Effects of Age and Amyloid Deposition on Aβ Dynamics in the Human Central Nervous System

Abstract

Objective

The amyloid hypothesis predicts that increased production or decreased clearance of amyloid beta (Aβ) leads to amyloidosis, ultimately culminating in Alzheimer’s disease (AD). Dynamic changes in human CNS Aβ levels may be altered by aging or AD pathology and contribute to the risk of AD.

Designs

In this study, hourly cerebrospinal fluid (CSF) Aβ concentrations were compared with age, PIB PET amyloid status and electroencephalography (EEG) and video recording data.

Results

Linear increases of CSF Aβ concentrations over time were observed in younger control participants and older Amyloid- participants, but not in older Amyloid+ participants. Significant CSF Aβ circadian patterns were observed in younger control participants; however circadian amplitudes were decreased in both Amyloid- and Amyloid+ older participants. Aβ diurnal concentrations were correlated to the amount of sleep, but not various awake activities.

Conclusions

Decreased linear rise of CSF Aβ levels associated with amyloid deposition, and decreased CSF Aβ diurnal pattern associated with increasing age disrupt the normal physiology of Aβ dynamics, and may contribute to AD.

Introduction

The amyloid-beta (Aβ) peptide has been implicated as a critical initiator of AD.¹ Pathologic studies of AD brain tissue demonstrate that extensive amyloid plaques associated with disrupted neuropil are deposited throughout the cortex. Aβ peptides are the primary component of amyloid plaques, which account for a thousand-fold increase of Aβ in AD brain. Increased brain Aβ in AD has been postulated to be caused by increased Aβ production or decreased clearance.^2,3 Thus, the study of Aβ in AD is likely to lead to a better understanding of AD patho-physiological changes, as well as normal Aβ physiology.

Aβ in the brain is produced predominantly by neurons by the cleavage of the amyloid precursor protein (APP) by beta and gamma-secretases. Aβ travels by diffusion and bulk flow to the cerebrospinal fluid (CSF) via interstitial fluid drainage pathways. CSF Aβ can be sampled as a biomarker of amyloidosis and can be used to diagnose and predict AD with 70–95% accuracy.^4,5 Compared to controls, concentrations of CSF Aβ42 are consistently decreased by approximately half in AD.⁶ Repeat measurements of CSF over months to years demonstrate stable CSF Aβ42 concentrations in AD,^7,8 but less stable levels of Aβ40.⁹ However, in younger healthy participants, hourly CSF sampling demonstrates highly dynamic and variable CSF Aβ40 and Aβ42 concentrations.¹⁰ Consistent with these dynamic CSF Aβ changes, microdialysis measurements of human brain tissue also demonstrates highly dynamic Aβ changes.¹¹

One hypothesis suggests that neuronal activity is responsible for Aβ dynamics.^12,13 Modulation of neuronal activity by electrical, pharmacological, and behavioral interventions has direct effects on CNS Aβ concentrations.^13,14 For example, increased stress and decreased sleep have both been demonstrated to increase Aβ concentrations in animal models.^15,16 Further, a recent study demonstrated the role of sleep on several synaptic markers, suggesting sleep may modulate the metabolism of a number of CNS proteins.¹⁷

Circadian rhythms have been described for a variety of biochemical, physiological and behavioral processes that occur over a 24-hour cycle.¹⁸ Examples of circadian rhythms include body temperature, circulating levels of hormones such as cortisol, and blood levels of ions such as sodium. CSF Aβ demonstrates a circadian pattern in healthy younger participants,¹⁰ however, the dynamics of Aβ in aging and AD are less well understood. Although prior reports indicate that Aβ42 is stable in AD, little is known about the effects of age and amyloid deposition on Aβ dynamics in the human central nervous system (CNS). As age is the largest risk factor for AD, understanding changes in the dynamics of Aβ with aging may inform about the patho-physiological processes which lead to amyloidosis and ultimately AD. Further, the effects of amyloidosis may reveal changes in CNS Aβ dynamics that are associated with the pathology of AD.

