Microscale Physiological Events on the Human Cortical Surface

Abstract

Despite ongoing advances in our understanding of local single-cellular and network-level activity of neuronal populations in the human brain, extraordinarily little is known about their "intermediate" microscale local circuit dynamics. Here, we utilized ultra-high-density microelectrode arrays and a rare opportunity to perform intracranial recordings across multiple cortical areas in human participants to discover three distinct classes of cortical activity that are not locked to ongoing natural brain rhythmic activity. The first included fast waveforms similar to extracellular single-unit activity. The other two types were discrete events with slower waveform dynamics and were found preferentially in upper cortical layers. These second and third types were also observed in rodents, nonhuman primates, and semi-chronic recordings from humans via laminar and Utah array microelectrodes. The rates of all three events were selectively modulated by auditory and electrical stimuli, pharmacological manipulation, and cold saline application and had small causal co-occurrences. These results suggest that the proper combination of high-resolution microelectrodes and analytic techniques can capture neuronal dynamics that lay between somatic action potentials and aggregate population activity. Understanding intermediate microscale dynamics in relation to single-cell and network dynamics may reveal important details about activity in the full cortical circuit.

Keywords: auditory stimulation; electrical stimulation; extracellular activity; human cortex; microelectrode.

Figure 1

PEDOT:PSS recording setup, fast events, and spectrum. (A) PEDOT:PSS 2-column 128-channel grid electrode layout (two rows of 64 channels separated by 50 μm) and dimensions. (B) Example of relative positioning. The larger orange circles and squares on the electrode array indicate large PEDOT contacts to orient the small electrode contacts (the 2-column grid, white arrowhead) between the two large squares. The adjacent clinical electrode is seen in blue (with blue arrowhead). The inset square is a zoomed-in view of the PEDOT:PSS electrode. (C) Photograph of microelectrode (white arrowhead) and clinical array (blue arrowhead) on the human cortical surface alongside reference electrodes (white and purple wires). (D) Example broadband recordings, from three participants (IP03, IP15, and IP18), of clinical leads (blue lines) and the PEDOT:PSS grid (black lines, bottom), with unitary events (red arrowheads, grid middle, and right traces). The geometry of the PEDOT:PSS grid in all three recordings was a 128-channel 2-column grid extending 3150 μm × 50 μm with a 50-μm pitch between sites. The time scale traces in D and E are the same. (E) Normalized power spectrograms of the simultaneous individual clinical and PEDOT:PSS recordings and one same-channel trace shown in D. Normalization was done by dividing each time step in power per frequency band by the mean power across all time windows per frequency band. Each LFP column traces in D match the spectrograms shown in E. (F) Individual average power spectra curves for PEDOT:PSS (black) and clinical (blue) channels for IP03, IP15, and IP18. Inset: Average values across participants (N = 14). Green boxes in D and E highlight similar dynamics between the clinical (top row in D and E separately) and PEDOT:PSS recordings (bottom row in D and E separately). Shaded regions indicate SEM.

Figure 2

Fast Type 1 events recorded from the surface of the human cortex. (A, i–iii) Example recording from participant IP37. In ii, the events in colored boxes from i are shown at higher resolution. (B, C) Example fast waveforms from the same participant, mapped to the circular grid locations across the PEDOT:PSS microelectrode array. The waveforms are color-coded to correspond to different clusters as determined in Kilosort using waveforms’ spatial locations on the PEDOT:PSS grids combined with spike times in a model (see Materials and Methods). The gray dots indicate good channels with low impedance on the circular PEDOT:PSS array. (D–F) Example recordings from the same clusters as in B and C shown as a raster plot across the microelectrode grid, as IEIs (E), and as auto-correlograms (F) for a subset of clusters with different time scales at ±0.2 s (left plot) and ±0.01 s (right plot). (G, H). Comparisons of half-peak widths versus event frequencies (spike rate; H) and half-peak widths versus trough-to-peak amplitudes (G) of the sorted waveforms across all participants. N = 24. Error bars indicate SEM.

