Columnar Localization and Laminar Origin of Cortical Surface Electrical Potentials

Abstract

Electrocorticography (ECoG) methodologically bridges basic neuroscience and understanding of human brains in health and disease. However, the localization of ECoG signals across the surface of the brain and the spatial distribution of their generating neuronal sources are poorly understood. To address this gap, we recorded from rat auditory cortex using customized μECoG, and simulated cortical surface electrical potentials with a full-scale, biophysically detailed cortical column model. Experimentally, μECoG-derived auditory representations were tonotopically organized and signals were anisotropically localized to less than or equal to ±200 μm, that is, a single cortical column. Biophysical simulations reproduce experimental findings and indicate that neurons in cortical layers V and VI contribute ∼85% of evoked high-gamma signal recorded at the surface. Cell number and synchrony were the primary biophysical properties determining laminar contributions to evoked μECoG signals, whereas distance was only a minimal factor. Thus, evoked μECoG signals primarily originate from neurons in the infragranular layers of a single cortical column.SIGNIFICANCE STATEMENT ECoG methodologically bridges basic neuroscience and understanding of human brains in health and disease. However, the localization of ECoG signals across the surface of the brain and the spatial distribution of their generating neuronal sources are poorly understood. We investigated the localization and origins of sensory-evoked ECoG responses. We experimentally found that ECoG responses were anisotropically localized to a cortical column. Biophysically detailed simulations revealed that neurons in layers V and VI were the primary sources of evoked ECoG responses. These results indicate that evoked ECoG high-gamma responses are primarily generated by the population spike rate of pyramidal neurons in layers V and VI of single cortical columns and highlight the possibility of understanding how microscopic sources produce mesoscale signals.

Keywords: auditory cortex; biophysical simulation; cortical column; neurophysiology; origins of ECoG.

Figure 1.

Stimulus-evoked cortical surface electrical potentials exhibit large peaks in the high-gamma range. a, Photomicrograph of an 8 × 16 μECoG grid (pitch, 200 μm; contact diameter, 40 μm) on the surface of rat A1. b, Top, Tone stimulus played during experimental recordings. Middle, z-Scored spectral decomposition of single-trial evoked cortical surface electrical potentials from a single electrode. Bottom, High-gamma component of single-trial evoked cortical surface electrical potentials indicated by horizontal dashed lines (middle). c, Trial-averaged evoked cortical surface electrical potential on one μECoG electrode in response to presentations of the best tuned frequency of that electrode. d, Trial-averaged neural spectrogram for the electrode shown in c in response to presentations of its best tuned frequency. Dashed vertical lines in c and d represent stimulus onset and offset. Red vertical lines in c and d correspond to the time window of extracted evoked response used for subsequent analysis. e, Grand-average (mean ± SE) z-scored amplitude as a function of frequency across all tuned electrodes (N = 333).

Figure 2.

Robust frequency tuning and high-resolution tonotopic maps from μECoG. a, b, FRA surfaces recorded from a μECoG array. Subplots correspond to responses of a single electrode and are organized according to electrode position on the grid/brain. In each subplot, pixels correspond to a stimulus frequency-intensity pairing and are colored according to the mean evoked z-score; Ηγ (a), tMuA (b). c, High-resolution tonotopic organization of multiple auditory cortical fields derived from Ηγ activity. Each pixel is color coded according to the best frequency of that electrode. The 8 × 16 μECoG array displayed here covered multiple auditory cortical fields (A1, PAF, and VAF) and the approximate boundaries are demarcated (black lines). d, Differential tuning at neighboring electrodes. FRAs are plotted for four electrodes (numbered as in c) and show that neighboring electrodes (1 vs 2; 3 vs 4) can have different response properties. e, f, Average normalized response surface for all electrodes with significantly tuned Ηγ (N = 333) and tMuA (N = 113) auditory responses. White line in each plot demarcates the FRA response boundaries. g, Across all tuned electrodes, the average (mean ± SE) FRA response boundaries for CSEP components (demarcated by colors) where similar. h, i, Distributions (25th–50th–75th percentiles) of best-frequencies (h) and bandwidths (i) for all tuned responses for CSEP components.

Figure 3.