In this study, we investigated CSF Aβ dynamics and the effects of aging and amyloidosis. Hourly CSF samples were collected from each participant, and continuous electroencephalography (EEG) and video recordings were obtained from a subset of study patients as diagrammed in Figure 1. CSF Aβ dynamics were modeled and compared to age, amyloid deposition, sleep, and awake behaviors. This study confirms CSF Aβ dynamics over time, describes associations between Aβ dynamics and sleep, age, and amyloid deposition, and provides insight into normal CNS Aβ changes over time and the effects of age and amyloid deposition.

Figure 1.

Diagram of a volunteer during a study of cerebrospinal fluid (CSF) amyloid-β (Aβ) dynamics. CSF was sampled from a lumbar intra-thecal catheter every hour for 36 hours while continuous Electroencephalogram (EEG) and video were recorded. CSF Aβ40 and Aβ42 concentrations were measured using enzyme-linked immunosorbent assay (ELISA), and analyzed over time for Aβ dynamics.

Materials and Methods

Study Design

This was a repeated measures case-control study conducted at the Washington University School of Medicine in St. Louis. Three groups of volunteers were enrolled for this study: 1) a case group with positive amyloid plaque by PIB PET imaging (Amyloid+)¹⁹; 2) an age-matched control group with negative amyloid plaque (Amyloid-); and 3) a younger control group (YNC). PIB binds to amyloid plaques in the brain, and the binding potentials of this compound for the prefrontal cortex, precuneus, lateral temporal cortex, and gyrus rectus were averaged to yield the mean cortical binding potential (MCBP) for each participant ²⁰. A MCBP of 0.2 or greater was considered amyloid plaque positive (Amyloid+), and a MCBP of less than 0.2 was considered amyloid plaque negative (Amyloid-).

Participants in the Amyloid+ and Amyloid- group were older than 60 years and were enrolled in the Washington University Alzheimer’s Disease Research Center (ADRC). Younger controls were between the ages of 18 and 60 years. All participants were in good general physical health and had no other clinical neurological diseases. Volunteers with active infections, bleeding disorders, or those who were treated with anticoagulants were excluded from this study. All human study protocols were approved by the Washington University Human Studies Committee and the General Clinical Research Center Advisory Committee. Informed consent was obtained from all participants.

Demographics of Study Participants

A total of 46 participants were analyzed in this study, among whom 53% were women and 47% men. The race composition of the study volunteers was 78% Caucasian and 22% African-American. The proportion of African-Americans was significantly higher in the YNC group as compared to the other two groups (72% in YNC versus 7% in Amyloid- and 9% in Amyloid+ group, p<0.05). However, no statistical difference in Aβ dynamics was found between races within the YNC group. The average ages for the younger normal controls (YNC) , the older amyloid negative (Amyloid- ) group, and the older amyloid positive (Amyloid+) group were 35.5±10.7 (N=20), 71.0±6.0 (N=15), and 76.7±7.7 (N=11) years, respectively. There was no statistical difference in age between the Amyloid+ and Amyloid- groups (p=0.77). ApoE genotypes were available for 28 participants, among whom 39% were found to have E3/E3 alleles (N=11), 35% E3/E4 (N=10), 14% E4/E4 (N=4), 7% E2/E3 (N=2) and 5% E2/E4 (N=1). The prevalence of one or more E4 allele was 67% in the amyloid+ and YNC groups, and 30% in amyloid- group. In the amyloid- group, 14 of the 15 participants were not clinically demented (CDR=0), and 1 participant had a CDR score of 0.5. In the amyloid+ group, 6 were clinically demented (5 with CDR=0.05, and 1 with CDR=1) and 5 were not (CDR=0).

Sample Collection

An intrathecal lumbar catheter was placed between 7:30 AM and 9:00 AM and sample collection started between 8:00 AM and 9:30 AM in all participants. Six milliliters of CSF were obtained each hour for 36 hours. CSF aliquots were frozen at -80°C immediately after collection in 1ml polypropylene tubes. Participants were encouraged to stay in bed and were allowed free choice of when to sleep, read, watch television, or talk throughout the study. Participants had meals served at 9:00 AM, 1:00 PM, and 6:00 PM.

