Plasma Amyloid-β (Aβ 42 ) Correlates with Cerebrospinal Fluid Aβ 42 in Alzheimer’s Disease

INTRODUCTION

There is a strong demand for fluid-based biomarkers for Alzheimer’s disease (AD) in both the research and clinical context, with applications in clinical diagnosis, developing AD-related drugs, and as screening tools in clinical trials, for example in the pre-screen of patients for amyloid positron emission tomography (PET) [1–8]. Quantifications of cerebrospinal fluid (CSF) biomarkers such as Aβ₄₂, total tau (t-Tau), and phosphorylated tau (p-Tau) proteins have been incorporated into standard diagnostic guidelines for AD [1, 2]. It has been demonstrated that the change in CSF Aβ₄₂ level happens 10–20 years earlier than the onset of clinical symptoms [3, 4]. The CSF Aβ₄₂ level was found to be significantly reduced in individuals with AD [5–9]. For inclusion of Aβ₄₂ analysis as the first step in a multiple process to screen preclinical AD, it is important to be able to measure it with a low cost and non-invasive method, such as blood analysis. However, the AD-related biomarkers in blood are very low abundant and thus ultra-sensitive assay technologies are needed to measure AD-related biomarkers in blood samples. This demand motivated several groups to develop ultra-sensitive assays. Several technologies for peripheral blood analysis have been developed, such as single-molecule array (Simoa), single-molecule counting (SMC), multi-analyte profiling (xMAP), and mass spectrometry (MS)-based quantification [10–15]. So far, the results have been inconsistent showing increased, unchanged or decreased plasma Aβ₄₂ concentrations in AD and no overall significant change upon meta-analysis [16]. The low levels of Aβ₄₂ in blood together with its short half life and matrix effects pose strong requirements on the methodologies [17].

An ultra-sensitive immunoassay with high sensitivity and low interference called immunomagnetic reduction (IMR) assay was developed to quantify blood Aβ₄₂ level [18, 19]. According to the results in a Taiwanese cohort, blood Aβ₄₂ level measured using IMR is approximately 15 pg/ml in healthy elderly individuals [20]. Interestingly, blood Aβ₄₂ levels are increased in early AD (specifically mild cognitive impairment (MCI) due to AD and mild AD in the two studies) [20, 21]. Importantly, higher blood Aβ₄₂ level in AD patients than healthy control was also observed in a study based on a US patient cohort [22]. The good consistency between the studies in Taiwan and US shows the high reliability and promising utilities of blood-based biomarkers for both research and clinical uses of AD using the IMR assay.

According to the published reports by independent groups, significant changes in Aβ₄₂ level in both CSF and in plasma in AD were found, but with an opposite change in Aβ₄₂ between CSF (reduced levels) and plasma (higher levels), which implies that there is an inverse correlation for Aβ₄₂ between CSF and plasma. However, the cohorts in the CSF studies were different from the above studies, precluding the possibility to directly correlate Aβ₄₂ levels between CSF and plasma in the same cohort. Therefore, in the present study, the levels of plasma Aβ₄₂ (measured with IMR) were directly compared and correlated with those in CSF (measured with ELISA) in the same cohort, that included 106 samples from 63 AD patients and 43 controls from two independent sites. In addition to examining the correlation in Aβ₄₂ level between CSF and plasma, possible factors contributing to the correlation are discussed.

MATERIALS AND METHODS

RESULTS

The demographic information of subjects enrolled at Sahlgrenska University (GOT) and Amsterdam (AMST) is given in Table 1. The mean values and standard deviations of CSF Aβ₄₂ level detected with conventional ELISA for each site are also shown in Table 1. The average levels of CSF Aβ₄₂ in AD groups at the two sites were 402.2±105.3 pg/ml (GOT) and 465.2±106.9 pg/ml (AMST), respectively. The CSF Aβ₄₂ levels for control groups of the two sites were 901.3±177.2 pg/ml (GOT) and 979.7±190.6 pg/ml (AMST), respectively. Combining subjects of the two sites, the CSF Aβ₄₂ level is 432.2±109.9 pg/ml for AD and 946.9±187.1 pg/ml for CONT, as given in Table 2. The combined AD group showed a significantly lower level of CSF Aβ₄₂ than the control group (p <  0.001).

The specificity of plasma Aβ₄₂ assay by IMR was clarified by spiking its similarity, Aβ₄₀, in sample contained Aβ₄₂ as interference material. The measured Aβ₄₂ on samples contained Aβ₄₂ only was 115.7 pg/ml. The measured Aβ₄₂ of another sample contained both Aβ₄₂ and Aβ₄₀ was quantified to be 104.6 pg/ml. The recovery rate of measured Aβ₄₂ by IMR between with and without Aβ₄₀ is 90.4% with an error range less than 10%.

