1. Computational and Systems Biology
  2. Neuroscience

Human interictal epileptiform discharges are bidirectional traveling waves echoing ictal discharges

  1. Elliot H Smith  Is a corresponding author
  2. Jyun-you Liou
  3. Edward M Merricks
  4. Tyler Davis
  5. Kyle Thomson
  6. Bradley Greger
  7. Paul House
  8. Ronald G Emerson
  9. Robert Goodman
  10. Guy M McKhann
  11. Sameer Sheth
  12. Catherine Schevon
  13. John D Rolston  Is a corresponding author
  1. Departments of Neurosurgery and Biomedical Engineering, University of Utah, United States
  2. Department of Neurology, Columbia University, United States
  3. Department of Anesthesiology, Weill Cornell Medicine, United States
  4. Department of Pharmacology & Toxicology, University of Utah, United States
  5. Department of Bioengineering, Arizona State University, United States
  6. Neurosurgical Associates, LLC, United States
  7. Hospital for Special Surgery, United States
  8. Lenox Hill Hospital, United States
  9. Department of Neurological Surgery, Columbia University Medical Center, United States
  10. Department of Neurological Surgery, Baylor College of Medicine, United States
https://doi.org/10.7554/eLife.73541
Share this article
Cite this article
  1. Elliot H Smith
  2. Jyun-you Liou
  3. Edward M Merricks
  4. Tyler Davis
  5. Kyle Thomson
  6. Bradley Greger
  7. Paul House
  8. Ronald G Emerson
  9. Robert Goodman
  10. Guy M McKhann
  11. Sameer Sheth
  12. Catherine Schevon
  13. John D Rolston
(2022)
Human interictal epileptiform discharges are bidirectional traveling waves echoing ictal discharges
eLife 11:e73541.
https://doi.org/10.7554/eLife.73541
  2. Download BibTeX
  3. Download .RIS

Abstract

Interictal epileptiform discharges (IEDs), also known as interictal spikes, are large intermittent electrophysiological events observed between seizures in patients with epilepsy. Although they occur far more often than seizures, IEDs are less studied, and their relationship to seizures remains unclear. To better understand this relationship, we examined multi-day recordings of microelectrode arrays implanted in human epilepsy patients, allowing us to precisely observe the spatiotemporal propagation of IEDs, spontaneous seizures, and how they relate. These recordings showed that the majority of IEDs are traveling waves, traversing the same path as ictal discharges during seizures, and with a fixed direction relative to seizure propagation. Moreover, the majority of IEDs, like ictal discharges, were bidirectional, with one predominant and a second, less frequent antipodal direction. These results reveal a fundamental spatiotemporal similarity between IEDs and ictal discharges. These results also imply that most IEDs arise in brain tissue outside the site of seizure onset and propagate toward it, indicating that the propagation of IEDs provides useful information for localizing the seizure focus.

Editor's evaluation

This manuscript describes the propagation patterns of electrical activity in the brains of patients with drug-resistant epilepsy. Specifically, the authors demonstrate that interictal spikes, commonly observed electrical events in epileptic patients, propagate in a similar manner to seizures, which are relatively uncommon and more difficult to capture. This suggests that interictal spikes could be used in surgical planning, improving the localization and treatment of epileptic networks.

https://doi.org/10.7554/eLife.73541.sa0

Introduction

While seizures are mostly unpredictable and rare, electrical recordings from people with epilepsy often show isolated epileptiform discharges between seizures (Alarcon et al., 1997; Jefferys and Avoli, 2012; Tatum et al., 2016). These IEDs are far more frequent, occurring up to several times per minute, and exhibit multidien variation in their frequency that correlates with seizure likelihood, making IEDs an attractive personalized biomarker for seizure risk (Baud et al., 2018). Beyond such temporal information about seizure occurrence, there is some evidence for overlap between cortical areas where seizures originate and those with more IEDs (Alarcon et al., 1997; Conrad et al., 2020; Marsh et al., 2010). Furthermore, some retrospective studies showed that removing brain areas with more IEDs improved surgical outcomes in patients with medically refractory epilepsy (Kim et al., 2010; Smart et al., 2012). Despite these findings, the long-debated relationship between IEDs and seizure generating tissue remains unresolved (Jefferys and Avoli, 2012; Paolicchi et al., 2000; Tonini et al., 2004; Vakharia et al., 2018).

Microelectrode array recordings in epilepsy patients have revealed the spatiotemporal features of ictal self-organization (Eissa et al., 2017; Martinet et al., 2017; Smith et al., 2020; Schevon et al., 2012; Smith et al., 2016). These studies reported two classes of recordings, one in which neuronal firing is recruited into the ongoing seizure, and another in which neuronal firing is relatively unaffected, despite seizure-like field potentials appearing on the same microelectrodes. These classes correspond to two dynamically evolving regions known as the ictal core and penumbra, respectively (Schevon et al., 2012). A slowly-propagating, narrow band of tonic action potential firing, the ictal wavefront (IW), delineates the transition between the core and penumbra (Martinet et al., 2015; Schevon et al., 2012; Trevelyan et al., 2006; Trevelyan et al., 2007). These dynamic seizure regions exhibit distinct spatial features. The slowly traveling IW repetitively emits rapidly traveling ictal discharges backwards, toward the seizure core (Figure 1—figure supplement 1, Figure 1—video 1). These discharges occur following the passage of the ictal wavefront and have thus been termed ‘post-recruitment’ discharges (Smith et al., 2016). In some patients, earlier ictal discharges, termed ‘pre-recruitment’, are also emitted outward, toward the penumbra (Martinet et al., 2017; Smith et al., 2020; Smith et al., 2016). To avoid confusion between ictal discharges and interictal discharges, we will from here on refer to ictal discharges as seizure discharges (SDs).

The discovery of these spatiotemporal features of human seizure activity inspired a computational model designed to explain the neuronal underpinnings of seizure dynamics from biophysical principles (Liou et al., 2020). After several induced seizures in this model, the network produces spontaneous seizures and IEDs with spatiotemporal dynamics. One prediction of the model is that the repeated barrages of traveling synaptic activity during SDs eventually coopt mechanisms of synaptic plasticity, biasing local tissue to propagate IEDs in similar directions as SDs, that is in the opposite direction of the slow propagation of seizure expansion. In this study we test the resultant hypothesis that IEDs have a predominant direction of propagation, towards the site of seizure onset; opposite the direction of seizure expansion. We found that the data supported this hypothesis, and further, that in the majority of participants, IEDs traveled bimodally on a linear axis with predominant and auxiliary sub-distributions whose directions, speeds, and proportions echoing those of SDs. Finally, we directly quantified the extent to which IED directions could be used to predict SD directions.

Results

IED detection in human microelectrode array recordings

To examine the spatiotemporal propagation of IEDs, we used a multi-institutional dataset of Utah-style microelectrode array (UEA; 10 × 10 microelectrodes in 4 × 4 mm grid, penetrating 1 mm) recordings from 10 epilepsy patients (two female, µ ±σ age: 29 ± 5.24 years) undergoing monitoring for neurosurgical treatment of medically refractory epilepsy (clinical details in Appendix 1—table 1). In order to capture seizures (2.2 ± 1.6 seizures recorded per participant; 22 total), we recorded data continuously throughout the patients’ monitoring periods (Figure 1A; 4.3 ± 2.4 days per participant; 43 total). Searching through weeks’ worth of microelectrode data, we detected 45,623 candidate IEDs across the 10 participants (4562.3 ± 5171.7 per participant) using an IED detection algorithm designed for microelectrode recordings, that operated on features of IEDs based on the American Clinical Neurophysiological Society’s definition, namely high-amplitude bursts of beta-range (20–40 Hz) local field potential (LFP) power occurring across multiple microelectrodes (Figure 1—figure supplement 2; Appendix1 - algorithm 1) (Tatum et al., 2016). We evaluated the positive predictive value of the algorithm against the ratings of two clinician experts yielding a µ ±σ precision of 0.86 ± 0.04. Interrater reliability (Cohen’s Kappa) was 44.9%, which is similar to that reported across a large multicenter study of IED ratings (Jing et al., 2020). Interrater reliability between the algorithm and board-certified neurologist was 61.1%. Using this algorithm, we detected an average of 0.43 ± 0.51 IEDs per minute. The UEA enabled us to record both LFP data and multiunit action potential firing (MUA) across high-density spatial grid during each IED. These features of an example IED are shown in Figure 1C–G.

Figure 1 with 5 supplements see all
Download asset Open asset
IEDs are traveling waves.

(A) A electron micrograph of an UEA and a picture of an UEA implanted next to an ECoG electrode. (B) A raster plot showing an example time course of semi-chronic microelectrode recording during an epilepsy patient’s hospital stay. Each gray dot represents the time of one IED (y-axis is arbitrary). (C) An example IED recorded across microelectrodes. Each gray line is the same IED recorded on a different microelectrode. The mean IED waveform is overlaid in white. (D) Mean spectrogram of the IED shown in (C) across microelectrodes. (E) A temporally expanded view of the IED shown in (C) color coded by when the IED occurs. Black dots indicate the location of the IED negative peaks for each microelectrode. (F) A raster plot of IED-associated MUA firing for the same IED as in (C) and the same timescale shown in (F). (G) IED voltage minima timings, color-coded as in (E), superimposed across the footprint of the UEA. A white velocity vector derived from the multilinear regression model is also shown on the UEA footprint. (H) A polar histogram showing the distribution of all IED traveling waves from which the IED in (C) was taken. See Figure 2—figure supplement 1 and Appendix 1-Algorithm 1 for IED detection and traveling wave classification details.

IEDs propagate in predominant and auxiliary directions

In order to determine whether the detected IEDs were traveling waves, and to measure wave speeds and directions, we fit a plane to the timings of both IED voltage extrema and MUA event times measured on each microelectrode using multi-linear regression (Liou et al., 2017). IEDs with regression slopes that were significantly different from zero were classified as traveling waves (permutation test against a distribution of 1000 spatially permuted timings; Figure 1H, Figure 1—figure supplement 3). Traveling wave speeds and directions were then derived from each significant model’s slope. Based on this operational definition, 30,278 IEDs (3027.8 ± 3190.0 per participant) were classified as traveling waves (66.4%). Example non-traveling IEDs and regression model betas are shown in Figure 1—figure supplement 4.