CSF Analysis

One ml of CSF from each collection hour was thawed, and Aβ40 and Aβ42 were measured by ELISA.¹⁰ Briefly, 2G3 (anti-Aβ40) and 21F12 (anti-Aβ42) antibodies were used as the capture antibodies, and biotinylated 3D6 antibody (anti-Aβ_1-5) was used as the detection antibody. Each sample was assessed in duplicate. All samples from each participant were measured together on the same ELISA plate to avoid inter-plate variation. To measure the effect of ELISA assay variability, we ran separate ELISA plates for Aβ40 and Aβ42 with a single CSF sample for both assays. The means of the intra-sample coefficient of variation for duplicates was 10.4% for Aβ40 and 5.9% for Aβ42. Similarly, total protein levels for each sample were measured using BCA assay. The mean of the intra-sample coefficient of variation for duplicates was 2% for total protein.

Sleep staging and Electroencephalography (EEG)

EEG data were collected for the YNC group. TrackIt ambulatory equipment (Lifelines Ltd) was used to record signals from six EEG electrodes (F3, F4, C3, C4, O1, & O2), chin electromyogram, LOC and ROC (left and right electrooculogram). This information was then imported and scored using Polysmith 6.0 (Nihon Kohden, Japan) using standard sleep scoring criteria. ²¹ Sleep stage (wake, REM sleep, N1, N2, or N3) was scored for each 30 second epochs. Total sleep time (including REM, N1, N2, and N3) was binned every hour and expressed as minutes of sleep per hour.

Video Recording

Participant activity was video recorded using a Logitech Quickcam installed on a laptop computer for the duration of the study. Recording began shortly after the intrathecal catheter was inserted and continued for a 36 to 48 hour period. Videos were reviewed, and participant activity was coded in 30 second intervals using Microsoft Excel 2007. Activities were rated as sleeping, talking, eating, reading, television, defecation/urination, writing, computer use, and sample draw and catheter manipulation (see eTable 1). The position of the participant was coded as upright (>60 degrees), partially upright (15-60 degrees), or flat (<15 degrees). After the videos were reviewed, they were quality-checked for accurate recording of the sample draw using patient charts. All video coders reviewed the same “test video” and Cohen's kappa coefficients were used to measure inter-rater agreement of video coding. The weighted kappa coefficients among three different coders were 0.77 to 0.79.

Statistical Analysis

All analyses were performed using SAS version 9.2 (Statistical Analysis Software, North Carolina). Graphs were plotted in GraphPad Prism version 4.03 for Windows (GraphPad Software, San Diego California USA). Linear changes and circadian rhythms of Aβ fluctuation over time were explored in this study. Both individual time-course data for each patient and group-averaged data were used in circadian pattern recognition. Group–averaged data was calculated as follows: (1) for each patient, mean Aβ40 and Aβ42 levels over a 36-hour period were calculated; (2) mean-adjusted Aβ40 and Aβ42 levels were estimated for each time point and expressed as percentage of the mean; (3) hourly serial values of mean-adjusted Aβ40 and Aβ42 levels were grouped based on their amyloid plaque status, and an hourly average was calculated for each group. Mean levels of Aβ40, Aβ42, and the Aβ42/ Aβ40 ratios over the 36-hour study period for each participant were compared between the YNC, Amyloid- and Amyloid+ groups using ANOVA.

Cosinor Analysis

Single cosinor analysis was used to analyze the patterns of 36-hour Aβ40 and Aβ42 levels in each participant. A cosine transformation was applied to the time variable using 24 hours as the default circadian cycle, and the PROC NLIN procedure in SAS was used to estimate the parameters of the circadian patterns for Aβ fluctuations. Mesor (midline of the Aβ oscillation), amplitude (distance between the peak and mesor) and acrophase (the time corresponding to the peak of the curve) were calculated for each patient and averaged within each group. ANOVA was used to assess the differences in mesor and amplitude among different groups. Similarly, group-averaged data was used to estimate the parameters of circadian rhythms in the three study groups.

Results

Decreased Aβ42 Variability in Alzheimer’s disease

To compare the effects of age and amyloid deposition on hourly fluctuations, the YNC group was compared to older Amyloid- and Amyloid+ groups, as determined by PET PIB.²⁰ Variability of Aβ in each patient was calculated as the standard deviation of serial Aβ measurements over time, and the 36-hr mean Aβ concentration was averaged by group (YNC, Amyloid-, Amyloid+) as shown in Tables 1 and 2.

