74, P < 0.001; little finger: F_3,66= 6.840, P < 0.001) indicated a reduction in RMSE with sessions across groups. The absence of a main effect of group (index finger: F_{1, 66} = 0.057, P = 0.812; little finger: F_1,66 = 0.014, P = 0.907) and interaction of group and session (index finger: F_3,66 = 0.192, P = 0.902; little finger: F_3,66 = 0.261, P = 0.853) indicated that the reduction in RMSE during the corresponding outer-ear stimulation was comparable between groups across sessions.

Figure 4.

Changes in RMSE of force during the first trial of practice with outer ear stimulation. Data in the tVNS (filled circles with solid lines) and sham (open squares with broken lines) groups are presented as a function of session. A: index finger force. B: little finger force. RMSE of force are normalized to the corresponding value in session 1. Data are means ± SD. n = 12 in each group. RMSE, root mean square error; tVNS, transcutaneous vagus nerve stimulation.

For simultaneous matching to a target pattern with individual finger forces, developing correlated force control across fingers would help simplify motor control. The peak value of the cross-correlation function between finger forces was determined to measure the temporal correlation between force trajectories for pattern A. The peak value was already close to the maximal possible value of 1.0 at the initial trial (0.952 ± 0.025, prepractice test in session 1), indicating little room for improvement (i.e., by only ∼5%). Nonetheless, the relative changes from the initial peak value were examined in each subject. When the overall changes across sessions were plotted, the increasing profile of the peak value in each group was a mirror-image trend of the decreasing profile of RMSE (Fig. 5A). With the main effect of the session (F_3,66 = 6.583, P = 0.001), the peak value increased by ∼3% across groups by the last session. The between-group difference in the slope for the early adaptations from sessions 1 to 3 was not significant (t = 1.717, P = 0.122) (Fig. 5B).

Figure 5.

A: peak value of cross-correlation function (CCF peak). B: slope of initial rapid change in cross-correlation peak from session 1 to session 3. CCF peak is normalized to the corresponding value in session 1. Data are means ± SD. n = 12 in each group. tVNS, transcutaneous vagus nerve stimulation.

Regarding the changes in the perceived difficulty of the trained pattern A task within each session, the VAS score declined immediately after each practice session by a comparable amount of 0.9 across groups (Pre: 4.81 ± 0.20, Post: 3.91 ± 0.24, when averaged across sessions, F_1,22 = 9.016, P < 0.001) without a significant effect of group or interaction. For the changes in the perceived difficulty across sessions, the VAS score in the prepractice test declined with sessions (F_4,88= 6.830, P = 0.016) without a significant effect of group or interaction. It indicates that the VAS score was comparable between groups, including at the initial (tVNS: 6.963 ± 0.711, sham: 5.892 ± 0.672; session 1) and last sessions (tVNS: 4.371 ± 0.565, sham: 3.317 ± 0.481; session 5). The difference in the VAS score from the first session was further analyzed to accommodate variable absolute values in the initial session. There was a main effect of the session (F_3,66 = 7.218, P < 0.001), showing a 2.58 point reduction (out of the 0–10 range) from the first to the last session. There was no significant effect of group or group × session interaction, confirming that the decline of VAS score with sessions was not different between groups.

Motor Adaptation Transfer

For the nontrained pattern (pattern B), the average change in RMSE from session 1 to session 5 was a nonsignificant reduction trend in index finger force (F_1,22= 3.353, P = 0.081) and a significant reduction by 14%, on average, in little finger force (F_1,22 = 8.792, P = 0.007, d = 0.537) across groups (Table 1). There was no significant group × session interaction in either index (F_1,22 = 0.214, P = 0.648) or little finger force (F_1,22 = 0.863, P = 0.364). To accommodate the intersubject variability in the pretraining value, RMSE was normalized to the session 1 value to yield the relative change. As a result, the normalized RMSE in session 5 was less than 1.0 (t = 2.961, P = 0.007). There was no significant difference between groups in the normalized RMSE for either index (tVNS: 0.854 ± 0.057, sham: 0.886 ± 0.072, t = 0.760, P = 0.263) or little finger force (tVNS: 0.860 ± 0.056, sham: 0.925 ± 0.055, t = 1.054, P = 0.201). No significant effect of group or session was found in the VAS score on the perceived difficulty for pattern B (session 1: 5.25 ± 0.42; session 5: 4.75 ± 0.44; when averaged across groups).