Summary statistics for the spatiotemporal features of IEDs are shown in Table 1. Mean IED speeds were on the same order as SDs before the passage of the ictal wavefront (Liou et al., 2017; Smith et al., 2016). Traveling waves were also detected from MUA, independent of LFP recordings, though at a slightly reduced rate (2215.7 ± 3237.6 per participant; 22,157 total; 48.6%; χ2 = 2957, p < 0.05). This result was expected, as LFP is a more reliable signal to record, and action potential firing during IEDs has previously been shown to be remarkably heterogeneous, particularly in areas further from the seizure onset zone (Keller et al., 2010). That there were significantly more IED traveling waves in UEA recordings that were eventually recruited into the seizure core, further supports the idea that more firing, closer to the seizure onset zone improves reliability of traveling wave detection with MUA (McNemar Test, χ21=2957, p < 10-6). We therefore focus our analysis on IED traveling waves measured from LFP minima in order to understand IED propagation across participants.

Table 1
Summary statistics for spatiotemporal features of the dataset.
ParticipantSeizure classN detected IEDsN (%) traveling waves (LFP)N traveling waves (MUA)Median speed (cm/s)Bimodal?N seizures
1recruited1,7611,567 (89.0)64020.9yes1
2recruited1,5321,217 (79.4)13159.2no3
3recruited17,98810,220 (56.8)10,38025.3yes1
4recruited2,8061,429 (50.9)28463.7yes1
5recruited2,1481,538 (71.6)1,131108.3yes2
6recruited3,5023,143 (89.7)200669yes2
7penumbral4,3482,132 (49.0)2,23676.8yes5
8penumbral8,3697,351 (87.8)4,977134.7no0
9penumbral834296(35.5)5080.5yes3
10penumbral2,3351,385 (59.3)32270.7yes4
Totals 45,62330,27822,157 22

Having determined the majority of IEDs met the criteria to be classified as traveling waves, we next sought to understand whether IEDs from each participant exhibited a predominant propagation direction. We therefore tested whether distributions of IED traveling wave directions deviated from a uniform circular distribution (Fisher, 1953), we found that each participant’s IED traveling wave distribution exhibited a dominant direction (Figure 2A; Hermans-Rasson Tests, 1000 permutations, all p < 10–3). These results show that many IEDs are traveling waves with predominant, consistent directions of travel in each participant.

Figure 2 with 1 supplement see all
Download asset Open asset
IED traveling wave distributions are non-uniform and bimodal.

(A) Polar histograms of IED traveling wave directions for all 10 participants. Each participant number is indicated in bold above and to the right of each histogram. (B) Classification index for bimodality of IED distributions across subjects. Criterion is indicated with a dashed line. (C) Difference in median angles of sub-distributions for bimodal IED distributions, with non-bimodal subjects omitted. See Figure 3 for bimodality classification details.

In addition to a predominant direction common to all participants, many participants appeared to have a second, auxiliary, distribution of IED directions. We therefore fit each participant’s IED distribution into a mixture of two circular normal sub-distributions (von Mises distribution). The mixture model was compared to a single von Mises distribution model by using permutation-based Kuiper tests (Figure 2—figure supplement 1; see Materials and methods). IED traveling wave distributions were thus classified as bimodal in 8 of the 10 participants (Figure 2B). The mean and ±s.d. angles between the two IED sub-distributions was 177.9 and 10.7 degrees, respectively (Figure 2C). These results show that IEDs also frequently propagate antipodally to their predominant direction, suggesting that IEDs may travel both directions on a linear track through a fixed recording site.

Spatial features of IED distributions echo ictal self-organization

We next sought to understand whether IED speed and direction related to the spatial self-organization of seizures. We hypothesize that spatial features of IED traveling waves would correlate with seizure propagation direction and those of seizure discharges. We therefore measured the spatial features of seizures first. Fast and slow spatial features of focal seizures were measured in both ictal LFP and MUA bands as in previous reports (Liou et al., 2017; Schevon et al., 2012; Smith et al., 2016). Both of these features were measured using the same multilinear regression framework used to measure IED speed and direction.

Following previous reports with microelectrode arrays, we confirmed that seizures could be divided into two classes based on ictal recruitment: ‘recruited’ and ‘penumbral’ (see Materials and methods; Khodagholy et al., 2015; Martinet et al., 2017; Merricks et al., 2021). Recruited tissue exhibited a slow expansion of tonic neuronal firing, the ictal wavefront (Figure 3D–E; Figure 3—figure supplement 1A-D), followed by rapidly traveling SDs. Penumbral tissue showed neither an IW nor repetitive SDs associated with phase-locked firing (Figure 3—figure supplement 1E-H). Six participants’ microelectrode arrays were recruited into the ictal core (10 seizures; Figure 3A–C), while the four remaining participants were penumbral (12 seizures).

Figure 3 with 1 supplement see all
Download asset Open asset
IEDs reflect ictal self-organization.

(A) Mean voltage recorded across the UEA at the start of a seizure. (B) Mean MUA firing rate across microelectrodes. (C) A raster plot ordered by time of recruitment and color coded to show the IW (blue) and ictal core (“recruited”, pink). (D) Slow firing rate dynamics on each microelectrode colored by time of maximum firing rate. (E) Times of maximum firing rate on each microelectrode superimposed on the footprint of the UEA, color-coded as in (D). (F) A polar histogram of IEDs and the direction of the IW. (G) Polar histograms showing probability densities of IEDs and SDs, and the direction of the IW. See Figure 4 for examples of classes of microelectrode array recorded ictal self-organization.

As predicted by our theoretical work (Liou et al., 2020), similar patterns of IED and SD propagation were apparent in the majority of seizures in participants with ‘recruited’ seizures. The majority of IEDs travelled opposite the direction of seizure expansion (i.e. the IW; direction difference from IEDs = 148.9 ± 17.2 degrees; median tests between IW and IED distributions, all p < 0.05; Figure 3F). Moreover, IEDs traveled in similar directions as SDs in these participants (example in Figure 3G; mean ± s.d. angle difference across participants = 23.7 ± 33.7 degrees). Direction distributions for IEDs, SDs, relative to the direction of the ictal wavefront are shown for all ‘recruited’ participants in Figure 4A–F, and direction summaries for these participants are shown in Figure 4G–H (raw directions shown in Figure 4—figure supplement 1). In the ‘penumbral’ category, where the tissue under the UEAs were not obviously recruited from adjacent cortex as in the ‘recruited’ category, we could not reliably detect or measure the direction of seizure expansion.

Figure 4 with 2 supplements see all
Download asset Open asset
IED and SD distributions in ‘recruited’ UEA recordings.

(A–F) Each lettered subpanel corresponds to one participant. Each polar histogram corresponds to one seizure. IED distributions are shown in gray and SD distributions are shown in black. Both IED and SD distributions are plotted relative to the direction of the IW (blue line). See Figure 5 for raw IED, SD, and IW directions. (G) Direction difference summaries for each seizure ordered and color coded as in (A–F). Dots indicate median directions and lines indicate standard deviations. (H) Median and standard deviation IED direction summaries relative to median SD directions. See Figure 5 for raw IW, IED, and SD directions.

In order to directly quantify information about the full distribution of SDs that is gained from observing IEDs, we measured the Kullback-Leibler Divergence (KLD) between IED and SD distributions for each participant. The µ ± σ KLD across these 10 seizures was 0.66 ± 0.56, and are shown next to each pair of distributions in orange in Figure 4. The KLD values between IED and SD distributions were all significantly greater than would be expected to occur by chance in each patient (permutation tests, all p < 0.01). These results indicate that less than one extra bit of information is needed to encode the direction of SDs with IEDs on average, suggesting that IED directions could be used to accurately predict SD directions.

Spatial features of IED sub-distributions predict those of SD sub-distributions

Finally, we sought to further understand geometric features of bimodal IED distributions and how they related to patterns of SD propagation. Such an understanding was only relevant for the ‘recruited’ participants with bimodal IED distributions (five participants, seven seizures). Using the same bimodality classification strategy as for IEDs, we found that five of the seven these seizures in these participants exhibited bimodal SD distributions, with similarly antipodal sub-distribution directions (Figure 5A–B; µ ± σ angle difference = 135.2 ± 24.6). Only participants with bimodal IED distributions had bimodal SD distributions, and only one participant, who had the least bimodal IED distribution, did not have clearly bimodal SDs.

Figure 5
Download asset Open asset
Correspondence between spatial features of IED and SD sub-distributions.

(A) Example IED sub-distributions from one participant. (B) Example SD sub-distributions from the same participant as in (A). Numbers of IEDs in each sub-distribution are displayed in the color of each sub-distribution. (C) Mean IED waveforms from each sub-distribution, color coded as in (A). (D) Mean firing rates from each IED sub-distribution, color coded as in (A). (E) Scatter and box plots for IED and SD propagation speeds, separated by sub-distribution across subjects, and color coded as in (A-B). Boxes indicate median, quartiles, and whiskers indicate 1.5 times the interquartile range. Speeds greater than 350 cm/s are not shown for display purposes. Asterisks indicate significant differences (p < 0.05). (F) Distribution summaries for IED and SD traveling wave directions, separated by sub-distribution across subjects, and color coded as in (A-B). For the two participants with more than one seizure (P4, P6), each seizure is shown separately in (E) and (F).

In order to determine whether IEDs in each sub-distribution came from neurophysiologically distinct IED populations, we tested for differences between IED waveforms and firing rates between IED sub-distributions. Such differences might indicate that each IED sub-distribution reflected a separate population of IEDs propagating across the footprint of the UEA. However, neither firing rates nor IED waveforms differed between IED sub-distributions (Figure 5C–D; cluster-based permutation tests, all p > 0.05), indicating that neither IED waveforms nor firing rates between the two sub-distributions could be statistically distinguished.

