Balance Perturbations in Simulated Low-Gravity Modulate Human Premotor and Frontoparietal Electrocortical Theta, Alpha, and Beta Band Spectral Power

We aimed to identify the influence of bodyweight unloading on healthy human electrical brain and muscle dynamics during standing balance. Using an underwater environment to simulate reduced gravity, we recorded high-density electroencephalography and lower limb electromyography from 10 individuals (5 female). During five standing balance conditions, we adjusted the water depth (1: on-land, 2: chest-level, 3: waist-level) and applied external fluid forces to the front of the body in the posterior direction (4: low and 5: high external fluid forces). Reduced bodyweight decreased ankle plantar flexor muscle activity and increased alpha band (8-13 Hz) spectral power from right premotor cortex. Underwater balance perturbations increased muscle activations surrounding the ankle and knee joints and increased theta band (4-8 Hz) spectral power from right premotor cortex. Concurrently, alpha and beta band (13-30 Hz) spectral power decreased from left prefrontal and right premotor cortices, along with reduced alpha band spectral power from left parietal cortex.


I. INTRODUCTION
N ASA (National Aeronautics and Space Administration) has announced the Artemis mission to land humans on the Moon by 2025 [1] and launch humans to Mars by late 2030 or early 2040 [2]. To make this possible, astronauts will encounter altered gravitational fields (i.e., microgravity during spaceflight, or 0.17-and 0.38-times Earth's gravity on our moon and Mars, respectively). Previous spaceflight records have documented post-flight gait and balance impairments among astronauts. Changes in the neural control of posture and movement from exposure to low-gravity environments can be attributed to interactive adaptations in tendon elasticity, loss of bone density, and muscle weakness due to tissue atrophy [3], [4], [5], [6], [7], [8], [9], [10].
Better understanding of neuromotor changes in response to reduced gravity can help to predict possible postural and locomotor instability following spaceflight [3] and assist in designing countermeasures [3], [11]. Prior studies have shown changes in brain structures [12], sensorimotor control, and cognition before and after microgravity exposure [13]. However, we know relatively little about how human electrical brain and muscle activation dynamics are influenced by bodyweight unloading in reduced gravity.
Gravity cannot be changed on Earth, but we can simulate reduced gravity by supporting body weight using assistive mechanical devices or buoyancy in water [14], [15], [16], [17]. Bodyweight support distributes loads among lower extremity joints and muscles by opposing gravitational and inertial loads that otherwise create unstable equilibrium in an upright posture [18], [19]. Underwater environments are beneficial for simulating low-gravity (less than Earth's) because they are relatively low-cost, and it is possible to simulate both weightlessness and partial gravity based on buoyancy forces.
In this study, we aimed to identify the influence of underwater bodyweight support on healthy human electrical brain and muscle dynamics during standing balance (Aim 1); see Fig. 1. The main objectives of this paper are to understand neural control of human standing balance. We also collected data during underwater treadmill gait because our broader aim is to better understand neural control of human gait and balance. We hypothesized that compared to standing on-land (Hypothesis H1.1), balance demands would decrease underwater, expressed by greater alpha and beta band spectral power from frontal, parietal, and central cortical regions. We also anticipated (Hypothesis H1.2) that ankle plantar flexor muscle activity would decrease when standing underwater at greater depths due to increased bodyweight support.
To cause postural instability in simulated low-gravity (Aim 2), we introduced balance perturbations using underwater jets. External fluid forces were applied to the anterior side of each subject, moving in the posterior direction; see Fig. 2. We hypothesized that greater sensory feedback from the surrounding fluid environment and increased balance demands from external fluid forces (Hypothesis H2.1) would increase theta band spectral power and decrease alpha and beta band spectral power from frontal, central motor, and parietal cortical regions. We also anticipated (Hypothesis H2.2) greater electrical muscle activations from ankle dorsiflexor, and knee and hip extensor muscles to maintain posture when applying external fluid forces in the backward direction.