CSEPs are anisotropically localized to a cortical column. a, Spatial distribution of weights from a regularized linear model of Hγ responses during the tone stimuli as a function of the other electrodes on the grid. Locations are all relative to the electrode used as the dependent variable in linear regression. Values are median across all N = 333 tuned (in Hγ) electrodes. b, Spatial distribution of normalized weights for tMuA. Values are median across all N = 113 tuned (in tMuA) electrodes. c, Median ± SD of normalized linear weights across all electrodes as a function of distance in the AP (solid lines, left axis) and DV (dashed lines, right axis) dimensions along the grid. Different frequency bands are demarcated with colors. Note inverted orientation of distances along x-axis for AP (black) versus DV (gray).

Figure 4.

Biophysical in silico cortical column reproduces in vivo observed μECoG response. a, Rendering of a random subselection of 626 neurons in the simulated column (∼2% of the total). Black, excitatory neurons; red, inhibitory. Circles represent somas, lines represent dendritic structures. The position of the simulated μECoG electrode relative to the column is shown above. b, Distribution of synapses from the thalamus along the depth axis of the simulated cortical column. c, Data from one simulated stimulation and prestimulus/poststimulus silence. i, Population spike rate of thalamic and background cortical spike trains activating synapses in the column. ii, Spike raster of all neurons in the column versus soma depth (y-axis). Note that differences in raster density in part reflect differences in neuron density across cortical layers. iii, Population spiking (fraction of neurons spiking in 1 ms) of biophysically detailed cortical neurons. iv, Cell-averaged spike rate of biophysically detailed neurons in each layer. Darker shades indicate deeper layers. d, CSEP computed by the Line Source Approximation from all neurons in the column during a 150 ms window centered around the 50 ms tone pip stimulation. e, Spectrogram of the CSEP in d. Top, Mean normalization. Bottom, z-Score normalization. f, Frequency content of CSEP during 10 ms centered at the response peak (indicated with dotted red lines in d and e). Top, Mean normalization. Bottom, z-Score normalization. Individual electrode averages from experimental results are in gray, black is grand average. Individual stimulus presentations from simulations are in pink, red is grand average. All traces are normalized to their respective maxima. g, Whisker plots (median, IQR, 95% CI) of correlation coefficients comparing the frequency content of experimental results and average simulation results for z-score and mean normalizations.

Figure 5.

In silico cortical column predicts experimentally observed relationship between response magnitude and frequency. a, Average z-score as a function of frequency in eight simulations with variable input amplitude. b, Average z-score as a function of frequency in the experimental data for six different stimulus amplitudes. c, Normalized response magnitude versus normalized response frequency for experimental data (black, mean ± SD) and for simulations (red). Each data point corresponds to the response frequency and magnitude associated with a distinct input magnitude (response magnitude increases monotonically with input magnitude). Circled point indicates the input magnitude used in Figures 4, 6, 7. Orange dashed line is unity.

Figure 6.

Evoked μECoG responses originate in infragranular layers. a, Contributions to the simulated CSEP from anatomic layers. Top to bottom, Cortical layers I through VI. The sum of these contributions is the total CSEP. b, Frequency content of the laminar contributions during stimulus peak. Layer V and VI contributions dominate the high-gamma peak. c, Magnitude at peak frequency of the CSEP contribution of each cortical layer versus number of neurons in the layer. d, Magnitude at peak frequency of the CSEP contribution of each cortical layer versus average distance of cell bodies in the layer from the recording electrode. e, Magnitude at peak frequency of the CSEP contribution of each cortical layer versus synchrony of somatic membrane potentials averaged over all pairs of neurons in the layer. f, Pie chart showing the relative importance of these three factors in a linear model of the high-gamma peak contribution magnitudes of anatomic layers.

Figure 7.

Evoked μECoG responses originate in sources 800–1400 μm below the surface. a, Proportional breakdown of segments by anatomic layer. Most slices contain segments from neurons in multiple cortical layers. Bars represent proportion of total segments in the slice, different slices not to scale. b, Total number of simulated neuronal segments in each 200 μm axial slice of the column. c, Contributions to the CSEP from 200 μm slices, organized by depth. Top, Cortical surface. The sum of these contributions is the total CSEP shown in Figure 4b. d, Frequency content of the slice contributions during stimulus peak, colored by slice depth. Slices containing somas of layer V neurons dominate the high-gamma peak. e, Magnitude at peak frequency of the CSEP contribution of each slice versus number of neuronal segments in the slice. f, Magnitude at peak frequency of the CSEP contribution of each slice versus average distance of segments in the slice from the recording electrode. g, Magnitude at peak frequency of the CSEP contribution of each slice versus average synchrony in the slice. h, Pie chart showing the relative importance of the three factors in our linear model of the high-gamma peak contribution magnitudes of the slices.