For this experiment, a group of 20 healthy young adults, comprising 8 males and 12 females, were carefully selected. All participants were thoroughly informed about the risks and requirements of the experiment, and they willingly agreed to participate by signing an informed consent form. The experimental protocol was reviewed and approved by the Ethics Committee of Shanghai University of Sports. Furthermore, all the participants had right-sided dominant legs, which was determined by asking them to kick a ball (van Melick et al., 2017). Exclusion criteria included: (1) any acute illness requiring hospitalization within the past 3 months; (2) the use of neuroactive drugs that affect the brain state; (3) any self-reported cardiovascular or cerebral diseases, neurological diseases, musculoskeletal disease, or any other disease that could affect upright standing; and (4) any contraindications concerning the use of tACS (e.g., mental-implanted devices in the brain). The basic profiles of the participants are presented in Table 1.

A double-blind, randomized, sham-controlled within-subjects crossover was completed, wherein participants were randomized to receive active or sham tACS: (1) fronto-parietal tACS at 6hz; (2) sham tACS (sham) with a minimum interval of 72 h between each session. Participants were required to visit 3 times. During the first visit (visit 1), also known as the screening visit, basic demographic information was collected, including age, sex, weight, height, ethnicity and living status. Participants also completed a cognitive test consisting of six blocks to determine their cognitive level of difficulty. The test had to be completed with an accuracy level of 0.6–0.8, and this level would be used in the subsequent visits. Of note, three participants had to be excluded from the study during visit 1 as their accuracy was deemed too high, which could have affected the results. Finally, of the 20 participants who were formally included in the study, the working memory test had an accuracy of 0.73 ± 0.06 as measured during visit 1. (2) Visits 2–3: all participants first performed a brief familiarization of the task. Single-task (cognitive single task × 3 or postural control single task × 2) and dual-task (cognitive task + postural control) × 2 were assessed before and after tACS. We numbered 20 participants and created a randomized list to counterbalance the order of assessments across participants. Each task lasted for approximately 90 s.

Additionally, we recorded resting-state EEG with eyes closed for 3 min before and after the stimulation intervention to investigate the effect of tACS on endogenous brain oscillations (Figure 1).

EEG was recorded with the Starstim^® (Neuroelectrics Inc., Barcelona, Spain) from 8 positions covering the left of the fronto-parietal cortex (F3, P3, Fz, CP2, F8, FC5, O2, AF3) with 3.14 cm² Ag/AgCl electrodes and digitalized with 24-bit resolution at a sampling frequency of 500 samples/s. EEG data were referenced to the earlobe.

Pre-processing analysis of EEG data involved using EEGLAB 19.0 and the Matlab open-source toolbox (Mathworks, Natick, United States) (Battaglini et al., 2020). Offline, eyes-closed resting EEG data were band-pass filtered between 0.05 and 40 Hz (Butterworth filter, order = 2) (Battaglini et al., 2020). The continuous data were segmented into 1 s epochs to obtain 180 epochs. All epochs were visually inspected to remove data segments contaminated with muscular or ocular artifacts conditions. Independent component analysis (ICA) was used to correct electrode artifacts when required.

To detect the brain θ power spectrum following θ-tACS, the cleaned epochs were then used to extract the FFT spectrum. Finally, the individual power values in the frequency range of interest were averaged for each participant and separately for the pre- and post-stimulation sessions. Hereafter, we referred to “θ power” to indicate the average of power values extracted in the frequency range between 4 and 8 Hz.

To examine the effects of tACS on phase-locked activity, the phase locking value (PLV) between channel F3 and channel P3 of the θ band was computed. The instantaneous phases of each channel were estimated by applying a Hilbert transformation to the source signals filtered into a band between 4 Hz and 8 Hz. The PLV was then computed as a function of the instantaneous phase difference between channels F3 and P3 using the following equation (Lachaux et al., 1999): P L V = 1 N ∑ n = 1 N e i θ 1 n − θ 2 n where N is the number of sampled time points, and θ₁ and θ₂ are the instantaneous phase values at time point n.

The PLV ranges between 0 and 1, where a value close to 0 indicates a random phase relationship and a value close to 1 indicates a fixed signal phase relationship. Finally, the PLV in θ band was averaged for each participant pre- and post-stimulation.

