Aggregation of β-amyloid (Aβ) in the brain begins to occur years before the clinical onset of Alzheimer’s disease (AD). Before Aβ aggregation, concentrations of extracellular soluble Aβ in the interstitial fluid (ISF) space of the brain, which are regulated by neuronal activity and the sleep-wake cycle, correlate with the amount of Aβ deposition in the brain seen later. The amount and quality of sleep decline with normal aging and to a greater extent in AD patients. How sleep quality as well as the diurnal fluctuation in Aβ change with age and Aβ aggregation is not well understood. We report a normal sleep-wake cycle and diurnal fluctuation in ISF Aβ in the brain of the APPswe/PS1δE9 mouse model of AD before Aβ plaque formation. After plaque formation, the sleep-wake cycle markedly deteriorated and diurnal fluctuation of ISF Aβ dissipated. As in mice, diurnal fluctuation of cerebrospinal fluid Aβ in young adult humans with presenilin mutations was also markedly attenuated after Aβ plaque formation. Virtual elimination of Aβ deposits in the mouse brain by active immunization with Aβ 42 normalized the sleep-wake cycle and the diurnal fluctuation of ISF Aβ. These data suggest that Aβ aggregation disrupts the sleep-wake cycle and diurnal fluctuation of Aβ. Sleep-wake behavior and diurnal fluctuation of Aβ in the central nervous system may be functional and biochemical indicators, respectively, of Aβ-associated pathology.

Aβ is secreted by neurons into the extracellular space of the brain in the interstitial fluid (ISF). Using in vivo microdialysis, ISF Aβ concentrations in APP transgenic mice have been found to be closely associated with brain synaptic and neuronal activity before Aβ plaque formation ( 14 , 15 ) and also to be related to subsequent amyloid plaque formation and growth in vivo ( 16 ). We previously reported in young APPswe transgenic and wild-type mice without Aβ plaques in the brain that ISF Aβ increases during wakefulness and decreases during sleep ( 17 ), similar to the diurnal fluctuation of cerebrospinal fluid (CSF) Aβ observed in humans ( 17 – 19 ). However, whether the sleep-wake cycle and Aβ fluctuation become disrupted after Aβ aggregation in the brain is not clear. Further, the causal relationship between changes in the sleep-wake cycle and changes in Aβ metabolism is not understood. Here, we characterized the amount and quality of sleep and the degree of diurnal Aβ fluctuation across two brain regions with different vulnerability to Aβ deposition before and after the onset of Aβ aggregation in a mouse model of β-amyloidosis. We also assessed CSF Aβ concentrations over 36 hours in humans carrying mutations that cause autosomal dominant forms of AD. Moreover, we examined whether preventing Aβ aggregation was sufficient to normalize the sleep-wake cycle and biochemical abnormalities in the mouse model.

Aggregation of the β-amyloid (Aβ) peptide in the extracellular space of the brain is one of the pathological hallmarks of Alzheimer’s disease (AD). Aβ is produced from amyloid precursor protein (APP) by sequential cleavage by the β- and γ-secretases ( 1 – 3 ) and exists as a soluble, monomeric form throughout life ( 4 ). Monomeric Aβ begins to aggregate in the human brain ~10 to 15 years before the clinical symptoms and signs of AD become apparent, by which time a substantial amount of neuronal and synaptic loss in several brain regions is present ( 5 , 6 ). In humans with Aβ aggregation in the brain who are clinically asymptomatic (preclinical AD), there is evidence of decreased functional connectivity in brain networks affected by amyloid deposition ( 7 , 8 ). There may be other functional changes associated with Aβ aggregation during preclinical AD ( 9 ). Identification of such factors that might be cognitive, behavioral, or physiological measures will be important so as to better assess functional impairment during this period of time as well as to assess responses to new disease-modifying therapies as they become available.

Chronological changes of absolute concentrations of ISF Aβ 42 and lactate in the hippocampus and striatum of mice and association between CSF Aβ 42 and amyloid plaque deposition in humans. ( A and D ) Absolute concentrations of ISF Aβ x –42 in the hippocampus (A) and striatum (D) of 3- and 9-month-old APPswe/PS1δE9 mice (n = 5 to 6 per group; Mann-Whitney test). ( B and E ) Absolute concentrations of ISF lactate in the hippocampus (B) and striatum (E) of 3-, 6-, and 9-month-old APPswe/PS1δE9 mice (n = 6 per group; one-way ANOVA, Tukey’s post hoc test). ( C ) Comparison of absolute values of CSF Aβ 42 in nonmutation carriers (NC), mutation carriers without amyloid plaque deposition (MC PiB − ), and mutation carriers with amyloid plaque deposition (MC PiB + ) (n = 4 per group; Kruskal-Wallis test). ( F ) Correlation between absolute concentrations of CSF Aβ 42 and amount of amyloid plaque deposition measured by mean cortical PiB binding potential (MCBP) (n = 12 paired measurement; Pearson’s correlation test). *P < 0.05. Values represent means ± SEM.

