Early intervention with electrical stimulation reduces neural damage after stroke in non-human primates

Abstract

For patients experiencing ischemic stroke, acute intervention offers the most critical therapeutic opportunity as it can reduce irreversible tissue injury and improve functional outcomes. However, currently available treatments within the acute window are highly limited and have strict patient selection criteria. Although emerging neuromodulation techniques have been proposed as a treatment for chronic stroke, acute stimulation is rarely studied due to concerns about exacerbating ischemia-induced electrical instability. Here, we demonstrate that acute cortical electrical stimulation, administered one hour post-stroke, provides neuroprotection in non-human primate brains. Using advanced electrophysiology and histology tools, we found that applying continuous theta burst electrical stimulation directly adjacent to the ischemic lesion significantly reduced neural activity in the surrounding tissue, as evidenced by lower electrocorticography signal power and c-Fos expression. This reduced depolarization was accompanied by decreases in neuroinflammation and infarct volume in the sensorimotor cortex. These findings suggest that acute electrical stimulation may serve as a safe and effective early intervention, offering a promising therapeutic strategy to improve outcomes in ischemic stroke.

The therapeutic effects of acute neuromodulation after stroke remain less studied. Here, authors demonstrate that early electrical stimulation after ischemic stroke reduces excessive neural activity and limits tissue damage in primates, supporting the potential of brain stimulation as an acute treatment strategy for stroke.

Introduction

Ischemic stroke is a major type of brain injury that results in high mortality and serious long-term disability for adults, especially in the aging population. Globally, over 7.6 million people suffer from ischemic stroke each year, causing significant health and economic burdens worldwide^1,2. An ischemic stroke happens when blood flow within the brain is interrupted, leading to a lack of oxygen supply, energy depletion, and subsequent neuronal death. Acute intervention within hours after stroke onset offers the most critical therapeutic opportunity to mitigate irreversible tissue injury, thus improving neurological and functional outcomes for stroke patients³. However, currently available treatments within the acute window are highly limited, and approved interventions such as the administration of tissue plasminogen activator (t-PA) and catheter-based thrombectomy often have strict patient selection criteria^3,4. During the past few decades, there has been a large amount of experimental research and clinical trials on neuroprotective drug therapies targeting acute ischemic stroke, with the aim to interrupt the ischemic cascade and thereby reduce neuronal death^5–7. However, most of the drugs failed to show consistent clinical efficacy when moving from animals to humans^8–10. Therefore, there is a pressing need to expand the therapeutic options for acute ischemic stroke and improve their translation from bench to bedside to help millions of stroke patients retain the maximum quality of life.

In recent years, neuromodulation paradigms such as electrical brain stimulation have been proposed as a promising treatment for ischemic stroke. Most of these stimulation paradigms target the subacute or chronic phase of stroke to promote neural plasticity and functional recovery, rather than reducing permanent ischemic damage^11–16. As a result, it might take months of treatment in conjunction with rehabilitative training for only a subset of patients to see positive results from these interventions^17–19. Meanwhile, comparing to the chronic implementation of electrical stimulation for stroke, the acute application of such strategy is still relatively rare, especially in larger animal models and humans due to the risk of causing adverse effects and greater tissue damage related to ischemia-induced electrical instability and spreading depolarizations (SDs). It has been widely reported that perilesional tissues are particularly susceptible to SDs, marked by intense neuronal depolarization waves that can lead to increased metabolic stress, neuronal swelling, and lesion progression^20–23. However, despite the concern about SDs, a few studies have successfully demonstrated the use of sensory or direct brain stimulation to exert neuroprotection and reduce tissue damage for rodents with acute stroke^24–28. Similar to neuroprotective drugs, these acute stimulation strategies aim to reduce the irreversible damage caused by the ischemic cascade within hours after stroke onset or after reperfusion by reducing inflammation, oxidative stress, excitotoxicity, and apoptosis, thus preserving the perilesional tissues surrounding the ischemic core^29,30. While these results from rodent studies are promising, the scale and anatomical differences between rodent and human brains pose significant challenges to the feasibility of clinical translation³¹. Therefore, we need a more comprehensive understanding of how electrical brain stimulation drives changes in the physiology of neuronal networks at scales comparable to that of the human brain before these strategies can be successfully translated from bench to bedside for acute stroke patients.

As a result, two major gaps in knowledge need to be addressed as we seek to implement brain stimulation paradigms for acute ischemic stroke: 1) The protective effect of electrical stimulation needs to be evaluated across large cortical areas in a more clinically relevant animal model. 2) We need to have a better understanding of the mechanisms underlying stimulation-induced changes, from both electrophysiological and cellular perspectives. In this work, we used a novel set of approaches capable of addressing these two gaps to investigate stimulation-induced neuroprotection. We combined a lesion-based toolbox^32,33, state-of-the-art neurophysiology techniques, and a range of histology markers to study the neuroprotective effects of cortical electrical stimulation following acute ischemic stroke in non-human primates (NHPs). We compared multiple aspects of physiological responses to stimulation from large areas (~ 3 cm²) of the macaque sensorimotor cortex for up to 4 hours after stroke onset to interrogate the mechanisms underlying any observed neuroprotection.

Here, we show that early electrical stimulation suppresses perilesional neural depolarization and attenuates cellular markers of inflammation, coinciding with smaller lesion volumes in the sensorimotor cortex of NHPs. The unprecedented insights gained from these experiments will inform the development of next-generation brain stimulation paradigms that can be used as an alternative treatment during acute stroke to minimize neural damage, reduce severe disabilities, and improve functional outcomes for stroke patients.