Table 1.

36-hr mean and standard deviation (SD) for amyloid- β (Aβ)42 (A) and Aβ40 (B) in three groups.

A
Group	Aβ42
	36-hr mean (pM)	36-hr SD (pM)
YNC (N=20)	213.8±80.6	54.2±25.5
Amyloid- (N=15)	226.2±160.4	42.5±22.6
Amyloid+ (N=11)	82.7±53.8**	14.9±9.9**

B
Group	Aβ40
	36-hr mean (pM)	36-hr SD (pM)
YNC (N=20)	2001.0±910.7	522.6±167.1
Amyloid- (N=15)	2590.2±1665.9	448.3±260.4
Amyloid+ (N=11)	2459.3±1187.7	425.1±309.2*

Values presented are mean±SD. Comparisons were made between Amyloid- and Amyloid+. Statistically significant differences were marked as ** (p<0.01).

Table 2.

Comparisons of linear rise and cosinor parameters for amyloid- β (Aβ)42 (A) and Aβ40 (B) among three groups using YNC as the reference.

A
Group	Aβ42	Aβ42 Circadian Pattern
	% Linear Rise/24hrs	Mesor (pM)	Amplitude (pM)	Amplitude/Mesor
YNC (N=20)	17.4%±18.8%	215.4±81.6	33.3±22.6	15.0%±7.2%
Amyloid- (N=15)	24.4%±23.5%	226.7±161.2	15.6±8.3**	8.3%±4.4%**
Amyloid+ (N=11)	7.3%±18.4% #	82.4±54.1**	6.3±5.3**	7.1%±2.7%**

B
Group	Aβ40	Aβ40 Circadian Pattern
	% Linear	Mesor (pM)	Amplitude (pM)	Amplitude/Mesor
YNC (N=20)	18.9%±22.5%	2012.9±917.8	317.6±187.2	16.9%±7.8%
Amyloid- (N=15)	23.5%±24.5%	2599.7±1682.6	180.3±115.5*	9.8%±6.2%**
Amyloid+ (N=11)	14.5%±13.8%	2441.2±1180.3	226.3±221.7	11.3%±9.5%

Values presented are mean±SD. Statistically significant differences were marked as * (p<0.05) and ** (p<0.01).

^#demonstrated a statistical difference (p<0.05) between Amyloid+ compared to PIB- and YNC combined.

As expected, CSF Aβ42 concentration was lower in the Amyloid+ group, compared to the Amyloid- group (Table 1, 63% lower, p<0.01). In addition, there was also a threefold decrease in the variability (standard deviation) of Aβ42 hourly changes in the Amyloid+ group compared to either the Amyloid- or YNC groups (p<0.01). There was a non-significant trend towards decreased variability in the cognitively impaired Amyloid+ group (n=5, Aβ40 260pM, Aβ42 9.7pM) compared to the cognitively normal Amyloid+ group (n=6, Aβ40 562pM p=0.11, Aβ42 19.2pM p=0.30). However, there was no difference in variability between the Amyloid- and YNC groups, indicating the reduction in Aβ42 and Aβ42 variability was due to amyloid status, but not age.

No difference was found in mean Aβ40 or mean Aβ40 variability (Table 2, p>0.05) between Amyloid- and Amyloid+ groups; however there was lower variability in the Amyloid+ group compared to the YNC group (p<0.05). Individual plots of Aβ highlight the significant decrease in hourly variability between Amyloid+, Amyloid-, and YNC groups (eFigure 1).

We explored CSF Aβ dynamics with respect to linear change over time and circadian rhythm. In the younger control group, a linear increase in the average Aβ concentration was observed over time (Figure 2 upper panel). Circadian patterns remained after the linear trend was removed (Figure 2 lower panel).

Figure 2.

Average mean-adjusted amyloid- β (Aβ)42 levels over time in participants <60 years (N=20). A linear increase and circadian pattern in Aβ42 over the duration of the study was observed (upper panel), and the Aβ42 circadian patterns remained after the linear trend was removed (p<0.05) (lower panel).

