As you lay yourself down to sleep, pray your neurons pick good memories to keep. A paper in the June 6 Science posits that during sleep, neurons reinforce freshly learned skills by sprouting new dendritic spines, and hence the number of synapses. Led by Wenbiao Gan at the Skirball Institute in New York, the study challenges the old view that sleep is when neurons prune synapses. Another sleep study, led by Jurgen Claassen at Radboud University in Nijmegen, the Netherlands, adds to the growing body of knowledge about how slumber affects brain levels of amyloid-β. In the June 2 JAMA Neurology, these researchers report that a single night of sleep deprivation prevents the drop in brain Aβ levels that normally occurs during a good night’s sleep. Both studies raise more questions than they answer, but one thing seems certain: When it comes to supporting memory, sleep is probably a good thing.

“During sleep, the brain is very active,” Gan told Alzforum. “It’s not really at rest; it’s doing a job. That’s really clear from our work.”

That healthy sleep promotes learning and memory has become household knowledge, and scientists increasingly accept that sleep may play a role in reducing the risk of Alzheimer’s disease. Exactly how the slumbering brain achieves these benefits, however, remains poorly understood. A widely accepted idea is that synapses weaken during sleep, which allows neurons to respond robustly and forge new connections during the waking hours (see Tononi and Cirelli, 2014). In line with that hypothesis, Giulio Tononi at the University of Wisconsin, Madison, reported that the number of dendritic spines drops when adolescent mice sleep (see Oct 2011 news story). Gan’s group observed spine elimination during sleep in three-week-old, i.e. pre-adolescent, mice (see Yang et al., 2012), but also that dendritic spines grow over a two-day period in response to learning (see Dec 2009 news story). For the current study, Gan looked specifically at whether learning-induced spines grew during subsequent waking or sleeping hours. 

First author Guang Yang and colleagues started off by pinpointing which spines grew in response to learning. Using transcranial two-photon microscopy, the researchers tracked the formation of spines on dendritic branches in the motor cortex of live mice after they learned to run forward on a spinning rod. Yang found that about a third of dendritic branches grew new spines a day after the training. When the researchers compared any given pair of branches from the same parent arbor, they found that typically one branch acquired more spines than its sibling. Interestingly, when researchers trained the same mice to run backward on the rod, the previously barren sibling branches sprouted new spines. This allowed them to focus in on branches that grew spines in response to specific learned tasks.

If the researchers deprived the mice of sleep for seven hours following rotarod training, then fewer spines grew on the more productive branches. The spines that grew in well-rested mice were also more likely to persist one day later, suggesting that sleep not only helps spines grow, but keeps them around. 

The Fruits of Learning.

A dendritic branch protruding from an apical tuft neuron sprouts a bounty of new spines (filled arrows) and loses a couple (open arrows) 24 hours after mice learn a new task. Sleep deprivation prevented these new spines from cropping up.

Gan and colleagues hypothesized that during sleep, the reactivation of neuronal circuits somehow stimulates spine growth. Using mice that express a calcium indicator, the investigators confirmed that neurons that fire during learning also fire again during sleep. When the researchers blocked this neuronal activation using an NMDA inhibitor following training, the branch-specific spine growth was blocked.

“Overall, this study provides for the first time the morphological evidence for sleep-dependent upregulation of synaptic function that is responsible for motor learning,” Igor Timofeev of Laval University in Quebec wrote in an email to Alzforum. “The morphological changes are extremely robust.” Timofeev was not involved in the study.

Marcos Frank of the University of Pennsylvania in Philadelphia called the study groundbreaking. “This is really strong evidence that sleep is a time when the brain is making connections, not breaking them,” he said. “The study really does challenge the dogma that has emerged in field, and it does so in a very convincing way.”

Does this phenomenon extend to other parts of the brain or to different types of memory? Frank said that depends on whether such memories are known to require sleep. Not all types of memory need sleep to sink in, Frank said; for example, fear conditioning or memories stored in the amygdala do not. However, some hippocampal-based memories are known to require sleep for optimal consolidation, and so could be subject to the same branch-specific spine growth during sleep as those in the motor cortex. Future studies using Gan’s technique could ultimately reveal whether this is the case.

Frank was also impressed by the data suggesting sleep helps maintain the persistence of spines. “It is possible that sleep is responsible not only for the growth of spines, but for maintaining the integrity of the network,” he said.

