Altered sleep homeostasis correlates with cognitive impairment in patients with focal epilepsy

Abstract

In animal studies, both seizures and interictal spikes induce synaptic potentiation. Recent evidence suggests that electroencephalogram slow wave activity during sleep reflects synaptic potentiation during wake, and that its homeostatic decrease during the night is associated with synaptic renormalization and its beneficial effects. Here we asked whether epileptic activity induces plastic changes that can be revealed by high-density electroencephalography recordings during sleep in 15 patients with focal epilepsy and 15 control subjects. Compared to controls, patients with epilepsy displayed increased slow wave activity power during non-rapid eye movement sleep over widespread, bilateral scalp regions. This global increase in slow wave activity power was positively correlated with the frequency of secondarily generalized seizures in the 3-5 days preceding the recordings. Individual patients also showed local increases in sleep slow wave activity power at scalp locations matching their seizure focus. This local increase in slow wave activity power was positively correlated with the frequency of interictal spikes during the last hour of wakefulness preceding sleep. By contrast, frequent interictal spikes during non-rapid eye movement sleep predicted a reduced homeostatic decrease in the slope of sleep slow waves during the night, which in turn predicted reduced daytime learning. Patients also showed an increase in sleep spindle power, which was negatively correlated with intelligence quotient. Altogether, these findings suggest that both seizures and interictal spikes may induce long-lasting changes in the human brain that can be sensitively detected by electroencephalographic markers of sleep homeostasis. Furthermore, abnormalities in sleep markers are correlated with cognitive impairment, suggesting that not only seizures, but also interictal spikes can have negative consequences.

Keywords: cognitive impairment; epilepsy; epileptogenesis; high-density EEG; sleep.

Figure 1

NREM sleep SWA power: topography and group differences between patients and controls.Top: SWA power topographies for average referenced EEG in controls and patients. (A) Controls mean SWA power. (B) Patients mean SWA power. (C) Difference in SWA power between patients and controls. (D) SPM T map for this difference. (E) SNPM T map for this difference. Middle: SWA power topographies for average-mastoid referenced EEG in controls and patients. (F) Controls mean SWA power. (G) Patients mean SWA power. (H) Difference in SWA power between patients and controls. (I) SPM T map for this difference. (J) SNPM T map for this difference. Bottom: Source space statistics showing increased SWA power in patients compared to controls. LL = left lateral; MF = medial frontal; RL = right lateral. All T maps are thresholded for display at uncorrected P < 0.05. A rainbow colour scale is used for SWA power, and a hot colour scale for T-values.

Figure 2

Topography of SWA power: comparison between NREM sleep and REM sleep.Top: NREM sleep and REM sleep SWA power topographies in controls. (A) Mean NREM sleep SWA power. (B) Mean REM sleep SWA power. (C) Difference in SWA power between NREM sleep and REM sleep. (D) SPM T map for this difference. (E) SNPM T map for this difference. Bottom: NREM sleep and REM sleep SWA power topographies in patients. (F) Mean NREM sleep SWA power. (G) Mean REM sleep SWA power. (H) Difference between NREM sleep and REM sleep. (I) SPM T map for this difference. (J) SNPM T map for this difference. All T maps are thresholded for display at uncorrected P < 0.05. A rainbow colour scale is used for SWA power, and a hot colour scale for T-values.

Figure 3

Homeostatic regulation of SWA.Top: Overnight decrease in SWA power (SWA homeostasis) in controls. (A) SWA power during first hour of NREM sleep (early night). (B) SWA power during last hour of NREM sleep (late night). (C) Difference in SWA power between early and late night. (D) SPM T map for this difference. (E) SNPM T map for this difference. Second row: Overnight decrease in SWA power (SWA homeostasis) in patients. (F) SWA power during first hour of NREM sleep (early night). (G) SWA power during last hour of NREM sleep (late night). (H) Difference in SWA power between early and late night. (I) SPM T map for this difference. (J) SNPM T map for this difference. Third row: NREM sleep spindle power topographies. (K) Mean spindle power in controls. (L) Mean spindle power in patients. (M) Difference in spindle power between patients and controls. (N) SPM T map for this difference. (O) SNPM T map for this difference. Bottom: Grouping of spindles by individual slow waves in controls (P) and patients (Q) for F7, Fz and F8 electrodes; upper row shows average slow wave morphology and lower row shows mean and standard error of spindle power at each time point. All T maps are thresholded for display at uncorrected P < 0.05. A rainbow colour scale is used for SWA and spindle power, and a hot colour scale for T-values.

Figure 4

Correlation between NREM sleep SWA and seizure frequency. (A) Positive slope of multiple regression between SWA power and seizure frequency. (B) SPM T map for this correlation. (C) SNPM T map for this correlation. (D) Linear fit for correlation maximum peak. (E) Positive slope of multiple regression between slow wave (SW) slope and seizure frequency. (F) SPM T map for this correlation. (G) SNPM T map for this correlation. (H) Linear fit for correlation maximum peak. All T maps are thresholded for display at uncorrected P < 0.05. A rainbow colour scale is used for regression slope values, and a hot colour scale for T-values.

Figure 5

Correlation between spikes during wake and changes in local SWA power during sleep.Top and middle: Whole-night sleep SWA power topographies in individual controls (A) and patients (B). Patients are ordered as in Table 1. In each patient, the maximum peak of statistical difference in local SWA power compared to the control group is displayed as a black star. A rainbow colour scale is used for SWA power values. Bottom: Correlation between spike frequency and local SWA power. (C) A high frequency of interictal spikes during the last hour of wakefulness preceding sleep was positively correlated with high local increases in SWA power (differential Z-scores) during subsequent NREM sleep. (D) This relationship remained significant after a log transform of spike frequency counts. In contrast, the frequency of spikes during sleep (E) or its log transform (F) were not correlated with focal increases in local SWA power.

Figure 6

Interictal spikes frequency during sleep and sleep slow wave homeostasis. (A) Negative slope of multiple regression between spike frequency and overnight decrease in sleep slow wave slope (slope homeostasis) in patients with epilepsy. (B) SPM T map for this negative correlation. (C) SNPM T map for this negative correlation. (D) Linear fit for correlation maximum peak. (E) Negative slope of multiple regression between log of spike frequency and overnight decrease in sleep slow wave slope. (F) SPM T map for this negative correlation. (G) SNPM T map for this negative correlation. (H) Linear fit for correlation maximum peak. All T maps are thresholded for display at uncorrected P < 0.05. A rainbow colour scale is used for regression slope values, and a cool colour scale for T-values.

Figure 7

Correlation between sleep markers and cognition. (A) Negative slope of multiple regression between IQ and spindle power in patients with epilepsy. (B) SPM T map for this negative correlation. (C) SNPM T map for this negative correlation. (D) Linear fit plot for correlation maximum peak. (E) Positive slope of multiple regression between learning and overnight decline of sleep slow wave slope (slope homeostasis) in patients with epilepsy. (F) SPM T map for this positive correlation. (G) SNPM T map for this positive correlation. (H) Linear fit plot for correlation maximum peak. All T maps are thresholded for display at uncorrected P < 0.05. A rainbow colour scale is used for multiple regression slope values, and cool and hot colour scales for negative and positive T-values.