Results

In this study, we induced controlled cortical ischemic lesions in the primate sensorimotor cortex using the photothrombotic technique^32–34, which generated focal infarcts by photo-activation of a light-sensitive dye (Rose Bengal) that interrupts local blood flow. The infarct size and location are controlled by setting a constant illumination intensity and aperture size through an opaque light mask placed above the cortical surface (Fig. 1a, b). In control monkeys D and E, we collected electrocorticography (ECoG) data through our customized multi-modal interface (Fig. 1b), which includes baseline before lesioning, during lesion induction, and up to 3 hours post lesioning to monitor network dynamics around the injury site. In stimulated monkeys F and G, we recorded 1 hour after lesioning and then applied electrical stimulation at approximately 8 mm medial to the lesion center on the ipsilesional hemisphere (Fig. 1b, c: blue arrow). The stimulation trains employed a theta burst paradigm with five 1 kHz pulses within each burst³⁵. Each stimulation block lasted 10 minutes, with recordings of spontaneous activity in between the blocks to track changes in neurophysiology as stimulation continued (Fig. 1d).

Fig. 1. Schematics of experimental procedures for stroke induction and stimulation.

a Illustration of the method to induce photothrombotic lesion in the NHP cortex. b Cortical view of the multi-model artificial dura used to record ECoG. Yellow circle indicates the area illuminated with light source to induce stroke. Blue arrow points to the stimulation electrode. c Illustration of the location of induced lesion and electrical stimulation. d Experimental timeline for the control (monkey D, E) and stimulated groups (monkey F, G). Scale bar indicates 30-min block length. Portions of this figure are reprinted with permission from Zhou, J. et al. (IEEE, 2022).

Reduction in ischemic lesion volume with acute cortical stimulation

To investigate the neuroprotective effects of electrical stimulation following acute ischemic stroke, we first quantified the extent of ischemic injury within the sensorimotor cortex through histological analysis. Specifically, we performed Nissl staining on mounted coronal sections of the brain to evaluate the amount of cell death and estimate lesion volumes in both control (monkey D, E) and stimulated animals (monkey F, G). The loss of Nissl substance at the infarct core led to distinct pale areas and well-defined boundaries on the stained sections (Fig. 2a). Using this identified lesion boundary, we applied edge detection and linear interpolation between sections to reconstruct the brain with the ischemic lesion in 3D space (Fig. 2b), including distinct anatomical features of the sensorimotor cortex such as the central sulcus. This reconstruction was used to estimate the lesion volumes in each animal (Fig. 2c). We found that in control monkeys D and E which did not receive stimulation post-stroke, the estimated lesion volumes were 35.3 and 28.4 mm³ respectively, while in stimulated monkeys F and G, the lesion volumes were 20.3 and 15.9 mm³ respectively. Notably, the ischemic lesions in the stimulated animals were consistently smaller than those in the controls, both in depth and medial-lateral width (Fig. 2c, d). As the stimulation pulses were delivered medially from the infarct core on the medial-lateral axis (Fig. 2b: blue arrow), these results suggest acute electrical stimulation post stroke may have reduced the infarct size by protecting the brain from ischemic injury progression during the 4 hours after lesioning.

Fig. 2. Histology analysis of ischemic damage and lesion size.

a One representative Nissl-stained coronal section is shown for each of the 4 monkeys. Lesion boundaries are marked by dashed lines. Medial and lateral regions outside of the boundary are considered perilesional (proximal to lesion) for all subsequent analysis. In each animal, consecutive coronal sections spanning the lesioned sensorimotor cortex were stained at 0.45 mm intervals. Although no statistics were performed on these images, lesion sizes appeared similar across both animals in the control and stimulation groups. b 3D reconstruction of the cortex and lesion (red) for monkey F. Blue arrow points to the location of stimulation electrode. c Estimated lesion volume comparison projected along the dorsal-ventral axis, onto the anterior-posterior and medial-lateral (ML) axes. Color scale represents lesion depth from the cortical surface. Scale bar indicates 2 mm length on the x and y axes. d Slice by slice comparison of lesion widths and depths for the control (monkeys D and E) and stimulated group (monkeys F and G). Source data are provided as a Source Data file. Portions of this figure are reprinted with permission from Zhou, J. et al. (IEEE, 2022).

Decrease of post-stroke neural activity by acute electrical stimulation

To further understand the physiological mechanisms underlying the reduced infarct volume as a result of acute electrical stimulation, we incorporated the histological results with the analysis of electrophysiology recordings by identifying the electrodes that overlapped with the reconstructed lesion. We registered the ECoG electrodes found to be within the anatomically defined lesion by overlaying the surgical image and the reconstructed cortex based on the location of and distance between sulci (Fig. 3a). After classifying the ipsilesional electrodes in each animal as either lesion or non-lesion, we computed the gamma band (30-59 Hz) signal power at each electrode, since gamma rhythms in the brain have been found to be correlated with neural activity levels and firing rates^36,37. By visualizing the gamma power changes on heatmaps (Fig. 3b), we found that in both control and stimulated monkeys, the lesion electrodes showed decreased gamma power as early as 10 minutes after ischemic lesioning, which persisted throughout the 3-hour recordings. This observation confirmed the location of ischemic injury and neuronal death caused by photothrombosis. In addition, we observed a gradual but large-scale downregulation of gamma power across the entire ipsilesional sensorimotor region in response to post-stroke stimulation for monkeys in the stimulation group (Fig. 3b and Supplementary Fig. 1, bottom). This was distinctly different from the ECoG activity in control monkeys, as gamma power at some of the non-lesion electrodes was elevated above the baseline at around 90 minutes post lesioning (Fig. 3b and Supplementary Fig. 1, top), suggesting hyperactivation of perilesional areas as a result of focal ischemia. Similar trends can also be seen from the bar plots for each of the control and stimulated monkeys. Here, gamma power at non-lesion electrodes increased from 50 minutes to 90 minutes post-stroke in both monkeys D and E (Fig. 3c). This increase was not observed in lower frequency ranges, such as theta band power (Supplementary Fig. 2). Meanwhile, for both stimulated monkeys, the perilesional gamma power at 70 to 110 minutes post stroke was significantly lower than their respective pre-stroke baseline (Fig. 3c, one-way ANOVA and Bonferroni corrections for multiple comparisons). Together, these results showed that gamma power was mildly suppressed across the injured sensorimotor cortex as a result of continuous electrical stimulation, suggesting that stimulation delivered from 60 minutes after stroke may prevent excessive neuronal activation and energy depletion caused by the ischemic cascade, thus slowing the lesion progression and reducing neuronal death.