Decreased Aβ Linear Rise with Amyloid Deposition

In order to compare the associations between the linear Aβ rise, age, and amyloid deposition, calculations of linear rise were expressed as the percent change over 24 hours for YNC, Amyloid-, and Amyloid+ participants. The average percent linear rise for Aβ42 per 24 hours was 17% for YNC, 24% for Amyloid-, and 7% for Amyloid+ (Table 3). The Amyloid+ group demonstrated a 66% lower Aβ42 linear rise when compared to the combined results of the Amyloid- and YNC groups (p<0.05). Furthermore, with increased amyloid deposition, as measured by the mean cortical binding potential of PIB, there was less linear rise in Aβ42 over 24 hours (Figure 3). Most participants with amyloid deposition (Amyloid+) had no significant Aβ42 linear rise. Conversely, most participants without amyloid deposition by PIB PET (Amyloid-) or those younger participants unlikely to have amyloid deposition by virtue of their age (YNC) had significant Aβ42 linear rise.

Figure 3.

Percent Aβ42 rise per 24 hours in each participant by mean cortical binding potential (MCBP) of PIB. In general, individuals without amyloid deposition (young normal controls, blue circles; and older cognitively normal controls, green squares) had significant Aβ42 linear rise, independent of age; while participants with amyloid deposition (red triangles) had lower Aβ42 linear rise (p<0.05).

For Aβ40, the average percent linear rise per 24 hours was 19% for YNC, 24% for Amyloid-, and 15% for Amyloid+ (Table 4). Although the Amyloid+ group had a lower average Aβ40 rise, the trend was not statistically different from the Amyloid- and YNC groups (p=0.20).

Decreased Aβ Circadian Patterns with Age and Amyloid Deposition

Cosinor analysis was used to assess the circadian patterns of Aβ dynamics in each individual participant. Mesor (midline of cosinor fit), amplitude (difference between mesor and peak of the cosinor fit) and amplitude-to-mesor ratio for each cosinor analysis were compared across groups with YNC as the reference group. Aβ42 circadian amplitudes were decreased by 53% in the older Amyloid- group and by 81% in the older Amyloid+ group compared to the younger normal control group (Table 3, p<0.01). In addition, Aβ42 circadian amplitude to mesor ratios were decreased by 45% in the older Amyloid- group and 53% in the older Amyloid+ group compared to the younger normal control group (Table 3, p<0.01). The Aβ40 cosinor amplitude and the amplitude to mesor ratio were 40% lower in the older Amyloid- group compared to the YNC group (Table 4, p<0.05). Thus, the largest decrease in circadian pattern was due to age, with amyloid deposition having a lesser effect.

To further explore the relationship between age and Aβ42 circadian rhythms, we plotted each participant’s Aβ42 circadian amplitude versus age (Figure 4). The amplitude of the Aβ42 circadian pattern was inversely correlated with age when the YNC and Amyloid- groups are included (r=-0.49, p<0.01) and when all three groups are included in the analysis (r=-0.61, p<0.01). After controlling for age, there was no significant difference in amplitudes between the Amyloid+ and Amyloid- or YNC groups (p=0.27). There was a non-significant trend towards decreased circadian pattern in the cognitively impaired Amyloid+ group (n=5, Aβ40 126pM, Aβ42 3.2pM) compared to the cognitively normal Amyloid+ group (n=6, Aβ40 309pM p=0.19, Aβ42 8.8pM p=0.08).

Figure 4.

Decreased Aβ42 circadian amplitude with increased age. Individual Aβ42 circadian amplitude was calculated for each participant and compared to the participant’s age. Age and Aβ42 circadian amplitude were negatively correlated (p<0.01). Young Normal Control (YNC) participants (blue circles) demonstrated the highest circadian amplitude, older amyloid negative participants (green squares) had decreased circadian amplitude and the older amyloid positive participants (red triangles) demonstrated the lowest circadian amplitudes.