The Amyloid Factor

In addition to sculpting delicate spines on dendritic branches, sleep also modulates levels of a peptide that has the power to destroy those spines—amyloid β. Poor sleep has been identified as a risk factor for Alzheimer’s disease, and some studies have reported that Aβ levels wax and wane to the beat of the circadian rhythm—rising during the day and falling during the night (see Bateman et al., 2007, and Huang et al., 2012). In mice, this rhythm has been linked to increased clearance of Aβ and reduced neuronal activation during sleep (see Oct 2013 news story on Xie et al., 2013; May 2014 conference story;  Sep 2009 news story; and Aug 2011 news story). However, no studies have directly linked Aβ flux to sleep in humans.

Claassen, first author Sharon Ooms, and colleagues attempted this by comparing amyloid levels in cerebrospinal fluid (CSF) after a night’s sleep to after a night spent awake. The researchers used an intrathecal catheter to sample CSF from 13 middle-aged men in the evening and then again in the morning after a night’s sleep. Another group of 13 men stayed awake all night, and were sampled throughout. Levels of Aβ42 dropped by an average of 6 percent overnight in men who slept, but stayed the same in those who pulled an all-nighter. Levels of tau and Aβ40 did not differ between the groups. Concentrations of Aβ42 varied widely between the 13 participants in each group, but Claassen said those who slept more soundly in the sleep group tended to have the largest dip in CSF Aβ in the morning.

The reduction in Aβ42 may have been modest, Claassen said, but when considered in light of findings that higher levels of the peptide trigger the formation of toxic oligomers, these small changes could have big impact down the road. “If you experienced partial sleep deprivation over a long period, that could lead to a gradual accumulation of Aβ oligomers,” Claassen told Alzforum. Whether sleep deprivation boosts Aβ42 production or lessens its clearance in people has yet to be determined, he added.

Claassen sees poor sleep as one of many possible risk factors for AD, and one that may hasten the onset of disease in some people. “If we can reduce this risk factor, we may not prevent AD, but we can perhaps postpone it,” he said.

"This well-designed and -analyzed study fills an important gap regarding the association between sleep and CNS Aβ dynamics in humans,” Andrew Lim of the University of Toronto wrote in an email to Alzforum. He added that more work is needed to understand how the effect translates to real-world situations such as partial sleep deprivation or shift work. “Whether this experimental effect seen in healthy middle-aged individuals is sufficiently large to be clinically significant vis a vis later-life Alzheimer's disease risk is unknown,” he wrote (see full comment below). 

David Holtzman of Washington University School of Medicine in St Louis wrote that the study helps confirm findings in animal models (see full comment below). “Overall, the authors should be applauded for going out of their way to do a very involved study that strongly suggests that sleep deprivation in humans results in acute elevation in CSF Aβ42, and supports the hypothesis that chronic sleep deprivation could lead to an acceleration of Aβ deposition.”—Jessica Shugart


  1. This paper by Ooms and colleagues studies the effect of sleep deprivation in humans on cerebrospinal fluid (CSF) Aβ40, Aβ42, tau, p-tau, and total protein in middle-aged individuals with lumbar catheters.

    It is found that CSF Aβ42 decreased in those who had normal sleep by the morning, reflecting when Aβ was being produced in the brain during sleep; whereas it did not decrease in those who were sleep deprived.  There was no change in tau, p-tau, or Aβ40. This is interesting and is very similar to results we observed in APP transgenic mice (Kang et al., 2009).  This supports that Aβ is dynamically regulated by the sleep-wake cycle.  There would be several things to follow up on in future studies.  It is not clear why Aβ40 did not also change with sleep deprivation in this study.  Since the research subjects who were not sleep deprived did not have lumbar catheter sampling between 10 p.m. and 9 a.m., whereas those who were sleep-deprived did, one wonders if this difference in sampling affected the results in some way.  The authors controlled for this by measuring total protein, but this should be reassessed in future studies.  It might also be useful to sample CSF just in the AM (without a lumbar catheter) to reflect the overnight period of Aβ production.  With a lumbar catheter in place, this probably affects the group that was attempting to get normal sleep.  Overall, however, the authors should be applauded for going out of their way to do a very involved study that strongly suggests that sleep deprivation in humans results in acute elevation in CSF Aβ42 and supports the hypothesis that chronic sleep deprivation could lead to an acceleration of Aβ deposition.


    . Amyloid-beta dynamics are regulated by orexin and the sleep-wake cycle. Science. 2009 Nov 13;326(5955):1005-7. PubMed.