A. EEG RESULTS
We identified three electrocortical source clusters from left prefrontal, right premotor, and left parietal cortices; see Fig. 3. Compared to standing on-land (Aim 1), only right premotor cortex alpha band electrocortical spectral power increased when standing underwater at chest-level ( Fig. 3, ST<CST, p = 0.043). No other electrocortical spectral power differences were identified between standing on-land and standing underwater at either waist or chest-level, without balance perturbations. Nor were electrocortical spectral power differences identified between standing underwater at waist and chestlevel, without balance perturbations (Fig. 3).
To understand the influence of balance perturbations on bodyweight supported postural control underwater (Aim 2), we applied low and high levels of external fluid forces to the body when standing underwater at waist-level. Compared to unperturbed standing underwater at waist-level, low external fluid forces applied to the body did not alter electrocortical spectral power. High external fluid forces applied to the body increased theta band spectral power from right premotor cortex (Fig. 3, WST<WST + High, p = 0.022). Alpha band spectral power decreased from right premotor and left parietal cortices (Fig. 3, WST>WST+High, p = 0.017, p = 0.02, respectively) and beta spectral power decreased from left Compared to standing underwater at chest-level (Aim 2), low external fluid forces applied to the body at waist-level increased theta band spectral power from left prefrontal cortex ( Fig. 3, CST<WST+Low, p = 0.028) and decreased alpha power (Fig. 3, CST>WST+Low, p = 0.009). High external fluid forces applied to the body increased theta power from right premotor cortex (Fig. 3, CST<WST+High, p = 0.011). Alpha and beta band spectral power decreased from left prefrontal and right premotor cortices (Fig. 3, CST>WST+High, alpha: p≤0.005, beta: p≤0.022), and alpha power decreased from left parietal cortex (Fig. 3, CST>WST+High: p = 0.033).
Compared to standing on-land (Aim 2), standing underwater at waist-level with low external fluid forces applied to the body did not alter electrocortical spectral power. High external fluid forces applied to the body increased theta band spectral power from right premotor cortex (Fig. 3, ST<WST+High, p = 0.028). Alpha band spectral power decreased from right premotor and left parietal cortices (Fig. 3, ST>WST+High, p≤0.024), and beta power decreased from left prefrontal and right premotor cortices (Fig. 3, ST>WST+High, p≤0.02).

B. EMG RESULTS
By measuring lower limb EMG along with electrical brain activity during standing balance in each condition, we simultaneously identified changes in electrical muscle activity during bodyweight supported balance, and in response to underwater balance perturbations.
When applying external fluid forces to the body while standing underwater at waist-level (Aim 2), low-level external fluid forces did not alter lower limb EMG activity compared to standing underwater at waist-level without external fluid forces. High external fluid forces increased EMG activity from biceps femoris and tibialis anterior muscles (Fig. 4, WST<WST+High, p≤0.003). Compared to standing underwater at waist-level with low external fluid forces applied to the body, high external fluid forces increased vastus lateralis, biceps femoris, and tibialis anterior EMG activity (Fig. 4, WST+Low<WST+High, p≤0.027).
Compared to standing underwater at chest-level (Aim 2), only tibialis anterior EMG activity increased when low external fluid forces were applied to the body while standing underwater at waist-level (Fig. 4, CST<WST+High, p = 0.006). High external fluid forces increased EMG activity from vastus lateralis and biceps femoris muscles (Fig. 4, CST<WST+High, p≤0.021).
No changes in lower limb EMG activity were identified between standing on-land and standing underwater at waistlevel with low external fluid forces applied to the body (Aim 2). Standing underwater at waist-level with high external fluid forces increased EMG activity from tibialis anterior (Fig. 4, ST<WST+High, p = 0.0025), but decreased medial gastrocnemius and soleus muscle activities compared to standing on-land (Fig. 4, ST>WST+High, p<0.001).