Human plasma contains various and abundant endogenous biomolecules, that may interfere the measurement of IMR assay on plasma Aβ₄₂. The chemicals in medicine used to treatment inflammatory diseases, viral and bacterial infections, cardiovascular disease, and AD may also interfere. Each of these common molecules and chemicals (10,000 mg/ml of hemoglobin, 600 mg/ml of conjugated bilirubin, 30,000 mg/ml of intra lipid, 200 mg/ml of uric acid, 500 IU/ml of rheumatoid factor, 60,000 mg/ml of albumin, 500 mg/ml of acetylsalicylic acid, 300 mg/ml of ascorbic acid, 1,000 mg/ml of ampicillin sodium, 100 ng/ml of quetiapine fumarate, 90 ng/ml of galantamine hydrobromide, 100 ng/ml of rivastigmine hydrogen tartrate, 1,000 ng/ml of donepezil hydrochloride, and 150 ng/ml of memantine hydrochloride) are spiked in same plasma sample containing 115.7 pg/ml Aβ₄₂ quantified by IMR assay, separately. The measured plasma Aβ₄₂ ranges from 104.3 pg/ml to 123.8 pg/ml, and the corresponding recovery rate of measured Aβ₄₂ by IMR between with and without interfering materials is within 90.1% to 107.0% with an error range less than 10%. Taken together, this finding indicates that the Aβ_40, biomolecules, drugs and chemicals listed above do not interfere with the assay of plasma Aβ₄₂ by IMR.

In order to validate the dilution linearity of assaying Aβ₄₂ by IMR, a dilution recovery study was performed by diluting a plasma sample containing Aβ₄₂ by factors of 5, 10, 20, 50, and 100 with PBS buffer. The measured Aβ₄₂ concentration of the original sample (un-diluted) is 1059.43 pg/ml spiked with synthetic Aβ₄₂ in plasma. The dilution recoveries for samples diluted 1:5, 1:10, 1:20, and 1:50 ranged from 97.2% to 108.9% with an error range less than 10%. The dilution recovery for sample diluted 1:100 was 115.3%. So, the highest linear dilution factor is 50 times.

The plasma Aβ₄₂ levels assayed with IMR in AD at individual and combined cohorts were 17.9±4.0 pg/ml (GOT), 17.9±3.9 pg/ml (AMST), and 17.9±4.3 pg/ml (combined). In controls, the plasma Aβ₄₂ levels assayed with IMR individual and combined sites were 13.7±0.7 pg/ml (GOT), 16.8±1.8 pg/ml (AMST), and 15.5±2.1 pg/ml (combined) for CONT. The AD group showed a higher level of plasma Aβ₄₂ than the control group (p <  0.001), as shown in Table 2. The increase in plasma Aβ₄₂ level in AD patients of these European cohorts assayed with IMR is consistent with Taiwan and US studies [18–20]. According to the results in Table 2, the ratio of Aβ₄₂ level in plasma to that in CSF was approximately 1.6% for controls and 4.1% for AD patients.

In Table 2, the opposite change in Aβ₄₂ levels in CSF and plasma between controls and AD patients is evidenced. The relationship between CSF Aβ₄₂ and plasma Aβ₄₂ of the 106 subjects is plotted in Fig. 1, showing a non-linear correlation between CSF Aβ₄₂ and plasma Aβ₄₂. Spearman correlations are used to analyse the CSF-Plasma Aβ₄₂ correlations in controls and in AD, separately. In controls, there was a weakly positive correlation between CSF Aβ₄₂ and plasma Aβ₄₂ levels (r = 0.186). However, in AD, there is a moderately negative correlation between CSF Aβ₄₂ and plasma Aβ₄₂ levels (r = –0.352). This points out that the plasma Aβ₄₂ level increases with the decreasing CSF Aβ₄₂ level in AD.

By combining the control and AD groups, all data points in Fig.1 are fitted to hyperbolic curve: OPEN IN VIEWER

