Delta oscillation coupled propagating fast ripples precede epileptiform discharges in patients with focal epilepsy

Abstract

Epileptiform spikes are used to localize epileptogenic brain tissue. The mechanisms that spontaneously trigger epileptiform discharges are not yet elucidated. Pathological fast ripple (FR, 200-600 Hz) are biomarkers of epileptogenic brain, and we postulated that FR network interactions are involved in generating epileptiform spikes. Using macroelectrode stereo intracranial EEG (iEEG) recordings from a cohort of 46 patients we found that, in the seizure onset zone (SOZ), propagating FR were more often followed by an epileptiform spike, as compared with non-propagating FR (p < 0.05). Propagating FR had a distinct frequency and larger power (p < 1e-10) and were more strongly phase coupled to the peak of iEEG delta oscillation, which likely correspond with the DOWN states during non-REM sleep (p < 1e-8), than non-propagating FR. While FR propagation was rare, all FR occurred with the highest probability within +/- 400 msec of epileptiform spikes with superimposed high-frequency oscillations (p < 0.05). Thus, a sub-population of epileptiform spikes in the SOZ, are preceded by propagating FR that are coordinated by the DOWN state during non-REM sleep.

Keywords: down state; epileptiform discharge; fast ripple; focal epilepsy; high-frequency oscillation; slow wave sleep.

Figure 1:. Fast ripples (FR) propagate in the seizure onset zone (SOZ).

(A1) Example of FR propagation (vertical line labelled “p*”) in the SOZ within the left cingulate gyrus of patient 4145 after band-pass filtering 200–600 Hz. FR propagate from the out-node (LPF3–3, blue) to the in-node (LPF3–1, red), as defined by the sign test, and between other nodes not examined here. Note that FR onset in contacts from another depth electrode precede this specific FR propagation event (green vertical line). (A2) The FR propagation event in A1 at increased temporal resolution. (A3) Normalized averaged time-frequency representation of signals triggered by the onset of propagating out-node FR recorded at the out-node (top) and in-node (bottom) during the recording duration. White vertical line and arrow show delay in the normalized signal that is incompatible with volume conduction. (B1) Corresponding unfiltered iEEG of same FR in A1, note that the FR in the out node (black vertical line) precedes epileptiform discharge onset and were not classified FR on spikes by our detector. Other FR propagation events are marked by vertical lines labeled p. (B2) The FR propagation event at increased temporal resolution. (C) Raster plot of FR onset times for the in- and out-nodes (Sign test, p<1e-10). FR propagated with a mean delay of 1.6 msec across a distance of 6.4 mm. The propagating FRs (|delay|<250 msec) are denoted by red ticks (top). The raster plot at the time corresponding to the example in A and B at increased temporal resolution (green box, bottom).

Figure 2:

(A) Summary plot of the FR propagation velocity of edges (i.e. out-node to in-node) demonstrating a trend towards a linear relationship between longer propagation (i.e., edge) distances and longer propagation delays in the SOZ:SOZ (A1,red), and NSOZ:NSOZ (A3, blue) edges. For NSOZ:SOZ (A2, cyan) edges the regression was significant indicating a constant propagation velocity across these edges. Note that increasing delay with increasing edge distance at relatively slow velocities is incompatible with FR volume conduction, although our results suggest the propagation velocities are not always uniform. (B) Summary vector plot, in MNI coordinates, of the individual edges, color coded as in (A), demonstrating significant FR propagation across all patients, arrows point from the out-node to the in node. Circles represent borders of the cerebrum. Abbreviations (AP: anterior [+] to posterior [−], DV: dorsal [+] to ventral [−], LR: left [+] to right [−], MNI: Montreal Neurological Institute Coordinates, mm: millimeters).

Figure 3:. Fast ripple (FR) frequency and power differs between propagating and non-propagating FR in the out- and in-node and is dependent on the mean delay and distance of propagation.

(A) Differences in FR frequency for out-node non-propagating FR (blue), out-node propagating FR (yellow), in-node non-propagating FR (red), and in-node propagating FR (cyan) in seizure onset zone (SOZ:SOZ) edges. (A1) Correlation of FR frequency with propagation delay showing mean values and standard error of the mean (s.e.m) at each edge (GLMM, four-way interaction, p<1e-6). (A2) Boxplot of the FR frequency distributions for the individual FR SOZ:SOZ event types irrespective of edge and propagation delay (ANOVA, p<1e-10, eta2=0.058). (B) Differences in FR power for the different FR types. (B1) Correlation of FR power with propagation distance showing mean values and standard error of the mean (s.e.m) at each edge. FR power differed in SOZ:SOZ edges (circles), NSOZ:NSOZ edges (square), and edges bordering the two territories (diamonds) as a function of FR propagation distance for the different FR types (GLMM, four-way interaction, p<0.01). (B2) Boxplot of the FR power distributions for the individual FR SOZ:SOZ event types irrespective of edge and propagation delay (ANOVA, p<1e-10, eta2=0.031). Abbreviations, GLMM:generalized linear mixed effects model, AU:arbitrary units, out p−:out-node no propagation, in p−: in-node no propagation, out p+: out-node propagation, in p+: in-node propagation.

Figure 4:. Fast ripple (FR) are coupled to delta oscillations and this coupling increases the probability of FR propagation between edges by a distinct mechanism.

