Granger causality relationships between local field potentials in an animal model of temporal lobe epilepsy

Abstract

An understanding of the in vivo spatial emergence of abnormal brain activity during spontaneous seizure onset is critical to future early seizure detection and closed-loop seizure prevention therapies. In this study, we use Granger causality (GC) to determine the strength and direction of relationships between local field potentials (LFPs) recorded from bilateral microelectrode arrays in an intermittent spontaneous seizure model of chronic temporal lobe epilepsy before, during, and after Racine grade partial onset generalized seizures. Our results indicate distinct patterns of directional GC relationships within the hippocampus, specifically from the CA1 subfield to the dentate gyrus, prior to and during seizure onset. Our results suggest sequential and hierarchical temporal relationships between the CA1 and dentate gyrus within and across hippocampal hemispheres during seizure. Additionally, our analysis suggests a reversal in the direction of GC relationships during seizure, from an abnormal pattern to more anatomically expected pattern. This reversal correlates well with the observed behavioral transition from tonic to clonic seizure in time-locked video. These findings highlight the utility of GC to reveal dynamic directional temporal relationships between multichannel LFP recordings from multiple brain regions during unprovoked spontaneous seizures.

Figure 1

The electrode placement visualized in the excised fixed brain with high-field MRI. The three panels show orthogonal slices from a three-dimensional, gradient echo MR image, acquired at 17.6 T (750 MHz): a coronal slice panel a), horizontal slice in panel b), and sagittal slice in panel c). The three image slices intersect at the tip of an individual electrode located in the right CA1 (see red arrows) ipsilateral to the site of injury. The panels also illustrate several additional electrode tracts (shown as black vertical lines in the coronal and sagittal slices).

Figure 2

An example of electrode tip placement validation using coronal sectioned histology. Perl staining shows the tips of five electrodes in the CA1 subfield of the hippocampus. The electrode to the left terminates in the stratum lacunosum-moleculare, the middle three electrodes terminate in the stratum radiatum and the electrode to the right terminates in the CA1 pyramidal cell layer. This histological staining validated the iron content (The Perl stain highlights iron in black) and electrode track and tip locations visualized in Figure 1. Sections were then counterstained with cresyl violet to visualize cellular morphology.

Figure 3

Example of GC analysis for a single (N=1) spontaneous temporal lobe seizure. The 1^st row contains cartoon plots of the results from the GC indices calculated across all seizures (N=15) and are overlaid on MRI images of the animals hippocampus to visualize GC index directional relationships (note that the location and shape of the arrows are not meant to imply specific physiological pathways or mechanisms). GC raw results for selected time windows in the example seizure (N=1) are presented in the 2^nd row. The 2^nd row plots depict the strength of directional relationships among each electrode pair for all 32 electrodes by color. The electrode pairings are arranged by anatomical area. The bottom row provides the electrographic activity from the example seizure (N=1) from 4 of the 32 electrodes, one from each of the four major subfields covered by the microelectrode arrays, R-CA1, R-DG, L-CA1, and L-DG. The highlighted time segments correspond to the GC analysis excerpts shown in rows 1 & 2. These traces include one electrode among the four major hippocampal areas observed. Colors indicate the magnitude of GC results, ranging from weak (blue) to strong (red). Directional information is presented by portraying the source location on the vertical axis and response location on the horizontal axis. Starting from the lower left and moving counter clockwise, the four quadrants of each plot represent within-left hemisphere, left-to-right cross hemisphere, within-right hemisphere, and right-to-left cross hemisphere directional relationships. The first column depicts the results from prior to the behavioral onset of seizure (PS) followed by six panels representing transitions (stages S1, S2a & S2b, S3, S4, S5) in the patterns during seizure. A time shifted surrogate analysis was carried out by applying GC to the time shifted data set. This surrogate analysis yields a GC threshold value of 3.66 for p < 0.01 (dark blue in the figure), below which GC results are not significant. Note: A supplementary high-resolution animation of the GC analysis of the example seizure (N=1) presented in this figure is available on the Neuroscience Methods website.

Figure 4

Continuous plots for the primary GC Indices for the same example seizure presented in Figure 2. Each window represents the GC Index (dark blue trace) for all of the electrode pairings from and to the areas specified on the label for each panel. For example, the top left plot (column 1, row 1) represents the GC Index or the sum of all GC results from the right DG to the L-CA1. The arrangement of the “from” and “to” areas in the subplots corresponds to the same relative arrangement in each of the GC plots shown in the 2^nd row of Figure 2. For each panel, seizure data from one of the corresponding “from” electrodes used in calculation of the GC index is plotted in light grey to give a temporal context to the GC index. Additionally, significance thresholds from GC analysis of a time shifted surrogate time series is shown with a red dashed line. The seizure occurs mainly in the middle 60 seconds, the same as in the supplemental video. The GC Index is increased after seizure for 11 of the indexes, the same for 4 indexes, and decreased for 1 index. Interestingly the sole decreased index is from L-CA1 to L-DG index (column 2, row 4). This is the same index implicated for having an abnormal GC relationship observed from the L-CA1 to the L-DG prior to seizure. The spikes between 60 and 70 seconds (most prevalent in the top left and bottom right) correspond to Stage 2 cross hemisphere directional relationships.

Figure 5

Non-directional mean (N=15 seizures) GC index of all channels pre seizure and through each of the five seizure stages. All of the directional relationships between all 32 electrodes have been averaged to produce a representation of a total nondirectional relationship. The strength of this non-directional GC index increases rapidly from the pre seizure period to Stage 1 and continues to increase until peaking during Stage 2. Following Stage 2 the non-directional index decreases during Stage 3 but briefly spikes once again in Stage 4 before terminating at the end of Stage 5. Note that this non-directional analysis shows no difference between stages 2 and 4. Overall, these results are consistent with general effective and functional connectivity assessments of synchronization during seizure found in the literature. Additionally, these results are consistent with the general dynamics observed during each of the fifteen seizures in our analysis. The average time length for each stage (N=15 seizures) is 8.4 +/- 4.4, 11.9 +/- 4.6, 11.7+/- 5.8, 6.9 +/-4.5, and 15.7 +/- 5.8 for seizures stages 1 though 5 respectively.