Fig. 3. Electrophysiological analysis of the ipsilesional cortex after stroke and stimulation.

a Electrode registration within the histologically defined lesion using 3D reconstruction and surgery photos of monkey G. Yellow circles indicate lesion electrodes. b Heatmaps of the changes in gamma ECoG power for one control (monkey E) and one stimulated animal (monkey G). Yellow circles indicate lesion electrodes. Blue circles indicate stimulation electrode. c Summary and statistics for gamma ECoG power of non-lesion electrodes within each individual animal at different time points post stroke (p.s.). Sample size is determined by the number of perilesional electrodes in each monkey (Monkey D: n = 23, Monkey E: n = 22, Monkey F: n = 15, Monkey G: n = 27). Gamma power increased significantly from baseline during sham stimulation periods in both control monkeys, but decreased continuously in both stimulated monkeys. (One-way ANOVA and Bonferroni corrections for multiple comparisons. ***P < 0.001, **P < 0.01, *P < 0.05). P-values for each time points in Monkeys D, E, F and G are, baseline-50min: 0.58, 0.45, 0.17, 0.088; baseline-70min: 0.94, 0.84, 0.014, 0.043; baseline-90min: 0.0031, <0.00001, 0.0002, <0.00001; baseline-110min: 0.0001, 0.98, <0.00001, <0.00001; 50–70 min: 0.17, 0.97, 0.85, 0.99; 50–90 min: <0.00001, <0.00001, 0.13, 0.20; 50–110 min: <0.00001, 0.18, 0.0091, 0.0007; 70–90 min: 0.034, <0.00001, 0.64, 0.33; 70–110 min: 0.0022, 0.53, 0.13, 0.0021; 90–110 min: 0.90, <0.00001, 0.84, 0.37. Error bars represent the mean ± standard error (SE). Source data are provided as a Source Data file.

Decrease of post-ischemia c-Fos activity due to acute electrical stimulation

To investigate the downregulation of neural activity and its potential neuroprotective effects, we went back to histological tools and performed immunohistochemistry staining of c-Fos, a common cellular marker of neuronal activity. c-Fos is a protein derived from the immediate early gene c-fos, which gets transiently expressed in neurons quickly following depolarization in the cerebral cortex, and can be detected reliably using immunohistochemistry up to many hours after intense neuronal activation in primates^38,39. Meanwhile, it was also widely reported that c-Fos expression can be temporarily elevated following focal ischemic stroke as part of the injury response^40,41. Therefore, we applied antibodies against c-Fos protein on the ipsilesional cortex of control (monkey C, D) and stimulated animals (monkey F, G). Monkey E was excluded from all immunostaining procedures and subsequent analyses due to consistently poor immunoreactivity across the sensorimotor cortex, regardless of the antibody tested. These sections were also co-stained with antibodies against the neuronal nuclear marker NeuN to establish precise lesion boundaries and identify neurons for the subsequent analysis. For each animal, we stained 4 coronal sections proximal to the lesion and another 4 sections at areas distal to the lesion to compare perilesional c-Fos to that of a remote region on the same hemisphere. For each proximal section, two images were taken at the medial and lateral sides of the ischemic core (Fig. 4a, pink squares). And for each distal section, two images were taken at similar cortical depth (Fig. 4a, blue squares). We saw that in control monkeys, there were high levels of c-Fos expression immediately adjacent to the lesion boundary (Fig. 4b and Supplementary Fig. 3, top left). By contrast, we saw very few neurons expressing c-Fos in stimulated monkeys proximal to the lesion (Fig. 4b and Supplementary Fig. 3: bottom left). Both control and stimulated monkeys had low levels of baseline c-Fos expression at distal regions (Fig. 4b and Supplementary Fig. 3, right). After calculating the percentage of NeuN-positive cells co-expressing c-Fos (Fig. 4b, white), we observed that control monkeys C and D showed significantly higher c-Fos expression in proximal regions compared to distal regions. However, this elevated c-Fos expression was not observed in the stimulated monkeys F and G (Fig. 4c, two-sample unpaired t-test). When combining results across animals, we found that the stimulated group exhibited significantly fewer c-Fos-positive neurons at the lesion boundary compared to the control group (Fig. 4d). These results suggest that electrical stimulation applied acutely after stroke reduced the intensity of neuronal depolarization and potentially prevented further cell death caused by overactivation of cortical neurons and excitotoxicity surrounding the ischemic core.

Fig. 4. Immunohistochemical analysis of c-Fos protein to quantify neuronal activation.

a Schematics of a coronal section illustrating confocal imaging locations at the proximal region in pink and distal region in blue. b Representative images of c-Fos and NeuN co-staining from monkeys D and F at the proximal and distal regions respectively. c Within-subject comparison of the percentage of neurons co-expressing c-Fos at the proximal versus distal regions. Significance was determined using a two-sided, two-sample unpaired t-test (***P < 0.001, n = 8 ROIs per region). P-values for Monkey C < 0.00001, Monkey D < 0.00001, Monkey F = 0.20, and Monkey G = 0.073. d Between-subject comparison of c-Fos expression in neurons proximal to the lesion, for the control (monkeys C, D) and stimulation group (monkeys F, G). Significance was determined with a two-sided, two-sample unpaired t-test (P < 0.00001, n = 16 ROIs per group). Error bars represent mean ± SE. Source data are provided as a Source Data file.

