Two male Rhesus monkeys (Macaca mulatta) were trained to sit in a primate chair. Under strict sterile conditions and propofol anesthesia (10 mg/kg/h), they then underwent two surgeries each. During the first surgery a headpost (Crist Instruments) was implanted on the skull and fixed using ceramic screws and dental cement which effectively functioned as an insulating layer to prevent current flow through the skin (see Figure 1A). At least six weeks later, both animals were trained in a passive fixation task. For this task they received a juice reward for fixating at a point on a screen in front of them. The task did not serve any experimental purpose; rather, it was to keep the monkeys alert during neural recordings. During the second surgery, a craniotomy above area PFG on the parietal convexity was drilled (Figure 1A) above the right hemisphere for monkey P and above the left hemisphere for monkey D, and a recording chamber was implanted over the craniotomy at approximately a 45° angle in relation to the mid sagittal line. This placement allowed an oblique entry into the parietal convexity we recorded only from this hemisphere. In the same surgery, a concentric ring electrode (CRE medical, Kingston, United States, outer ring diameter: 10 mm, center disk diameter 2 mm) was implanted directly on the skull to target the same PFG area. A second concentric ring electrode (same dimensions) was implanted on the other side of the head to target the same brain region on that side. This montage allowed for focused as well as unfocused stimulation, both described below.

Each of the 2 Monkeys was housed in an enclosure (3x3x2.5 m W x L x H) – with two other monkeys – which contained swings and perches and bedding on the floor. The layout of the enclosure was changed weekly and monkeys received toys which were scattered around. They had daily contact with researchers and care takers. Around experiment days they received a total of 1 Liter of liquid (water and juice reward); otherwise their liquid intake was not restricted. All procedures adhered to the National Institutes of Health guide for the care and use of laboratory animal, the EU directive 2010/63/EU and were also approved by the animal ethics committee for laboratory experiments at KU Leuven (ethics approval number: P126/2017).

Figure 2, shows group results of unfocused 10 and 40 Hz stimulation from 0.25 to 3 mA (10 Hz n = 90, 40 Hz n = 73); higher stimulation intensities tended to increase entrainment and spike rates (PLV – 10 Hz – 1 mA: p < 0.001, 2 mA: p < 0.001, 3 mA: p < 0.001; 40 Hz – 0.5 mA: p < 0.05, 1 mA: p < 0.001, 2 mA: p < 0.001, 3 mA: p < 0.001; spike rate – 10 Hz – 3 mA: p < 0.001; 40 Hz – 3 mA: p < 0.05; Wilcoxon sign rank; see figure for full statistics). 10 and 40 Hz entrainment were similar across stimulation intensities (Supplementary Figures 3A,B; Wilcoxon sign rank, see Figure for full statistics). A linear mixed model [PLV ~ stimIntensity + (1|subject)] showed that entrainment increased with increasing stimulation intensity (entrainment-10 Hz z = 54.49 p < 0.001, 40 Hz z = 19.97 p < 0.001) (Supplementary Figures 2A,B).

We calculated spike rate changes (Δ spike rate) by subtracting the stimulation-OFF rates from the stimulation-ON rates. For both 10 and 40 Hz higher intensities tended to cause spike rate increases. Furthermore, a linear mixed model [spikeRate ~ stimIntensity + (1|subject)] analysis showed that spike rate increase depended, for both 10 and 40 Hz, on stimulation amplitude (spike-rate-10 Hz z = 54.49 p < 0.001, 40 Hz z = 19.97 p < 0.001) (Supplementary Figures 2C,D).

Figure 3 shows the effects of focused stimulation at amplitudes from 0.25 to 4 mA (2 upper panels 10 Hz n = 47 and 2 lower panels 40 Hz n = 25). At 2 mA neural activity became entrained to the stimulation (10 Hz – 2 mA: p < 0.05, 3 mA: p < 0.05, 4 mA: p < 0.05; 40 Hz – 2 mA: p < 0.01, 3 mA: p < 0.05, 4 mA: p < 0.01; Figures 3A–F for 10 Hz and M-R for 40 Hz). However, in contrast to unfocused stimulation this montage did not affect spike rate, even at 4 mA (Wilcoxon sign rank, see Figures 3G–L,S–X for full statistics). Furthermore, 40 Hz focused stimulation caused comparable levels of entrainment to the focused 10 Hz (Supplementary Figure 3B, Figure contains full statistics).

At amplitudes where entrainment was observed in both montages (i.e., 2 and 3 mA) the level of neural entrainment was higher in the unfocused montage (Supplementary Figure 3C, 3 at 2 mA p < 0.001, at 3 mA, p < 0.001, panel 4 at 2 mA p < 0.001, at 4 mA p < 0.001 see figure for full statistics; Wilcoxon sign rank). This is likely due to the stronger electric field values in the brain with the unfocused montage (Figures 4B,C at 10 Hz p < 0.001, at 40 Hz p < 0.001, Wilcoxon sign rank; see Figure for full statistics).