Figure 3

Classifiable events can be identified across recordings and participants. (A) Recording from participant IP25, mapped to the circular grid. Waveform color-coding matches channel color-coding with cyan to magenta (channels 1–64) and green to yellow (channels 65–128) indicating the four separate 32-channel amplifier banks used in the recording (Bank A: 1–32; Bank B: 33–64; Bank C: 65–96; Bank D: 97–128). Only the channels that fulfilled the criteria are shown as nongray dots. (B) Example recording (electrodes that fulfilled criteria represented as nongray dots) from participant IP15 showing activity changes in time, mapped to the 2-column electrode array. (C) Zoomed view of the red box in B showing Type 2 and 3 events. (D) Example recordings of events at different time points (for IP15) showing only a subset of electrode channels. Repeated events through time for two different event examples are shown (orange and blue ovals). (E) Example recordings (IP11) of a detected IID, indicated by a green rectangle, along with Type 2 and 3 events (circled in gray). (F, G). Stepwise criteria to detect Type 2 and 3 events, with template waveforms shown in G. (1) The first step detected absolute peaks that are greater than 25 μV, as shown in example traces from a single channel and case (IP15). Gray vertical lines denote detected cross-threshold peaks, green lines the thresholds. Red arrows show the events that fulfill all four selection criteria. (2) The next step detected peaks that correlate with the templates ≥0.80. (3) Selected events had a second derivative greater than 2 in the time period at onset, indicated by the shaded gray box. (4) Finally, waves were kept if they had no large deflections before event onset, that is, within 100 ms before crossing the 25-μV threshold, indicated by the green boxes. (H) Average waveforms per participant (N = 36) for the different event classes. The dotted lines indicate the zero-crossing value, in microvolts, to illustrate the number of positively (two left columns) and negatively (two right columns) deflecting detected waveforms. All voltage data shown were low-pass filtered using the high cut-off frequency of 500 Hz.

Figure 4

Temporal and spatial spread of Type 2 and 3 events. (A) Example recording from participant IP17 illustrating methods for determining event spatial spread (area). (B, i–iv) In cases where multiple channels detected one event, we compared the detected waveforms across channels, area covered (μm²), speed of event propagation across channels (m/s), and distance traveled across the electrode array (μm). Box plots and confidence intervals around the mean values per measure per participant are indicated by whisker bars; center line indicates median, N = 36. (C) Event time span (s). White dots are the 2-column grids, while black dots are the circular grid recordings per participant. Asterisks indicate P < 0.05, Wilcoxon rank-sum test. (D) Images of the circular PEDOT:PSS grid electrode (D, i, ii), Utah array (Diii–iv), implanted electrode (arrowhead) overlaid by a clinical grid (Div), the implanted electrode (Dv), and a laminar array electrode (Dvi). (E, i–iii) Top: diagrams of the clinical surface electrodes, PEDOT:PSS array, Utah laminar arrays (1.0- and 1.5-mm depths), and laminar arrays showing relative sampling geometry of the different systems. Bottom: (E, i) Significantly more Type 2 and 3 events per second occur in the PEDOT:PSS array recordings than the clinical recordings (P < 0.001; N = 9; Wilcoxon rank-sum test). (E, ii) Significantly more Type 2 and 3 events are detected per second in the shallower Utah array recordings than the deeper recordings (P < 0.001; N = 4 each Utah array depth; Wilcoxon rank-sum test). (E, iii) More Type 3 events are detected per second in superficial laminar contacts as evidenced by the average negative correlation between the channel number and Type 3 counts across the array (rho = −0.48 ± 0.38; significantly different from zero; P = 0.0313; Wilcoxon signed rank test; N = 9). (F) Example average detected Type 2 and 3 events per participant with n = number of detected events. The dotted lines indicate the zero-crossing value in microvolts to illustrate the number of positively and negatively deflecting detected waveforms. In B, C, and E, the number of events were first averaged per contact and then averaged per patient. Each dot in B, C, and E represents this average for an individual participant.

