This prospective, longitudinal study demonstrates that reduced sleep spindle and SW density predict development of PD-MCI. In addition, cognitive performance correlates cross-sectionally with spindle density, SW density, and SW-spindle co-occurrence percent in PD. Based on ROC curve analysis, fewer than 7.74 spindles/min during N2 is a good predictor of PD-MCI development. However, the best predictor was the combined model including spindle density, SW density, CCS, and LED. Because spindle density and SW are modifiable,9,10 our findings suggest potential not only as predictive biomarkers of PD-MCI, but also as therapeutic targets for cognitive improvement in PD.
This study supports a growing body of evidence linking sleep spindles with cognition in physiological aging, MCI, AD, and PD.7,8,11,12 Pioneering intracellular recordings established that sleep spindle oscillatory activity originates in the thalamo-cortical loop.13 Specifically, thalamic reticular neurons are implicated in generation of sleep spindles through reciprocal synaptic interactions with corticothalamic cells.13,14 Based on the active systems consolidation theory, the physiological function of spindles is to consolidate declarative and non-declarative memory through synaptic plasticity.15 The current study’s findings support this proposed mechanism by showing relationships between sleep spindles and cognition in PD both cross-sectionally and longitudinally. Thus, abnormal spindle activity in PD may alter mechanisms of brain plasticity and contribute to cognitive decline.8 Indeed, neuroimaging studies have demonstrated structural and functional alterations in the thalamus of patients with PD.16 Therefore, this study provides evidence that sleep spindles may serve as a biomarker for PD-MCI risk, supporting prior work showing that sleep spindles predict dementia in PD.8
Exploratory analyses revealed that relationships between spindle density and cognitive performance were predominantly driven by correlation with visuospatial function and a trend towards correlation with memory. These results are consistent with previous studies and our understanding that visuospatial deficits contribute to early cognitive decline and dementia onset in PD.17,18 Furthermore, PD patients with cognitive deficits have reductions in acetylcholinesterase activity, particularly in parietal areas,19,20 as well as decreases in choline acetyltransferase activity in reticular thalamic nuclei.21,22 Therefore, cholinergic deficits may disrupt the corticothalamic feedback loop, affecting the mechanism by which sleep spindles are generated.8 This speculation is further bolstered by work showing that increased uptake of cholinesterase inhibitors in healthy elderly subjects resulted in increased N2 sigma activity.23
In addition to sleep spindles, other sleep architectural features have been related to cognition in health and disease. In particular, slow wave sleep (N3) and slow wave activity (0.5-4.0 Hz) are important for synaptic potentiation and neural synchrony.24 Further, glymphatic function, which has been proposed to clear neurotoxins from the brain, is increased during slow wave sleep, suggesting a potential mechanism for these relationships.25,26 Our group previously showed that slow wave sleep and delta spectral power (1–4 Hz) during N3 correlates with cognitive performance in PD.6 Since SWA is composed of two independent components, SW (< 1 Hz)4 and delta waves (1–4 Hz),5 the current study expands upon our prior work showing that SW density predicts longitudinal development of PD-MCI and is correlated with cognition cross-sectionally in PD. This relationship appears to be driven by a trend towards correlations with executive function and attention/working memory. These results fit well with our understanding of the critical role played by slow waves in cognitive processing. This knowledge is based on both the sleep homeostasis hypothesis, which recognizes the importance of slow wave sleep for synaptic pruning/downscaling, and the active system consolidation theory, which suggests that slow wave sleep fosters interactions between the hippocampus and the neocortex.27,28 Furthermore, imaging studies have demonstrated that slow wave sleep results in increased metabolism of the prefrontal cortex, a critical region for executive function and attention/working memory.29,30
The findings that SW density predicts longitudinal development of PD-MCI is in contrast to findings by Latreille et al.8 that SWA did not predict development of PD-dementia. One potential explanation for this apparent discrepancy is that the current study investigated SW (< 1Hz), while Latreille and colleagues evaluated oscillations in the range of 0.1-4Hz. Our reason for evaluating SW morphology at a lower frequency was based on findings of a prior cross-sectional study showing that lower delta frequencies were correlated with cognition in PD.31 Further, SWA, through its key nightly role in glymphatic clearance, may have an acute influence on cognition that is not maintained over time unless the SWA is also maintained. Because both studies only evaluated sleep EEG at baseline, it is unknown if the SWA remained consistent longitudinally, or waned over time. This highlights the potential importance of interventions, such as exercise, that can enhance or maintain slow wave sleep.32
Because both spindles and slow waves are related to cognitive function, we also investigated how these oscillations co-occur in PD. In intracranial recordings of human mesial temporal lobe (MTL), hippocampal sharp-wave ripples, spindles, and slow oscillations during sleep interact in a coordinated manner.33,34 These fine-tuned temporal interactions play a crucial role in memory consolidation through brain plasticity, which can be measured by SW-spindle coupling. In fact, SW-spindle phase-amplitude coupling facilitates memory consolidation in young and older people and declines with physiological aging.35,36 Interestingly, impairment in this coupling has been described in AD.37,38 However, to our knowledge, no prior publications investigate SW-spindle coupling in PD. Taking advantage of this cohort with comprehensive neurocognitive assessment and polysomnography, we found that SW-spindle co-occurrence percent was significantly related to cognitive performance in participants with PD. In further exploring these relationships, executive function, memory, and visuospatial function were also correlated with SW-spindle co-occurrence percent. Further, the percent SW-spindle co-occurrence predicted the longitudinal development of PD-MCI. However, this relationship did not persist after adjusting for age and duration of disease.