Differences between speeds and proportions of IED sub-distributions, corresponding to those we previously showed in SD sub-distributions (Smith et al., 2016), would support a learned relationship between IEDs and SDs, as predicted by the centripetal pattern of learning in the theoretical model (Liou et al., 2020). To address this question, we tested for differences in speed and relative size of SD and IED sub-distributions. Speeds were significantly different between IED sub-distributions within each participant (Figure 5E; Mann-Whitney U, all p < 10–4). Speeds were also significant between SD sub-distributions in five of the seven seizures (Figure 5E; Mann-Whitney U, p < 0.04 in five seizures; p > 0.62 in two seizures). Finally, the proportion of IED directions in each sub-distribution predicted the direction of each SD sub-distribution in four of the five participants (two sample proportion tests, χ2 > 347.6, p < 10–6). More pre-recruitment discharges occurred in a fifth patient with nearly equivalent proportions of IEDs across sub-distributions. Importantly, the directions of significant differences in these spatial features corresponded across IED and SD sub-distributions. These results show that when IED and SD distributions were bimodal, their spatial features were similar, underscoring the extent to which IEDs mimic spatiotemporal features of ictal self-organization.

Discussion

Our results, using microelectrode array recordings in patients with epilepsy, show that IEDs are traveling waves that echo – and are predictive of – the propagation patterns of SDs, along a parallel axis to the direction of expansion of the seizure core. It is likely that the frequent barrages of coordinated activity seen during seizures—SDs—potentiate propagation pathways through neocortex that are revisited during IEDs. This conclusion is corroborated by the predictions of our theoretical work, where a computational model incorporating spike-timing dependent plasticity and realistic connectivity between inhibitory and excitatory cells self-organized to produce IWs, SDs, and IEDs that echo through the pathways potentiated by the strong, repeated barrages of SD activity, antipodal to the IW direction (Figure 6; Liou et al., 2020; Nguyen et al., 2020). Therefore, these results suggest that IEDs originate from tissue immediately eccentric to the seizure core and propagate inwardly across the seizure onset zone, indicating their potential use for localizing the seizure source (Liou et al., 2020; Liou et al., 2019).

Figure 6
Download asset Open asset
Schematic of how IEDs relate to ictal self-organization.

A schematic of spatial features of ictal self-organization is shown on the left. Schematics illustrating the correspondence between spatial features (speed, directions, and proportion) of IEDs and SDs is shown on the right.

The empirical results reported here extend our understanding of the geometric properties of epileptic tissue, beyond the model predictions, in showing that IEDs travel largely bidirectionally on a linear axis. The bidirectional propagation of IEDs is similar to the bidirectional traveling waves we have previously observed during seizures (Liou et al., 2017; Smith et al., 2016). The bidirectional pattern of traveling waves during seizures emanated from a slowly expanding, motile source of ictal activity—the IW—passing through a fixed recording site (Smith et al., 2016). The data presented here show that IEDs travel in similarly oriented, bimodal distributions, even in the absence of an IW or ictal self-organization.

The IED directions we report are not perfectly antipodal to the IW. These differences in mean angles may be due in part to error when fitting a plane to the IW, which discounts the propagating wave’s curvature, or biological variability, such as changes in brain state and arousal. Furthermore, much of the deviation in the summary statistics is driven by one participant (P3). Rigorous statistical analysis of cluster bimodality, the resulting mean angles from that analysis, and KLD between IEDs and SDs all support the major result of the paper, that IEDs and SDs travel bidirectionally along a similar path. The dominant direction of travel is largely in the opposite direction of seizure expansion.

The theoretical model predicted that the directional preferences of IEDs were learned from SDs via spike-timing dependent plasticity (Bi and Poo, 1998). While we cannot address the specific learning mechanism with this dataset, in participants whose UEAs were recruited into the seizure core from adjacent cortical tissue, we showed correspondences between several spatial properties of IEDs and SDs. The speeds, directions, and relative sizes of IED sub-distributions echoed those of SDs. Moreover, predominant IED and SD directions opposed the directions of the ictal wavefronts in ‘recruited’ UEAs. These relationships were unable to be determined from UEAs that were not recruited into the seizure core (‘penumbral’ recordings). Together, these results suggest that spatiotemporal biases exist in epileptic tissue. Whether spatiotemporal biases in IEDs arise from learning during SDs or vice versa remains to be determined. While the theoretical model indicates that several seizures must occur before IEDs begin to form, electrographic discharges, similar to IEDs, often appear before seizures in animal models of epilepsy (Staley et al., 2011).

While we show that the majority of IEDs are traveling waves whose directions overlap with those of SDs, it is important to recognize the small spatial scale of the recordings analyzed here. Beyond the small area of brain we recorded from, there is some sampling bias inherent in these recordings, as we sought to place the microelectrode arrays as close to the site of seizure onset as possible, and within the area of resection. Additional, more eccentric populations of IEDs could be propagating from distant areas that are connected to the seizure onset zone, though not necessarily from adjacent tissue on the cortical surface (Gelinas et al., 2016). Higher density ECoG that spans a larger cortical territory than the UEA would be useful in gaining more context on where IEDs arise, and how IEDs propagate across the cortical surface. On the other hand, there is currently no evidence that the ictal wavefront can be detected without action potential recordings, though the time of seizure recruitment can be roughly estimated on each ECoG electrode from high-frequency LFP (Smith et al., 2020; Weiss et al., 2013). Precise IED propagation patterns are also difficult to measure with the relatively low sampling density of ECoG, and minor shifts in the relative orientation of the ECoG grid and microelectrode array could impugn the accuracy of comparing traveling wave directions between the two modalities. Animal studies using calcium indicators capable of imaging neuronal activity across large cortical territories may overcome these limitations (Liou et al., 2019; Wenzel et al., 2017). Such animal studies are also poised to understand how learning and plasticity contributes to the geometric relationships reported here.

While these MEA recordings inform fundamental geometric relationships between IED propagation and ictal self-organization, several MEAs flanking the seizure onset zone may be required to accurately triangulate the SOZ from traveling wave directions. Additional spatial context may also inform our understanding of ‘penumbral’ tissue and its relevant spatial biases (Smith et al., 2020). Future work will therefore focus on translating these microelectrode array results to a more clinically relevant spatial scale. For example, using ECoG (Khodagholy et al., 2015Khodagholy et al., 2015; Viventi et al., 2011) with vector field or convolutional methods (Muller et al., 2016), or examining propagation of source-localized IEDs in stereo-EEG along white matter tracts. Such approaches will be useful for linking the micro results reported in this paper to the coarser spatial resolutions and broader coverages encountered with typical intracranial recordings. Recordings with a broader spatial scale could also inform how IEDs propagate in a larger epileptic network, including patients whose seizures have a common onset site: the mesial temporal lobe (MTL). Three of the participants in this study (P1, P2, and P3) were classified as having MTL seizure onsets from their clinical reports, yet still exhibited similar IED and SD traveling wave dynamics. Despite MTL onsets, these patients’ seizures spread to the lateral temporal cortex, and eventually to the microelectrode array site in all three cases. We would speculate that this lateral seizure spread in neocortex could establish similar spatiotemporal biases to those in the other patients, yet a more widespread recording technology with higher resolution than ECoG may be required to full reveal the dynamics of MTL to neocortex seizure spread. It is noteworthy that two of these patients had larger differences between median IED and SD direction distributions than the others. Integrating these multi-scale geometrical understandings of how IEDs relate to the seizure onset zone could then provide an additional piece of information to inform diagnosis and treatment of medically refractory epilepsy, and potentially enable localizing the seizure source without having to directly observe seizures.

Materials and methods

Key resources table
Reagent type (species) or resourceDesignationSource or referenceIdentifiersAdditional information
Software, algorithmMATLABhttps://www.mathworks.com/products/matlab.htmlSCR:001662 
Software, algorithmNPMKhttps://github.com/BlackrockNeurotech/NPMK (Torab, 2014)
 
Software, algorithmIED analysis codehttps://github.com/elliothsmith/IEDs (Smith, 2022) 
Software, algorithmCircular statistics toollboxhttps://github.com/circstat/circstat-matlab (Berens et al., 2019) 
Software, algorithmfitmvmdisthttps://github.com/chrschy/mvmdist 
Software, algorithmhrtesthttps://github.com/cnuahs/hermans-rasson (Cloherty, 2020) 

Participants, ethics statement, and data

Request a detailed protocol

The data for this study were acquired from Utah-style microelectrode arrays (UEAs) that were implanted in 10 human patients across two surgical sites who were undergoing neurophysiological monitoring for surgical treatment for medically refractory seizures. Clinical details for all participants are shown in Appendix 1—table 1. The Institutional Review Boards at the University of Utah (IRB_00114691) and Columbia University Medical Center (IRB-AAAB6324) approved these studies. All participants provided informed consent prior to surgery for implantation of the clinical electrocorticography (ECoG) electrodes and UEA (10 × 10 electrodes in 4 × 4 mm, penetrating 1 mm). Methodological details of surgical implantation of UEAs into human epilepsy patients area described in detail in House et al., 2006. During implantation of ECoG electrodes, UEAs were pneumatically inserted into areas that were most likely to be in the seizure onset zone, and therefore most likely to be resected. Electrophysiological data were pseudodifferentially amplified by 10 and acquired at 30 kilosamples per second using a neural signal processing system (Blackrock Microsystems, Salt Lake City, UT) semi-chronically, that is throughout the duration of the participants’ hospital stays. Patients were weaned off their antiepileptic medication during that time, as dictated by their individual clinical courses. Throughout the manuscript, numerical quantities are presented as mean ± standard deviation (µ ±σ).

