Fifteen healthy right-handed adults (five men and 10 women) aged 27 ± 7 years (mean ± SD), range 20–41, volunteered to participate in this study. As lateralization of the vestibular system depends on handedness (Dieterich et al., 2003), only right-handed participants were selected to decrease the interindividual variability and enable better interpretation of the results. The Edinburgh test of manual laterality was used to confirm handedness (Oldfield, 1971). Participants who reported a history of otologic, neurologic, cardiovascular, orthopedic, or traumatic illnesses were excluded.

The experimental protocol was performed in two parts: first, the motor threshold (T) to the GVS was determined during standing, and then GVS was applied at an intensity of 1 or 1.5 T when participants walked at their comfortable speed on a treadmill.

The complete procedure to determine the motor threshold to GVS is described in Mikhail et al. (2021), but briefly participants stood with their head facing forward, eyes closed, arms along the body, and wearing flat shoes on a force platform. The distance between the left and right medial malleoli was 50% of pelvic width, which was measured using calipers at the level of the greater trochanter. GVS was applied 10 times in random order for each of the following intensities: 0 (control), 1, 1.5, 2, 2.5, 3, 3.5, and 4 mA, with a minimum interval of 5 s between each stimulus. A recruitment curve was constructed using the response evoked by the GVS on the accelerometer signal. The intensity at which 50% of the responses exceeded the mean of background signal+1 SD was identified as an initial threshold (Figure 1A). Next, to obtain a more precise threshold, we stimulated at each 0.1 mA around the initial threshold to obtain a final threshold.

Participants were stimulated at 1 and 1.5 T (30 stimuli each) while walking on a treadmill at a comfortable speed (mean = 1.03 ± 0.06 m/s; Figure 1B). A pressure sensor was placed underneath the right heel to determine the gait cycle onset. The GVS was delivered at the onset of the stance phase of the right leg (10 ms after the pressure was increased under the right heel). Each participant took part in two sessions separated by at least 2 days. In one session, the cathode and anode were placed over the right and left mastoid process, respectively. In the other session, the electrode placement was reversed, with the cathode placed behind the left ear and the anode behind the right ear. The order of the sessions was randomized between the participants.

Data are expressed as means ± standard error of the mean. The Shapiro–Wilk test was used to test the normal distribution of the data. Two-way repeated-measures analysis of variance (ANOVA) was used to compare amplitude values of different configurations (RCathode vs. LCathode) and intensity (1 and 1.5 T). The effect of GVS on the cycle duration, heel stance duration, and EMG duration was assessed at 1 and 1.5 T using a one-way repeated-measures ANOVA. Tukey’s or Holm-Sidak multiple comparison tests were used to compare the control, stim, and next cycles. T test was used to investigate the effect of cathode configuration (right × left) on the response of each muscle to GVS at each intensity. p < 0.05 was considered statistically significant for all tests. We used Cohen’s d (Cd) to investigate effect size (small, ≤0.2; medium, 0.2–0.8; and large, ≥0.8; based on Lakens, 2013). All statistical analyses were conducted using the Prism software version 9.3.1 (GraphPad Software, Inc., San Diego, CA).

Each participant took part in two sessions. In each session, a RCathode or a LCathode configuration was used. First, to determine the intensity of stimulation, a recruitment curve was performed for all participants, and the motor threshold (T) for vestibular response was identified, as illustrated in Figure 1A for a representative participant. For this participant, the threshold for RCathode was obtained at 2.3 mA. For the group, the mean thresholds were 2.6 ± 0.19 for RCathode and 2.06 ± 0.2 for LCathode. Next, two distinct intensities were used to evoke vestibular responses during gait: 1 and 1.5 T.