III. DISCUSSION
Mobile high-density EEG and waterproof, wireless EMG allowed us to study the effects of simulated reduced gravity on healthy human electrical brain and muscle dynamics during underwater standing balance. Our hypotheses were partially confirmed. Among conditions, we identified electrocortical responses from left prefrontal, right premotor, and left parietal cortical regions. Standing underwater at waist-level did not show changes in electrocortical spectral power (H1.1), but ankle plantarflexor muscle activity decreased from lateral and medial gastrocnemius and soleus muscles (H1.2). While standing underwater at chest-level, right premotor alpha band spectral power increased and medial gastrocnemius muscle activity decreased (H1.1). At the same time, tibialis anterior muscle activity increased (H1.2). Taken together, although bodyweight support underwater reduced ankle plantarflexor muscle activity compared to standing on-land (H1.2), only right premotor electrocortical spectral power was altered by greater levels of bodyweight support (H1.1).
When applying underwater balance perturbations, our hypotheses were also partially confirmed. While standing underwater at waist-level, low external fluid forces applied to the body did not alter electrocortical spectral power (H2.1) or lower limb electrical muscle activations (H2.2). In response to high external fluid forces, theta band spectral power from right premotor cortex increased, and alpha and beta band spectral power from right premotor and left prefrontal cortices decreased, along with reduced alpha power from left parietal cortex (H2.1). Concurrently, tibialis anterior, biceps femoris, and vastus lateralis muscle activations increased,

FIG. 4. EMG Results. Boxplots display average percent change in electrical muscle activity from ST (Standing on-land). CST is standing underwater at chest-level, WST is standing underwater at waist-level, +Low is low external fluid force, +High is high external fluid force. Two-tailed t-tests compared changes from ST (zero mean) (+ is, p < 0.05 and +++ is p < 0.01). One-way ANOVA post-hoc pairwise comparisons between conditions ( * is p<0.05 and
* * * is p<0.01).

A. AIM 1: REDUCED PREMOTOR ELECTROCORTICAL PROCESSING & ANKLE PLANTARFLEXOR MUSCLE ACTIVITY IN SIMULATED LOW-GRAVITY
Reduced ankle plantar flexor muscle activity while standing underwater compared to standing on-land aligns with previous underwater gait studies [33]. Harrison and colleagues [34] found that standing underwater at waist and chest-levels corresponded to approximately 30% and 50% bodyweight support, respectively. Because plantar flexor muscles support bodyweight [35], which is offset by upward buoyancy forces underwater [21], electrical muscle activity expectedly decreased in simulated reduced gravity. Greater alpha band electrocortical spectral power from right premotor cortex during standing balance underwater at chest-level compared to standing on-land suggests that greater levels of bodyweight support reduced premotor cortical processing, likely due to lesser balance demands. Prior studies have shown that alpha and beta band spectral power from prefrontal, somatosensory, and motor cortices adjust to changes in task difficulty, which is indicative of cortical engagement [30], [31], [36]. During robotically guided walking, alpha and beta band spectral power increased with the level of assistance [30], [31], suggesting reduced cortical engagement. In contrast, faster gait speeds reduced alpha and beta spectral power [36], suggesting greater sensorimotor cortical processing.
No differences in electrocortical spectral power were identified when standing at waist-level compared to standing onland. It is possible that bodyweight support when standing underwater at waist-level was not sufficient to cause perceptual differences. However, we also observed greater inter-subject variability when standing underwater at waist-level compared to chest-level. Because the lower body was underwater, but the upper body was above water, when standing underwater at waist-level, contrasting sensory feedback could disrupt neural coupling between the upper and lower body, leading to varied subject perceptions [37].