Alleviating neuroinflammatory response through acute electrical stimulation

To understand the role of electrical stimulation in reducing excitotoxicity and neuronal death caused by acute ischemic lesion, we evaluated the neuroinflammatory response surrounding the ischemic core by quantifying microglia activation and accumulation in the peri-infarct region. Glial cells including microglia are known to play a significant role in the ischemic cascade and glutamate excitotoxicity⁴². Therefore, we applied immunostaining against ionized calcium-binding adapter molecule 1 (Iba1), a widely used intracellular microglial marker⁴³, on ipsilesional sections of the sensorimotor cortex. Again, these sections were co-stained for NeuN to define the lesion boundaries and select regions of interest (ROIs) for all subsequent image analysis. For monkeys C, D, F, and G respectively, we stained four coronal sections per animal with two proximal and another two distal to the lesion. For the proximal regions, we took four ROIs per section at the medial and lateral sides of the ischemic core (Fig. 5a). For the distal regions, we randomly sampled four ROIs per section at similar cortical depths. We observed stronger expression of Iba1 and greater number of Iba1-positive cells in proximal ROIs of the control animals (Fig. 5b, Supplementary Fig. 4a). Notably, we saw a greater percentage of perilesional microglia with activated morphology in the controls, as characterized by larger cell body area and reduced ramification (Fig. 5b: white arrows). After comparing the binarized ROIs from the proximal versus distal regions in each animal, we found significantly greater level of microglia activation at the lesion boundary only in control animals, reflected by greater percentage coverage of Iba1 staining (Fig. 5c: Monkeys C and D, two-sample unpaired t-test, P < 0.001). When combining the results across animals, we found that the stimulated group had a significantly lower level of microglia activation compared to the control group at proximal, perilesional regions (Fig. 5d). Through cell segmentation, we also found that microglia density was higher at proximal regions only in control monkeys (Supplementary Fig. 4b), and the combined proximal microglia density was higher for the control group than for the stimulation group (Supplementary Fig. 4c).

Fig. 5. Immunohistochemical analysis of Iba1 protein to quantify microglia activation.

a Confocal image of a representative perilesional section from monkey D stained for Iba1 (red) and NeuN (cyan). Dashed line represents the lesion boundary. White squares represent two sample ROIs used for subsequent analysis. For each monkey, 2 sections per region (proximal and distal) were stained and 4 ROIs were imaged per section (n = 8 ROIs per region). Proximal and distal regions were compared within individual animals. This experiment was independently repeated in 2 animals per group, with similar results observed across both monkeys in each group. b Sample ROIs showing Iba1 immunoreactivity from one control and one stimulated monkey at the proximal and distal regions with respect to the lesion. White arrows point to examples of morphologically activated microglia. c Within-subject analysis of microglia activation as measured by the percentage coverage of Iba1-positive pixels, at proximal versus distal regions for each animal. Significance was determined using a two-sided, two-sample unpaired t-test (***P < 0.001, n = 8 ROIs per region). P-values for Monkey C < 0.00001, Monkey D = 0.00012, Monkey F = 0.26, Monkey G = 0.18. d Between-subject comparison of perilesional microglia activation for the control (monkeys C and D) and stimulation group (monkeys F and G). Significance was determined with a two-sided, two-sample unpaired t-test (P < 0.00001, n = 16 ROIs per group). Error bars denote mean ± SE. Source data are provided as a Source Data file.

Combining these observations with the smaller lesion volumes, reduced c-Fos expression, and downregulated neural activity in animals receiving stimulation, our findings indicate that post-stroke electrical stimulation can alleviate the inflammatory response and excessive neuronal depolarization in the perilesional cortex, thus exerting neuroprotection by potentially reducing ischemia-induced excitotoxicity. Given this protective effect, cortical electrical stimulation warrants further investigation as a promising therapeutic intervention in the acute phase post-stroke to minimize the irreversible neuronal damage caused by the ischemic cascade.

Discussion

In this study, we demonstrated potential neuroprotective effects of stimulation acutely following ischemic stroke by integrating advanced electrophysiology and histology techniques with an NHP stroke model. Our results suggest that early electrical stimulation decreases the extent of neuronal cell death by reducing peri-infarct depolarization and inflammation in the sensorimotor cortex of NHPs. These findings highlight the promise of electrical stimulation as a therapeutic strategy to protect the brain and minimize tissue damage during the acute window. Such an approach could play a pivotal role in addressing the global burden of stroke, given that infarct size significantly impacts not only patient mortality but also the functional outcomes of stroke rehabilitation strategies⁴⁴.

We utilized the photothrombotic method to produce focal ischemic lesions in NHPs. In comparison to other interventions for generating infarcts⁴⁵, our method is less surgically challenging and allows for more robust control of the location and size of infarcts across animals by implementing the same aperture, intensity, and duration of light illumination as demonstrated in our previous work³². This technique enabled us to generate controlled focal lesions in the sensorimotor cortex of NHPs while simultaneously collecting ECoG recordings from the affected brain regions to monitor changes in neural activity, making it a valuable and reproducible model for studying acute stroke pathophysiology and evaluating neuroprotective interventions in a large-brain system. Importantly, cortical infarcts are commonly observed in several clinical subtypes of ischemic stroke, including cardioembolic events and large-artery atherosclerosis involving distal vessels, both of which are frequently associated with significant neurological deficits such as aphasia⁴⁶. Multiple studies report that 30 to 50% of stroke patients experience cortical damage, and that approximately 15% of the overall stroke population present with cortical-only lesions^47–49, highlighting the cortex as a critical therapeutic target for acute intervention. Moreover, even in strokes that primarily affect deeper brain regions such as the white matter, cortical function can still be impacted due to axonal injury⁵⁰, network disconnection⁵¹, secondary neuronal degeneration and apoptosis⁵². Electrophysiology studies in humans have also shown that both cortical and subcortical stroke can lead to hyperexcitability of the ipsilesional cortex during the acute window^53,54, which may contribute to spreading depolarization, injury progression, or even seizure activity⁵⁵. Therefore, by suppressing cortical overactivation early during acute ischemia, our stimulation paradigm offers a targeted strategy to limit neural damage across a broad range of stroke subtypes and establishes a foundation for future investigations in other clinical stroke cases. Beyond stroke, this novel intervention also holds promise for other neurological conditions characterized by cortical hyperexcitability and continuous depolarization, including traumatic brain injury⁵⁶, underscoring the broader therapeutic relevance and translational potential of our findings.