HD-tACS was delivered via the StarStim8 device, which is a hybrid wireless neurostimulation system for simultaneous EEG and tACS, controlled by the Neuroelectrics Instrument Controller (NIC 2.0; http://www.neuroelectrics.com/products/software/nic2/). We used 7 PIS-TIM Ag/AgCl electrodes with a 1 cm radius for stimulation. The optimal electrode placement and current intensity were developed using Stimweaver^® optimization technique simulations (Miranda et al., 2013; Ruffini et al., 2014) based on the main stimulation target areas (the left dorsolateral prefrontal cortex and left posterior parietal cortex) of this study. The montage injects a total current of 2 mA with the target En-field was set to +0.25 V/m over each target region, and 0 V/m over the remaining regions.

The tACS was delivered with the participant seated and resting for 20 min, including a 30 s fade in and a 30 s fade out. Aim to enhance the synchronization of θ oscillations between LDLPFC and LPPC (Polanía et al., 2012), the modeling of the tACS resulted in the placements of electrodes on the F3: 784 μA, P3: 831 μA, Fz: 384 μA, CP2: −402 μA, F8: −481 μA, FC5: −522 μA, O2: −594 μA of the 10–10 EEG placement system (Figure 4). The stimulation frequency was 6 Hz with a 0° relative phase. Sham tACS utilized the same montage yet the current was only applied for only 1 min (30 s fade in and 30 s fade out). All participants were asked to complete a questionnaire regarding the side effects and blinding efficacy of tACS at the end of each session (Fertonani et al., 2015).

Normality and homogeneity of variance were tested for all outcomes using the Shapiro–Wilk test and the Levene’s test. Independent samples t-tests were used to compare the difference in each outcome data before the stimulation intervention.

To investigate whether there was a dual-task interference effect, we used the one-way ANOVA to analyze behavioral data of single- and dual-task before the stimulation intervention at 2 visits (tACS or sham).

Primary analyses utilized two-way (stimulation condition × time) repeated-measures ANOVAs to examine the effects of tACS on working memory performance (ACC, RT, IES), θ power of rs-EEG (F3_PSD, P3_PSD) and PLV of θ band. Similar analyses were applied to the secondary outcomes including the performance of upright standing (V_ML, V_AP, and Vcop). Paired t-tests were used within groups to analyze the differences in each outcome before and after stimulation. ANOVA effect sizes are denoted using the bias η_p²; Cohen’s d is used to show t-test effect sizes (where Cohen’s d < 0.19 is a weak effect, 0.20–0.49 is a low effect, 0.50–0.79 is a medium effect, and >0.80 is a high effect.)

To study how the tACS influences behavioral performance, the Pearson correlation analysis was used to examine the correlation between the change of resting-state EEG (∆F3_PSD, ∆P3_PSD, ∆PLV) and the change of behavioral outcomes (ΔACC, ΔRT, ΔIES, ΔV_ML, ΔV_AP, ΔV_cop) after different stimulus conditions (post-pre).

Finally, the Mann–Whitney U test was used to examine the side effects of tACS, and the Fisher exact test was used to examine blinding efficacy.

The statistical software was SPSS 25.0 with significance α = 0.05.

When exploring the dual-task interference, we found no significant difference in working memory performance in either the upright posture or the seating conditions (p > 0.05) (see Table 3).

With respect to exploring the dual-task interference effects of postural control, as the data of single-task V_ML and single- and dual-task V_AP did not satisfy a normal distribution, a non-parametric correlation sample Wilcoxon signed-rank test was performed on this part of the results; we found that compared with those under the single-task, the V_ML (z = −2, p = 0.045) and V_AP (z = −2.9, p = 0.004) under dual-task conditions were significantly reduced by 0.4% and 8.0%, respectively (Table 3). These results indicate that there was less body sway during the concurrent execution of the working memory task; that is, working memory did not negatively interfere with upright postural control.

No interaction was noted between the effects of the two groups before and after stimulation (stimulation condition × time) on single-task working memory ACC (F_1,19 = 1.376, p = 0.225, η_p² = 0.068), RT (F_1,19 = 0.019, p = 0.893, η_p² = 0.001) and IES (F_1,19 = 0.483, p = 0.496, η_p² = 0.025) (Table 2). There were also no interactions for the effects on dual-task working memory ACC (F_1,19 = 0.069, p = 0.795, η_p² = 0.004), RT (F_1,19 = 0.417, p = 0.526, η_p² = 0.021), and IES (F_1,19 = 0.588, p = 0.453, η_p² = 0.030) (Table 2). A paired t-test was used to analyze the within-group data and found a 5.3% significant decrease in working memory RT under dual-task conditions after tACS (t = −3.157, p = 0.005, Cohen’s d = 0.742, Figure 5). In contrast, no significant difference was observed in the RT after sham stimulation (t = −1.279, p = 0.216, Cohen’s d = 0.3). The ACC and IES during the dual task and the ACC, RT, and IES during the single task were unaffected by any stimulation (Table 2).

