Fluctuations of CSF amyloid-β levels

The amyloid hypothesis implicates amyloid-β (Aβ) as necessary in the pathogenesis of Alzheimer disease (AD). There are strong genetic, biochemical, and animal model data that support the importance of Aβ in AD.^1,2 Due to Aβ deposition as insoluble aggregates in plaques, Aβ levels are increased by 100-fold and greater levels in AD brain vs controls. Thus, understanding Aβ metabolism both prior to and after Aβ deposition is likely to lead to insights into AD pathogenesis. Soluble Aβ is continuously produced by cleavage from amyloid-precursor protein (APP) predominantly by neurons throughout life. Due to evidence supporting the amyloid hypothesis, multiple studies have evaluated Aβ as a therapeutic target and as a biomarker for AD.

Measurement of Aβ in the human CSF is likely to provide insight into CNS Aβ metabolism. Aβ₄₂ levels are significantly decreased in patients with dementia of the Alzheimer type (DAT) vs controls.³ However, there is significant variability in reported values and in sensitivity and specificity in regard to the clinical diagnosis.^4,5 In studies of very early DAT or mild cognitive impairment (MCI), CSF Aβ₄₂ levels are decreased and appear to predict further progression in DAT severity.^6,7 A recent study assessed CSF Aβ₄₂ in combination with a new molecular imaging technique that allows for determination of the presence of amyloid in the brain. This study showed that the level of CSF Aβ₄₂ (sampled in the fasting state at the same time of day) strongly correlates with the presence or absence of amyloid in the brain regardless of clinical diagnosis.⁸

Prior studies have indicated that Aβ₄₂ levels remain stable in DAT individuals when CSF samples are compared over an average interval of 10 or 18 months.^4,5 However, studies have not evaluated levels of different CSF Aβ species with frequent sampling to determine its variability over 24 hours or longer. As the development of drugs that decrease Aβ production or increase clearance are currently being developed, it will be important to understand the normal variability of CSF Aβ over time. Given the growing use of Aβ as a biomarker in prognostic, diagnostic, and therapeutic research studies, a better understanding of the stability and time course of this biomarker will be useful in the interpretation and design of research studies.

Methods.

Results.

Aβ levels in human CSF had significant variability over 12 (n = 1), 24 (n = 3), and 36 (n = 11) hours of sampling (figure 1, A through C, and table). In most participants, the levels for Aβ_1-x, Aβ₄₀, and Aβ₄₂ varied by 50% or more over several hours and the maximum Aβ levels were greater than 200% the minimum values over the entire study. There were no significant differences in average Aβ_1-x, Aβ_1-40, or Aβ_1-42 minimums, maximums, or means between the younger (20 to 45 years) and older (46 to 80 years) groups in this small group of participants.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1. Human CSF Aβ levels fluctuate over hours. (A-C) The levels of human CSF Aβ_1-x (triangle), Aβ₄₀ (square), and Aβ₄₂ (circle) over 36 hours are shown in three typical individual participants (A, B age 20 to 45, C age 46 to 80). Aβ levels had significant fluctuations, changing >50% within 6 hours, and >100% over 12 hours. (D-F) Mean levels of Aβ_1-x, Aβ₄₀, and Aβ₄₂ averaged across all participants at each sample time are shown as a percent of the average Aβ level. Aβ levels demonstrate troughs at 0 and 25 hours with peaks at 14 and 23 hours.

Aβ₄₀ and Aβ₄₂ were correlated (p ≤ 0.05) to Aβ_1-x in all samples (figure 2) except in the samples of three older participants (n = 3, average age 67 years). For significant correlations, the average correlation was r = 0.77 for Aβ_1-40 and r = 0.68 for Aβ_1-42 compared to Aβ_1-x. Aβ_1-42 was correlated to Aβ_1-40 in all participants except in two older participants (n = 2, average age 65.5 years). For significant correlations, the average correlation of Aβ₄₀ to Aβ₄₂ was r = 0.67.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2. CSF Aβ₄₀, Aβ₄₂, and Aβ_1-x are correlated over time and mean levels vary by time of day. The CSF Aβ_1-x vs Aβ₄₀ and Aβ₄₂ and Aβ₄₀ vs Aβ₄₂ are shown for three participants. Aβ₄₀ and Aβ₄₂ are correlated with Aβ_1-x, and Aβ₄₂ is correlated with Aβ₄₀. Linear regression of each indicates a strong correlation between each pair of human Aβ species assessed in CSF (p < 0.0001).

