Distinct SOZ-associated waveform biomarkers and stimulation response profiles
DOI: Pending DOI
Synopsis
Drug-resistant epilepsy patients undergo neurosurgical implantation of stereotactic electrodes to localize seizure-generating brain regions, also called the seizure onset zone (SOZ). However, the SOZ is elusive, as evidenced by inconsistent outcomes from its surgical removal. Single-neuron recordings have proved useful in developing a more precise understanding of where seizures originate, which may lead to more efficacious treatments for epilepsy. In our recently published work, we examined cell-type-specific neuronal responses to low-frequency single-pulse electrical stimulation. We found exciting preliminary evidence that action potential waveforms within the clinically determined SOZ are distinctly wider and have distinct firing patterns in response to stimulation. Here, we further substantiate the persistent SOZ-associated action potential waveform changes in an expanded human single-neuron dataset.
Background
Neurostimulation is an increasingly common component of surgical epilepsy treatment that can be used to probe cortical excitability,1–3 assess functional connectivity,4–6 map epileptic networks,7,8 and define the eloquent cortex and irritative zones.9–11 These evaluative techniques rely on observing mesoscale signals such as cortico-cortical-evoked potentials.12–14 Despite ubiquitous probing and recording, the efficacy of surgical outcomes remains stable. While passive single-neuron recordings of epileptiform activity have enriched our understanding of seizure-generating mechanisms,15,16 we recently discovered tantalizing correlates of single-neuron responses to single-pulse electrical stimulation and examined neurons recorded within the SOZ.17 Actively probing the SOZ revealed distinct response properties, with SOZ neurons exhibiting shallower suppression than non-SOZ neurons.
Furthermore, by classifying putative cell types, we can better specify the dynamic activity of inhibitory and excitatory neurons with spatiotemporal precision during a seizure’s evolution.18,19 Recent work has shown that as seizures develop, fast-spiking interneurons lag behind principal cells, signifying the failure of feedforward inhibition.20 Other work has used single-neuron activity to help define seizure recruitment, positing a ‘seizure-responsive’ state in mesial temporal structures, which may refine surgical outcomes.21 In our recent paper, we implemented cell-type classification in human microelectrode recordings from cortical, subcortical, and clinically-defined SOZ regions.17 Importantly, we observed significantly wider SOZ waveforms compared to non-SOZ waveforms. Taking advantage of the fine-grained resolution of microelectrode recordings allowed us to observe a potential biomarker of epileptogenic tissue.
Methods
We conducted stimulation on implanted macroelectrode contacts while recording from nearby microelectrodes. Electrical stimulation was delivered in a monopolar configuration using single biphasic pulses (cathodic followed by anodic) at intervals of 3–3.5s. Each pulse had an amplitude of 3 mA, a phase duration of 200 µs, and a 100-µs inter-phase interval.4,6 Each contact was stimulated approximately ten times, approximately 800 trials per patient, taking around an hour to complete. The protocol has multiple constraints to prevent excessive stimulation of any single cortical region to reduce the likelihood of overlapping electrical fields and limit potential habituation effects. Intracranial brain activity was recorded using a 256-channel data acquisition system that allows us to record up to 192 macroelectrode channels and four microelectrode bundles. Stimulation was delivered using a 96-channel stimulation. Electrophysiological data from microwires was sampled at 30 kHz with a white matter depth electrode contact as a reference. Single-neuron recordings were preprocessed using the spike-sorting software Offline Sorter.
Waveforms were classified into putative cell types using shape metrics (i.e., trough-to-peak). Firing rates of were calculated using MATLAB psth functionality and were z-scored. To examine single-unit responses to stimulation, we characterized numerous features of stimulation suppression including baseline frequency, suppression amplitude, and suppression duration. We also sought to understand the effect of stimulation on firing from various distances. We calculated the Euclidean distance at which firing rate responses transitioned from evoked to suppressed, by fitting sigmoidal curves to mean firing rates. This procedure yielded a model of firing rate versus stimulation distance.
Results
Intracranial electrical stimulation was applied to a total of 4,329 SEEG contacts in 30 participants (16 female, 36 ± 10.4 years of age). In total, 228 isolated units were identified in recordings from 76 microwires (2.5 ± 0.6 microwires per patient). We classified 191 (83.8%) as putative principal cells (PCs) and 37 (16.2%) as putative interneurons (INs) and found that INs had significantly faster TTPs (z = 9.6, p < .001) and FWHM (z = 4.6, p < .001), compared to PCs.