We also investigated whether ISF Aβ 42 behaved similarly to Aβ 1–x and Aβ x –40 in both hippocampus and striatum. At 3 months, ISF Aβ x –42 showed diurnal fluctuation, which was lost by 9 months in APPswe/PS1δE9 mice (fig. S6). PBS-treated mice showed loss of diurnal fluctuation at 9 months similar to that seen in 9-month-old untreated APPswe/PS1δE9 mice (fig. S7, A, C, E, and G). In contrast, APPswe/PS1δE9 mice actively immunized with Aβ 1–42 had normal diurnal fluctuation of ISF Aβ x –42 (fig. S7, B, D, F, and H) and a marked decrease of amyloid plaques in the brain (fig. S8). Also notable was that the absolute amount of ISF Aβ 42 decreased between 3 and 9 months in the hippocampus in APPswe/PS1δE9 mice ( Fig. 7A ). In contrast to Aβ, concentrations of ISF lactate, which does not aggregate in the brain, continue fluctuating at 6 months with no change in absolute concentrations through 9 months ( Fig. 7 , B and E). Absolute concentrations of CSF Aβ 42 in humans also decreased in those with Aβ plaque formation. CSF Aβ 42 was highest in nonmutation carriers and lowest in PiB + mutation carriers, and the amount of CSF Aβ 42 inversely correlated with the amount of fibrillar Aβ in the brain as measured by amyloid imaging ( Fig. 7 , C and F).

Lactate is an indicator of neuronal activity both in vitro and in vivo that is increased during wakefulness and decreased during sleep ( 15 , 21 , 22 ). We measured ISF lactate in the hippocampus and striatum of APPswe/PS1δE9 mice and found differences across the dark and light phases. To investigate whether lactate concentrations could be a biological indicator of wakefulness irrespective of Aβ pathology in the brain, we compared ISF lactate concentrations and the amount of wakefulness in APPswe/PS1δE9 mice at different ages. ISF lactate concentrations showed a significant correlation with the amount of wakefulness at 3, 6, and 9 months (fig. S3). Additional analysis of the amplitude of diurnal fluctuation of lactate as assessed by cosinor analysis ( 23 ) showed a decrease in amplitude by 9 months ( Fig. 6 ) corresponding to an increase in wakefulness. To further investigate a potential causal relationship between changes in Aβ metabolism and changes in brain neuronal activity, we compared the chronological changes in correlation between ISF Aβ and ISF lactate in relation to Aβ accumulation. Hippocampal concentrations of ISF lactate and ISF Aβ correlated at 3 and 6 months, but the correlation was lost by 9 months (fig. S4, A to C). This was validated in PBS-treated APPswe/PS1δE9 mice, where there was no correlation between ISF lactate and ISF Aβ at 9 months (fig. S5A). In contrast, APPswe/PS1δE9 mice actively immunized with Aβ 1–42 maintained a significant correlation between concentrations of ISF lactate and ISF Aβ in the hippocampus and striatum at 9 months (fig. S5, B and D). These data demonstrate a strong relationship between the sleep-wake state and the neuronal activity even after the acquisition of Aβ pathology and suggest that the cause of ISF Aβ fluctuation disruption is more likely due to the biochemical effect of the formation of Aβ plaques resulting in sequestration of ISF Aβ rather than a change in neuronal activity associated with the sleep-wake cycle.