GVS applied during gait did not induce any adverse events. However, strict guidelines were given with the participants prior to data collection. Notably, they had to always fix a point in front of them and refrain from turning their head to the side. Pilot testing indicated that when participants turned their head while stimulation was applied, they briefly became disoriented and had to hold on to the rails to prevent a fall. Because the participants were young and healthy, they could focus on the task, and no fall occurred. Therefore, the participants walked looking ahead and did not feel any discomfort. Some participants had tingling sensations and dry skin/itching behind their ears at the end of the session, as reported previously by others (Utz et al., 2011). However, these inconveniences were mild and brief.

Each participant walked at their preferred speed on a treadmill (mean = 1.03 ± 0.06 m/s). Figure 1B shows the ongoing EMG pattern and heel strike during gait in a representative subject. The black traces represent the EMG activity when GVS is applied 10 ms after the onset of the gait cycle. It shows alternation between the TA and SOL in both the left and right legs, as well as the heel contact period (heel stance). The onset of the right heel contact determines the onset of the right stance phase and was referred to as the onset of the gait cycle. The gray trace shows the EMG and heel contact activity when no GVS was applied. The stimulated trace (black) is superimposed over the control trace (gray) and enables the identification of the responses evoked by GVS.

Figures 2A,B details the response pattern observed in a representative participant for both RCathode and LCathode configurations. Although the responses varied within the group (see Supplementary Material), LLRs were the main responses observed for the right SOL, the right TA, and the left TA across intensities and configurations. Therefore, these were the main responses that were further assessed (Table 1). SLR was the main response observed in the left SOL and, accordingly, was further analyzed.

Responses were facilitatory in both left and right SOL in an RCathode configuration, and they were inhibitory in both SOL in an LCathode configuration. Observing the same responses in both legs during gait is different from what would be expected by applying bipolar, binaural GVS while standing with the head facing forward. In the latter condition, responses between the left and right SOL are generally reversed with the opposite polarity (Lund and Broberg, 1983; Pastor et al., 1993; Fitzpatrick and Day, 2004). Observing similar responses on both sides was previously observed during gait, which suggests it might be inherent to the locomotor task (Iles et al., 2007). Except for the left SOL, most responses observed during gait were LLRs, which again contrasts with responses evoked during standing, with the MLR being the most analyzed. This finding suggests that longer latency responses may be a feature of vestibular responses during gait, as was proposed previously (Iles et al., 2007).

While LLRs were observed in the RSOL, only SLRs were observed in the LSOL. One explanation might be that L SOL EMG rapidly decreases at the offset of stance to give way to swing and to TA contraction (Simonsen, 2014). As EMG activity is absent during swing, it might not be possible to see a response at later latencies.

The nature of the dominant vestibular responses (SLR, MLR, and LLR) in the TA was not as clear as in the SOL, and a larger variability was observed depending on the intensity and GVS configuration. The apparent stability in the responses observed in SOL could reflect the preferred connectivity of vestibulospinal neurons to extensor muscles during gait, as was shown for treadmill walking in cats (Orlovsky, 1972; Matsuyama and Drew, 2000). Furthermore, as we are stimulating at the onset of the stance of the right leg, we are in the double support phase where balance is more of a challenge (Nielsen, 2003; Simonsen, 2014). The weight transfer implies coordination between the right SOL EMG, which is rapidly increasing to accept the body weight at the onset of stance, and the left SOL EMG, which is also increasing for push off at the offset of stance and then quickly decreasing to start the swing phase. This could explain the more robust and repeatable responses obtained in SOL, as clear responses are mainly observed in muscles actively engaged in a balance task (Wardman et al., 2003; Fitzpatrick and Day, 2004).