IED detection and signal processing

Request a detailed protocol

In order to detect IEDs from continuous data recorded on each UEA channel, we developed a simple algorithm for detecting IEDs across a microelectrode array (Appendix 1 - Algorithm S1). For each channel on each UEA, we first resampled the data at 400 samples per second and zero-phase filtered the data between 20 and 40 Hz using a fourth-order Butterworth filter. We then detected any peaks in the absolute amplitude of this signal that were greater than eight times the standard deviation of the remainder of the recording segment (2-hr median duration; Figure 1—figure supplement 3A). In order to remove redundant detections, those following any other detection by less than 250ms were discarded. Only detections that occurred within the same 250ms window across at least 10 electrodes on the UEA were retained for further analysis (Figure 1—figure supplement 3B). Spectrograms of IEDs were generated via the continuous wavelet transform following the methods used in Schevon et al., 2012; Smith et al., 2020, and colored with the turbo color map (Mikhailov, 2021).

Multiunit action potentials (MUA) were detected on each microelectrode by filtering each channel between 0.3 and 3 kilohertz and detecting peaks in the filtered signal less than –4 times its root mean square. The times of these peaks were retained for further analysis. Example retained detections are shown in Figure 1—figure supplement 3C,E.

We employed several post-detection processing steps to ensure the quality of this expansive data set and reject artifacts. First, temporally outlying voltage extrema were removed in order to constrain extrema detection into a temporally focused window (approximately 50 ms duration) around the time of IED detection and to exclude broken microelectrodes or those without IED signal. Next, we excluded any discharges with outlying amplitudes, defined as double the interquartile range of the distribution of IED voltage ranges (Figure 1—figure supplement 3F,G).

In order to validate the detection algorithm’s ability to detect real IEDs, we calculated the algorithm’s positive predictive value (precision) against ratings of experienced clinicians (C.A.S. and J.D.R.) on a dataset of 78 random IED detections, including 9 of those that did not pass our post-detection quality assurance steps described in the previous paragraph. Algorithmic positive predictive value (precision) was defined as the ratio of the number of true positive detections to the number of total positive detections. Cohen’s Kappa was also calculated to evaluate inter-rater reliability against chance.

Traveling wave measurement

Request a detailed protocol

In order to measure IED traveling wave speed and direction, we fit a plane to the timings of IED voltage minima, and MUA event times, using ordinary multilinear regression, regularized via the absolute deviation of the signal (Figure 1—figure supplement 4). This methodology is described in detail and validated for measuring traveling waves during ictal discharges in Liou et al., 2017. Briefly, the regression model for each IED yielded three coefficients, describing the best-fit plane to the timing of IEDs across the UEA in spacetime. Traveling wave direction was determined by the gradient direction of the plane and speed was defined as the inverse of that gradient norm. Each IED was operationally defined as a traveling wave if its model significantly deviated from a plane with zero slope. Statistical significance for this measure was determined by a permutation test in which the model was reevaluated 1,000 times with the microelectrode spatial locations randomly permuted. Differences in IED speed across and within participants were tested using a two-way Kruskal-Wallis Test with a participant factor (10 levels; one for each participant) and a signal factor in which the two levels were the speed measurements derived from LFP and MUA. Only MUA times from 50ms before and after the median time of the LFP negative peak were included in the regression model (Liou et al., 2017). Post-hoc pairwise comparisons were carried out using Dunn’s test (Dunn, 1964). The significance criterion was chosen as 0.05 for all of these tests.

Directional statistics

Request a detailed protocol

Polar histograms were plotted using 18 bins. Circular normal distributions were fit using the circular statistics toolbox (Berens, 2009). These distributions defined by two parameters, μ and κ, which describe the central angle and concentration of the distribution, respectively. μ and 1/κ are analogous to the mean and variance parameters that define a standard normal distribution. Directional statistics were carried out using modified functions from the circular statistics toolbox (Berens, 2009). These modifications were such that statistical significance was evaluated using permutation tests from which p-values were derived by comparing the circular test statistic with a distribution of circular test statistics from 1000 permuted datasets. As an example, testing for differences between IED and SD means would involve comparing the test statistic from the true data to a distribution of 1000 test statistics in which the measurement categories were permuted. The significance criterion was chosen to be 0.05. Hypotheses that within-participant IED propagation directions were non-uniform, were tested with Hermans-Rasson tests of circular non-uniformity, again with 1000 permutations (Landler et al., 2018).

Bimodality and sub-distributions

Request a detailed protocol

In order to determine whether two, unimodal distributions better fit the ostensibly bimodal IED and SD distributions we observed, we first fit von Mises Mixture (vMM) Models to overall distributions of IED and SD directions using the Matlab function fitmvmdist (https://github.com/chrschy/mvmdist; Schymura, 2016). Overall IED direction distributions were then clustered into two component vMM distributions using the Matlab function cluster. We did not observe distributions that appeared to have more than two modes and therefore set an upper limit on the number of hypothesized clusters, h, at two. These vMMs yielded three parameters for each sub-distribution, h∈N, such that h≤2: the sub-distribution means, μh , concentration parameters, κh , and probability densities, θh .

Rather than assuming these vMM models better fit overall IED and SD distributions, we assessed whether the overall distribution or each vMM sub-distribution better fit the distributions defined by μh and κh . The permutation-based Kuiper tests used to assess goodness-of-fit were carried out as follows. We first estimated μh , κh , and θh for both the overall and vMM sub-distributions. We then carried out permutation-based Kuiper tests, to compare empirical distributions of 60 randomly sampled IED directions to theoretical circular normal distributions derived from μh and κh from both the original and vMM sub-distributions. We repeated this procedure 1,000 times in order to create a permutation distribution. In this way, we were able to measure the extent to which randomly sampled IED angles deviated from theoretical circular distributions defined by the overall and vMM parameters. This is akin to cross-validating vM parameters and choosing h corresponding to the highest log-likelihood, yet in a model-free way. Comparing Kuiper test statistics to a permutation distribution, rather than zero (the null hypothesis of a uniform distribution), makes the tests more conservative and allows us to determine whether the distribution cannot be determined to be non-normal (Louter and Koerts, 1970). We then defined our circular bimodality index as the minimum difference between the Kuiper test statistic for the overall distribution and each vMM sub-distribution. Positive bimodality indices thus indicated that overall traveling wave distributions were better modeled as two vMM sub-distributions, and negative bimodality indices indicated that overall traveling distributions were better modeled as a single von Mises distribution. The Matlab functions for implementing these classifications are highlighted in the online code repository.

Seizure characterization

Request a detailed protocol

In order to study the propagation patterns of IEDs relative to seizures, it was necessary to quantify spatial features of SDs and seizure expansion for each recorded seizure. These measures have also been described in previous publications (Smith et al., 2020; Schevon et al., 2012; Smith et al., 2020; Smith et al., 2016). The ictal wavefront (IW) is the slowly expanding edge of the seizure representing the spatial signature of failure of feedforward inhibition, and therefore defines recruitment of the tissue surrounding an electrode into the ictal core (Schevon et al., 2012). This biomarker of seizure recruitment and expansion has thus far only been detected with recordings of multiunit firing rates (Smith et al., 2016), although can also be detected on single microelectrodes by observing widening of action potential waveforms (Smith et al., 2020; Merricks et al., 2015). We followed methods from our previous manuscripts to generate multiunit firing rates, that is filtering the broadband data between 300 and 3000 Hz and detecting any peaks larger than the median absolute value of the signal divided by 0.6745 (Quiroga et al., 2004).

To find seizures, we examined time periods in the microelectrode array recordings that corresponded to the seizure times reported in the clinical monitoring reports and confirmed that the microelectrode array LFP exhibited the characteristic high amplitude, rhythmic discharging associated with seizures. We then looked for tonic multiunit firing spreading across the array that would suggest the presence of an ictal wavefront, and phase locked multiunit bursting associated with ictal discharges. We detected the ictal wavefront feature of seizures by smoothing the firing rates on each microelectrode with a 250 ms Gaussian kernel and fitting a multilinear regression model to the peaks of these slow firing rate estimates across the UEA (Liou et al., 2017; Smith et al., 2016). We detected ictal discharges by detecting peaks in multiunit firing rates, calculated with a 25 ms Gaussian kernel, as we have previously (Liou et al., 2017; Smith et al., 2016). We quantified the propagation direction and speed of ictal discharges using the same methods used to determine IED traveling wave speeds and directions, as in Liou et al., 2017. IED speeds and directions were measured in a manner that was blinded to each microelectrode array’s recruitment classification. We defined the presence of ictal phase-locked firing with a Hermans-Rasson test of circular uniformity on MUA action potential times across microelectrodes relative to the phase of the mean LFP recorded across the UEA, similarly to Schevon et al., 2012. Using these measures, we operationally defined two patterns of ictal self-organization, based on observed patterns in these fast and slow spatial features of seizures: ‘recruited’ seizures were operationally defined as those with a significant ictal wavefront multilinear regression model and significant phase-locked multiunit firing (Figure 3—figure supplement 1A-D). ‘Penumbral’ seizures were operationally defined as those seizures in which we were unable to detect an ictal wavefront on the microelectrode array (Fig Figure 3—figure supplement 1E-H). Similar classifications of adjacent and non-adjacent recruitment have been reported by other groups (Martinet et al., 2015; Schevon et al., 2012). For seizures that were associated with secondary generalization, we only included discharges up to the clinically defined point of secondary generalization in order to constrain our study to the dynamics of focal seizure onset and spread.

Comparing IED and SD sub-distributions

Request a detailed protocol

Mean firing rates were estimated by binning MUA event times into one-hundred 10 ms bins across microelectrodes. For distributions that were determined to be bimodal, we used cluster-based permutation tests to test for differences between median IED waveforms and firing rates between IEDs from the two component vMM distributions for each participant. Permutation tests consisted of carrying out t-tests on each sample in the IED waveforms and each bin in the IED firing rates. The t-statistics for each sample or bin were then compared to permutation distributions of t-statistics for each sample or bin in which the sub-distribution labels were shuffled. If any of these p-values exceeded the specified alpha value of 0.05, we considered that sample significant, and included it with adjacent significant samples in a temporal cluster. In order to control for multiple hypothesis testing, we used cluster-based correction over temporal clusters. This procedure involved finding the largest temporal clusters of significant samples or bins in the permutation distributions and comparing those clusters to the largest temporal clusters of significant samples or bins in the real data.