Neural Adaptation

There was no significant effect of group or session on RMT or AMT (Table 1). There were three outlier datasets that included erroneously high MEP amplitude (>70% Mmax), which were excluded from the analysis. No significant main effect or interaction was found for MEP amplitude in either muscle, indicating that corticospinal excitability was not affected by the intervention at any session or time in either muscle (Fig. 6). For SICI, there were no significant main effects or interactions except for a main effect of group in FDI (F_1,88 = 12.135, P = 0.008, d = 0.717) (Table 2), indicating that outer ear stimulation did not induce SICI adaptation. Background EMG was at noise level across trials with no significant effect of group, session, or time.

Figure 6.

MEP amplitude in FDI (top row) and ADM (bottom row) before and after practice in session 1 (left column) and session 5 (right column) in the tVNS and sham groups. MEP amplitude is normalized to the amplitude of maximal M-wave (Mmax). Data are means ± SD. n = 12 in each group. ADM, abductor digiti minimi; FDI, first dorsal interosseous; MEP, motor evoked potential; tVNS, transcutaneous vagus nerve stimulation.

Table 2.

MEP amplitude for SICI before and after practice during the first and last sessions in the tVNS and Sham groups

	tVNS		Sham
	Pre	Post	Pre	Post
SICI for FDI
Session 1	0.501 ± 0.262	0.548 ± 0.246	0.622 ± 0.279	0.697 ± 0.300
Session 5	0.588 ± 0.352	0.429 ± 0.215	0.744 ± 0.275	0.769 ± 0.213
Average	0.517 ± 0.271		0.708 ± 0.261**
SICI for ADM
Session 1	0.655 ± 0.343	0.598 ± 0.258	0.794 ± 0.531	0.680 ± 0.235
Session 5	0.567 ± 0.444	0.497 ± 0.203	0.688 ± 0.386	0.612 ± 0.316
Average	0.579 ± 0.319		0.693 ± 0.376

Normalized MEP amplitude in SICI relative to the value during single-pulse TMS. Pre, before practice; Post, after practice; Average, average across sessions and times; **P < 0.01, main effect of group. Means ± SD n = 12 in each group. ADM, abductor digiti minimi; FDI, first dorsal interosseous; MEP, motor evoked potential; SICI, short-interval intracortical inhibition; tVSN, transcutaneous vagus nerve stimulation.

DISCUSSION

The primary finding is that the tVNS group had attenuated RMSE reductions with training sessions compared with the sham group, while the in-session reduction was comparable across groups. In addition, RMSE for the nontrained task was reduced after the training completion in similar amounts across groups. The attenuated RMSE reductions in the tVNS group indicate that applying tVNS at nonspecific timings throughout training compromises motor adaptation in healthy humans. The absence of a group or session effect on MVC force supports that the difference in the submaximal task was not influenced by a potential confounding effect of muscle strength. The mirror-image trend of the adaptation profiles of RMSE and peak in cross-correlation function between fingers across groups implies a different trend of developing strategies for synchronizing finger forces between groups. The comparable reductions of VAS scores of the perceived difficulty across groups indicate that the cognitive benefit of skill improvement was large enough for healthy participants to perceive, while the attenuation of the improvement rate due to the nonspecific tVNS cannot be captured by the perceived difficulty.

The compromised across-session adaptation in the tVNS group corroborates previous findings in healthy mice that received invasive VNS at nonspecific timings during training (4). In their comprehensive examination of the VNS timing effect during motor training, applying VNS immediately after the successful trials of the visuomotor reach-grasp-retrieve task enhanced the speed and amount of motor adaptation across training sessions compared with the control. This facilitative effect with timing-specific VNS in healthy mice aligns with those in disabled rats (5–10, 25). In contrast, applying VNS at nonspecific timings during the training compromised the early adaptation compared with the control in healthy mice (4). In healthy mice (4) and humans, the delayed adaptation in early training sessions can be overcome in later training sessions, probably because the improvements with the same training become gradual and reach a ceiling. As a compromising effect in poststroke rats, the facilitative effects of timing-specific VNS after successful trials were partially attenuated when extra VNS were applied at random intervals during training (25). The attenuation of error reductions with nonspecific tVNS in this human study corroborates these findings with nonspecific VNS in animals (4, 25). There is some uncertainty as to whether tVNS activates the same mechanisms in the brain or the same degree of activation as VNS. Nonetheless, the results suggest that training-induced motor adaptations can be compromised if stimulation is applied at nonspecific timings, whether tVNS or VNS, across animals and humans.

