Could gauging Aβ and tau accumulation in healthy older adults be as simple as measuring their brain activity during sleep? Researchers led by Matthew Walker, University of California, Berkeley, think it might. In the June 17 Journal of Neuroscience, they describe distinct EEG patterns that associate with each of these pathologies. Less slow-wave activity during non-REM sleep tracked with greater Aβ load in the cortex, while weak coupling between slow oscillations and an EEG phenomenon called sleep spindles correlated with greater tau burden in the medial temporal lobe. While the study doesn’t address cause and effect, people who reported losing sleep time in midlife had more Aβ and tau pathology in later life. “It is interesting, from a scientific point of view, that you can separate these [Aβ and tau] signatures,” said Kristine Yaffe, University of California, San Francisco.

  • Reduced slow-wave brain activity during non-REM sleep correlated with Aβ, but not tau.
  • Weak coupling of slow oscillations and sleep spindles correlated with tau, but not Aβ.
  • Historically in volunteers, sleep disturbance correlated with Aβ first, then tau.

Many studies have linked disturbed sleep to Aβ, and more recently with tau (e.g., Aug 2017 conference news; Mar 2018 news; Jan 2019 news). Previously, Walker and colleagues had reported that during non-REM sleep loss of slow-wave activity (SWA), specifically frequencies lower than1 Hz, predicted levels of Aβ deposits in healthy, older adults (Jun 2015 news). Recently, researchers at Brendan Lucey’s and David Holtzman’s labs at Washington University School of Medicine, St. Louis, reported that reduced SWA correlated even more so with tau (Jan 2019 news). 

“Essentially, we had found a sleep biomarker for the amount of amyloid in the brain,” Walker told Alzforum. “But we didn’t know if tau pathology came with its own unique sleep signature.”

Aβ and Tau Fingerprints. In healthy, older adults, the proportion of slow-wave activity during non-REM sleep correlated with Aβ (top left), but not tau (top right). Conversely, decoupling of slow oscillations and sleep spindles correlated with tau (bottom, right), but not Aβ (bottom left). [Image courtesy of Winer et al., J Neurosci, 2019.]

To test this idea, first author Joseph Winer analyzed data from 101 cognitively healthy participants in the Berkeley Aging Cohort Study. All had undergone Pittsburgh compound B (PiB)-PET imaging to measure Aβ accumulation and flortaucipir-PET to measure tau tangles. Within 11 months of the tau PET scan, 24 women and seven men, at an average age of 76, spent a night at a sleep lab with electrodes attached to their scalps to track their brain waves.

The sleep recordings confirmed that less SWA correlated with greater Aβ.  burden. Winer also determined that the association held after adjusting for age, sex, sleep apnea risk, and the interval between EEG and PET scans (see image). Moreover, the SWA signature distinguished Aβ-positive from Aβ-negative individuals, as determined by a PiB uptake threshold. However, in contrast to the work from Lucey’s and Holtzman’s labs, the amount of SWA correlated poorly with tangle load and could not distinguish tau-positives from tau–negatives.

“We became really interested when we saw that the slow-wave measure was not associated with tau,” Winer said (image above). He noted that 20 percent of the subjects in the WashU study had mild cognitive impairment. In those cases, tau might already have spread into cortical regions where it could start affecting slow-wave activity, he suggested.

To look for a specific tau signature, the authors turned to the medial temporal lobe (MTL), a site of early tau buildup. Here, during non-REM sleep, slow oscillations couple with sudden bursts of high-frequency waves known as sleep spindles. The authors reasoned this coupling might be affected by tau pathology because it is orchestrated by hippocampal SWA, which previously was found to be disrupted in a mouse model of tauopathy (Witton et al., 2014). Indeed, Winer found that individuals with weaker slow oscillation-spindle coupling harbored more tau in the MTL (see image above). This association held after adjusting for age, sex, sleep apnea risk, and interval differences between EEG and PET scans. The degree of coupling also distinguished between low- and high-tau individuals, but not between Aβ positives and Aβ negatives.

“We found a double dissociation, where each signal is uniquely sensitive to one pathology but not the other. That is what makes the findings so powerful,” said Walker. Sigrid Veasey, University of Pennsylvania, Philadelphia, was impressed, particularly by the tau signature. “The nice thing is that synchronization between SWA and sleep spindles reflects connectivity, which is a measure of functionality,” she said. In fact, coupling of slow oscillations with spindles has been proposed to play a role in hippocampal memory processing during sleep, so uncoupling could be how tau disrupts memory, noted Winer and Walker (Diekelmann and Born, 2010). 