RCathode GVS prolonged the duration of the ongoing gait cycle by increasing the burst duration of muscles located on the anodic side. This more important effect on the anodal side echoes previous findings: when applying GVS while standing, the main behavioral response is a deviation of the body (lean) toward the anode (Britton et al., 1993; Fitzpatrick and Day, 2004). Furthermore, the GVS also caused participants walking overground to deviate their trajectory toward the anode (Bent et al., 2000). Although the effect of GVS on EMG burst duration was not assessed in the latter study, the stimulation might have prolonged the EMG on the anodic side and contributed to the deviation. Moreover, the RCathode GVS also influenced the subsequent gait cycle and shortened its duration. This may reflect a compensatory strategy aimed at recovering a normal length cycle during gait (Eng et al., 1994; Schillings et al., 2000; Nachmani et al., 2020). The involvement of brainstem networks in determining burst and cycle duration was also observed during fictive locomotion in neonatal mice. Removal of the brainstem decreases the burst duration (Jean-Xavier and Perreault, 2018). In parallel, in patients with chronic bilateral vestibular loss gait initiation is characterized by a lower maximum displacement of the center of pressure n the first- and second-steps compared with control participants (Sasaki et al., 2001).

In the current study, RCathode GVS induces prolonged burst duration on the anodic side, which might be explained by altered perception/sensation induced by GVS. Bilateral bipolar transmastoid GVS causes a whole body sway directed toward the ear with the anode, irrespective of the orientation of the head in the yaw plane (Pastor et al., 1993). Fitzpatrick and Day (2004) introduced a vector summation model, suggesting that the brain interprets bilateral bipolar GVS predominantly as a “rotation toward the side of the cathode,” and that in reaction to this perception, the evoked postural response (body sway) is directed toward the anode. Hence, participants would perceive that they were falling toward the cathodic side, i.e., on the right as they are putting their right foot on the ground. As a reaction, the participant would have a postural response toward the left side (anode), in the subsequent step. This reaction could explain the longer EMG bursts on the anodic side (left), while the cathode was behind the right ear. The opposite GVS configuration (LCathode-RAnode) would not be in congruence with the postural reaction during the stimulated cycle, because the participant will perform a step with the left leg following stimulation, which is on the cathode side.

An alternative explanation for the difference in the effects triggered by the left and right cathode is the vestibular dominance of the right-handed participants recruited. As the vestibular system is lateralized (Dieterich et al., 2003), the right cathode might have a more preponderant effect on the vestibular system in the right-handed participants. Although this is only a hypothesis, it rests on previous data (Fink et al., 2003) showing that the right and left cathode stimulation activated different parts of the brain in right-handed participants and that different regions are activated by GVS in right- and left-handers (Kirsch et al., 2018). A similar asymmetry could occur caudally in vestibular projections to the spinal cord, which would lead to more dominant responses when the cathode is on the right side. Further studies could shed light on this hypothesis.

Behavioral responses observed during GVS are due to the activation of the peripheral vestibular system. GVS was found to predominantly activate or inhibit thick irregularly firing afferent fibers from both semicircular canals and otoliths (Goldberg et al., 1982, 1984), as well as hair cells (Norris et al., 1998; De Waele et al., 2002; Cheng et al., 2010; Gensberger et al., 2016). Cathodal current increases action potential discharge, whereas anodal currents silence vestibular afferents (Angelaki and Perachio, 1993; Straka and Dieringer, 2000). The different afferent signals will then be processed centrally, weighted, and integrated with other sensory inputs. The information would be transmitted to the lower limb either directly through the vestibulospinal tract or indirectly through the reticulospinal tract (Wilson et al., 1979; Kennedy et al., 2004). These descending pathways then modulate MN excitability through polysynaptic or disynaptic interneuronal networks (Gossard et al., 1996). Therefore, segmental interneurons that regulate MN excitability and are activated or inhibited by changes in vestibular afferent discharge could contribute to these SLR, MLR, and LLR responses. Notably, Renshaw cells provide recurrent inhibitory input to MNs and respond to descending vestibulospinal input (Pompeiano, 1988). Increased activity of vestibular afferents (following cathodal stimulation) could facilitate the motoneuron pool by inhibiting Renshaw cells (Kennedy et al., 2004). Another potential mechanism is the modulation of reciprocal inhibition and presynaptic inhibition interneurons through both vestibulospinal and reticulospinal pathways (Manzoni, 1988; Iles and Pisini, 1992; Kennedy et al., 2004).