In order to test for different directions of overall distributions and sub-distributions of IEDs and SDs, we used Watson-Williams multi-sample tests for equal means. To test for differences in IED speed between different sub-distributions, we used within-participant Mann-Whitney U tests. In order to test differences in the proportion of IEDs from predominant and auxiliary sub-distributions we used two-sample proportion tests. We employed a significance criterion of 0.05 for all of these tests.

Finally, in order to directly quantify the reduction in uncertainty about the direction of SD travel that can be estimated from observing IED directions, we calculated the Kullback-Leibler Divergence (KLD) between distributions of IED and SD directions (Kullback and Leibler, 1951). We estimated the KLD as the expectation of the logarithmic difference between discrete distributions of IED and SD directions on the half-closed interval between 0 and 2π. We interpret the KLD as measuring the information that can be gained about SD directions from observing IED directions. We also tested for whether these KLD values were significantly greater than would be expected by chance by comparing the true KLD values against a distribution of 1000 KLD values generated from directions drawn at random from both IED and SD distributions.

Acknowledgements

Request a detailed protocol

Thanks to the Schevon and Rolston labs.

Appendix 1

Appendix 1 - Algorithm 1. Algorithm for detecting IEDs from microelectrode recordings.
Input: A matrix of microelectrode voltage recordings, V(c, t), where measured voltage is a function of time, t, and which microelectrode array channel it was recorded on, c.
Output: A vector of IED times, I(t).
1: for each c, do:
2: filter data between 20 and 40 Hz (non-causal, 4th order Butterworth filter)
3: detect peaks, p, in filtered signal greater than eight times the standard deviation of the beta power in the data.
4: discard peaks that occur within 250ms of a preceding detection
5: for each p, co-occurring within 250ms, across more than 10 microelectrodes.
6: find all local minima of V(c, t) in time, across all c.
7: bin local minima in time, across all c.
8: convolve the resulting histogram with a modified Heaviside function,H(t):{0,0≥n<0.410t,0.4≥n<0.5−t,0.5≥n<1
9: find absolute minima of V(c, t) within the bin containing the most local minima.
Appendix 1—table 1
Clinical details for research participants.
ParticipantAgeSexEpileptogenic zoneUEA implant sitePathologyOutcome
119femaleright posteriror lateral temporalRight posterior temporal, 1 cm inferior to angular gyrusNon-specificEngel 1 a at >2 years
232femaleleft inferior temporal lobeinferior temporal gyrus, 2.5 cm from temporal poleone neuronal loss; lateral temporal nonspecificEngel 1 a at 55 months
332maleRight mesial temporal loberight inferior frontal gyrusNormal hippocampusEngel1a at 2.5 years
428maleleft dorsal posterior prefrontal cortexleft posterior middle frontal gyrusMild reactive astrogliosis, patchy microgliosis, Chaslin’s marginal sclerosisEngel 3 a at 32 months
526maleleft middle subtemporalleft posterior inferior temporal gyrusDiffusely infiltrating low grade glioma, IDH-1 negativeEngel 4 a at 2 years, 5 months
625maleLeft mesial temporal lobeLeft middle temporal gyrus, 3 cm posterior to temporal poleNon-specificEngel 1 a at 7 months
730maleright middle inferior temporal gyrusright middle temporal gyrusMild astrocytosisEngel 1 a at 12 months
830maleright lateral and mesial temporal lobe; nonlesionalRight middle temporal gyrus, 4 cm posterior to the temporal poleMesial temporal sclerosisEngel 2 at 22 months
930maleleft supplementary motor arealeft supplementary motor areaN/A (multiple subpial transections performed)Engel three at >2 years
1039maleleft frontal operculumleft lateral frontal, 2 cm superior to Broca’s AreaNonspecificEngel 1 a at >2 years

Data availability

Raw data is available upon establishment of a data use agreement with Columbia University Medical Center as required by their Institutional Review Board (IRB). Data from human subjects was analyzed, from which the dates of implants can potentially be reconstructed. This is especially true for a study like this one, in which chronic recordings were carried out for the full duration of the patients' hospital stays. Sharing these data widely could therefore expose private health information of participants, which is why a data use agreement is required by the IRB. Interested Researchers should contact Dr. Schevon to get the data use agreement process started with the Columbia University Medical Center IRB. Analysis code is upload to GitHub: https://github.com/elliothsmith/IEDs, (copy archived at swh:1:rev:218f6e60e8f0b32d4e16a88298475618d5c09589). We have included preprocessed data files for all IEDs, hosted online at OSF: https://osf.io/zhk24/. Data files include LFP, MUA event times, and traveling wave model coefficients for all detected IEDs.

The following data sets were generated
    1. Smith EH
    2. Liou JY
    3. Merricks EM
    4. Davis TS
    5. Thomson K
    6. Greger B
    7. House PA
    8. Emerson RG
    9. Goodman RR
    10. McKhann II GM
    11. Sheth SA
    12. Schevon CA
    13. Rolston JD
    (2021) Open Science Framework
    ID osf.io/zhk24/. Human interictal epileptiform discharges are bidirectional traveling waves echoing ictal discharges.
    https://osf.io/zhk24/

References

    1. Dunn OJ
    (1964) Multiple Comparisons Using Rank Sums
    Technometrics: A Journal of Statistics for the Physical, Chemical, and Engineering Sciences 6:241.
    https://doi.org/10.2307/1266041
    1. Fisher R
    (1953) Dispersion on a Sphere
    Proceedings of the Royal Society A 217:295–305.
    https://doi.org/10.1098/rspa.1953.0064
  13. Book
    1. Jefferys JGR
    2. Avoli M
    (2012) Interictal Epileptiform Discharges in Partial Epilepsy
    In: Noebels J, Avoli M, Rogawski M, Olsen R, Delgado-Escueta A, editors. Jasper’s Basic Mechanisms of the Epilepsies. Oxford University Press. pp. 213–227.
    https://doi.org/10.1093/med/9780199746545.003.0017
  29. Software
    1. Mikhailov A
    (2021)
    Turbo, An Improved Rainbow Colormap for Visualization
    Google AI Blog.

Decision letter

  1. Matthew Shtrahman
    Reviewing Editor; University of California, San Diego, United States
  2. John R Huguenard
    Senior Editor; Stanford University School of Medicine, United States
  3. Brian Litt
    Reviewer

Our editorial process produces two outputs: (i) public reviews designed to be posted alongside the preprint for the benefit of readers; (ii) feedback on the manuscript for the authors, including requests for revisions, shown below. We also include an acceptance summary that explains what the editors found interesting or important about the work.

Decision letter after peer review:

Thank you for submitting your article "Human interictal epileptiform discharges are bidirectional traveling waves echoing ictal discharges" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, and the evaluation has been overseen by a Reviewing Editor and John Huguenard as the Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: Brian Litt (Reviewer #1).

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

If possible, please address all of the reviewer concerns discussed below. A summary of the essential revisions are listed here.

Essential revisions:

1)Please describe how you validated automated IED detections. If possible, please quantify the positive predictive value for this method for a random subset of IEDs.

2) The angle between the SDs (ictal wavefronts) and IEDs was not identical (exactly opposed). In addition, the SD distributions were not exactly antipodal. Please address your interpretation of the data and whether SDs and IEDs follow the same path and what might account for the sizable angle between their observed paths.

3)Please discuss the sizable proportion of IEDs (~35%) that were not classified as traveling waves.

4)Please describe how seizures were defined.

5)Please elaborate on the implications of antipodal IED distributions.

6)Please provide more detail about how you quantified differences in the IED waveforms for the cluster-based permutation test.

7) Please comment on whether the "mixture" von Mises distribution has more adjustable parameters than the single von Mises distribution model. If so, does this bias your quality of fit towards the bimodal distribution for IEDs?

8) Please further clarify the terms used for the different discharges examined and

discussed in the manuscript.

9) Please compare IED propagation measured from the ECoG array to the IED propagation vector produced from the UEA.

10) Please address whether the resulting surgical resections in recorded patients agreed with the axis predicted by the IED propagation measured by the UEA. If possible, please discuss whether the success of the surgery correlated with the UEA-predicted focus.

11) Please comment on the relationship between the neocortical propagation measured by the UEA on the cortical surface and the ability to accurately localize deep foci such as in MTLE. How well does the bidirectional IED hypothesis apply to mesial temporal lobe seizures where the ictal onset is far from the neocortex?

12) Please include whether patients were weaned from their AEDs during these recordings.

13) Please discuss how the IED distributions for subjects with multiple seizures were calculated.

Reviewer #1 (Recommendations for the authors):

1)How did you validate automated IED detections? Could you report on positive predictive values for a random subset of detections?

2)You write in the abstract and discussion that IEDs traverse the same path as ictal discharges, but the angle between the SDs and IEDs was 24 degrees, and the IEDs weren't exactly opposite the ictal wavefront (150 degrees). Do you think the SDs and IEDs really follow the same path, and if so, what accounts for the sizable angle between their observed paths?

Reviewer #2 (Recommendations for the authors):

1. The terminology used to describe the various types of epileptiform activity described in this paper should be laid out in a clear way from the beginning. Although many of the features are defined in prior work, they are critical to understanding the results. Specifically, the introduction of SDs focuses on their propagation following the IW, but "pre-recruitment" SDs are a prominently analyzed group in Figure 5. It would be helpful to include a schematic figure like that at the top left of 6A earlier in the paper. It may also be useful to define two types of SDs (anterograde and retrograde?), and include them on the schematic figure. Perhaps "recruited" and "penumbral" could be illustrated here as well (or in a separate subpanel).

