Semantic Processing During Sleep
Thomas Andrillon, Andreas Trier Poulsen, Lars Kai Hansen, Damien Léger, and Sid Kouider
(see pages 6583–6596)
When sleeping, people are behaviorally unresponsive to most environmental stimuli. Nevertheless, salient stimuli can cause arousal. Moreover, several studies have demonstrated that sensory stimuli evoke cortical activity during sleep and that some semantic processing occurs. The extent to which stimuli are processed, how processing of most stimuli is suppressed, and whether suppression differs across sleep stages remain unclear, however.
To investigate these questions, Andrillon et al. recorded human electroencephalographic responses to words presented at different stages of the sleep–wake cycle. Participants were instructed to categorize spoken words as ‘animals’ or ‘objects’ by moving their left or right hand. Because preparing a motor response involves lateralized neural activity, the presence of a “lateralized readiness potential” (LRP) indicated that participants processed auditory information sufficiently to prepare a response. When participants were awake, large LRPs were recorded over motor cortex. Although behavioral responses disappeared after subjects fell asleep, stimulus-evoked LRPs were detected during light sleep, suggesting semantic processing of words continued.
LRP amplitude during light sleep was correlated with the amplitude of the P200, a positive deflection that is thought to reflect activation of primary sensory cortex. In contrast, P200 and LRP amplitude were negatively correlated during deep non-REM sleep, and LRPs were generally undetected. During this stage, the P200 was followed by a large negative deflection, the N500, thought to reflect cortical silencing. Because N500 and LRP amplitudes were negatively correlated, the authors suggest that stimuli that strongly activate sensory cortex during deep non-REM sleep induce down states, thus preventing further processing.
Interestingly, words that had been presented during wakefulness evoked LRPs during REM sleep, but novel words did not. This suggests that the extent of stimulus processing during REM sleep depends on stimulus familiarity. In addition, although LRP amplitude was positively correlated with the complexity of ongoing neural activity during light sleep, it was negatively correlated with complexity during REM sleep. Therefore, the authors propose that dreams, which are associated with high neural complexity, interfere with processing of sensory stimuli during REM sleep.
These results indicate that whether external stimuli are processed or suppressed depends on the sleep cycle and the nature of the stimuli. This stage-dependent processing may allow people to monitor the environment periodically during the night without unnecessarily disrupting sleep.
DNA Methylation in Dyskinesia
David Figge, Karen L. Jaunarajs, David G. Standaert
(see pages 6514–6524)
Parkinson's disease is characterized by tremors and slowed movement resulting from loss of midbrain dopaminergic neurons. In early stages of the disease, treatment with the dopamine precursor levodopa reverses motor symptoms; but eventually, levodopa treatment begins to produce involuntary jerking movements (dyskinesia). This levodopa-induced dyskinesia (LID) subsides if the drug is discontinued, but reappears when treatment is reinstated, suggesting it stems from long-lasting changes in striatal circuits. These changes are thought to be induced by the continued loss of dopaminergic input, combined with fluctuations in extracellular dopamine resulting from the pharmacokinetic properties of levodopa.
Changes in the expression of numerous genes have been identified in animal models of LID. These changes are reversed after cessation of levodopa treatment, however, suggesting they are involved in the expression, rather than the maintenance of LID. The immediate return of dyskinesia when treatment is reinstated suggests that longer-lasting modifications occur. These may include epigenetic modifications, such as DNA methylation and demethylation.
To test this hypothesis, Figge et al. examined DNA methylation in the striatum of rats treated with levodopa or saline after dopaminergic neurons had been killed. Methylation of ∼27,000 DNA regions was altered by dopamine depletion alone, and additional differences emerged after treatment with levodopa. Methylation of ∼4,000 DNA regions differed by at least 5% between levodopa- and saline-treated rats. Interestingly, while levodopa appeared to reverse many methylation changes produced by dopamine depletion, it exacerbated other changes. Notably these changes included demethylation near several genes involved in synaptic plasticity and/or previously shown to be upregulated in LID.
Because levodopa treatment was associated preferentially with decreased rather than increased methylation, and because demethylase expression was increased in the striatum of levodopa-treated rats, the authors asked whether administering methionine, which increases methylation, would affect LID. Indeed, methionine reduced motor impairment in levodopa-treated rats. Conversely, a methyltransferase inhibitor, which reduced methylation in the striatum, exacerbated LID.
These results suggest that extensive changes in DNA methylation produced by loss of dopamine and treatment with levodopa contribute to the development and expression of LID. Reversing these changes might ameliorate LID in Parkinson's disease patients.
Footnotes
This Week in The Journal is written by Teresa Esch, Ph.D.