Association of plasma and cortical beta-amyloid is modulated by APOE ε4 status

Abstract

Background

APOE ε4’s role as a modulator of the relationship between soluble plasma beta-amyloid (Aβ) and fibrillar brain Aβ measured by Pittsburgh Compound-B positron emission tomography ([¹¹C]PiB PET) has not been assessed.

Methods

Ninety-six Alzheimer’s Disease Neuroimaging Initiative participants with [¹¹C]PiB scans and plasma Aβ_1-40 and Aβ_1-42 measurements at time of scan were included. Regional and voxel-wise analyses of [¹¹C]PiB data were used to determine the influence of APOE ε4 on association of plasma Aβ_1-40, Aβ_1-42, and Aβ_1-40/Aβ_1-42 with [¹¹C]PiB uptake.

Results

In APOE ε4− but not ε4+ participants, positive relationships between plasma Aβ_1-40/Aβ_1-42 and [¹¹C]PiB uptake were observed. Modeling the interaction of APOE and plasma Aβ_1-40/Aβ_1-42 improved the explained variance in [¹¹C]PiB binding compared to using APOE and plasma Aβ_1-40/Aβ_1-42 as separate terms.

Conclusions

The results suggest that plasma Aβ is a potential Alzheimer’s disease biomarker and highlight the importance of genetic variation in interpretation of plasma Aβ levels.

1. Introduction

Alzheimer’s disease (AD) is the most common type of dementia, affecting an estimated 5.4 million Americans. At present, there are no treatments that can stop its progression. However, worldwide research efforts are being conducted to identify improved methods to prevent, diagnose, and treat this disease [1]. Objective measures of biological or pathogenic processes, termed biomarkers, can help in the evaluation of disease risk or prognosis. To date, no reliable biomarkers for AD in peripheral blood have been found [2].

AD is characterized by declining memory and cognition. Amnestic mild cognitive impairment (MCI) is a clinical condition thought to be a prodromal stage of AD, in which an individual has cognitive problems not normal for his/her age, but are not severe enough to interfere significantly with daily life activities. An estimated 14–18% of individuals aged 70 years and older have MCI, and approximately 10–15% of these individuals with MCI will progress to dementia, mostly AD, each year [3, 4].

Accumulation of beta-amyloid (Aβ) fragments into amyloid plaques in the brain is one of the defining pathologies of AD. Attempts to monitor the presence and/or progression of amyloid deposition have primarily focused on measurements of Aβ in the brain and cerebrospinal fluid (CSF). The level of CSF Aβ_1-42 has been shown to be a sensitive biomarker for detection and diagnosis of AD [5–7]. Positron emission tomography (PET) imaging techniques with ligands such as Pittsburgh Compound-B ([¹¹C]PiB) [8] and [¹⁸F]florbetapir [9, 10], which bind fibrillar Aβ plaques with high affinity are being studied for their efficacy in predicting and diagnosing AD and have shown some promise [11–13].

Identifying a peripheral biomarker of central Aβ deposition may help in the diagnosis and treatment of the disease at earlier stages. Measuring soluble Aβ levels in plasma would provide an easy method to study Aβ species, as the procedure is minimally invasive and relatively inexpensive. The utility of plasma Aβ as a potential AD biomarker has been assessed in previous studies, but the results have been inconsistent [14–17]. Possible reasons for the inconsistent results could be the use of different enzyme-linked immunosorbent assays and platforms, and timing of sample collection in relation to the stage of disease progression. Therefore, additional studies are needed to fully characterize the utility of plasma Aβ measures as sensitive and effective biomarkers of AD.

Genetic factors, such as the APOE (apoliprotein E) gene, may play a role in amyloid accumulation and the development of AD. The APOE gene is expressed as three variants: ε2, ε3, and ε4. The APOE ε4 allele is the strongest genetic risk factor of late-onset AD and confers a dose-dependent increase in AD risk of approximately four-fold in carriers compared to non-carriers [18–20]. The ε4 allele is also associated with increased fibrillar Aβ [21] and decreased soluble plasma Aβ_1-42 [22] in a dose-dependent manner. The APOE gene codes for the apoE protein, which is essential for maintaining blood brain barrier (BBB) integrity [23]. The apoE4 form of the apoE protein, coded for by the ε4 allele, has been associated with reduced Aβ clearance from the brain [24] and plasma [25] and with impaired tight junction integrity [26].