2. The prominent clinical relevance of the analysis presented in this manuscript is that IEDs could be used to locate the seizure focus in pre-surgical planning. As briefly indicated in the discussion, the UEA is likely too small to localize a seizure focus – it produces one vector, which at best could define the line along which the focus falls. To identify the seizure focus, greater spatial sampling is needed. The authors point out that the simultaneously recorded ECOG is insufficient to measure IW propagation as it cannot detect MUAs and its poor spatial sampling make estimation of IED propagation low-fidelity. Nevertheless, since LFP was used to record IED propagation in the UEA, it would be useful to analyze IED propagation from the ECoG array to compare the vector produced, particularly by neighboring electrodes.

3. If the ECoG analysis fails to produce a vector that match the UEA vectors, it may be necessary to implant two UEAs. The intersection of the IED propagation vectors would hypothetically localize the seizure focus (in 2D). Obviously this experiment is outside the scope of this study, but worth mentioning in the discussion.

4. Did the resulting surgical resections in recorded patients align with the axis predicted by the IED propagation at the UEA? Did the success of the surgery relate to the removal (or not) of the UEA-predicted focus?

5. Synaptic connectivity is probably important determinant of whether vector points to focus. Furthermore, UEA recordings are performed on the cortical surface. In the most straightforward application of the proposed strategy, seizures would start on cortical surface and propagate locally. However, many seizures begin in deep structures. Some discussion of the relationship between neocortical propagation and localization of a deep focus (e.g. MTLE) is warranted.

6. Were patients weaned from their AEDs during these recordings. One important advantage of analyzing IEDs vs seizures for surgical planning is that they are very common. Thus, it might be possible to perform the mapping in patients without having them in withdrawal from their AEDs, which might improve predictive value.

7. In Figure 4A-F, for the subjects with multiple seizures, the IED distribution appears to be different for each seizure. What IED distributions are plotted – only those that had occurred up to the time of that seizure? Maybe I missed it, but I couldn't find this info in the results or figure caption…

Reviewer #3 (Recommendations for the authors):

The Utah array is an excellent tool to study spatiotemporal dynamics in and around a seizure core. The strength of such a technique is unfortunately also the weakness. Since it is implanted on the cortex, the modeling of IED and seizure discharges has a potential for bias due to the location of the implant. The model can be strengthened by comparing the MUA and LFP of the model with their single-unit data of IEDs from hybrid electrodes of mesial temporal lobe implants. The authors do mention the limitation of their study, but it would help if they discuss this some more, especially since only two patients in Table S1 had mesial temporal onset and the rest were lateral or neocortical onsets. They should not generalize that this model is applicable to all seizure types.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Human interictal epileptiform discharges are bidirectional traveling waves echoing ictal discharges" for further consideration by eLife. Your revised article has been evaluated by John Huguenard as the Senior Editor, and a Reviewing Editor.

The manuscript has been improved but there are some remaining issues that need to be addressed, as outlined below:

The majority of reviewer concerns have been adequately addressed. However, there remain a few points that require clarification or additional submission materials. The Reviewing Editor has drafted this concise list to facilitate your revision.

Essential revisions:

1) Multiple reviewers raised the question regarding "the sizable angle between… SDs and IEDs" and the fact that the "IEDs weren't exactly opposite the ictal wavefront". You state that "the angle discrepancies to which the reviewer refers are statistically insignificant and can be explained by expected biological variability". On the other hand you state that based on Smith et al. 2016, you don't expect the IED to be strictly antipodal to IW. Finally, while the KLD analysis is useful it does not seem to fully address the reviewers' questions regarding what appears to be a (statistically) significant deviation (almost 2 standard deviations) of the IED/IW angle from 180 degrees. A single paragraph in the discussion summarizing and clarifying the various points of why SD, IEDs, and IWs may not be strictly aligned or antipodal would improve the manuscript.

2) Regarding Reviewer 2, major point #7 "In Figure 4A-F, for the subjects with multiple seizures, the IED distribution appears to be different for each seizure. What IED distributions are plotted – only those that had occurred up to the time of that seizure?" The reviewer is asking why the shape of the distribution, and not just the rotation, of IED distribution varies for different seizures in the same patient. Are you analyzing/including a different subset of IEDs for each seizure? If so, why? If not, why does the distribution change?

3) Regarding Reviewer 2, major point #2, we strongly suggest including the figure showing the "qualitative information about the location of grids, arrays, and IED vectors". We realize that there are a number of caveats and limitations in the interpretation of this qualitative figure that require further explanation, but including the figure would greatly enhance the clarity of the manuscript

4) Please refer to https://submit.elifesciences.org/html/eLife_author_instructions.html#revised for instructions on how to submit source data for each figure. There do not seem to be source data included for the majority of figure plots. Also the data repository on OSF would greatly benefit from a description of the different data file types, the different variable types within the structures, and the sampling rate for the IED traces.

https://doi.org/10.7554/eLife.73541.sa1

Author response

Reviewer #1 (Recommendations for the authors):

1) How did you validate automated IED detections? Could you report on positive predictive values for a random subset of detections?

We have validated the algorithm against the gold standard (board-certified neurologist) and calculated the PPV for a subset of 78 IEDs across patients. Two clinicians who were adept at identifying IEDs evaluated groups of 10 microelectrode array channels along with a raster of firing across channels for 10 s segments around each IED (CAS and JDR). We calculated the positive predictive value against the ratings of the board-certified neurologist (CAS). The PPV for the neurosurgeon was 0.83, whereas the PPV for the algorithm was 0.89. Calculating the PPVs for both clinicians in reference to the algorithm yielded 0.89 for both the neurosurgeon and neurologist. We also calculated Cohen’s Kappa to measure interrater reliability against chance performance, and report a value of 44.9%, which is similar to kappa values (48.6%) reported in a large multicenter study of IED interrater reliability (Jing et al., 2020). Cohen’s Kappa between the algorithm and Neurologist was 61.1%. These results are reported in the Results section on page 4, lines 16-20, and the method of validation is described in the methods on page 13, starting on line 8.

2) You write in the abstract and discussion that IEDs traverse the same path as ictal discharges, but the angle between the SDs and IEDs was 24 degrees, and the IEDs weren't exactly opposite the ictal wavefront (150 degrees). Do you think the SDs and IEDs really follow the same path, and if so, what accounts for the sizable angle between their observed paths?

In testing the hypotheses that arose from the model predictions, we assumed that a tolerance of approximately π/4 would be appropriate to report that IEDs and SDs were similar directions, as we showed there is not a directly antipodal relationship between pre and post -recruitment SDs (Smith et al., 2016). However, the reviewer is correct to point out that such a heuristic is insufficient. We therefore carried out statistical tests to determine the extent to which the similarity of these directions might arise by chance. This test relied on comparing the true KLD between IED and SD distributions to a permutation distribution comprised of randomly sampled directions from both distributions. This test is described in the methods on page 16, starting on line 15. All of these tests were highly significant (all p < 0.01), indicating that while the IED/SD distributional overlap is not perfect, it is significantly unlikely to occur by chance in each patient. We have reported these results on page 8, on lines 9-12.

It's also important to note that the mean angle differences the reviewer points to represent those across participants, where two patients had larger differences, but the majority had mean angle differences near zero, as well as mean IED angle differences from the IW direction near 180 (Figures 4 G-H). This many similar angle differences, across all but 2 patients (in which the angle is only slightly larger) is highly unlikely to occur by chance. The two patients with larger angles also were classified as MTL onset, which we have mentioned in the final paragraph of the discussion (Page 11, sentence starting on line 45).

Reviewer #2 (Recommendations for the authors):

1. The terminology used to describe the various types of epileptiform activity described in this paper should be laid out in a clear way from the beginning. Although many of the features are defined in prior work, they are critical to understanding the results. Specifically, the introduction of SDs focuses on their propagation following the IW, but "pre-recruitment" SDs are a prominently analyzed group in Figure 5. It would be helpful to include a schematic figure like that at the top left of 6A earlier in the paper. It may also be useful to define two types of SDs (anterograde and retrograde?), and include them on the schematic figure. Perhaps "recruited" and "penumbral" could be illustrated here as well (or in a separate subpanel).

We have edited the second paragraph of the introduction to make this clearer, and have added a supplementary figure (Figure 1—figure supplement 1) similar to that in (Smith et al., 2020) in order to clarify the seizure domains of interest. In that paragraph, we have added quotations to bring the reader’s attention to the pre-recruitment and post-recruitment discharge terms that correspond to the reviewer’s idea of anterograde and retrograde discharges. We showed previously that discharges that precede the wavefront largely travel in the same direction as the IW, discharges that follow the IW largely travel antipodally (Smith et al., 2016).

2. The prominent clinical relevance of the analysis presented in this manuscript is that IEDs could be used to locate the seizure focus in pre-surgical planning. As briefly indicated in the discussion, the UEA is likely too small to localize a seizure focus – it produces one vector, which at best could define the line along which the focus falls. To identify the seizure focus, greater spatial sampling is needed. The authors point out that the simultaneously recorded ECOG is insufficient to measure IW propagation as it cannot detect MUAs and its poor spatial sampling make estimation of IED propagation low-fidelity. Nevertheless, since LFP was used to record IED propagation in the UEA, it would be useful to analyze IED propagation from the ECoG array to compare the vector produced, particularly by neighboring electrodes.

While we agree with multiple elements of this critique, there are numerous reasons why we believe that a full analysis of IED propagation patterns in ECoG is beyond the scope of this study. A primary issue is that the brain sampled by a planar Utah array can be readily mapped to 2-dimensional space, as in this study. However, the brain beneath a grid or sampled by SEEG is not flat—the sampling is no longer along a rigid 2-dimensional plane, given the brain’s gyri and sulci, so Euclidean distances are not an appropriate metric for relating recording locations to brain activity. We are currently working to understand mesoscale patterns in IED propagation on sEEG and ECoG using methods that are more appropriate for: (1) the larger spatial scale, (2) complex 3D patterns of IED travel involved with sEEG/ECoG (e.g. travel across Euclidean, geodesic, or axonal lengths), and (3) more complicated IED waveforms recorded on ECoG and sEEG. There are further methodological uncertainties inherent in spatiotemporal alignment between ECoG and UEA IED propagation vectors that made us uncomfortable with making claims about the correspondence between ECoG and MEA IED propagation. Temporally, alignment was not carried out similarly across all 10 participants (given the different recording locations and approximately a decade between some recordings). Spatially, it is nearly impossible to precisely align the MEA and ECoG from intraoperative photos and CT/MRI co-registration due to ‘brain drop’ or potential spatial translation/rotation of ECoG grids after surgery. Physical memory in the UEA wire bundle could add systematic bias to angle estimation from CT/MR co-registration.