No significant interaction was observed for V_ML (F_1,19 = 0.971, p = 0.337, η_p² = 0.051), V_AP (F_1,19 = 3.602, p = 0.074, η_p² = 0.167) or Vcop (F_1,19 = 1.820, p = 0.194, η_p² = 0.092) during single-task. There were also no interactions for V_ML (F_1,19 = 0.344, p = 0.565, η_p² = 0.020), V_AP (F_1,19 = 0.046, p = 0.833, η_p² = 0.002) or Vcop (F_1,19 = 0.038, p = 0.848, η_p² = 0.148) during dual-task (Table 2). Paired t-tests revealed no significant differences in postural control performance before or after both stimulations (Table 2).

RM_ANOVA revealed no interaction for spectral power changes of θ bands in channel F3(F_1,19 = 1.163, p = 0.294, η_p² = 0.58) or channel P3(F_1,19 = 0.179, p = 0.677, η_p² = 0.009). Paired t-tests failed to identify regions with a significant power difference following any stimulation conditions (Table 2).

RM_ANOVA revealed an interaction for PLV between channel F3 and channel P3 (F_1,19 = 5, p = 0.038, η_p² = 0.208). Post-hoc testing indicated that the PLV increased after tACS compared with that in the sham group. A paired t-test was used to analyze within-group data and found a significant increase in PLV after tACS (p = 0.010, Cohen’s d = 0.637, Figure 6). However, no significant difference was observed after the sham stimulation (p = 0.644, Cohen’s d = 0.105).

A moderate correlation was observed between ΔF3_PSD and ΔDT_RT (r = −0.515, p = 0.02), as well as between ΔP3_PSD and ΔDT_RT (r = −0.483, p = 0.031) in the tACS session (Figures 7A,B), indicating that improvements in working memory performance corresponded to an increase in the power spectra of channels F3 and P3. We also found a significant correlation between ΔV_AP and ΔF3_PSD in the single-task condition (r = 0.458, p = 0.042); that is, an increase in body sway velocity in the anteroposterior direction was significantly associated with an increase in power in the band of the F3 channel (Figure 7C). However, no significant associations were observed for other behavioral outcomes (p > 0.05).

A moderate correlation was observed between ∆PLV and ∆ST_RT (r = 0.455, p = 0.02) in the tACS session (Supplementary Table S1). No significant associations were observed for other behavioral outcomes (p > 0.05).

Eight adverse events (itching, tingling, burning sensation, warmth, fatigue, metallic/iron taste, phosphene, and others) with five severity levels (none, mild, moderate, considerable, and strong) were reported in the side-effect questionnaire.

None of the participants reported serious or severe adverse events associated with tACS stimulation. Only three participants reported considerable itching, and one reported considerable tingling in the tACS group. The Mann–Whitney U test showed no significant difference in side effects between the two groups (p > 0.05) (Supplementary Table S2).

For blinding efficacy analysis, we only included guesses from the first visit because participants’ guesses on subsequent visits may have been influenced by subjective guesses from the first visit. Fisher’s exact test revealed no significant difference in the blinding efficacy between the two stimulation conditions (p = 0.103) (see Table 4).

Neuropsychological studies have demonstrated that the frontalparietal network and brain region θ phase synchronization play an important role in the processing of WM. The tACS, as a tool for brain oscillatory modulation, can effectively elicit working memory-related neural oscillatory activity and improve WM performance (Polanía et al., 2012; Violante et al., 2017). In the present study, our results suggest that the WM reaction significantly improved after tACS. However, of note, the improvement in working memory in this study only occurred in conditions where upright postural control was performed (dual task). In contrast, WM performance in the seated position (i.e., single task) showed no significant change. This is not consistent with hypothesis 1, possibly because the demanding condition of limited resources under the dual-task condition is more sensitive to weak changes in the cognitive performance of healthy subjects. This may be due to differences in the brain mechanisms between single- and dual-task WM (Ozdemir et al., 2016). Our tACS montage, an HD-tACS with higher spatial accuracy than the traditional sponge tES montage, may modulate the brain regions of dual-task WM (Alam et al., 2016).

Contrary to our initial experimental hypothesis, the current study did not show significant differences in postural control performance between the two stimulation conditions. Moreover, postural control performance did not deteriorate in the dual-task condition, but instead appeared to improve while simultaneously performing WM, which was also unexpected. We speculate that this might be due to the attentional focus shift, namely the dual-task performance in the present study, which increases the automatic processing of posture (McNevin and Wulf, 2002; Riley et al., 2003) and decreases body weight sway by shifting the focus of attention from standing performance (internal focus) to the execution of a working memory task (external focus) (Wulf et al., 2001). However, redundant eye movements may also have an impact when performing visual cognitive tasks (Bonnet and Baudry, 2016; Bonnet et al., 2021). Previous reports have shown a synergistic relationship between the postural and visual systems, with the central neural system possibly guiding a more stable postural state to complete visual-cognitive tasks (Bonnet and Baudry, 2016; Bonnet et al., 2021).