Figure 4A shows good agreement between model estimated values electric potential values and the measured electric potential in monkey P. Figures 4B,C shows the model estimated electric fields at the unfocused recording sites was significantly larger than at the focused sites (Wilcoxon sign rank 10 Hz p < 0.001, 40 Hz p < 0.001). A linear mixed model analysis showed that entrainment increases with increasing electric field strength for the unfocused montage at both 10 [PLV ~ eField + (1|subject); p < 0.001] and 40 [PLV ~ eField + (1|subject); p < 0.001] Hz (Figures 4D,E). A similar effect was found for spike rate [spikeRate ~ eField + (1|subject)] but only with 10 Hz (p < 0.01) stimulation (Figures 4F,G). Overall, focused stimulation did not show this electric field dependency on entrainment and spike rate effects. Spike rate at 40 Hz did increase with increasing electric field strength (p < 0.001), although this might be driven by an outlier (see Supplementary Figure 4 for all statistics).

Figure 5 shows the fMRI activations caused by unfocused ECS stimulation at 10 Hz and 3 mA in monkey D. ECS activated a relatively restricted region under the stimulation electrodes (white arrows in Figure 5), but also in the Lateral and Superior Temporal Sulcus (STS, blue arrows). Note that the activations were located slightly lower and more anterior in the left hemisphere, consistent with the lower and more anterior position of the stimulation electrode on that side. Supplementary Figure 5 shows a similar figure for monkey P. However, except in one session, ECS at 1 and 2 mA did not evoke activations (data not shown). These results demonstrate that ECS at 3 mA can activate the target area under the electrodes in addition to a series of other cortical sites. Furthermore, using fMRI it is possible to visualize this activity.

This study provides the first in-vivo evidence that ECS generates strong E-field ranging from 1 to 30 V/m (depending on distance to stimulation-electrode) in the macaque brain (Figures 5D–G). We expect lower values in humans due to the thicker skull. On the other hand, DBS (Huang et al., 2017; Lafon et al., 2017) and TES (Khatoun et al., 2019b; McIntyre et al., 2004) where, respectively, 100 V/m and 0.15 V/m are generated at the same 1–2 mA intensity.

Neural response to ECS was linear with higher stimulation intensities causing larger responses (Supplementary Figure 2). This revealed a wide range of neuromodulation from small entrainment [similar to TES (Krause et al., 2019; Ozen et al., 2010)] levels at low stimulation intensities to robust spike rate increase [similar to DBS (McIntyre et al., 2004)] with high intensities (Figures 2, 3).

The linear neural response is in line with many electric modulation paradigms (Asamoah et al., 2019; Vieira et al., 2020; Vöröslakos et al., 2018). Although we expect that further increase would truncate this linearity; however this was not tested. It should be noted that some electric neuromodulation studies (mostly using direct current) have reported (partially) non-linear effects (Batsikadze et al., 2013; Vimolratana et al., 2023). It remains unclear how the non-linearity relates to this study given the particular conditions of this work.

Observed effects between focused and unfocused montage were mostly similar although there were some differences. Most notably the unfocused montage induced stronger E-fields and spike rate as well as entrainment responses depended on stimulation amplitude and E-field strength (Figures 4D–G; Supplementary Figure 2). This relationship was not as clear for the focused montage (Supplementary Figure 4) which tends to generate weaker and more focused E-fields in the brain (Khatoun et al., 2018a, 2018b). Recording positions may sometimes have been sub-optimal in relation to the stimulation electrode which targeted a small brain area. It is thus possible that the focused montage also causes amplitude dependent effects in the stimulation intensity range we investigated.

40 Hz unfocused stimulation appeared to cause entrainment at lower stimulation intensities (in Figures 2B,L entrainment starts at 0.5 mA for 40 Hz and 1 mA for 10 Hz). A direct comparison however showed this frequency difference was not significant (Supplementary Figures 3A,B). Notably, Monkey D seems to have a lower response threshold as compared to Monkey P (see stimulation amplitude 0.5 mA in Supplementary Figures 6, 7 and Figure 2). This is in line with other non-invasive neuromodulation where differences between individuals have been reported (Hsu et al., 2016; Santarnecchi et al., 2016; Zanto et al., 2021). It would be interesting to test whether other brain regions would respond differently. For example, in the hippocampus where gamma band plays a central role in memory encoding (Lisman and Jensen, 2013) 40 and 10 Hz stimulations may elicit differential responses.

A crucial advantage of our ECS approach in monkeys is that we could chart the effects of ECS throughout the brain by means of an fMRI study, where we only used the unfocused montage. Numerous previous studies using intracortical stimulation during fMRI have furnished invaluable insights into the organization of cortical networks underlying face (Moeller et al., 2008), body (Premereur et al., 2016), attention (Ekstrom et al., 2008), and 3D shape processing (Van Dromme et al., 2016). We observed localized activations in the cortex immediately below the stimulation electrodes rather than the diffuse activations that may have been expected from the modeling [Figure 1A (inset)]. This observation is highly similar to the results of a previous TMS study in monkeys (Romero et al., 2019), which measured highly localized spiking responses in a 2 by 2 by 2 mm volume of cortex under the TMS coil despite the widespread electric field effects predicted by modeling. A thresholding phenomenon may explain why the activations were localized under the stimulation electrode (note that we also did not observe any fMRI activations at 1 or 2 mA.)