The interparticipant hourly average of Aβ levels suggests a sinusoidal pattern over the time period measured (figure 1, D through F). Aβ levels increased from 0 to 14 hours, with peaks at 14 and 23 hours and troughs at 0 and 25 hours in the levels of Aβ_1-x, Aβ₄₀, and Aβ₄₂.

Discussion.

We found hour-by-hour variability in human CSF Aβ, which has implications in diagnostic and therapeutic biomarker studies. The biologic variability of Aβ in CSF has been studied by others using repeated lumbar punctures over months to years. In a prior study,⁴ large individual fluctuations were noted over several years in Aβ_1-40 and Aβ_1-42; however, only Aβ_1-42 had an average change (decrease) across subjects. Another study⁵ reported a high correlation (r = 0.90) of CSF Aβ_1-42 on repeat LP in subjects with DAT. It may be that CSF Aβ_1-42 variability is decreased in patients with AD pathology and amyloid plaques, but has higher fluctuations in individuals without plaques. In this study, sampling of CSF was frequent enough to detect hourly changes in Aβ CSF levels. The significant variability due to these fluctuations may make detecting changes in Aβ levels difficult in acute or chronic therapeutic studies. However, a study of treatment for 4 weeks with a selective M1 muscarinic agonist was able to demonstrate an overall decrease in CSF Aβ levels in 19 subjects with AD, despite an increase of Aβ in 3 subjects and no change in 2 subjects.¹¹ Prior longitudinal studies have been restricted to DAT and to sampling an average of over a year apart.^4,5 In the DAT group, Aβ_1-42 levels remained stable and low; however, there was significant Aβ_1-40 fluctuation. Due to the fluctuations observed, we recommend that lumbar punctures assessing Aβ levels be performed at the same time of day, preferably under similar conditions, such as early morning fasting. Alternatively, direct assessment of Aβ synthesis and clearance rates, while more labor intensive, can now be performed.¹²

There is a significant correlation of Aβ_1-x to Aβ₄₀ and Aβ₄₂, as well as Aβ₄₀ to Aβ₄₂. The Aβ species 40 and 42 are tightly correlated to each other in hourly measurements and this likely indicates similar processes in the production and clearance of each from human CSF. The relationship may be altered in the presence of Aβ deposition in plaques, leading to a decreased level of Aβ₄₂ and a decreased Aβ 42/40 ratio reported in studies.³ In addition, in several older participants, there was not a significant correlation of Aβ₄₂ to Aβ₄₀ or to Aβ_1-x. This may reflect a disturbance in clearance of Aβ₄₂ relative to Aβ₄₀ and may be a precursor or an indicator of AD pathology. Future studies will measure these correlations in AD participants and controls and will also follow participants for conversion to AD, to determine if this change is predictive of clinical onset of dementia.

A sinusoidal pattern of Aβ levels was observed across participants. This may be time of day or activity dependent. There is recent evidence from cellular and animal studies that Aβ release is dependent on synaptic activity^13,14 and this may account for much of the variability in CSF levels that we observed. In recent animal studies using microdialysis in APP transgenic mice, we have seen similar dynamic variability over hours in Aβ levels in brain interstitial fluid as seen here in human CSF (J. Kang and D.M. Holtzman, unpublished observations). Future studies using sleep and behavior monitoring and collecting CSF at different or longer time durations may address this hypothesis. The increase in Aβ levels during the first 5 to 10 hours in most participants may be due to a catheter interaction with CSF Aβ, a rostral-caudal Aβ gradient in CSF, another artifact, or biologic fluctuation. A catheter artifact is not likely because three participants had decreasing or stable Aβ levels over the first 5 hours, and the catheter volume is only 3.3% of the sample collected each hour. An Aβ gradient from ventricular to lumbar CSF is possible; however, Aβ levels returned close to starting values later during the study in most participants. Further, in studies in which we have collected 25 to 35 mL of lumbar CSF from individuals over 15 to 30 minutes at the time of lumbar puncture, we have not observed a rostral-caudal gradient in CSF Aβ levels. Sample handling artifact is not likely, given all samples were frozen at −80 °C immediately and analyzed by ELISA on thawing. A probable reason for the significant variation of Aβ levels over hours may be due to dynamic changes in the production or clearance rate of Aβ in the CNS.

Acknowledgment

The authors thank the participants for their time and effort and Eli Lilly for providing anti-Aβ antibodies.