    View all comments by David Holtzman
  2. This well-designed and -analyzed study fills an important gap in our understanding of the association between sleep and CNS amyloid β (Aβ) dynamics in humans. There is a growing body of epidemiological evidence relating sleep disturbance to cognition and dementia in older persons. One hypothesis suggests that sleep deprivation/disruption may affect the production/clearance of Aβ. Prior experimental work had indicated that to be the case in rodents. One previous study in humans had looked at the temporal dynamics of CSF Aβ, but this was an observational study, limiting the ability to draw conclusions about causality. In contrast, this work by Ooms and colleagues used a randomized interventional design, providing strong evidence that at least in healthy middle-aged individuals, perturbations in sleep can, in principle, attenuate the normal nocturnal decline in CSF Aβ levels. Particular strengths include its randomized interventional design, careful and repeated serial measurements of CSF Aβ within the same individuals, and direct quantification of sleep.

    Much work remains to be done in this area.  Whether this experimental effect seen in healthy middle-aged individuals is sufficiently large to be clinically significant vis-à-vis later-life Alzheimer's disease risk is unknown.  Moreover, total sleep deprivation is unusual in the real world, and it is unclear whether partial sleep deprivation and sleep fragmentation, which are much more common, would have the same effects.  In addition, whether there is an important effect of circadian disruption (which is commonly seen in the context of shift work) on amyloid dynamics remains uncertain.

    Notwithstanding the need for additional work, this study by Ooms and colleagues bridges an important gap between prior experimental studies in rodent models and human epidemiological studies, and will provide a good foundation to gauge the importance of sleep and amyloid dynamics to clinical Alzheimer's disease risk.

    View all comments by Andrew Lim
  3. David Holtzman and Andrew Lim raise important points for future studies on this topic. Some further thoughts for those who would like to engage in a follow-up study:

    • Despite the overnight stay in a hospital-like environment (clinical research lab) and the presence of the spinal catheter, the 13 subjects who were allowed to sleep normally achieved sufficient sleep (according to our observatinos, and as seen on EEG). Nonetheless, the EEG revealed differences in the time spent in deep sleep (S3), which may have been influenced by the research setting.
    • In a future study, our advice would be to withhold nightly sampling in the sleep-deprived group. We performed these samples because we wanted to pinpoint the timing of changes in Aβ overnight. Our study clearly showed that there were no relevant Aβ dynamics between midnight and morning. Therefore, future studies could keep the sampling scheme exactly the same in the two groups (normal sleep vs. sleep deprivation). 
    • We would also advise to extend the morning sampling until noon. We stopped sampling at 10 a.m., but also saw the largest decline (in the normal sleep group) at 10 a.m. It would be interesting to know if Aβ declines even further later in the morning (i.e., the maximum decline could be at 11 a.m.).

Make a Comment

To make a comment you must login or register.


News Citations

  1. Sleep Trims Synaptic Spines in Adolescents
  2. Persistence of Dendrites Leads to Lifelong Memories
  3. From ApoE to Zzz’s—Does Sleep Quality Affect Dementia Risk?
  4. Glymphatic Flow, Sleep, microRNA Are Frontiers in Alzheimer’s Research
  5. Sleep Deprivation Taxes Neurons, Racks Up Brain Aβ?
  6. Brain Activity and Aβ—The Interstitial Plot Thickens

Paper Citations

  1. . Sleep and the price of plasticity: from synaptic and cellular homeostasis to memory consolidation and integration. Neuron. 2014 Jan 8;81(1):12-34. PubMed.
  2. . Sleep contributes to dendritic spine formation and elimination in the developing mouse somatosensory cortex. Dev Neurobiol. 2012 Nov;72(11):1391-8. Epub 2012 Jul 13 PubMed.
  3. . Fluctuations of CSF amyloid-beta levels: implications for a diagnostic and therapeutic biomarker. Neurology. 2007 Feb 27;68(9):666-9. PubMed.
  4. . Effects of age and amyloid deposition on aβ dynamics in the human central nervous system. Arch Neurol. 2012 Jan;69(1):51-8. PubMed.
  5. . Sleep drives metabolite clearance from the adult brain. Science. 2013 Oct 18;342(6156):373-7. PubMed.

Further Reading


  1. . Neuroscience. Memories--getting wired during sleep. Science. 2014 Jun 6;344(6188):1087-8. PubMed.

Primary Papers

  1. . Effect of 1 night of total sleep deprivation on cerebrospinal fluid β-amyloid 42 in healthy middle-aged men: a randomized clinical trial. JAMA Neurol. 2014 Aug;71(8):971-7. PubMed.
  2. . Sleep promotes branch-specific formation of dendritic spines after learning. Science. 2014 Jun 6;344(6188):1173-8. PubMed.