Mechanisms of stimulation-induced neuroprotection

An hour following stroke induction, we applied repeated electrical stimulation directly adjacent to the lesion on the ipsilesional cortex. The stimulation train contains 5 Hz bursts of biphasic pulses, similar to the theta burst stimulation (TBS) pattern of transcranial magnetic stimulation (TMS) protocols that are widely adopted in the clinic⁵⁷. However, one major difference is that our stimulation paradigm used five pulses at 1 kHz in contrast to the three 50–100 Hz pulses per burst used in traditional TBS protocol. Stimulation via high frequency pulses at greater than around 200 Hz has been shown to have an inhibitory effect on neuronal firing rates^58,59. While there are many possible mechanisms of action underlying this inhibitory effect, it is likely that these high frequency pulses contribute to recurrent hyperpolarization⁶⁰, or generate a transient reduction of excitatory currents in the stimulated neurons, thereby decreasing the excitability and firing rates of postsynaptic cells⁵⁹. Given that our ECoG recordings showed decreasing gamma power during and after stimulation, and that gamma activity in ECoG has been shown to correlate with neuronal firing³⁶, it is likely that applying theta bursts of electrical stimulation on the cortex suppressed peri-infarct neural activity and network excitability through similar mechanisms. It is important to note that high-frequency band power reflects synchronized firing of neuronal populations and has been associated with both synaptic activity⁶¹ and action potential generation⁶². Therefore, the observed reduction in gamma power likely reflects a global suppression of neuronal output, encompassing both somatic and axonal effects. Moreover, postsynaptic excitability has been shown to play a major role in excitotoxic cell damage after ischemia^63,64, potentially explaining the protective effect of our inhibitory stimulation paradigm.

Importantly, the observed decrease in c-Fos immunoreactivity in stimulated monkeys suggests that the downregulation of ECoG activity is not a manifestation of the harmful cortical spreading depression, given that an increase in c-Fos expression has been shown to correlate with sustained depolarizations and the subsequent spreading depression in the presence of focal ischemia⁶⁵. Furthermore, both c-fos mRNA and c-Fos protein expression have been reported as reliable markers for glutamate toxicity, correlating with excitotoxic cell death in cultured neurons^66,67. These works and the lack of c-Fos positive neurons we found in stimulated animals support our hypothesis that electrical stimulation reduces peri-infarct excitotoxicity, likely through reducing neuronal hyperexcitability. Combined with the reduction in lesion size and the electrophysiology results discussed above, our findings provide evidence that electrical stimulation applied early after stroke onset produced an inhibitory and neuroprotective effect, without exacerbating tissue damage attributed to spreading depolarization as previously concerned²³, making this stimulation protocol a safe and promising treatment option for acute ischemic stroke.

To better understand the cellular response to electrical stimulation, we analyzed Iba1 immunoreactivity to assess microglia-mediated neuroinflammation in the perilesional tissues. Since the microglial response during acute stroke involves both cell migration to the infarct and morphological activation through cell body enlargement and process thickening⁶⁸, we used two common measures to quantify inflammation: percent area of Iba1 immunoreactivity and Iba1-positive cell density, accounting for both microglia accumulation and activation within a given region⁶⁹. Based on these measures, the lack of significant microglial response near the lesion boundary in stimulated monkeys suggest that electrical stimulation was able to mitigate neuroinflammation for tissues undergoing the acute ischemic cascade. Given the important role of activated microglia in detecting and propagating the ischemia-induced excitotoxicity^42,68, the decreased microglial activation in stimulated tissues may suggest an attenuation of excitotoxic injury as a result of our acute intervention. Nevertheless, additional experiments are essential to further investigate the interplay between microglial response and other components of the ischemic cascade, such as glutamate signaling and oxidative stress following ischemia and electrical stimulation, to establish the precise mechanisms through which acute stimulation induces neuroprotection.

Together, the reduced c-Fos activity and microglial response near the lesion boundary in stimulated monkeys suggest that electrical stimulation was able to mitigate both cortical depolarization and neuroinflammation for tissues undergoing the acute ischemic cascade, leading to reduced neural damage. These findings are consistent with prior studies in rodents, where cortical stimulation via bipolar electrodes^27,30 and cathodal transcranial direct current stimulation (C-tDCS)^25,26 were shown to decrease tissue damage by inhibiting peri-infarct excitability, neuroinflammation, and apoptosis during acute stroke. Furthermore, Our findings also align with the hypothesized neuroprotective mechanisms of many pharmacological agents designed for acute ischemic stroke, which aim at attenuating excitotoxicity due to energy depletion by inhibiting neuronal excitability^5,6,63,70. Compared to pharmacological treatments, electrical stimulation may offer distinct advantages, particularly for patients who experience severe side effects or allergic reactions to these medications. Future studies could extend the post-stimulation monitoring to determine the optimal dose and duration for stimulation while minimizing potential side effects.

Comparison with alternative brain stimulation techniques

To understand the spatial distribution of our stimulation effects and compare it with other widely adopted neuromodulation techniques, we simulated the electric field in the underlying cortical tissues. The simulation results suggest that the small electrode size (0.25 mm diameter) and current amplitude (60 μA) in our protocol generated a highly localized electric field (Supplementary Fig. 5), starting at approximately 50 mV/mm directly beneath the electrode and decreasing rapidly to 1 mV/mm near the lesion boundary (~3 mm² cortical area stimulated at above ~10 mV/mm). Although this field strength and spatial distribution are substantially smaller than those typically used in TMS therapies at motor threshold (~30 cm² cortical area stimulated at ~60 to 100 mV/mm)^71,72, our perilesional stimulation is specifically designed to modulate cortical excitability without inducing widespread neuronal depolarization. Additionally, the observed cellular response also extended beyond the immediate stimulation site, affecting regions as far as the lateral side of the ischemic lesion, as suggested by the changes in ECoG activity and c-Fos expression. This broader physiological effect is likely attributed to the dense cortical connections that facilitate large-scale excitability changes. This explanation is supported by the extensive studies on intracortical microstimulation (ICMS), where electrodes delivering the stimulation current have a much smaller surface area^73–75. For instance, stimulating the thalamic region in NHPs with a microelectrode at similar current amplitudes has been shown to suppress neural activity in its visual cortex projection, likely through synaptic inhibition mechanisms⁷⁵. These studies suggest that even a highly localized current source can produce widespread neural effects, supporting the large-scale inhibitory modulation observed with our ECoG stimulation paradigm. Future studies should investigate whether this neuroprotective effect arises specifically from localized stimulation or can also be achieved using more global neuromodulation techniques, thereby facilitating the clinical translation of our findings.