Potential neural mechanisms underlying the (t)VNS-induced neuromotor plasticity include cortical excitation and arousal through activation of noradrenergic neurons in the locus coeruleus (2, 3, 23, 44, 45) and cholinergic neurons in the basal forebrain (4, 46) as well as the involvement of GABA (47). In a calcium imaging study in healthy mice, some neurons showed either transient activation or suppression via VNS, without a change in the overall discharge rate of the neuron population in the primary motor cortex (4). In healthy humans (22, 48), MEP amplitude was unchanged after a 60-min of an isolated tVNS session without motor training. The absence of a group × time or group × session interaction in MEP amplitude in the present study thus extends the absence of persistent modulation after the isolated tVNS to the paired tVNS with motor training. The current absence of SICI adaptation to the paired tVNS with training for two weeks in healthy humans is not consistent with increased GABA after a 3-mo VNS intervention in patients with partial seizures (47). Note that this comparison includes differences in the type of VNS, the duration of the intervention, and the type of subjects. Previous studies in healthy humans reported mixed results on the effect of tVNS on GABA-associated measures: increased GABA via SICI assessment (22), reduced GABA via a behavioral measure and readiness potential (20), and no GABA change via brain oscillation assessment (49). The methodological approaches for tVNS studies are variable in terms of the parameter of tVNS (e.g., 20 Hz in Ref. 22), side of tVNS, and hemisphere for TMS testings (e.g., different SICI adaptation between hemispheres in Ref. 22). The inconsistent observations may imply that tVNS-induced neuromodulation may be variable and sensitive to the methodological approaches. It is possible that we would have observed different results if we had used a different approach. As this is a new area of investigation, a number of future studies will be required to gain a systematic understanding of the effect of tVNS (and paired motor training) on neuromotor adaptations. The current study provided a unique case of pairing tVNS and motor training in healthy humans, in which there were no persistent adaptations in corticospinal excitability or GABA-mediated cortical inhibition at least in the tested condition.

The comprehensive study on healthy mice found that the in-session skill improvement was not influenced by the VNS timing (4). The current absence of a significant difference in the in-session RMSE reduction between groups is in line with that finding in mice. It suggests that the acute skill improvement within practice sessions would not be the primary contributor to the compromised adaptation with the nonspecific tVNS. The comparable reductions in RMSE measured in the first practice trial between groups further suggest that examining the adaptations specific to the practice condition may obscure the effect. By finding that cholinergic inhibition prevents VNS-induced neuromodulation and adaptation enhancement in healthy mice, Bowles et al. (4) suggested that VNS-enhanced motor adaptation is likely mediated by cholinergic neuromodulatory systems that reinforce use-dependent plasticity and reinforcement signaling. Most studies on adding tVNS during motor training do not compare the effect on the in-session improvement. While the underlying neural mechanisms are still uncertain and could be more complex, our similar findings in healthy humans provide a supportive insight into the importance of future explorations into the potential effects on various adaptation processes and associated neural mechanisms.

After completing the full training, the reduction of RMSE for pattern B indicates that the visuomotor adaptations to the trained force-tracking task are transferrable to a nontrained force-tracking task. The comparable RMSE reductions for pattern B across groups may suggest that the attenuated motor adaptation due to nonspecific tVNS does not influence motor transfer, at least for the employed protocol in healthy young adults. The result is understandable, considering that the RMSE for pattern A was comparable across groups at the end of the full training. Possibly, the motor transfer can be influenced by tVNS if it is examined during early adaptation phases when the tVNS effect appears to be greater.

Since tVNS is an emerging new intervention for potentially altering motor recovery or motor learning, especially in humans, there are many knowledge gaps in both basic and applied aspects. The current study has advanced the knowledge of motor adaptations and transfer in the case of applying tVNS at nonspecific timings during motor training in healthy young adults. This basic understanding of healthy humans is expected to serve as the basis for expanding the research of tVNS to clinical populations and other conditions in the future.

Conclusions

In conclusion, applying tVNS at nonspecific timings during visuomotor skill training with fingers attenuates motor adaptations across training sessions while not influencing in-session adaptation in healthy humans. Transfer of motor adaptation does not appear to be influenced by this attenuation.

DATA AVAILABILITY

Data will be made available upon reasonable request.

GRANTS

This work was supported by grants from the National Institute of Neurological Disorders and Stroke of the National Institutes of Health (award number: 1R03NS106088-01A1) and the McCamish Parkinson’s Disease Innovation Program at Georgia Tech and Emory University.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

M.A.S.P. and M.S. conceived and designed research; M.A.S.P. performed experiments; M.A.S.P. analyzed data; M.A.S.P. and M.S. interpreted results of experiments; M.A.S.P. and M.S. prepared figures; M.A.S.P. and M.S. drafted manuscript; M.A.S.P. and M.S. edited and revised manuscript; M.A.S.P. and M.S. approved final version of manuscript.