(A1) Example of FR-delta coupling in the unfiltered iEEG recorded from the SOZ. Arrows point to individual FR. (A2) Trace shown in A1 after filtering in the delta (2–4 Hz) band with instantaneous phase color coded. (B) Among the 46 patients, illustration of the aggregated nodes with statistically significant FR-delta phase coupling (Rayleigh test, p<0.001) in the SOZ (A1) and NSOZ (A2). The size of the node indicates the Rayleigh Z value, a measure of the strength of coupling. The color of the node is the mean phase angle of coupling. (C) Three-dimensional bar plot illustrating the percentage of edges with FR mutual information > 0 and significant propagation (Sign Test, p<0.005) as a function of the out-node and in-node FR-delta phase coupling strength. Strength was measured as the Rayleigh test p-value for edges in the SOZ (blue), between NSOZ and SOZ or SOZ and NSOZ (green), and NSOZ (yellow). An increased sign test Z value (i.e. propagation measure) was positively correlated with higher out- and in-node FR-delta coupling strength (Rayleigh Z) in the SOZ (GLMM, p<1e-9). (D) Polar histograms quantifying pooled FR occurrence in relation to the phase of the delta wave. Note FR were measured only from edges with statistically significant propagation (p<0.005), and FR independent of a delta wave were excluded. FR-delta phase angle distributions were compared in the out- and in-node in limbic regions (black), frontal lobe (light blue), parietal lobe (red), temporal lobe (dark blue), and occipital lobe (green) for non-propagating events (D1) and propagating events (D2) in the SOZ. Note the difference in the scale of the polar histograms between D1 and D2. In limbic regions, the FR-delta coupling angle was best predicted by the interaction of whether the FR was recorded from the out- or in-node, and whether the FR event was propagated. (Bayesian mixed-effect regression model for circular data, relative Bayes Factor 2 of 2). In the frontal and parietal SOZ, the FR-delta coupling angle was best predicted by out- or in-node status than by the propagation status or the interaction (relative Bayes Factor ~2 of 2).

Figure 5:. Fast ripple (FR) propagation and modulation to delta in the seizure onset zone (SOZ) precede epileptiform spikes in the intracranial EEG (iEEG) at the out node.

(A) Superimposed example of FR triggered average of iEEG signals (white traces, right axis modulation strength, see cross modulation in methods) and normalized averaged time-frequency representation in the FR band (color axis, normalized power). The peak-to-peak amplitude of iEEG signals (white traces) corresponds to the strength of modulation with FR events that are temporally aligned, with onset at 250 ms, and modulated as shown by the normalized averaged time-frequency representation. To demonstrate cross-modulation, the in-node averaged iEEG signals were triggered by the FR events in the out-node. Displayed are non-propagating FR events (n=138) in the out-node (A1) and the in-node (A2), and propagating FR events (n=47) in the out-node (A3) and the in-node (A4) in an SOZ:SOZ edge patient 4145. (B) Example average fitting oscillations & one over f (fooof) derivations for the non-propagating out- (B1) and in-node (B2) iEEG signals, and propagating out-node (B3) and in-node (B4) iEEG signals corresponding to those used to derive A1–4. Blue line indicates aperiodic fit, red line indicates full model fit. Note that the wavelet transform averages FR power and thus does not accurately portray FR onset, but in individual traces FR onset always preceded the epileptiform spike (C-D). Summary plot of comparisons of derived average fooof parameters calculated between the non-propagating events (np) and propagating (p) events in the out-node (C) and in-node D) in the 29 SOZ:SOZ edges of 6 patients shown in distinct colors. Asterisk next to the plot title indicates statistical significance (paired t-test, p-adjusted<0.05).

Figure 6:. Irrespective of fast ripple (FR) propagation, FR probability is highest preceding epileptiform spikes with superimposed HFOs (HFO spike) and HFOs preceding these spikes most often occur in the seizure onset zone (SOZ).

(A) Normalized histograms of the latency between FR events and HFO spikes in individual contacts pooled across contacts and averaged across patients (N=46). Error bars show standard error of the mean (s.e.m). FR on spike shown in black were arbitrarily assigned a latency of zero. The histogram in A1 was calculated using latencies measured from single HFO spike events (n=84,067), whereas A2 was calculated using single FR events (n=39,616). Note that although the probability of FR occurrences is highest before HFO spikes, only a small proportion of HFO spikes are preceded by FR. (B) Tests of skewed HFO spike-FR latency distribution in individual patients. The proportion of patients with a non-uniform HFO spike-FR latency distribution (χ2 test, Holms-Bonferoni, p-adjusted<0.05) was over 75% (B1), and in over 50% of patients FR were most probable within 400 ms preceding an HFO spike. For these calculations FR superimposed on spikes were excluded. (C) The mean proportion of the FR in the SOZ, compared to the non-SOZ, among FR preceding HFO spikes (<400 ms, yellow) and all other FR (blue). FR preceding HFO spikes occurred in the SOZ more often than the other FR (paired t-test, n=37, p<1e-5, cohen’s d=0.97). Error bars indicate s.e.m.

Figure 7:. Graphical illustration of the mechanisms of fast ripple (FR) propagation.

In the seizure onset zone, FR generating nodes receive inputs from delta oscillation drivers (green circle). The strength of coupling between delta and FR events is shown by the thickness of the green arrow. The FR occur near or at the peak of the delta oscillation corresponding to the DOWN state. The two nodes (black circles) with FR propagation across the connecting edge (black line) exhibit stronger FR-delta coupling at the out- and in-node as compared with the node with non-propagating FR. At the out-node with FR propagation, the propagated FR is higher power and higher frequency as compared with a non-propagated FR. At the out-node a propagated FR event precedes an epileptiform discharge likely through increased neuronal excitability overcoming an inhibitory restraint mechanism. Alternatively, the epileptiform discharge may be triggered by the delta driver corresponding to the DOWN state. The propagated FR event at the out-node reaches the in-node at a relatively slow velocity through poly-synaptic conduction. At the in-node the FR is relatively lower in power and higher in frequency. FR propagation in NSOZ:SOZ, SOZ:NSOZ, and NSOZ:NSOZ edges involves diverse unresolved mechanisms.