The mechanism whereby SW-spindle coupling might interact with cognitive performance is not known. This has been explored with intracellular recordings in a mouse model of tauopathy, which demonstrated that MTL tau deposition impairs generation of hippocampal sharp-wave ripples,39 which also interact temporally with slow oscillations and spindles. This disruption of sharp-wave ripples could contribute to the relationship between tau pathology and impaired hippocampal-related memory performance.39,40 It is unclear how or if the accumulation of alpha-synuclein may similarly affect these processes via disruption of the SW-spindle coupling, potentially contributing to cognitive decline associated with PD.41 However, one study suggested that disruption of glymphatic clearance in the A53T synuclein mouse model of PD led to increased accumulation of alpha synuclein, suggesting that glymphatic function may also be important in this context.42 In light of the findings reported here, a testable hypothesis appears: disruption of the SW-spindle coupling represents a novel pathway through which alpha-synuclein pathology interferes with cognition.
This study has several strengths, including the longitudinal design, large sample size at baseline, comprehensive qEEG analysis of NREM sleep derived from laboratory-based polysomnography, employment of a comprehensive assessment of cognitive function designed in accordance with the MDS Task Force recommendations for PD-MCI diagnosis, and use of conservative statistical methods. However, the study also has limitations. First, not all PD participants in the baseline cohort underwent a follow-up cognitive assessment. This may have affected study conclusions since participants more at risk for cognitive impairment could contribute more to study attrition. In fact, when comparing baseline characteristics for the 41 participants who joined the longitudinal study to participants who only completed baseline, the longitudinal cohort was significantly younger (p = 0.037), had higher baseline sleep efficiency (p = 0.016), lower LED (p = 0.005), higher CCS (p = 0.0059), higher SW-spindle co-occurrence % (p = 0.039), and trend toward higher spindle density (p = 0.068). This may have resulted in underestimation of the strength of predictive effects. The longitudinal sample size was further reduced because we only analyzed participants characterized as PD-normal cognition at baseline in the longitudinal study (N = 26). Another potential limitation is that PD participants continued to take their usual medications, including dopaminergic medications and medications known to affect sleep. Participants who remained PD-NC were taking significantly more medications than those who developed PD-MCI, but adjustment for this association did not change the outcomes for the cross-sectional analysis. Interestingly, once SW density was adjusted for use of medications that affect sleep, this qEEG parameter predicted development of PD-MCI. Furthermore, this study failed to account for the first night effect (poor sleep in unfamiliar setting) on the sleep outcomes. However, our prior research found that first night effect does not negatively affect sleep in PD.43 Finally, because of the correlational nature of the current study, the direction of the observed effects isn’t clear. Future interventional studies may address this limitation.
In conclusion, this comprehensive investigation of NREM sleep qEEG in PD demonstrates that SW density, sleep spindle density, and SW-spindle co-occurrence percent correlate with cognitive performance. Additionally, lower spindle density and lower SW density are predictive of longitudinal development of PD-MCI and could therefore be potential modifiable markers of cognitive decline. These findings represent a step forward in our understanding of the pathophysiology underlying cognitive decline in PD. However, much remains to be learned and future studies should include sleep qEEG analysis and comprehensive cognitive assessment combined with neuroimaging and investigation of glymphatic clearance. Considering the prevalence of cognitive decline among people with PD, this is a critical research area.