The fMRI study showed that ECS activates areas directly under the electrodes as well as at remote locations (Figure 5; Supplementary Figure 5). Interestingly these remote activations were localized implying they were not immediately caused by the generated electric field in the brain or conduction through the cerebrospinal fluid. Rather, they were activated indirectly by the areas that were directly activated by the stimulation. Future studies investigating the behavioral effects of ECS will also benefit from ECS-fMRI for the interpretation of results.

DBS and ICS are effective therapies; however, due to their high cost and invasive nature, they are only considered in advanced disease stages (Kennedy et al., 2011; Krack et al., 2019; Lozano et al., 2019; Tsubokawa et al., 1993). These considerations are minimized in the portable ECS system (Kravalis and Schulze-Bonhage, 2020; Schulze-Bonhage et al., 2023) where electrodes are implanted in the subgaleal space between the skin and the skull. Through technological advancement ECS implantation will likely require only an incision in the scalp under local anesthesia thereby siginificantly reducing cost, risk and discomfort to the patient. As a result ECS could be considered at a much earlier disease stage than more invasive neuromodulation methods. In a current clinical trial epilepsy patients receive subthreshold stimulation with the option to self-administration; current results show a promising development (Schulze-Bonhage et al., 2023).

On the other hand, TES has been investigated as a noninvasive treatment for a range of neurological and psychiatric disorders (Mehta et al., 2015; Palm et al., 2016). Its advantage is that it can be applied at a much earlier disease stage. However, it is not yet an accepted therapy, possibly because of reproducibility issues caused, among others, by weak E-fields in the brain (Asamoah et al., 2019; Lafon et al., 2017). In a subpopulation of these patients ECS can be applied more effectivity. More studies (animal models and clinical trials) are needed to fully explore the potential of ECS as a neuromodulation therapy. Intersectional short duration pulses (ISP) or interfering fields (IFS) are novel neuromodulation approaches which may bypass neurons in superficial layers (Grossman et al., 2017; Vöröslakos et al., 2018). In a recent computational study (Khatoun et al., 2021) we showed that these novel approaches combined with ECS allow for stronger and more focused E-fields in subcortical regions than TES.

Our ECS approach combined with extracellular recordings showed that moderate stimulation intensities (1–2 mA) induce robust neural entrainment without changing the spike rates of the neurons. This observation opens the possibility to study the neural and behavioral effects of pure neural entrainment at different frequencies in a cortical area under the stimulation electrode. Moreover, this robust entrainment means that applying in-phase or out of phase ECS in two distant but interconnected areas may provide a critical test of the communication by coherence hypothesis (Fries, 2015), which states that interareal communication is improved if two areas oscillate together. With further carefully designed clinical trials we can better understand how to use the different montages to target very specific brain region (e.g., in DBS) or induce a more network response as the fMRI experiment showed.

A limitation of ECS is that it requires surgery, possibly under local anesthesia. As such, it could not be used to probe neural mechanisms in healthy volunteers. Furthermore, our ealier work shows that insulation can prevent current flow between implanted electrodes and prevent stimulation of scalp and cervical nerves. Nevertheless, it can still stimulate dura nerves (Lv et al., 2014). By choosing the right stimulation intensity or approach (i.e., ISP and IFS) this co-stimulation can be mitigated while achieving strong E-fields in the brain.

In this study we implanted the electrodes on the monkey’s skulls and conducted experiments during a period of approximately nine months per monkey. In this period the electrodes functioned and caused no obvious damage to the skull or neural tissue. For ECS to function as a long duration neuromodulatory approach, stimulation electrodes will be implanted for prolonged periods of time (years). This work however does not provide direct insight into this aspect of ECS. However, deep brain stimulation has been applied as a therapy for decades (Delgado et al., 1952; Sironi, 2011), it is well tolerated, reversible and unknown to cause brain damage (Krack et al., 2019; Nuttin et al., 1999). Previous rodent studies have charted damage as a response to stimulation (Grossman et al., 2017). Future animal studies as well as ongoing clinical trials may give deeper insights into tissue responses to prolonged stimulation of skull and brain.

We note that for the fMRI section of this study the monkeys were anesthetized. Anesthesia generally may induce inhibition and reduced excitation of neural activities (Kitamura et al., 2003). Activity may show increased correlation in functional areas while activity between functional areas may be decreased (Bettinardi et al., 2015; Ordek et al., 2012). We therefore expect the fMRI results to differ from awake responses. Notably, neural activity recorded under anesthesia remains active and informative. Further studies can determine the extent to which awake responses may be different from anesthetized responses in ECS.