While our electrical stimulation approach requires access to the cortical surface, similar neuromodulatory effects could potentially be achieved noninvasively using existing technologies. For example, a range of TMS paradigms, including the intermittent theta-burst stimulation (iTBS) protocol recently approved by the FDA⁷⁶, have demonstrated clinical efficacy in treating neurological disorders affecting various brain regions. These findings highlight the potential to develop new TMS paradigms that replicate the electric field distribution and stimulation-induced effects in our current study, paving the way for next-generation therapies targeting acute stroke. On the other hand, existing TMS protocols with reported inhibitory effects can be reevaluated in acute stroke settings to test whether the neuroprotective outcomes are reproduced. For instance, continuous TBS (cTBS) has been shown to reduce cortical excitability in humans by inducing synaptic depression^77–79. Recent clinical trials also showed that suppressing the contralesional neural activity with cTBS improved motor recovery in chronic stroke patients⁸⁰, highlighting its safety and relevance as a therapeutic option for acute stroke. In addition to various TMS paradigms, tDCS also represents a promising noninvasive alternative, which generates sub-threshold cortical electric fields (~ 1 mV/mm) while retaining the ability to modulate neuronal spiking over large cortical areas⁸¹. Various tDCS protocols have already been tested in rodents for acute neuroprotection²⁵ and in humans for stroke rehabilitation⁸², demonstrating their strong translational potential for use in acute settings to mitigate inflammation and protect the brain following ischemia.

Limitations

For lesion volume estimation and electrophysiology analysis, we compared results from control monkeys D and E to stimulated monkeys F and G, all of which were lesioned using identical illumination parameters known to produce predictable infarct sizes³². However, for immunohistochemistry (IHC) analysis of biomarkers such as NeuN, control monkey E exhibited poor immunofluorescence signals over the sensorimotor cortex. Despite this, Nissl staining revealed consistent cell density and a clear lesion boundary, suggesting that the lack of immunoreactivity was likely due to perfusion fixation issues or subsequent tissue handling steps prior to immunostaining. For example, it has been shown that perfusing the injured brain with fixatives could lead to incomplete and asymmetrical fixation, thus decreasing the immunoreactivity in affected tissues⁸³. Unlike NeuN staining, Nissl staining is less susceptible to such perfusion issues, as the dye binds directly to nucleic acids within cells, making it more robust across different fixation and processing conditions^84,85. To address this staining challenge, we added monkey C as an additional control and excluded monkey E from all IHC analysis.

It should also be noted that photothrombotic infarcts do not capture all aspects of acute stroke in humans. For example, this technique generates narrow penumbras and is largely constrained to the cortical layers, thereby excluding stroke subtypes that primarily involve subcortical or white matter lesions⁸⁶. Nevertheless, for stroke patients with infarcts that extend into the white matter, suppressing cortical depolarization with our stimulation paradigm could still mitigate the ischemia-induced ionic imbalance and high extracellular glutamate, potentially preventing them from propagating into the affected white matter and thus reducing the excitotoxic damage of oligodendrocytes and myelin⁸⁷. Meanwhile, in cases involving deep white matter stroke, noninvasive methods like TMS may be preferable due to their ability to generate electric fields that reach greater depths and volumes⁸⁸, potentially preventing oligodendrocyte excitotoxicity and subsequent cell death that are central to ischemic white matter injury⁸⁹. Future research should explore how stimulation-induced effects vary with infarct size and location, focusing on optimizing parameters such as stimulation strength and duration to improve outcomes for larger and deeper infarcts. Such efforts will validate the generalizability of our findings and tailor interventions to patients with varying injury severities.

While our study was limited by a small number of monkeys in both the control and stimulation groups, it is important to recognize the inherent challenges associated with conducting stroke research involving NHPs, particularly those suitable for combined electrophysiology and histology analysis. These NHPs are highly valuable but scarce resources due to ethical and logistical considerations, thus constraining our sample size. Despite this limitation, the unique technology and promising results from this study not only provide valuable insights into the effects of acute stimulation on ischemic damage, but also open many exciting opportunities for future investigation regarding acute stroke intervention in large-brain models. In particular, applying the lesioning toolbox and electrical stimulation to monkeys chronically implanted with our multi-modal interface⁹⁰ can allow us to study the impact of stimulation well beyond the acute window. Combined with the sensorimotor behavioral capabilities of NHPs, these experiments will deepen our understanding of the mechanisms underlying stimulation-induced neuroprotection and its effect on functional recovery from days to months after stroke onset. More importantly, investigating the chronic effect of localized electrical stimulation using a reproducible lesioning toolbox in NHPs, which closely resemble human brain anatomy and immune response, can help bridge the gaps between rodent and human studies that hindered the success of past clinical trials on neuroprotective therapy^9,10,31.

Methods

Animals subjects and surgical procedures

All animal procedures were approved by the University of Washington Institutional Animal Care and Use Committee and were in accordance with the National Research Council’s Guide for the Care and Use of Laboratory Animals. All our animals were housed and maintained at the Washington National Primate Research Center (WaNPRC) accredited by the American Association for Assessment of Laboratory Animal Care (AAALAC). This study used five adult macaques (Control group: monkey C - Macaca mulatta, 10.3 kg, 16 years, female; monkey D - Macaca nemestrina, 12.8 kg, 14 years, female; monkey E - Macaca nemestrina, 13.10 kg, 14 years, female; Stimulation group: monkey F - Macaca nemestrina, 13.8 kg, 14 years, female; monkey G - Macaca nemestrina, 14.6 kg, 7 years, male) that were part of the Tissue Distribution Program (TDP) at the WaNPRC, which aims to conserve and fully utilize the NHPs no longer needed for other experiments.