Sleep-wake patterns and diurnal fluctuation of ISF Aβ in 9-month-old PBS-treated and Aβ 42 -immunized APPswe/PS1δE9 mice. ( A and G ) Sleep-wake patterns in 9-month-old PBS-treated (A) and Aβ 42 -immunized (G) APPswe/PS1δE9 mice across 2 days (two light-dark periods) shown as minutes awake per hour. ( D and J ) Comparison of minutes awake per hour between the dark and the light periods in each group (n = 5 to 6 per group; two-tailed t test). ( B and H ) Diurnal fluctuation of ISF Aβ 1–x in the hippocampus of 9-month-old PBS-treated (B) and Aβ 42 -immunized (H) APPswe/PS1δE9 mice across 2 days presented as percent average of absolute values of ISF Aβ 1–x . ( E and K ) Comparison of percent average of absolute values of ISF Aβ 1–x in the hippocampus between the dark and the light periods (n = 5 to 6 per group; two-tailed t test). ( C and I ) Diurnal fluctuation of ISF Aβ 1–x in the striatum of 9-month-old PBS-vaccinated (C) and Aβ 42 -vaccinated (I) APPswe/PS1δE9 mice across 2 days. ( F and L ) Comparison of percent average of absolute values of ISF Aβ 1–x in the striatum between the dark and the light periods (n = 5 to 6 per group; two-tailed t test). *P < 0.05; ***P < 0.001. Values represent means ± SEM.

Aβ plaque deposition in the hippocampus and striatum of 9-month-old APPswe/PS1δE9 mice treated with PBS or immunized with Aβ 42 . ( A to D ) Representative brain sections of the hippocampus (A and C) and striatum (B and D) of mice from each group stained with HJ3.4 antibody to visualize Aβ-immunoreactive plaques (Aβ-IR). ( E and F ) Amount of Aβ deposition in the PBS-treated mice and Aβ 42 -vaccinated mice are shown with amount of Aβ deposition in 6- and 9-month-old APPswe/PS1δE9 mice in the hippocampus (E) and striatum (F) (n = 5 to 6 in each group; two-tailed t test). ***P < 0.001. n.s., not statistically significant. Values represent means ± SEM. Scale bar in (A), 500 μm.

To investigate whether Aβ aggregation is responsible for the changes in sleep amount and quality as well as the attenuation of the diurnal fluctuation of ISF Aβ, we actively immunized APPswe/PS1δE9 mice starting at 1.5 months and then monthly with subcutaneous injections of synthetic Aβ 1–42 or phosphate-buffered saline (PBS) and compared the patterns of ISF Aβ and the sleep-wake cycle at 9 months. PBS-treated APPswe/PS1δE9 mice showed a pattern of Aβ plaque deposition similar to untreated 9-month-old APPswe/PS1δE9 mice ( Fig. 4 , A and B). Diurnal fluctuation of ISF Aβ was absent in the hippocampus of PBS-treated animals, and the mice had strongly disrupted sleep-wake patterns ( Fig. 5 , A, B, D, and E). In contrast, APPswe/PS1δE9 mice actively immunized with Aβ 1–42 showed markedly decreased Aβ deposits in the brain ( Fig. 4 , C and D) and exhibited both a normal sleep-wake cycle and a diurnal fluctuation of ISF Aβ at 9 months ( Fig. 5 , G to L). During the light period, PBS-treated mice were awake 29.6 ± 4.1 min/hour, whereas Aβ 42 -vaccinated mice were awake 17.2 ± 2.4 min/hour (P = 0.0256) ( Fig. 5 , A, D, G, and J). Wild-type littermates had a normal sleep-wake cycle and diurnal fluctuation of endogenous Aβ through 9 months (fig. S1, G to L).

In addition to a loss of diurnal fluctuation of ISF Aβ seen in APPswe/PS1δE9 mice, we also observed attenuation of the diurnal pattern of Aβ in the CSF of humans with mutations in the presenilin (PS) gene ( Fig. 3 ). These individuals also had Aβ deposition as detected by amyloid imaging with Pittsburgh Compound B (PiB). Cosinor analysis is a statistical method to fit a (co)sine wave to data sets obtained over time to capture a circadian oscillation pattern. Cosinor analysis was used to assess the diurnal patterns of CSF Aβ dynamics in mutation + , PiB − ; mutation + , PiB + ; and age-matched groups without a PS mutation (mutation − ) with mean-adjusted group average data after linear trend subtraction ( 19 ). Significant cosinor patterns were found in both the mutation − and the mutation + , PiB − groups (P < 0.05), but not in the mutation + , PiB + group (P > 0.05) ( Fig. 3 , A to F). Results from mutation carriers who are PiB − showed an attenuation of diurnal fluctuation of CSF Aβ 42 , in which the degree of attenuation is in between that seen in noncarriers and mutation carriers who are PiB + ( Fig. 3 , A to C).