Overall, we found that stimulation caused widespread suppression in neuronal firing rates with 174 (76.3%) units significantly modulated by stimulation, 91.4% of those units exhibiting firing rate suppression post-stimulation. When comparing between putative cell-type responses, we observed that INs had higher baseline firing rates (trial-averaged firing rate during a 1-s pre-stimulation time window [1.11s–0.01s]; z = 2.8, p = .0053) and post-stimulation firing rates (trial-averaged firing rate during a 1-s post-stimulation time window [0.01s–1.11s]; z = 2.5, p = .014), but PCs had greater magnitude of suppression (lowest instantaneous trial-averaged firing rate during suppression during a 2.5-s post-stimulation time window [0.01 s–2.51 s]; z = 2.3, p = .021). Finally, between cell-type features, we found a significant positive association between suppression amplitude and duration only for PCs (t(90) = 2.6, p = .012). These findings emphasize PC-specific magnitude of suppression from stimulation.
Euclidean distance analyses revealed that stimulation from local sites can cause evoked firing at distant sites. We observed evoked firing in 40 (17.5%) units occurring approximately 200 ms after stimulation, before becoming characteristically suppressed. Interestingly, the units that exhibited evoked firing had significantly longer recovery rates back to baseline firing rates (1.13s ± 0.5), compared to units that did not exhibit this effect (0.87s ± 0.6; z = 2.04, p = .041).
27 of our 228 single-neuron recordings were in the clinically determined SOZ. These units clustered into a distinct group in action potential waveform metric space, with a significant increase in TTP duration (0.52 ± 0.15ms), compared to non-SOZ units (0.45 ± 0.12ms; z = 2.3, p = .02, Fig.1A-D). SOZ units also exhibited statistically smaller suppression thresholds (z = 3.5, p = .0004) and quicker recovery rates (time between the minimum suppression amplitude and the instantaneous firing rate crossing the upwards 2 SD threshold below the baseline frequency, time window [0.01s–2.51s]; z = 1.97, p = .049) compared to non-SOZ units, implying diminished responses to single-pulse electrical stimulation in the SOZ.
Finally, we wanted to develop our analysis to corroborate the hypothesis that SOZ neurons have tonically wider action potential waveforms. First, we examined a held-out data set of Utah Array recordings that found core SOZ waveforms to have significantly wider FWHM than penumbral waveforms (Fig.1E). Next, we looked at a resection case, where wider waveforms in the SOZ were ablated, leading to seizure reductions (Fig.1F). Finally, we analyzed a large dataset of single-neuron recordings without neurostimulation to corroborate our finding. From these recordings, we isolated 1,218 single-neurons. Our simple waveform shape cluster method classified 274 (22.5%) as INs and 994 (77.5%) as PCs. 204 (16.7%) of these units were recorded within the clinically determined SOZ (Fig.1G). The waveform shape clustering also revealed a distinct group of 81 wider waveforms. We applied the Uniform Manifold Approximation and Projection (UMAP) dimensionality reduction technique to better visualize action potential waveform clusters (Fig.1H). Tonically, wider action potential waveforms appeared as a distinct cluster on the UMAP embedding, providing intriguing evidence of global and local properties of wider waveforms. When we projected the SOZ neurons onto the embedding, it revealed that 20% of the wider waveform cluster were recorded from the SOZ.
Discussion
Neuromodulation is being increasingly used in the detection and diagnosis of SOZ for patients with drug-resistant epilepsy. The clinical relevance of this work lies in its potential to make SOZ localization more objective, precise, and biologically grounded. Rather than relying only on macroelectrode-level signals or clinician-defined boundaries, single-neuron features such as waveform shapes and distinct responses to stimulation could provide additional evidence to more precisely distinguish epileptogenic from non-epileptogenic tissue, that cannot be fully captured by standard recordings. In practice, these biomarkers may be especially valuable in patients with unclear, multifocal, or distributed seizure networks, where surgical decisions are difficult and outcomes remain especially variable. Clinically, these measures could help refine SOZ localization and improve decisions about surgical resection or targeted neurostimulation. If validated in larger cohorts, waveform features and stimulation-evoked firing patterns may complement existing SEEG methods by distinguishing epileptogenic from non-epileptogenic tissue at the cellular level.
For future research, these results support further investigation into how putative PCs and INs contribute to seizure generation and stimulation response. The finding that wider waveform neurons form a distinct cluster and are enriched in the SOZ raises important questions about whether these cells represent a specific pathological population or altered excitatory-inhibitory dynamics. Future studies should test whether these single-neuron biomarkers predict surgical outcomes, optimize stimulation parameters, and improve personalized neuromodulation strategies.