B. AIM 2: INCREASED PREMOTOR AND FRONTOPARIETAL ELECTROCORTICAL PROCESSING & LOWER LIMB MUSCLE ACTIVITY DURING UNDERWATER BALANCE PERTURBATIONS.
External fluid forces applied to the body while standing underwater required additional cortical processing and lower limb muscle activation to maintain balance. Previous studies have shown increased muscle activity from tibialis anterior, rectus femoris, and soleus muscles during anterior-posterior directional perturbations while standing on a force plate onland [38], [39]. We identified increased biceps femoris, vastus lateralis, and tibialis anterior muscle activations in response to high external fluid forces applied to the front of the body in the posterior direction, suggesting that greater hip and knee extension, and ankle dorsiflexion were required to maintain balance. Even during the application of high external fluid forces to the body, medial gastrocnemius and soleus muscle activation decreased compared to standing on-land. This emphasizes the role of these plantarflexor muscles in bodyweight support, which was offset by unloading due to underwater buoyancy.
In response to high external fluid forces applied to the body, increased theta band spectral power and decreased alpha and beta power from right premotor cortex aligns with prior studies that showed increased mid-central theta band spectral power during balance perturbations and reduced alpha and beta band spectral power from central cortical regions during lower limb sensorimotor processing [27], [36], [40]. Frontoparietal cortical regions showed similar alpha power decreases from left prefrontal and left parietal cortices, and reduced beta power from left prefrontal cortex, attributed to attention, error monitoring, and motor adjustments [41], [42]. These results identify increased cortical processing to maintain underwater standing balance in response to external perturbations.

C. STUDY LIMITATIONS AND FUTURE DIRECTIONS
Our study sample was limited to healthy young adults to better our understanding of healthy human electrical brain and muscle dynamics during standing balance underwater. Because our results also have applications for rehabilitation settings, future studies could benefit from broader samples.
We did not directly quantify external fluid forces applied to the body due to limitations of the underwater jet controls in our experimental setup. Instead, we expressed low and high relative forces based on the underwater treadmill settings. Nevertheless, we identified electrocortical and myoelectric changes between the relative force conditions, suggesting that more sensitive analyses based on direct external fluid force measurements may be worthwhile.
Although we aimed to simulate a reduced gravity environment, underwater bodyweight support differs from microgravity, which includes additional factors that can affect human electrical brain and muscle dynamics. In particular, microgravity space environments alter the distribution of human body fluids, including cerebrospinal fluid [43], [44]. In addition, greater external forces due to water viscosity are not expected in microgravity spaceflight environments, likely altering sensorimotor processing during human movement.
Here, we focused our analysis on standing balance, but future studies should examine electrical brain and muscle dynamics during motion to better understand the effects of gravitational and drag forces on human neuromotor control.

IV. CONCLUSION
During standing balance in a simulated reduced gravity underwater environment, we identified reduced cortical processing from premotor cortex. These results suggest that fewer cortical resources are needed to maintain balance in reduced gravity, which may lead to changes in motor coordination following spaceflight.
We also showed that cortical engagement and muscle activations increased while maintaining balance when exposed to balance perturbations from external fluid forces. Based on these findings, it may be possible to develop training interventions to mitigate neuromotor changes that occur during exposure to microgravity. Increasing balance demands and sensorimotor feedback under controlled conditions, could be used to stimulate spinal and supraspinal neural pathways.
This work emphasizes possibilities for using electrical brain and muscle activation measures as powerful tools for tracking human neuromotor control in reduced gravity.

A. PARTICIPANTS
Ten healthy young subjects (age: 23.8 ± 4.39, weight: 67.65 ± 16.11 kg, height: 170.2 ± 13.34 cm, five male and five female) were recruited for this study. Subjects were free from any cognitive or lower limb impairments. All procedures were approved by Texas A&M University Institutional Review Board (IRB2021-0625DCR). Each subject gave informed consent before the experiment.

B. MEASUREMENTS
Subjects were instrumented with 64-channel wireless electroencephalography (EEG) sensors (LiveAmp, Brain Products, Germany) to measure electrical brain activity at 500 Hz sampling rate (Fig. 1). We recorded subject-specific 360degree electrode locations using a portable 3D scanner (Artec Leo, Artec3D, Luxembourg) and extracted the x,y,z coordinates of the EEG electrode's location. Fourteen wireless and waterproof electromyography (EMG) sensors (Wave, Cometa, Italy) were located bilaterally on lower limb muscles: rectus femoris, vastus lateralis, biceps femoris, tibialis anterior, lateral gastrocnemius, medial gastrocnemius, and soleus muscles. We placed two inertial measurement units (IMU) on the feet for later analyses.