Chronological changes in sleep-wake patterns and diurnal fluctuations of ISF Aβ in APPswe/PS1δE9 mice. ( A to C and G to I ) Diurnal changes of ISF Aβ 1–x in APPswe/PS1δE9 mice at 3, 6, and 9 months across 2 days shown as percent average of 2 days of absolute values of ISF Aβ 1–x in the hippocampus (A to C) and striatum (G to I). ( D to F and J to L ) Comparison of percent average of 2 days of ISF Aβ 1–x between dark and light periods in the hippocampus (D to F) and striatum (J to L) for each age group (n = 6 to 8 per group; two-tailed t test). ( M and N ) Absolute concentrations of ISF Aβ 1–x in the hippocampus (M) and striatum (N) of 3-, 6-, and 9-month-old APPswe/PS1δE9 mice [n = 6 to 8 per group; one-way analysis of variance (ANOVA), followed by Tukey’s post hoc test]. *P < 0.05; **P < 0.01; ***P < 0.001. Values represent means ± SEM.

APPswe/PS1δE9 mice ( 20 ) at 3 months of age, before Aβ deposition begins, displayed a diurnal fluctuation of ISF Aβ in the hippocampus and in the striatum ( Fig. 1 , A, D, G, and J) and a normal sleep-wake cycle ( Fig. 2A ), similar to that seen in wild-type littermates (fig. S1, A to F). APPswe/PS1δE9 mice begin to deposit substantial amounts of Aβ plaques in the hippocampus by 6 months of age, whereas striatal deposition is not detectable until 9 months (fig. S2). At 6 months, diurnal fluctuation of ISF Aβ was disrupted in the hippocampus ( Fig. 1 , B and E) but maintained in the striatum ( Fig. 1 , H and K). There was also a trend for an increase in wakefulness and a decrease in sleep during the light phase ( Fig. 2 , B and G to I) at 6 months. At 9 months, when Aβ plaques were increased to a greater extent in hippocampus and now were present in striatum (fig. S2, E and F), loss in diurnal fluctuation of ISF Aβ was observed in both brain regions ( Fig. 1 , C, F, I, and L). There was also a marked disruption of the sleep-wake cycle with significantly increased wakefulness and decreased rapid eye movement (REM) and non-REM (NREM) sleep ( Fig. 2 ). Absolute concentrations of ISF Aβ were decreased in the hippocampus and remained unchanged in the striatum ( Fig. 1 , M and N).

Discussion

Here, we found an Aβ accumulation–associated disruption of the sleep-wake cycle and loss of diurnal fluctuation of ISF Aβ in a mouse model of AD amyloidosis. Similar findings, namely, loss of diurnal fluctuation of CSF Aβ associated with amyloid deposition, were also seen in humans with mutations that cause autosomal dominant AD. These changes were not seen in age-matched wild-type mice in ISF nor in age-matched humans lacking PS mutations in CSF, suggesting that overexpression of APP/PS1 transgenes associated with autosomal dominant AD in mice or some form of Aβ aggregation was responsible. The findings that active immunization with Aβ 42 prevented the changes in sleep disruption as well as diurnal fluctuation of Aβ in mice and that aged littermates maintained both sleep-wake cycle and diurnal fluctuation of Aβ strongly suggest that Aβ accumulation rather than overexpression of transgenes is responsible for these changes.

The sleep-wake cycle is a fundamental property of the brain. Diurnal fluctuation of Aβ is a physiologic finding observed in brains of mice and humans that occurs before the development of Aβ plaque deposition. We found that the diurnal fluctuation of ISF Aβ was disrupted sequentially in line with a hierarchical deposition of Aβ plaques in different brain regions. Because changes in Aβ fluctuation occurred before the changes in sleep quality and perturbation of neuronal activity, the results suggest that the changes in ISF Aβ fluctuation are likely due to a biochemical change in Aβ metabolism induced by plaque formation rather than changes in the sleep-wake cycle itself. Using mouse models of β-amyloidosis, we previously showed that sleep deprivation and administration of the neuropeptide orexin, which regulates arousal and wakefulness, acutely increased ISF Aβ, and chronic sleep deprivation strongly increased Aβ plaque formation (17). Blocking orexin receptors acutely and chronically decreased ISF Aβ and plaque formation (17). These results suggested the possibility that sleep disruption and disorders might be a risk factor for the development of Aβ deposition and possibly AD. Emerging evidence in humans suggests that this may be the case (24, 25), although additional longitudinal studies with biomarkers are required because studies with elderly participants could not assess a cause-effect relationship between AD pathology and changes in sleep. Although disrupted sleep has the potential to lead to Aβ aggregation, our current results suggest that once Aβ aggregates, some of the damage it induces in the central nervous system leads to dysregulation of the sleep-wake cycle. Previous reports on AD patients and mouse models of Aβ amyloidosis support the possibility that the presence of Aβ-related pathology in the brain is associated with a disrupted sleep-wake cycle or circadian rhythm by affecting molecules including orexin, melatonin, and associated brain regions (26–32). Thus, there could be a positive feedback loop between the sleep-wake cycle and the Aβ metabolism. The early increase in wakefulness possibly initiated by the aggregation of Aβ may accelerate Aβ accumulation, which may lead to further neuronal dysregulation and increase sleep-wake cycle abnormalities.