Surgical procedures and induction of focal ischemic lesions

Using standard aseptic technique, the five macaques (monkeys C, D, E, F, G) were anesthetized with isoflurane and placed in a stereotaxic frame. The animals’ temperature, oxygen saturation, heart rate, and electrocardiographic responses were monitored throughout the procedure. Bilateral craniotomies and durotomies (25 mm diameter) were performed using stereotaxic coordinates that target the sensorimotor cortices⁹¹. To protect the exposed brain, a transparent silicone artificial dura (monkey C) or a semi-transparent multi-modal artificial dura (monkey D to G) fabricated using previously described methods^90,92 was implanted bilaterally on top of the sensorimotor cortex, offering optical and electrical access during subsequent procedures. For animals in both the control (monkeys C, D, E) and stimulation groups (monkeys F, G), we induced ischemic lesions using the photothrombotic technique³², which produces focal infarcts by photo-activation of a light-sensitive dye (Rose Bengal). Upon illumination, the intravenously administered dye produced singlet oxygen that damaged endothelial cell membranes, causing platelet aggregation and interrupting local blood flow. To control the infarct size and location, we placed an opaque silicone mask on top of the artificial dura. In monkey C, the mask contains circular apertures of different diameters (0.5, 1.0, and 2.0 mm) to induce control lesions in various sizes. In monkey D to G, the mask has a single aperture located in the center (diameter 1.5 mm) and was placed on one hemisphere (monkeys D, F, and G: left hemisphere; monkey E: right hemisphere). After the mask was in place, each animal was injected with Rose Bengal (20 mg/kg) for 5 minutes as we started illuminating the ipsilesional cranial window for 30 minutes through the aperture using an uncollimated white light source.

Histology, lesion reconstruction and size estimation

At around 4 hours after the stroke was induced, animals were deeply sedated and transcardially perfused with phosphate buffered saline (PBS) followed by 4% paraformaldehyde (PFA). The brains were harvested and post-fixed by immersion in 4% PFA for 24 to 48 hours. A coronal block containing the lesioned region was dissected using a custom matrix and then stored at 4 °C in 30% sucrose in PBS. To prepare for staining, the block was frozen and sectioned into 50 μm thick coronal sections using a sliding microtome (Leica). Sliced sections were stored at 4 °C in PBS with 0.02% sodium azide. To evaluate the extent of ischemic damage and neuronal death, we mounted a rostrocaudal series of 50 μm coronal sections with approximately 0.45 mm separation and then performed Nissl staining on the mounted sections using Thionin acetate. Nissl-stained sections were scanned and registered using a custom script in MATLAB (2019b, MathWorks) for alignment and three-dimensional reconstruction. The registered images were smoothed and binarized so that lesion boundaries on each slice can be identified and visualized within the reconstructed cortex. The widths and depths of each lesion within the coronal sections were calculated based on the detected boundary and image resolution. Linear interpolation was used between sections to estimate the lesion volume in each animal.

To combine histological information on the ischemic injury with electrophysiology recordings, we overlayed the surgical image taken for each animal on top of its reconstructed cortex based on the location of sulci and other distinct anatomical features. The overlayed images allowed us to register the ECoG electrodes that fell within the estimated lesion area and classify them into lesion and non-lesion groups respectively for the subsequent electrophysiology analysis.

Electrophysiology recording and electrical stimulation

All electrophysiology recordings and electrical stimulation were performed with Grapevine Nomad processors, four Nano front ends (Ripple Neuro, Salt Lake City, UT), and our customized large-scale multi-modal interface. The design and characterization of this interface with 32 embedded ECoG electrodes can be found in our previous work^90,92. A skull screw close to the midline and anterior to the ipsilesional cranial window was used as an electrical ground for subsequent recordings. All animals were transitioned from isoflurane to urethane anesthesia prior to ECoG recordings and stayed on urethane until the end of the experiment to allow reliable monitoring of neural activity. In control monkeys D and E, we collected ECoG data bilaterally at 30 kHz sampling frequency, including 30 minutes of baseline before photothrombotic lesioning, 30 minutes during illumination, and up to 3 hours post lesioning to monitor the extent of neuronal damage and network dynamics around the injury site. In stimulated monkeys F and G, we followed the same recording timeline for baseline and illumination periods and recorded spontaneous neural activity for 1 hour after lesioning. We then applied electrical stimulation through a single ECoG electrode (0.25 mm diameter) approximately 8 mm medial to the lesion center on the ipsilesional (left) hemisphere. We delivered the stimulation trains in 6 blocks lasting 10 minutes each, with 2-minute recordings of spontaneous activity in between the blocks to track changes in neurophysiology as stimulation continued. The stimulation trains had a 5 Hz burst frequency and 5 biphasic charge-balanced pulses at 1 kHz within each burst. The stimulation amplitude was 60 μA and pulse width was 200 μs per phase with 50 μs inter-phase interval.

Electrophysiology data analysis

ECoG signal power calculation was conducted in MATLAB (R2022b, MathWorks). After downsampling the signal from 30 kHz to 1 kHz, the signals were notch filtered at 60, 120, and 180 Hz. Channels with a power spectral density that did not exhibit the expected 1/f curve were excluded from further analysis (1 channel was excluded for monkeys D and G, 2 channels for monkey E, and 14 channels for monkey F). We then filtered the downsampled signals into different frequency bands including theta (4–8 Hz) and gamma bands (30–59 Hz). We split the filtered signals into two-minute blocks separated by ten minutes across the pre-stroke baseline, post-stroke, and post-stimulation recording periods. We calculated the average signal power of each channel for each two-minute window by squaring the signal and dividing by the elapsed time. To compare power across each two-minute time window, we conducted one-way ANOVA test with Bonferroni corrections for multiple comparisons (family-wise error rate of 0.05). Pairwise comparisons of power distributions were made between baseline and each subsequent two-minute window and between 50 min post-stroke and each window following stimulation.