Cosinor analyses were also conducted on the group-averaged data for both Aβ40 and Aβ42 in three groups (Figure 5). The range in average Aβ levels over time before cosinor transformation was approximately 40% of the mean (Figure 5). Aβ40 circadian patterns were found by cosinor curve fit (p<0.01) in all three groups, however an Aβ42 cosinor pattern was only found in the YNC group. Similar to individual cosinor analysis, the YNC group had higher circadian amplitudes in both Aβ40 and Aβ42. The CSF Aβ40 peaks (acrophase) occurred at approximately 10PM and troughs occurred at approximately 10AM for the YNC, Amyloid- and Amyloid+ groups. Similarly, CSF Aβ42 amplitudes reached a maximum at 10PM and a minimum at 10AM for the YNC group.

Figure 5.

Cerebrospinal fluid Aβ circadian amplitude decreases with age. Cosinor fit for group-averaged Aβ42 (upper panel) and Aβ40 (lower panel) levels over time. The Aβ42 circadian amplitude decreased with age and amyloidosis. Aβ40 circadian amplitude was lower in the older groups.

Aβ Increases during Wakefulness and Decreases during Sleep

Previous animal studies have demonstrated a direct relationship between wakefulness and increases in Aβ.¹⁶ Here, we assessed the correlation between Aβ levels and total sleep in a subset of participants who had EEG recordings (N=12) in the YNC group (Figure 6). As expected, a circadian pattern was identified in the average total sleep time (p<0.01). Interestingly, the wake time peak occurred at 4PM, while the CSF Aβ peak occurred 6 hours later at 10PM. The sleep time peak occurred at 4AM, while the CSF Aβ trough occurred 6 hours later at 10AM. A 6-hour delay from sleep to change in Aβ is expected, as there is a 6-hour lag from the time of labeling to the time of detection of labeled Aβ in the lumbar CSF.²² Thus, both CSF Aβ40 and Aβ42 were inversely correlated with sleep after a 6-hour delay.

Figure 6.

Cerebrospinal fluid (CSF) amyloid- β (Aβ) circadian pattern follows sleep after a 6 hour delay. Circadian rhythms of average total sleep (minutes in sleep per hour, green triangles) and Aβ circadian pattern in a subset of Young Normal Control participants (YNC, N=12). A delay of 6 hours was observed between the maximum of sleep and minimum of Aβ levels.

Individual Activity and Aβ Levels

As individual behaviors may influence neuronal activity and therefore influence Aβ production, we assessed correlations between individual activities and hourly CSF Aβ levels. Video-rated activities included catheter manipulation, CSF sample draw, computer use, defecation or urination, eating, reading, sleeping, talking, watching television, and writing (eTable 1). A subset of younger participants who were monitored with video analysis were included in the analysis (N=9). We assessed correlations between each activity and CSF Aβ40 and Aβ42 levels after a 6-hour delay due to an expected 6-hour lag time. However, there were no significant correlations between individual behaviors and CSF Aβ levels (IrI<0.1). Correlations remained low when assessed with no delay between activity and Aβ levels.

Relationship of Aβ and Total protein

There was no hourly correlation between total protein and Aβ40 (r = −0.56 to 0.54, N=30, mean=0.06, SD=0.28). Furthermore, there was no correlation for total protein and Aβ40 in individual YNC, Amyloid- and Amyloid+ group.

The average levels of Aβ40 and total protein for each person over time were also calculated, and no correlation was found between them (r=0.15, p=0.44). As expected, average total protein levels were significantly lower in young controls compared to the older groups (with a mean of 540.6 μg/ml for YC, 696.4 μg/ml for Amyloid- and 656.1 μg/ml for Amyloid+ groups, p=0.014). Total protein levels were averaged by group and a cosinor fit was applied. There was no significant circadian pattern in total protein for any of the three groups (p>0.05, eFigure 2). Thus, CSF Aβ dynamics appear to be independent of CSF total protein changes.

Discussion

The relationship between soluble Aβ concentrations in the human CNS and sleep/wake cycles reveals novel insights into the normal physiology of brain and Aβ metabolism. In healthy participants, the peak-to-peak magnitude of the circadian pattern was 30%, which is significant compared to other circadian biological rhythms such as temperature with a peak-to-peak magnitude of ~2%. We report that normal circadian patterns of CSF Aβ were disrupted by increasing age, the largest risk factor for AD, and hourly CSF Aβ dynamics and the Aβ rise were attenuated with amyloid deposition. This ‘flat-line’ CSF Aβ42 was a characteristic finding with amyloid deposition, but was not found in older non-amyloid burdened individuals. These findings suggest the normal physiologic patterns of CSF Aβ are dynamic and circadian, while amyloid and aging diminish normal CSF Aβ dynamics to a flat-line.