The associations between SWA and Aβ, and between sleep spindles and tau, jibe with a recent report by Ricardo Osorio, New York University, and Andrew Varga, Icahn School of Medicine at Mount Sinai, New York. These researchers found that in cognitively healthy, older adults, tau in the CSF correlated with spindle density and other spindle properties, but not with SWA (Kam et al., 2019). “It is largely consistent with what we have seen," said Varga, but he cautioned that given the discrepancy with the data from the Lucey and Holtzman groups, the jury is still out on whether these sleep patterns could be used as surrogate markers for Aβ and tau.

Osorio doubts it. “I have trouble seeing this becoming more specific or more available than PET,” he said. Yaffe agreed and questioned whether an overnight sleep study would be substantially cheaper or more desirable for patients than a PET scan. However, Walker noted that sleep-wave tracking devices are getting smaller, cheaper, and easier to use. In theory, both the SWA and spindle signals can be monitored with just two electrodes—one placed slightly above the middle of the forehead, and a reference, often placed behind the ear. This can be done with portable EEG units. Although limited in the data they acquire, headsets for in-home use cost about $500. These devices monitor brain waves that can identify sleep problems, and can even improve sleep quality by delivering bursts of “pink” noise (similar to white noise but with a different frequency signature) when they detect slow oscillations.

Could sleep EEG patterns predict future Aβ or tau pathology? Longitudinal analyses might tell. In the meantime, the authors asked 95 of the volunteers, with an average age of 78, how many hours they usually slept during their 40s, 50s, 60s, and 70s. To jog their memories, the questionnaire prompted them to recollect life circumstances, such as jobs, births, and moves. Subjects whose sleep time dropped from their 40s to 50s, or from their 60s to 70s, carried greater Aβ burdens later in life than those whose sleep increased in those decades. On the other hand, subjects who reported spending less time sleeping in their 60s versus 50s had slightly greater tau burden than those who slept more. “This is very exploratory,” acknowledged Winer, but he found it interesting that the associations were sequential, with the tau link surfacing after the first sign of the Aβ correlation, consistent with the time course of neuropathology.

Veasey and Yaffe cautioned that people are often terrible at estimating their sleep time, particularly retrospectively across such long time spans. Veasey added that the data are highly variable and overlap extensively between those reporting increases and decreases in sleep. Yaffe said that without comparing these data to current sleep reports, it is unclear whether the historical data add predictive value.—Marina Chicurel


  1. This paper from Winer et al. provides a convincing contribution to the growing literature that objective and subjective measures of sleep are associated with Alzheimer’s disease pathology. In this case, lower amplitude of slow-wave activity (specifically in the 1Hz and below frequency range) was associated with amyloid burden as measured with amyloid PET, and disrupted coupling between slow oscillations and sleep spindles was related to tau pathology in the medial temporal lobe, also measured with PET.

    Furthermore, participants completed a questionnaire reporting their estimated average sleep duration in each decade of life. Focusing on sleep duration in midlife and older age (decade intervals from the participants’ 40s to 70s) the researchers determined that self-reported changes in sleep duration were associated with both amyloid and tau pathology. While self-reported measures may have some limitations, it is still interesting to note that midlife decreases in sleep duration (i.e., when participants were in their 50s) appeared to be associated with later life amyloid pathology. This finding underscores the notion that midlife may be a critical time period when trajectories of brain pathology are determined, and may be a time frame when interventions should be implemented to ensure beneficial effects on trajectories of aging. Studies are needed to prospectively test whether treating sleep dysfunction attenuates the accumulation of Alzheimer’s disease pathology. 

Make a Comment

To make a comment you must login or register.


News Citations

  1. New Ties between AD and the Stages, Waves, and Molecules of Sleep
  2. Does Daytime Drowsiness Foreshadow Aβ Accumulation?
  3. Another Reason to Catch Some Zzzs: Sleep Regulates Tau Release
  4. Does Amyloid Disturb the Slow Waves of Slumber—and Memory?
  5. Tau, More than Aβ, Affects Sleep Early in Alzheimer’s

Paper Citations

  1. . Disrupted hippocampal sharp-wave ripple-associated spike dynamics in a transgenic mouse model of dementia. J Physiol. 2014 Dec 5; PubMed.
  2. . The memory function of sleep. Nat Rev Neurosci. 2010 Feb;11(2):114-26. Epub 2010 Jan 4 PubMed.
  3. . Sleep oscillation-specific associations with Alzheimer's disease CSF biomarkers: novel roles for sleep spindles and tau. Mol Neurodegener. 2019 Feb 21;14(1):10. PubMed.

Further Reading

No Available Further Reading

Primary Papers

  1. . Sleep as a Potential Biomarker of Tau and β-Amyloid Burden in the Human Brain. J Neurosci. 2019 Aug 7;39(32):6315-6324. Epub 2019 Jun 17 PubMed.