DISCUSSION

To our knowledge, this is the first report showing that plasma Aβ₄₂ levels are negatively correlated with CSF Aβ₄₂ levels in AD patients. In contrast, a previous study recently reported that both plasma and CSF Aβ₄₂ levels dropped in the dementia stage when assayed using ultrasensitive digital ELISA methodology (Simoa assay) [12]. This means that Aβ₄₂ level in plasma and in CSF assayed with Simoa showed a slightly positive correlation. The opposite finding of an inverse (negative) correlation between CSF and plasma Aβ₄₂ in the current study might be related to the different designs of technological platforms used. The Simoa method [12] is based on the sandwich assay and thus relies on the binding of two antibodies (capture and detection antibody) to measure Aβ₄₂ molecules in body fluids [12, 28], with the first antibody used to capture the N-terminal of Aβ₄₂, whereas the second antibody binds to the C-terminal domain of Aβ. Because plasma Aβ₄₂ is frequently bound to carrier proteins in blood, such as albumin or lipoproteins [29], this may induce a potential stereoscopically obstacle for two antibodies to associate with one Aβ₄₂ molecule simultaneously, with loss of some plasma Aβ₄₂ signal by using two antibodies in sandwich method. In contrast, the IMR method is a single-antibody immunoassay. The antibody specifically capturing C-terminal (a.a. 37–42) of Aβ₄₂ is anchored on the magnetic nanoparticles to detect Aβ₄₂ molecule. Based on the design of IMR assay, it ideally has a higher possibility to capture and detect Aβ₄₂ molecule when C-terminal of Aβ₄₂ is exposed in various conformations, such as isolated, complex or oligomeric forms. This may explain the different signal for IMR in comparison with the sandwich-based immunoassay when detecting Aβ₄₂. Another explanation for the differing results is that the antibody used in the IMR experiments is a polycloncal antibody that has been reported to react with several 25–85 kDa bands of unknown identity at western blot of human plasma according to the commercial vendor. It was evidenced that IMR Aβ₄₂ assay is specific for Aβ₄₂ in the presence of Aβ₄₀ in present study. Whether these are made up from oligomerized Aβ₄₂ or other anti-Aβ-reactive proteins remains to be examined.

In addition to assay methodologies, we propose the following hypothesis from the biology point of view to explain the observed negative correlation between plasma Aβ₄₂ and CSF Aβ₄₂ levels in this study. CSF Aβ₄₂ shows a significant reduction in AD patients as reported in many studies [5–9], probably caused by aggregation and deposition of Aβ₄₂ in brain or a defect of Aβ₄₂ clearance which leads to lower amount of Aβ₄₂ molecules transport to CSF [30]. In contrast, when plasma Aβ₄₂ levels are measured by the IMR technique, there is an increase in AD [18, 20–22], probably related to the different transportation systems to move Aβ₄₂ from the brain to the CSF and to move Aβ₄₂ from the brain to the peripheral blood. The following is our hypothesis to emphasize the impact of independent transportation system on increasing plasma Aβ₄₂ levels in AD.

It is known that some clearance systems exist for transportation of Aβ₄₂ from brain to peripheral blood in order to clean out toxic Aβ₄₂ [31]. The blood-brain barrier could play a role as a barrier to allow transportation of monomeric and soluble forms of Aβ₄₂ to the peripheral blood. The ratio of Aβ₄₂ in plasma as compared with CSF is small (<5%; Table 2), a figure similar to a previous study [32], which may indicate that only small portion of Aβ₄₂ born in brain can reach peripheral blood. But in AD, the clearance systems might keep working to transport more Aβ₄₂ to peripheral blood to avoid more Aβ₄₂ accumulating in brain. This may be one of the reasons that Aβ₄₂ level increased in peripheral blood of AD patients in the present study. An alternative explanation for the inverse correlation between plasma and CSF Aβ₄₂ levels in AD patients is the difference in composition of the matrix. The total protein concentration in blood plasma is approximately 50–70 g/l for an adult. The blood plasma is abundant in albumin (which constitute 50–60% of blood plasma proteins) that may carry substantial amounts of Aβ₄₂ [33] as compared with the amount of free Aβ₄₂. Because albumin is highly soluble in blood matrix, albumin-Aβ₄₂ complexes may be more prone to stay in a soluble form, which may prevent Aβ₄₂ from oligomerization and aggregation in blood. When Aβ₄₂ is transported from brain to peripheral blood in AD, it is thus conceivable that higher levels of soluble Aβ₄₂ in blood plasma are quantified by the IMR assay than by a sandwich immunoassay. On the other hand, the total protein level in CSF is less than 1% of that in plasma, with albumin levels around 230 mg/l as compared to 50–70 g/l in plasma [34]. In this environment with low levels of carrier proteins, aggregation-prone Aβ₄₂ molecules may relatively more easily contact with each other to form insoluble aggregates, which could reduce levels quantified by conventional sandwich immunoassays. In contrast, IMR assay shows better sensitivity and consistence when quantifying ultra-low concentration of plasma Aβ₄₂ from various and independent cohorts than conventional ELISA. In order to clarify the matrix effect in above interpretation, and to study the divergence between methodologies, a study of quantification of CSF Aβ₄₂ and plasma Aβ₄₂ by IMR assay form the same subjects is necessary.

DISCLOSURE STATEMENT

Authors’ disclosures available online (https://www.j-alz.com/manuscript-disclosures/17-0784r1).