Abstracts

Rationale: The integration and strengthening of memory is hypothesized to require the precise temporal coordination of cardinal sleep rhythms – slow oscillations (0.5-2 Hz), sleep spindles (8-15 Hz), and ripples (70-120 Hz) – across at least three brain regions implicated in their generation – medial orbitofrontal cortex (OFC), thalamus, and hippocampus. We used intracranial electrophysiological recordings and a validated behavioral task to investigate the spatial and temporal orchestration of sleep oscillations and predict human sleep-dependent memory consolidation.

Methods: 19 patients with drug refractory epilepsy undergoing surgical evaluation with intracranial electrodes in the OFC, thalamus, and hippocampus were prospectively recruited. Slow oscillations, spindles, and ripples were detected from multiple contacts in each region using previously validated methods. We used either far-field (slow oscillations) or bipolar (spindles and ripples) referencing and ignored events that occurred ±0.5s of epileptiform spikes. Normalized cross-correlation histograms across events and regions were estimated over an interval of ±2s of the slow oscillation down-state in the OFC. The modulation of spindle and ripple probabilities relative to a baseline interval was estimated in 40ms window ±3 s of the OFC down-state. A subset of patients were trained and tested on a motor sequence typing (MST) task ~1 hour before sleep and ~1 hour after waking the following morning. Slow oscillation, spindle, and ripple rates in each region, and across all combinations of rhythms and regions (19 hemispheres from 14 patients) were tested in an elastic net regression model to predict overnight memory consolidation.

Results: Slow oscillations, sleep spindles, and ripples were temporally coordinated within and across regions (Fig. 1). On average, the down-state of slow oscillations in the OFC preceded the thalamus by 6 ms and hippocampus by 21 ms. Relative to the down state of slow oscillations in the OFC, sleep spindles in the thalamus were maximally upregulated at 0.1 s, in the hippocampus at 0.23 s, and in the OFC at 0.25 s. Relative to the down state of slow oscillations in the OFC, ripples in the thalamus were maximally upregulated at 0.325 s, in the OFC at 0.4 s, and in the hippocampus at 0.575s. A regression model with 7 predictors optimally predicted memory consolidation (Fig. 2). Rate of slow oscillation coupling between OFC and thalamus, ripple rate in the OFC, rate of coupled ripples between the OFC and hippocampus, and rate of slow oscillation-ripple coupling in the OFC positively correlated with memory consolidation. In contrast, rate of spindle-ripple coupling in the thalamus, rate of OFC spindle and thalamic ripple coupling, and rate of slow oscillation-spindle-ripple coupling in the thalamus anticorrelated with memory consolidation. Together, these predictors accounted for 45% of the variance in overnight memory consolidation.


Conclusions: These findings provide experimental evidence in humans that the precise coordination of sleep rhythms across medial OFC, hippocampus, and thalamus contributes to the determination of which memories will be retained and which forgotten.

Funding: R01EB026938 and R01NS115868.