The CSF Aβ circadian pattern was strongly correlated with sleep after a 6-hour delay. For the group averaged data (Figure 2), the peak of wakefulness at 4PM occurred 6 hours prior to the maximum levels of Aβ at 10PM. These findings were consistent with a 6-hour delay observed from the time of labeling Aβ in the human CNS to the appearance of labeled Aβ in the lumbar CSF ²². Taken together, these findings suggest that wakefulness precedes and causes increases in CSF Aβ, while sleep decreases CSF Aβ, generalizing findings from murine models¹⁶ to humans.

Circadian amplitudes were approximately two times higher for both Aβ40 and Aβ42 in the younger group compared to the older Amyloid+ and Amyloid- groups. Furthermore, there was a significant inverse correlation between circadian amplitude and age. However, circadian amplitudes were not significantly different between those with and without amyloid deposition, suggesting that increased age primarily affects Aβ circadian rhythms. Possible causes for the loss of correlation between sleep and CSF Aβ include impaired Aβ transport or clearance mechanisms from the brain to the CSF³ or Aβ production no longer being modulated by sleep.

In this study, we observed a linear increase in CSF Aβ levels over time.

The Aβ linear rise has been observed by multiple pharmaceutical groups and several academic groups who perform serial CSF collection studies. The Aβ linear rise was not different between the younger and older amyloid negative groups, but was significantly decreased in the amyloid positive group. This finding indicates that amyloidosis or associated disease processes attenuate the rise in Aβ and this attenuation is independent of age. We postulate that this steady increase in Aβ may be caused by interrupted sleep and cumulatively increased stress in our study participants, who received hourly CSF and blood sampling and had evidence of interrupted sleep by EEG. Animal studies indicate that stress¹⁵ and sleep deprivation¹⁶ can both increase Aβ in the CNS. Alternatively, the CSF Aβ rise may be caused by changes in CSF flow pathways as a result of CSF sampling which increase lumbar CSF Aβ levels to concentrations approximating brain or subarachnoid CSF levels. Amyloidosis may be associated with or cause impairments in the clearance of Aβ to the CSF, thus blocking the normal Aβ rise.

Loss of dynamic patterns was more pronounced in Aβ42 compared to Aβ40. More selective loss of dynamics in Aβ42 may be due to its greater propensity to aggregate and deposit in amyloid plaques. Studies of Aβ generation indicate brain Aβ is dynamic over minutes to hours^11,14 and is circadian in animal models¹⁶. Studies of single measures of CSF Aβ42 demonstrate low CSF Aβ42 in the presence of amyloid deposition²⁴. In this study, we found decreased CSF Aβ42 dynamics in the presence of amyloid deposition. Taken together, these results suggest that the dynamic changes of brain Aβ42 concentrations may be buffered by amyloid plaques that serve as a pool of Aβ42 species to both decrease CSF Aβ42²⁴ and buffer dynamic changes in CSF Aβ42 concentrations.

CSF Aβ has been successfully utilized as a diagnostic ⁶, prognostic ⁵, and therapeutic biomarker²⁵ . Our results are consistent with reports of decreased and stable levels of Aβ42 in AD⁸, with higher variability in CSF Aβ40⁹, and highly variable and dynamic Aβ changes in younger normal controls¹⁰. The range in average Aβ levels over time before cosinor transformation was approximately 20 to 40% of the mean (Figure 5), indicating that sampling time can significantly affect test results in both younger controls and older participants. Therefore, sampling at consistent times is helpful in making comparisons of CSF Aβ between patients and groups, especially for Aβ measurements in controls. These findings provide insight into the normal dynamic changes of the Aβ protein in the human CNS, as well as the effects of aging and amyloidosis as they relate to Alzheimer’s disease. Further research into the mechanisms that contribute to the age and amyloid related changes in Aβ dynamics may offer novel therapeutic approaches for Alzheimer’s disease.