Vestibular signals can also be integrated into the ongoing gait pattern. The Central pattern generator (CPG) enables a specific pattern of muscle activation that leads to flexion/extension and left–right alternation (Rossignol et al., 2006). It also adapts the locomotion to incoming sensory feedback (notably from the treadmill belt). This intrinsic spinal locomotor network might have a gating effect on the potential modulation of locomotion by vestibular afferents (Guillaud et al., 2020). This possibility might underlie the small effect size we observed in the prolonged duration of the gait cycle, especially when compared to the larger effect size of vestibular responses induced in the muscles (SLR, MLR, and LLR). As vestibular signal relays information about head movement, such gating might be used to reduce the influence of head movement on the locomotor pattern, as suggested by others (Pflieger and Dubuc, 2004).

A limitation of this study is that the gait cycle was only assessed using heel sensors on one side (the right side); thus, the duration of the gait cycle itself was not assessed on both legs, which prevents direct assessment of symmetry. Moreover, the participants were tested only with eyes open and at comfortable walking speeds. As the modulation of EMG by GVS is larger with eyes closed and at slower speed, a wider bracket of speeds should be tested, and both eyes open and eyes closed conditions should be investigated.

Only right-handed individuals were tested, which prevents extrapolation of our conclusion. Future studies should include both right-handed and left-handed individuals. Moreover, our cohort was young, and the results might be different in older adults, where GVS sensitivity is generally increased (Jahn et al., 2003; Dalton et al., 2014; Peters et al., 2016). Furthermore, although the lengthening of muscle activity on the anodal side underlies the lengthening of cycle duration, the data do not inform on the specific contribution of the vestibular system to the muscular changes observed. In subsequent steps, other techniques such as time and frequency correlation approaches (see Blouin et al., 2011) should be used to clarify the vestibulo-muscular coupling during the gait cycle.

Another limitation is that the recruitment curve was taken during standing and not during gait. Although the latter approach would have enabled a better approximation of the excitability of the neuronal networks during gait, it was technically more of a challenge due to time constraints. Indeed, a large amount of stimulation was necessary to construct this curve. During standing, we can apply this stimulation relatively quickly in a randomized manner. During gait, it takes longer, as we do not want the stimulation to be applied in consecutive cycles, and we offered frequent resting periods to participants to prevent fatigue.

The main effect of a short-duration GVS burst applied at the onset of the right gait cycle was the prolongation of EMG activity on the left side (anodic side) without changing EMG activation on the right side. Repeating this stimulation at each cycle might enable the lengthening of the gait cycle over the course of a sequence of locomotion. This approach might be useful in a therapeutic/training paradigm with people with asymmetrical gait, for example, after a stroke (Verma et al., 2012). Indeed, asymmetrical gait is an important issue in the rehabilitation of patients with neurological disorders, and one that has received much attention is poststroke gait impairment (Patterson et al., 2010; Verma et al., 2012; Beyaert et al., 2015). Gait asymmetry is reflected by a significant difference in the length of the left and right gait cycles (Kim and Eng, 2003; Patterson et al., 2010; Wonsetler and Bowden, 2017). During therapeutic interventions, the restoration of symmetrical gait (Harris-Love et al., 2001; Lindquist et al., 2007; Kahn and Hornby, 2009) is an important goal, but it is rarely fully achieved. Studies have demonstrated that asymmetrical gait can be reduced in patients with stroke by using asymmetrical somatosensory cues, such as those provided by a split-belt treadmill (Reisman et al., 2007; Malone and Bastian, 2014). Our results suggest that asymmetrical vestibular cues might effectively address gait asymmetry as well. With a configuration where the cathode is on the nonparetic side, repeatedly applying a low-intensity GVS burst at the onset of the gait cycle could lengthen EMG bursts on the paretic side and increase cycle duration. Future studies should also determine if a short burst, a longer burst, or continuous GVS (such as nGVS) with an RCathode configuration might be appropriate in that regard.