Nevertheless, we have included supplementary qualitative information about the location of grids, arrays, and IED vectors for reviewers to judge. Figure 4—figure supplement 2 shows electrode localization information for each participant, a schematic of the implant around the UEA, and rotated IED distributions to match each UEA orientation. While this is all qualitative information, as we cannot precisely align IED distributions based on intraoperative photos, the axis of IED travel bisects the SOZ in most cases.

Furthermore, we have done some preliminary analysis of peri-ictal IEDs on ECoG for one of the penumbral participants, and can share one patient’s results in Author response image 1.

Author response image 1
Download asset Open asset
For one participant whose UEA was not recruited into the seizure examined gradients in the beta phase across the ECoG grid space during the IEDs preceding seizures.

We evaluated the divergence of these gradients in order to determine sources and sinks in patterns of IED traveling waves, hypothesizing that the location of the ictal core will be a sink.

This figure shows that there is a ‘sink’ IED propagation pattern across an ECoG grid that co-locates with the SOZ. In upcoming work, we aim to translate these methodologies to sEEG, which is more commonly used in contemporary clinical practice.

Finally, while we recognize the importance of the clinical implications of this work, we would categorize this study as basic science with great relevance for a particular clinical population. We believe it is essential to translate the understandings in this manuscript to a clinically useful scale, and are taking steps to do just that to be reported in future publications. We have added more text to the discussion to clarify the directions that could be taken in order to improve patients’ lives.

3. If the ECoG analysis fails to produce a vector that match the UEA vectors, it may be necessary to implant two UEAs. The intersection of the IED propagation vectors would hypothetically localize the seizure focus (in 2D). Obviously this experiment is outside the scope of this study, but worth mentioning in the discussion.

This is an astute observation by the reviewer. We would refer the reviewer to the schematics on the previous page showing the axis of IED travel bisects the SOZ. We have also added text to the last paragraph of the discussion mentioning the considerations involved with estimating seizure spread from IEDs recorded on MEAs on page 11, line 28.

4. Did the resulting surgical resections in recorded patients align with the axis predicted by the IED propagation at the UEA? Did the success of the surgery relate to the removal (or not) of the UEA-predicted focus?

Appendix 1 – Table 1 shows all the clinical details for each participant. While information about outcomes is in the table there was insufficient statistical power to carry out across subjects analyses (power = 0.14 for a McNemar test). Three patients had outcomes that were Engle 3/4. Arrays in each of those patients were near the SOZ and angles of IED travel bisected the SOZ in each case, however it must be reiterated that it is difficult to infer much from these maps due to uncertainty in spatial alignment of arrays and classification of the SOZ.

5. Synaptic connectivity is probably important determinant of whether vector points to focus. Furthermore, UEA recordings are performed on the cortical surface. In the most straightforward application of the proposed strategy, seizures would start on cortical surface and propagate locally. However, many seizures begin in deep structures. Some discussion of the relationship between neocortical propagation and localization of a deep focus (e.g. MTLE) is warranted.

The third reviewer also suggested we discuss the implications of mesial temporal lobe onsets, so we have added text to the discussion on page 11, starting on line 36 discussing this point directly.

6. Were patients weaned from their AEDs during these recordings. One important advantage of analyzing IEDs vs seizures for surgical planning is that they are very common. Thus, it might be possible to perform the mapping in patients without having them in withdrawal from their AEDs, which might improve predictive value.

Patients were weaned from their AEDs during this study, as dictated by each patient’s clinical course. The IED rate reported in this study may therefore reflect a slightly higher rate than would be expected when patients are fully on AEDs, however the analysis period also includes earlier segments of data that are within the period of IED weaning. We have added a relevant sentence to the methods on page 12, line 25.

7. In Figure 4A-F, for the subjects with multiple seizures, the IED distribution appears to be different for each seizure. What IED distributions are plotted – only those that had occurred up to the time of that seizure? Maybe I missed it, but I couldn't find this info in the results or figure caption…

We thank the reviewer for the opportunity to clarify this point. The figure caption states that each distribution is plotted relative to the direction of the ictal wavefront (IW), which is why all of the IWs point to zero. This is to allow for comparison between IEDs and SDs for each seizure, relative to each seizure’s predominant geometrically organizing feature: the IW. Raw IED and SD distributions, and raw IW directions are plotted in Figure 4—figure supplement 1. We have edited the figure caption for 4A-F in order to clarify both points.

Reviewer #3 (Recommendations for the authors):

The Utah array is an excellent tool to study spatiotemporal dynamics in and around a seizure core. The strength of such a technique is unfortunately also the weakness. Since it is implanted on the cortex, the modeling of IED and seizure discharges has a potential for bias due to the location of the implant. The model can be strengthened by comparing the MUA and LFP of the model with their single-unit data of IEDs from hybrid electrodes of mesial temporal lobe implants. The authors do mention the limitation of their study, but it would help if they discuss this some more, especially since only two patients in Table S1 had mesial temporal onset and the rest were lateral or neocortical onsets. They should not generalize that this model is applicable to all seizure types.

We thank the reviewer for the recommendations and have expanded discussion of these points in the final discussion paragraph. Another reviewer also asked us to discuss non-neocortical onset seizures, so we have discussed this on page 11, starting on line 36.

References:

Jing J, Herlopian A, Karakis I, Ng M, Halford JJ, Lam A, Maus D, Chan F, Dolatshahi M, Muniz CF, Chu C, Sacca V, Pathmanathan J, Ge W, Sun H, Dauwels J, Cole AJ, Hoch DB, Cash SS, Westover MB. 2020. Interrater Reliability of Experts in Identifying Interictal Epileptiform Discharges in Electroencephalograms. JAMA Neurology 77:49–57. doi:10.1001/jamaneurol.2019.3531

Liou J, Smith EH, Bateman LM, Bruce SL, McKhann GM, Goodman RR, Emerson RG, Schevon CA, Abbott L. 2020. A model for focal seizure onset, propagation, evolution, and progression. eLife 9:e50927.

Liou J-Y, Baird-Daniel E, Zhao M, Daniel A, Schevon CA, Ma H, Schwartz TH. 2019. Burst suppression uncovers rapid widespread alterations in network excitability caused by an acute seizure focus. Brain 142:3045–3058. doi:10.1093/brain/awz246

Schevon CA, Weiss SA, McKhann G, Goodman RR, Yuste R, Emerson RG, Trevelyan AJ. 2012. Evidence of an inhibitory restraint of seizure activity in humans. Nat Commun 3:1060. doi:10.1038/ncomms2056

Smith EH, Liou J, Davis TS, Merricks EM, Kellis SS, Weiss SA, Greger B, House PA, McKhann II GM, Goodman RR. 2016. The ictal wavefront is the spatiotemporal source of discharges during spontaneous human seizures. Nature communications 7:1–12.

Smith EH, Merricks EM, Liou J-Y, Casadei C, Melloni L, Thesen T, Friedman DJ, Doyle WK, Emerson RG, Goodman RR, McKhann GM, Sheth SA, Rolston JD, Schevon CA. 2020. Dual mechanisms of ictal high frequency oscillations in human rhythmic onset seizures. Scientific Reports 10:19166. doi:10.1038/s41598-020-76138-7

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Essential revisions:

1) Multiple reviewers raised the question regarding "the sizable angle between… SDs and IEDs" and the fact that the "IEDs weren't exactly opposite the ictal wavefront". You state that "the angle discrepancies to which the reviewer refers are statistically insignificant and can be explained by expected biological variability". On the other hand you state that based on Smith et al. 2016, you don't expect the IED to be strictly antipodal to IW. Finally, while the KLD analysis is useful it does not seem to fully address the reviewers' questions regarding what appears to be a (statistically) significant deviation (almost 2 standard deviations) of the IED/IW angle from 180 degrees. A single paragraph in the discussion summarizing and clarifying the various points of why SD, IEDs, and IWs may not be strictly aligned or antipodal would improve the manuscript.

We thank the reviewers for engaging with this issue. We have added an additional paragraph to the discussion to address this issue directly. The new paragraph begins on page 10, line 30. and is pasted here:

“The IED directions we report are not perfectly antipodal to the IW. These differences in mean angles may be due in part to error when fitting a plane to the IW, which discounts the propagating wave’s curvature, or biological variability, such as changes in brain state and arousal. […] The dominant direction of travel is largely in the opposite direction of seizure expansion.”

2) Regarding Reviewer 2, major point #7 "In Figure 4A-F, for the subjects with multiple seizures, the IED distribution appears to be different for each seizure. What IED distributions are plotted – only those that had occurred up to the time of that seizure?"

The reviewer is asking why the shape of the distribution, and not just the rotation, of IED distribution varies for different seizures in the same patient. Are you analyzing/including a different subset of IEDs for each seizure? If so, why? If not, why does the distribution change?

The same IEDs go into each plot, but the IED directions are rotated by the ictal wavefront for each seizure. The reason they look slightly different is because all the rotated angles are fitting into the same 18 bins as the original data, so the differences in angles per bin very slightly changes the shape of the distribution. For example in figure 4E, the larger IED bin closer to zero in the right panel appears to get split into two smaller bins in the left panel. It may help the reviewer to compare the IED distributions in figure 4 to the raw distributions that are shown in the figure supplement.

3) Regarding Reviewer 2, major point #2, we strongly suggest including the figure showing the "qualitative information about the location of grids, arrays, and IED vectors". We realize that there are a number of caveats and limitations in the interpretation of this qualitative figure that require further explanation, but including the figure would greatly enhance the clarity of the manuscript.

We agree that this figure does aid the manuscript clarity, despite its many uncertainties. We have included an edited version as Figure 4—figure supplement 2.