Disclosures
Nothing to disclose by any author(s).
This project was funded by the Neurtex Brain Research Institute.
References
1. Feys, O. et al. Variability of cortico-cortical evoked potentials in the epileptogenic zone is related to seizure occurrence. Ann. Clin. Transl. Neurol. 11, 2645–2656 (2024).
2. Freestone, D. R. et al. Electrical probing of cortical excitability in patients with epilepsy. Epilepsy Behav. EB 22 Suppl 1, S110-118 (2011).
3. Matsumoto, R., Kunieda, T. & Nair, D. Single pulse electrical stimulation to probe functional and pathological connectivity in epilepsy. Seizure 44, 27–36 (2017).
4. Davis, T. S., Rolston, J. D., Bollo, R. J. & House, P. A. Delayed high-frequency suppression after automated single-pulse electrical stimulation identifies the seizure onset zone in patients with refractory epilepsy. Clin. Neurophysiol. 129, 2466–2474 (2018).
5. Frauscher, B. et al. Stimulation to probe, excite, and inhibit the epileptic brain. Epilepsia 64, S49–S61 (2023).
6. Kundu, B. et al. A systematic exploration of parameters affecting evoked intracranial potentials in patients with epilepsy. Brain Stimulat. 13, 1232–1244 (2020).
7. Bernabei, J. M. et al. Quantitative approaches to guide epilepsy surgery from intracranial EEG. Brain 146, 2248–2258 (2023).
8. Mouthaan, B. E. et al. Single Pulse Electrical Stimulation to identify epileptogenic cortex: Clinical information obtained from early evoked responses. Clin. Neurophysiol. 127, 1088–1098 (2016).
9. David, O., Bastin, J., Chabardès, S., Minotti, L. & Kahane, P. Studying Network Mechanisms Using Intracranial Stimulation in Epileptic Patients. Front. Syst. Neurosci. 4, (2010).
10. Alarcon, G., Nawoor, L. & Valentin, A. Electrical Cortical Stimulation: Mapping for Function and Seizures. Neurosurg. Clin. N. Am. 31, 435–448 (2020).
11. Suller Marti, A. et al. Extraoperative electrical stimulation mapping in epilepsy presurgical evaluation: a proposal and review of the literature. Clin. Neurol. Neurosurg. 214, 107170 (2022).
12. van ’t Klooster, M. A. et al. Time-frequency analysis of single pulse electrical stimulation to assist delineation of epileptogenic cortex. Brain J. Neurol. 134, 2855–2866 (2011).
13. Mouthaan, B. E. et al. Single Pulse Electrical Stimulation to identify epileptogenic cortex: Clinical information obtained from early evoked responses. Clin. Neurophysiol. Off. J. Int. Fed. Clin. Neurophysiol. 127, 1088–1098 (2016).
14. Hays, M. A. et al. Cortico-cortical evoked potentials in response to varying stimulation intensity improves seizure localization. Clin. Neurophysiol. Off. J. Int. Fed. Clin. Neurophysiol. 145, 119–128 (2023).
15. Schevon, C. A. et al. Evidence of an inhibitory restraint of seizure activity in humans. Nat. Commun. 3, 1060 (2012).
16. Smith, E. H. et al. The ictal wavefront is the spatiotemporal source of discharges during spontaneous human seizures. Nat. Commun. 7, 11098 (2016).
17. Cowan, R. L. et al. Cell-type-specific responses to single-pulse electrical stimulation of the human brain. Brain Stimulat. 18, 1266–1278 (2025).
18. Pouille, F., Watkinson, O., Scanziani, M. & Trevelyan, A. J. The contribution of synaptic location to inhibitory gain control in pyramidal cells. Physiol. Rep. 1, e00067 (2013).
19. Merricks, E. M. & Schevon, C. A. Human Single-Neuron Recordings in Epilepsy. in Jasper’s Basic Mechanisms of the Epilepsies (eds Noebels, J. L., Avoli, M., Rogawski, M. A., Vezzani, A. & Delgado-Escueta, A. V.) (Oxford University Press, New York, 2024).
20. Merricks, E. M. et al. Aberrant fast spiking interneuronal activity precedes seizure transitions in humans. medRxiv 2024.01.26.24301821 (2024) doi:10.1101/2024.01.26.24301821.
21. Agopyan-Miu, A. H. et al. Cell-type specific and multiscale dynamics of human focal seizures in limbic structures. Brain J. Neurol. 146, 5209–5223 (2023).