C. EXPERIMENTAL PROTOCOL
Ten subjects completed standing balance tasks barefoot on an underwater treadmill (HYDROWORX 1200). Study conditions (Fig. 2) included five different standing conditions, (1) standing on-land (ST), (2) standing underwater at waist-level (WST, ∼30% bodyweight [34]), (3) standing underwater at chest-level (CST, ∼50% bodyweight [34]), (4) standing underwater at waist-level with low-level external fluid forces applied in the posterior direction (WST + 30%, 30% jets from the underwater treadmill setup), and (5) standing underwater at waist-level with high external fluid forces applied in the posterior direction (WST + 70%, 70% jets from the underwater treadmill setup). We also collected underwater treadmill locomotion conditions for future analyses. The waist-level water depth corresponded to the superior iliac crests of the pelvis, and the chest water depth was set to the level of the xiphoid process on the sternum for each participant. Each condition lasted 3-minutes, and conditions were presented in randomized order among participants.

D. EEG SIGNAL AND STATISTICAL ANALYSIS
EEG data were analyzed in MATLAB with a customized pipeline based on EEGLAB functions [45]; see Fig. 5. We modified the subject 3D head model axis in Artec Studio 16 (Artec 3D, Luxembourg) and loaded subject-specific channels location for each EEG electrode on the head. We applied a 1 Hz high-pass filter and did re-referencing by estimating the true average of the EEG channels. Then, we rejected noisy channels. We performed ASR (Artifact Subspace Reconstruction) [46] on the EEG channel data at a 20 standard deviation cutoff using a clean standing baseline condition for each subject. After we cleaned large-amplitude artifacts using ASR, we concatenated all five study conditions and used CCA (Canonical Correlation Analysis) [47] to remove highfrequency noise, such as muscle artifacts. Next, we rejected bad frames from the continuous time series data.
After pre-processing the data by filtering and stepwise channel and frame corrections to remove noise, we performed an adaptive mixture independent component analysis (AM-ICA) [48], [49] to estimate brain sources from channel data. During AMICA, the standing and underwater treadmill walking data were included to provide sufficient data points for analysis. We estimated brain source locations using dipole fitting and retained components with residual variance below 0.15. We then separated the standing conditions after rejecting noise components based on the ICLabel toolbox [50] and visual inspections.
We used a k-mean clustering approach (k = 10 standard deviation = 3) to group similar brain components among subjects based on brain component locations, orientations, power spectra, and scalp maps. We calculated individual components' power spectral density and used a 1/f aperiodic fit to flatten the power spectrum. Finally, the average flattened power spectrum among electrocortical source clusters was calculated and compared among conditions. We also calculated the average spectral power within each frequency band for each study condition.
Spectral power differences among study conditions were assessed using non-parametric bootstrap-based ANOVA in EEGLab. The Shapiro-Wilk test was used to check the normality of the average power of each frequency band. During statistical analysis, outliers were removed using boxplot methods, but the figures were created with all data points. We used a one-way ANOVA with Bonferroni corrections for multiple comparisons between conditions in R. All statistical analyses in this study used α = 0.05.

E. EMG SIGNAL AND STATISTICAL ANALYSIS
EMG data were analyzed in MATLAB with a customized pipeline. We applied a band-pass filter to the EMG data (30-500 Hz), full-wave rectified the data, low-pass filter (4 Hz) and calculated the root mean square. Average EMG amplitudes of each condition were calculated. Then, we calculated the percent change from the standing condition.
Statistical analysis was conducted in the R language. We tested data normality using the Shapiro-Wilk test, and outliers were removed during statistical analysis using boxplot methods. To understand the difference between standing conditions, we used two-tailed t-tests compared to a zero mean (ST condition). To compare across conditions, we used a one-way ANOVA with Bonferroni corrections.