4) Please refer to https://submit.elifesciences.org/html/eLife_author_instructions.html#revised for instructions on how to submit source data for each figure. There do not seem to be source data included for the majority of figure plots. Also the data repository on OSF would greatly benefit from a description of the different data file types, the different variable types within the structures, and the sampling rate for the IED traces.

We apologize for this omission. We were under the impression that the uploaded code and OSF database were sufficient to reproduce figures. We have remedied this misunderstanding by uploading source data (IED directions) for figures 2 and 4, which are the main data figures in the manuscript to the github repository. We have also added a description of the data types in the OSF repository.

https://doi.org/10.7554/eLife.73541.sa2

Article and author information

Author details

  1. Elliot H Smith

    1. Departments of Neurosurgery and Biomedical Engineering, University of Utah, Salt Lake City, United States
    2. Department of Neurology, Columbia University, New York, United States
    Contribution
    Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Software, Supervision, Validation, Visualization, Writing - original draft, Writing – review and editing
    For correspondence
    e.h.smith@utah.edu
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-4323-4643

  2. Jyun-you Liou

    Department of Anesthesiology, Weill Cornell Medicine, New York CIty, United States
    Contribution
    Conceptualization, Investigation, Methodology, Project administration, Writing – review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-4851-3676

  3. Edward M Merricks

    Department of Neurology, Columbia University, New York, United States
    Contribution
    Investigation, Methodology, Resources, Writing – review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-8172-3152

  4. Tyler Davis

    Departments of Neurosurgery and Biomedical Engineering, University of Utah, Salt Lake City, United States
    Contribution
    Data curation, Investigation, Resources, Writing – review and editing
    Competing interests
    No competing interests declared

  5. Kyle Thomson

    Department of Pharmacology & Toxicology, University of Utah, Salt Lake City, United States
    Contribution
    Conceptualization, Funding acquisition, Investigation, Methodology, Resources
    Competing interests
    No competing interests declared

  6. Bradley Greger

    Department of Bioengineering, Arizona State University, Tempe, United States
    Contribution
    Data curation, Resources
    Competing interests
    No competing interests declared

  7. Paul House

    Neurosurgical Associates, LLC, Murray, United States
    Contribution
    Data curation, Resources
    Competing interests
    is affiliated with Neurosurgical Associates, LLC. The author has no financial interests to declare

  8. Ronald G Emerson

    Hospital for Special Surgery, New York, United States
    Contribution
    Resources
    Competing interests
    No competing interests declared

  9. Robert Goodman

    Lenox Hill Hospital, New York, United States
    Contribution
    Resources
    Competing interests
    No competing interests declared

  10. Guy M McKhann

    Department of Neurological Surgery, Columbia University Medical Center, New York, United States
    Contribution
    Resources
    Competing interests
    reports fees from Koh Young, Inc

  11. Sameer Sheth

    Department of Neurological Surgery, Baylor College of Medicine, Houston, United States
    Contribution
    Investigation, Resources, Writing – review and editing
    Competing interests
    consulting for Boston Scientific, Abbott, Neuropace, Zimmer Biomet

  12. Catherine Schevon

    Department of Neurology, Columbia University, New York, United States
    Contribution
    Conceptualization, Data curation, Funding acquisition, Investigation, Methodology, Project administration, Resources, Software, Supervision, Validation, Writing – review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-4485-7933

  13. John D Rolston

    Departments of Neurosurgery and Biomedical Engineering, University of Utah, Salt Lake City, United States
    Contribution
    Conceptualization, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Writing – review and editing
    For correspondence
    john.rolston@utah.edu
    Competing interests
    reports fees from Medtronic, Inc
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-8843-5468

Funding

National Institutes of Health (NINDS R21 NS113031)

  • Elliot H Smith
  • Catherine Schevon
  • John Rolston

National Institutes of Health (NINDS K23 NS114178)

  • John Rolston

National Institutes of Health (S10 OD018211)

  • Catherine Schevon

National Institutes of Health (R01 NS084142)

  • Catherine Schevon

American Epilepsy Society (JIA)

  • Elliot H Smith

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Ethics

Human subjects: The Institutional Review Boards at the University of Utah (IRB_00114691) and Columbia University Medical Center (IRB- AAAB6324) approved these studies. All participants provided informed consent prior to surgery for implantation of the clinical and research electrodes.

Senior Editor

  1. John R Huguenard, Stanford University School of Medicine, United States

Reviewing Editor

  1. Matthew Shtrahman, University of California, San Diego, United States

Reviewer

  1. Brian Litt

Publication history

  1. Preprint posted: April 29, 2021 (view preprint)
  2. Received: September 2, 2021
  3. Accepted: January 19, 2022
  4. Accepted Manuscript published: January 20, 2022 (version 1)
  5. Accepted Manuscript updated: January 24, 2022 (version 2)
  6. Version of Record published: February 3, 2022 (version 3)

Copyright

© 2022, Smith et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 3,917
    Page views
  • 385
    Downloads
  • 9
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Elliot H Smith
  2. Jyun-you Liou
  3. Edward M Merricks
  4. Tyler Davis
  5. Kyle Thomson
  6. Bradley Greger
  7. Paul House
  8. Ronald G Emerson
  9. Robert Goodman
  10. Guy M McKhann
  11. Sameer Sheth
  12. Catherine Schevon
  13. John D Rolston
(2022)
Human interictal epileptiform discharges are bidirectional traveling waves echoing ictal discharges
eLife 11:e73541.
https://doi.org/10.7554/eLife.73541

Categories and tags

Research organism

Further reading

    1. Computational and Systems Biology

    Shared enhancer gene regulatory networks between wound and oncogenic programs

    Swann Floc'hlay, Ramya Balaji ... Stein Aerts
    Research Article Updated

    Wound response programs are often activated during neoplastic growth in tumors. In both wound repair and tumor growth, cells respond to acute stress and balance the activation of multiple programs, including apoptosis, proliferation, and cell migration. Central to those responses are the activation of the JNK/MAPK and JAK/STAT signaling pathways. Yet, to what extent these signaling cascades interact at the cis-regulatory level and how they orchestrate different regulatory and phenotypic responses is still unclear. Here, we aim to characterize the regulatory states that emerge and cooperate in the wound response, using the Drosophila melanogaster wing disc as a model system, and compare these with cancer cell states induced by rasV12scrib-/- in the eye disc. We used single-cell multiome profiling to derive enhancer gene regulatory networks (eGRNs) by integrating chromatin accessibility and gene expression signals. We identify a ‘proliferative’ eGRN, active in the majority of wounded cells and controlled by AP-1 and STAT. In a smaller, but distinct population of wound cells, a ‘senescent’ eGRN is activated and driven by C/EBP-like transcription factors (Irbp18, Xrp1, Slow border, and Vrille) and Scalloped. These two eGRN signatures are found to be active in tumor cells at both gene expression and chromatin accessibility levels. Our single-cell multiome and eGRNs resource offers an in-depth characterization of the senescence markers, together with a new perspective on the shared gene regulatory programs acting during wound response and oncogenesis.

    1. Cancer Biology
    2. Computational and Systems Biology

    Comprehensive characterization of tumor microenvironment in colorectal cancer via molecular analysis

    Xiangkun Wu, Hong Yan ... Li Liang
    Research Article

    Colorectal cancer (CRC) remains a challenging and deadly disease with high tumor microenvironment (TME) heterogeneity. Using an integrative multi-omics analysis and artificial intelligence-enabled spatial analysis of whole-slide images, we performed a comprehensive characterization of TME in colorectal cancer (CCCRC). CRC samples were classified into four CCCRC subtypes with distinct TME features, namely, C1 as the proliferative subtype with low immunogenicity; C2 as the immunosuppressed subtype with the terminally exhausted immune characteristics; C3 as the immune-excluded subtype with the distinct upregulation of stromal components and a lack of T cell infiltration in the tumor core; and C4 as the immunomodulatory subtype with the remarkable upregulation of anti-tumor immune components. The four CCCRC subtypes had distinct histopathologic and molecular characteristics, therapeutic efficacy, and prognosis. We found that the C1 subtype may be suitable for chemotherapy and cetuximab, the C2 subtype may benefit from a combination of chemotherapy and bevacizumab, the C3 subtype has increased sensitivity to the WNT pathway inhibitor WIKI4, and the C4 subtype is a potential candidate for immune checkpoint blockade treatment. Importantly, we established a simple gene classifier for accurate identification of each CCCRC subtype. Collectively our integrative analysis ultimately established a holistic framework to thoroughly dissect the TME of CRC, and the CCCRC classification system with high biological interpretability may contribute to biomarker discovery and future clinical trial design.

    1. Computational and Systems Biology
    2. Neuroscience

    Flexible control of representational dynamics in a disinhibition-based model of decision making

    Bo Shen, Kenway Louie, Paul W Glimcher
    Research Article

    Inhibition is crucial for brain function, regulating network activity by balancing excitation and implementing gain control. Recent evidence suggests that beyond simply inhibiting excitatory activity, inhibitory neurons can also shape circuit function through disinhibition. While disinhibitory circuit motifs have been implicated in cognitive processes including learning, attentional selection, and input gating, the role of disinhibition is largely unexplored in the study of decision-making. Here, we show that disinhibition provides a simple circuit motif for fast, dynamic control of network state and function. This dynamic control allows a disinhibition-based decision model to reproduce both value normalization and winner-take-all dynamics, the two central features of neurobiological decision-making captured in separate existing models with distinct circuit motifs. In addition, the disinhibition model exhibits flexible attractor dynamics consistent with different forms of persistent activity seen in working memory. Fitting the model to empirical data shows it captures well both the neurophysiological dynamics of value coding and psychometric choice behavior. Furthermore, the biological basis of disinhibition provides a simple mechanism for flexible top-down control of the network states, enabling the circuit to capture diverse task-dependent neural dynamics. These results suggest a biologically plausible unifying mechanism for decision-making and emphasize the importance of local disinhibition in neural processing.