Figure 5

Manipulations alter rates of Type 1–3 events. (A) Effects of auditory stimuli on Type 2 and 3 events. Left: Recording from participant IP22 showing neural dynamics during repeated auditory stimuli in the averaged evoked potentials across trials. Right: Raster plots of Type 3 events occurring across all trials with auditory stimulus trial times indicated by blue vertical lines. (B) Left: PSTHs of each event type binned per 25 ms and then averaged across all patients (N = 11). For the line plots, shaded error regions are all SEM. (C) Effects of pro-convulsant medication and cold saline on Type 1–3 events and IIDs. PSTH of log binned activity following the administration of pro-convulsant medication (methohexital or alfentanil; left) or cold saline application (right) averaged across patients. Type 1 (top), Type 2 (middle), and Type 3 events (middle) as well as IIDs (bottom) were binned every 100 ms and normalized relative to baseline by dividing each bin by the average baseline activity. The dotted red line is the baseline normalization value. (D) Histogram of activity changes with 5-min bins. Error bars are SEM, N = 9 for pro-convulsant medication, N = 8 for cold saline application. Asterisks indicate significant difference from baseline with P < 0.001, Wilcoxon rank-sum test. (E) Electrical stimulation changes in the rate of Type 1–3 events. Recording from participant IP07 showing a wave of activity (Type 3 event) following a train of stimuli delivered at 60 Hz for 1–2 s during clinical mapping, with the channels laid out along a single line (left) and time-voltage curves mapped relative to the 2-column grid electrode layout (activity in space, right). Blue and orange arrowheads indicate the same pair of channels as in the left time-voltage plots versus the right time-voltage plots relative to the electrode layout. The red arrowhead indicates the direction of stimulation relative to the electrode layout, though the bipolar stimulation was 1–2 cm away. (F) PSTHs of binned activity around the time of stimulation (bar below figure) averaged across patients (N = 9, n > 5 trials per participant). Type 1 (top), Type 2 (middle), and Type 3 (bottom) events were binned every 100 ms and normalized relative to 1 s before stimulation. The dotted red line is the normalized baseline value. Note: the y-axis is in log-scaled increments. (G) Histograms of activity in 0.5-s bins before and after stimulation. P-values reflect differences between each 0.5-s bin per event type, N = 9; asterisks indicate significant difference from baseline with P < 0.001, Wilcoxon rank-sum test. Error bars are all SEM.

Figure 6

Temporal relationships between event Types. (A, B) Concurrent Type 1, 2, and 3 events across participants and channels in the simultaneous high-pass filtered data (between 250 and 6000 Hz, black lines) and the low-pass filtered data (at 500 Hz, gray lines). Type 1 events are circled in red. In A, Type 1 events are isolated from Type 2 and 3, and in B, a Type 1 event is on a different channel than the Type 2 event. (C) Left: Average cross-covariance plots between Type 1, 2 (gray line), and 3 (black line) events occurring simultaneously on different channels (N = 24). Right: Average cross-covariance plots between Type 2 and 3 event occurring simultaneously on different neighboring channels (N = 36). Shaded lines are standard error. (D) Average covariance from the 50 ms before each event to same interval after the event. Asterisks indicate P < 0.001; N = 24; Wilcoxon rank-sum test. Error bars are SEM. (E) Binned maximum time lag between the Type 1 event and the Type 2 and 3 events for all human recordings (two right plots) and the binned maximum time lag between the Type 2 and 3 events. (F) Median maximum lag in covariance between Types 1 and 2 (gray box plot), Types 1 and 3 (black box plot), and Types 2 and 3 with confidence bounds. Asterisk indicates that distribution differs significantly from zero, P = 0.0004. Wilcoxon signed rank test.