Based on these findings, it is challenging to further investigate whether tACS influences postural control by improving working memory. As standing performance was already better in the dual-task situation (compared to single-task postural control performance), even if tACS had a positive effect on postural control during a working memory task, it may not have been well assessed. Future studies could select more difficult or real-life postural control tasks, such as throwing (Zhuang, 2021) or dynamic postural control tasks (standing on a translation platform, walking, and obstacle crossing) (Bogost et al., 2016; Lin and Lin, 2016; Ozdemir et al., 2016; Nóbrega-Sousa et al., 2020). In addition, as biological aging and age-related conditions appear, our cognitive function and brain mechanisms become increasingly crucial in preserving our ability to maintain balance (Manor et al., 2010; Manor and Lipsitz, 2013); future work could focus on older adults (Rizzato et al., 2021) to prevent ceiling effects.

Correlation analyses revealed a relationship between the change in EEG θ power and WM response time, i.e., an increase in θ power after tACS corresponded to an accelerated response to a WM task in the present study, which in part supports a link between the modulation of endogenous neural processes and changes in behavior. Available evidence suggests that the primary mechanisms by which tACS modulates the brain are entrainment of endogenous rhythms at the frequency of stimulation (Zaehle et al., 2010; Herrmann et al., 2013) and induction of synaptic changes via spike-timing-dependent plasticity (Zaehle et al., 2010; Vossen et al., 2015). The fronto-parietal in-phase θ-tACS in the present study may modulate endogenous θ oscillations in the brain with exogenous θ oscillations, allowing behaviorally relevant neural oscillatory networks (i.e., fronto-parietal θ phase synchronization) to be driven synchronously (Reinhart and Nguyen, 2019).

PLV analyses showed increased θ phase synchronization between frontal and parietal brain regions after tACS, as opposed to sham. The PLV is an indicator of phase synchronization across cortical regions, and previous studies have shown that phase synchronization can alter spike-time-related plasticity (Gregoriou et al., 2009; Wang, 2010). We speculate that external modulation of θ phase synchronization improves WM possibly due to neuroplasticity changes in functional connectivity. Intervention in the temporal synchronization patterns of large-scale human brain activity via tACS has the potential to enhance the postsynaptic effects of spiking impulses in one region on another, ultimately improving neural communication related to working memory capacity (Reinhart and Nguyen, 2019). However, contrary to our hypothesis, no significant changes in θ power values were observed after tACS in the present experiment. Studies that have investigated the effect of tACS on brain oscillatory power reported inconsistent results. Some studies reported a decrease in endogenous power values after tACS (Pahor and Jaušovec, 2017), while others found an increase in power values. Additionally, some studies found no change (Wischnewski et al., 2016) possible reason for the mixed evidence above is due to the heterogeneity of the variables (variables vary in terms of amplitude, power, and relative ratios) (Wischnewski et al., 2016; Kleinert et al., 2017; Pahor and Jaušovec, 2017; Battaglini et al., 2020).

In the current study, the transcranial electrical stimulation (tES) protocol used a high-definition stimulation montage design (HD-tACS). To date, few studies have used HD-tACS (Reinhart and Nguyen, 2019; Klírová et al., 2021), rendering the application of HD-tACS in this study a possible limitation. HD-tES has the advantage of higher spatial accuracy than the traditional sponge tES montage, while the latter produces more diffuse currents throughout the brain (Datta et al., 2009; Ruffini et al., 2014; Alam et al., 2016). However, the more concentrated the current pattern produced by HD-tES, the less modulation of relatively distant brain regions of the target function may be achieved, ultimately resulting in weakening of the modulation effect (Hill et al., 2019). Overall, the effects of focused current patterns produced by HD-tACS must be investigated in detail. Furthermore, the physiology of a participant’s head is quite variable (Truong et al., 2013; Opitz et al., 2015) and the method of locating the stimulation target area using the same EEG cap may ultimately result in the actual stimulation site deviating from the ideal stimulation target area. Even more, high definition montage designs can amplify this limitation (Mikkonen et al., 2020). To this end, magnetic resonance imaging can be used to determine individual target areas (Datta et al., 2012; Mikkonen et al., 2020; Klírová et al., 2021).