Immunohistochemistry

To evaluate the neuronal activation and neuroinflammatory response near the ischemic lesion, we performed immunostaining with antibodies against c-Fos protein and the microglia/macrophage-specific calcium-binding protein Iba1 respectively. Coronal sections around the lesion in monkeys C, D, F, and G were co-stained with either neuronal nuclear protein NeuN and c-Fos or NeuN and Iba1. Monkey E was excluded for all immunohistochemistry analysis due to lack of immunoreactivity at the cranial window. To replace monkey E, we used tissues from monkey C as an additional control. For this animal, we specifically selected coronal sections containing a single lesion comparable in size to those observed in monkeys D to G. To prepare tissues for staining, coronal sections were first rinsed in 1x PBS and incubated in 1% NaBH₄ solution for 1 hour to reduce background autofluorescence, after which they were washed with PBS and incubated in normal donkey serum (NDS) blocking solution (10% NDS and 0.1% triton-X100 in PBS) overnight at 4 °C. Sections were then incubated in primary antibodies including either rabbit anti-c-Fos (Abcam ab190289, 1:500 dilution) or goat anti-Iba1 (Abcam ab5076, 1:1000 dilution) plus mouse anti-NeuN (Millipore Sigma MAB377, 1:500 dilution) in NDS blocking solution at 4 °C for ~72 hours. Sections were then rinsed in PBS and incubated in secondary antibody mixtures containing either donkey anti-rabbit antibody (Invitrogen #A10042, 1:500 dilution) or donkey anti-goat antibody (Invitrogen #A-11057, 1:500 dilution) conjugated with Alexa Fluor 568, plus donkey anti-mouse antibody conjugated with Alexa Fluor 488 (Invitrogen # A-21202, 1:500 dilution) and DAPI at 4 °C overnight. This was followed by rinsing sections in PBS for 5 times and incubating in 1:1 Glycerol-PBS solution for 10 minutes. Sections were mounted onto slides using DABCO mounting media and later imaged using a Nikon A1R HD25 laser scanning confocal microscope and Plan Apo 20x objective, acquired at a resolution of 1024 × 1024 pixels.

c-Fos and microglia activation analysis

To estimate the level of c-Fos expression within neurons, we performed NeuN and c-Fos co-staining on four proximal tissue sections per animal, collected from regions adjacent to the lesion. For each of these proximal sections, we acquired two confocal images (Nikon A1R, 20x objective) on either side of the lesion boundary. Additionally, we stained another four sections distal to the lesion (at least 20 mm away) within the same hemisphere and randomly sampled two regions of interest (ROIs) per section. Custom algorithms in MATLAB were then used for image analysis, including binarization with adaptive thresholding and image segmentation, applied separately to the NeuN channel (cyan) and c-Fos channel (red). Neuronal c-Fos expression in each ROI was quantified by calculating the percentage of NeuN-positive cells that co-expressed the c-Fos protein.

To evaluate the level of microglial reactivity, we stained 2 sections per animal proximal to the lesion. For each section, we imaged 4 ROIs within 1 mm from the lesion boundary using a 20x objective through the Nikon A1R laser scanning confocal microscope. For each ROI we took z-stack images with 20 μm thickness and created a maximum intensity projection. The projections were converted to 8-bit and inverted so that dark pixels represent Iba1 signals. We then used ImageJ to denoise each ROI by subtracting the background and removing pixel intensity outliers. The ROIs were converted to binary by setting the threshold to one standard deviation below the average pixel intensity and filtering out objects below 150 pixels in size. The percentage area covered by dark pixels, a widely used Iba1 reactivity measure^69,93, was then calculated to estimate the microglia activation level. To quantify microglial density, we used the same binarized images and counted the number of segmented cells in each ROI to compute the average microglia density at the lesion boundary for each animal. Similar imaging and quantification steps were performed for regions distal to the lesion on the same hemisphere of each animal (4 ROIs per section) to compare microglial density and activation within each of the control and stimulated monkeys.

Statistics and reproducibility

A total of five monkeys were used in this study, with three animals assigned to the control group (Monkeys C, D, E) and two to the stimulation group (Monkeys F and G). Given the limited number of animals and the challenge of NHP research, statistical tests were performed within each individual animal. Because of this, the experiments were not randomized, and the investigators were not blinded to group allocation during data collection and analysis. For Nissl-stained tissues, lesion volume estimates were derived from serial coronal slices and their 3D reconstructions. No statistical comparisons were conducted on the Nissl-stained sections or lesion volumes. For immunohistochemistry analysis, proximal and distal regions were compared within each monkey using n = 8 ROIs per region and two-sided unpaired t-test. Similar results were independently observed across both animals in the control and stimulation groups. The only histological data excluded was the immunohistochemistry sections from Monkey E as discussed above. For neurophysiology data, electrodes with abnormal spectral characteristics were excluded from the power analysis. Perilesional ECoG signal power was compared across different time points after stroke using one-way ANOVA with Bonferroni correction for multiple comparisons, with sample size determined by the number of non-lesion electrodes. All electrophysiological and histological findings were independently reproduced in both animals from the control and stimulation groups.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

J.Z., K.K., and A.Y. designed the research and experiments in this manuscript; J.Z., K.K., and A.Y. performed the experiments and collected the data. J.Z. analyzed the data. J.Z., K.K., and A.Y. wrote and revised the manuscript.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

All data supporting the findings of this study are included within the article and the Supplementary Information. Source data are provided with this paper.

Code availability

All custom MATLAB codes and scripts for electrophysiology signal processing, immunohistochemistry image analysis, and figure generation are available at: https://github.com/jzhou33/stroke_estim.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-025-61948-y.