Observations from this study also demonstrate that at least the initial changes in sleep and Aβ fluctuation are likely due to Aβ aggregation and not the effect of aging. Recent human studies assessing CSF noted that diurnal fluctuation of CSF Aβ in young adults was attenuated in older adults who had a mean age of 73.4 years. Whether this attenuation was due to aging or Aβ aggregation was not clear (19). The oldest APPswe/PS1δE9 mice used in this study only reached the equivalent of middle age, and wild-type littermates at the same age did not have sleep-associated impairments. Because the mouse model we are using expresses mutations found in humans with dominantly inherited AD who begin to have AD pathology as young adults, we further investigated presymptomatic individuals within autosomal dominant AD families. In those young subjects (mean age of 42.4 years) with and without Aβ pathology in the brain, we demonstrated that the aggregation of Aβ in the brain is associated with the attenuation of diurnal fluctuation of Aβ in human CSF. This attenuation was not present in age-matched siblings that lacked these mutations. This suggests that changes in brain Aβ metabolism associated with amyloid plaque formation induce attenuation in the fluctuation of CSF Aβ independent of an effect of aging. Although the change in diurnal fluctuation in CSF Aβ in presenilin carriers with versus without amyloid deposition is significant, it is small, suggesting that it may be difficult to be used as a biomarker. However, if the changes in sleep quality and amount seen in APPswe/PS1δE9 mice are also present in humans, this may provide a useful quantitative and functional endophenotype during the period of preclinical AD that can be assessed in response to therapeutic intervention.

Aβ aggregation was also associated with disruption of homeostatic fluctuation of neuronal activity in the brain. Before Aβ plaque deposition, there was a strong correlation between concentrations of ISF Aβ and ISF lactate, suggesting neuronal activity–associated release of Aβ within the brain (14, 15, 33). The correlation between ISF Aβ and ISF lactate, however, was lost after substantial accumulation of Aβ in the brain. Sequential loss of correlation between ISF lactate and ISF Aβ with a decrease in absolute amount of ISF Aβ indicates that changes in the equilibrium between ISF Aβ and Aβ plaques may cause dissociation between ISF Aβ concentrations and neuronal activity. The decrease in absolute concentrations of ISF Aβ and CSF Aβ in humans with Aβ plaque formation suggests that soluble ISF and CSF Aβ is being sequestered by amyloid plaques, consistent with other studies (4, 34). On the other hand, lactate, which does not aggregate within the brain, did not show changes in absolute concentration in ISF. The significant decrease in the fluctuation of lactate in both hippocampus and striatum by 9 months in APPswe/PS1δE9 mice occurred in concert with the changes in the sleep-wake cycle, the most prominent change of which was a marked increase in wakefulness during the light phase by 50%, a time when the animals would otherwise be sleeping most of the time. This increase in wakefulness may be very damaging to the brain in an additive fashion to other Aβ-linked pathways of damage, because many studies have shown the important function of sleep to learning, memory, synaptic plasticity, and risk for other medical disorders (35–38). It is possible that the Aβ-induced changes to sleep are due to local cortical and hippocampal Aβ-induced changes to synaptic activity and excitability occurring throughout affected brain regions (39–41). Notably, at all ages assessed in our studies, APPswe/PS1δE9 mice had no phenotypic or electroencephalogram evidence of seizures. Thus, the disruption of the sleep-wake cycle, which may be due to synaptic alterations, was not secondary to seizures.

Sleep-wake patterns of the human subjects who participated in this study were not investigated. Sleep changes in young adults with autosomal dominant AD before or after Aβ pathology in the brain will be important in future studies to determine whether similar changes in the sleep-wake cycle are present because Aβ pathology develops in the preclinical stages of disease. Because changes in the sleep-wake cycle and Aβ fluctuation in both mice and humans were all in the presence of presenilin mutations, it will be important in the future to assess whether the same changes occur in the absence of such mutations. Thus, it will also be interesting to determine whether changes in the sleep-wake cycle are present in the preclinical stages of late-onset AD, which would have important implications both diagnostically and for therapeutic assessment.