To our knowledge, only four studies have investigated the relationship of soluble plasma Aβ and fibrillar brain Aβ as measured by [¹¹C]PiB [22, 27–29]. The first study [27] did not identify any relationships between plasma Aβ_1-40 and Aβ_1-42 levels and [¹¹C]PiB binding. In the other studies, inverse correlations were observed between plasma Aβ_1-42 and [¹¹C]PiB uptake [22, 28] and between Aβ_1-42/Aβ_1-40 and brain amyloid [28, 29]. However, none of these studies examined the potential influence of genetic variation in AD-related genes (e.g., APOE ε4) on relationships between peripheral and central markers of Aβ. Furthermore, these studies only included regional measures of [¹¹C]PiB uptake rather than voxel-based mapping across the whole brain, which may have limited the findings as extracting information from spatially large regions may dilute or obscure relevant results that are spatially constrained.

In the present report, we studied the associations among [¹¹C]PiB brain uptake, soluble plasma Aβ measurements, and APOE ε4 genotype status in 96 participants from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) to determine whether the relationship of soluble plasma Aβ measures and fibrillar brain amyloid was influenced by APOE ε4 status. First, we used the average regional [¹¹C]PiB uptake across four target brain regions known to have amyloid deposition in AD as a quantitative phenotype in regression analyses. We then conducted whole-brain, voxel-wise regression analyses to identify spatially-specific clusters in which APOE ε4 genotype modulated the association of plasma and brain PET measurements of Aβ.

2. Methods

2.1. Alzheimer’s Disease Neuroimaging Initiative

Data used in the preparation of this report were obtained from the ADNI database (http://adni.loni.ucla.edu). The ADNI was initiated in 2003 as a $60 million, 5-year public-private partnership by the National Institute on Aging, the National Institute of Biomedical Imaging and Bioengineering, the Food and Drug Administration, private pharmaceutical companies and nonprofit organizations. ADNI’s primary goal is been to test whether serial magnetic resonance imaging (MRI), PET, other biological markers, genetics, and clinical and neuropsychological assessments can be combined to detect and measure the progression of MCI and early AD. Determining sensitive and specific markers of very early AD progression can aid researchers and clinicians in developing new treatments and monitoring their effectiveness, as well as lessen the time and cost of clinical trials.

Michael W. Weiner, MD, Veterans Affairs Medical Center and University of California-San Francisco, is the Principal Investigator of this initiative. ADNI is the result of the efforts of many co-investigators from a broad range of academic institutions and private corporations. As part of the initial phase of ADNI, more than 800 participants, ages 55 to 90, were recruited from over 50 sites across the USA and Canada, including approximately 200 cognitively healthy older individuals (healthy control or HC) to be followed for three years, 400 people with MCI to be followed for three years and 200 people with early AD to be followed for two years. Further information about ADNI can be found in [30] and at http://www.adni-info.org.

The study was conducted after Institutional Review Board approval at each site. Written informed consent was obtained from all study participants or their authorized representatives.

2.2. Participants

Data from ninety-six participants in the ADNI cohort were evaluated. Participant selection was based on the availability of the following data: [¹¹C]PiB PET scans, plasma measurements of Aβ_1-40 and Aβ_1-42 at time of PET scan, and APOE ε4 genotype data. At the time of PET scan, 22 participants were in the AD group, 52 in the MCI group and 22 in the HC group. The participants included 89 non-Hispanic Caucasians, two non-Hispanic African Americans, two non-Hispanic Asians, two Hispanic Caucasians, and one Caucasian participant of unknown ethnicity. Additional demographic information about the included sample is presented in Table 1.

Table 1.

Sample characteristics

	AD (n=22)	MCI (n=52)	HC (n=22)	p value*
Initial [11C]PiB scans at baseline/12-month/24-month visit	3/13/6	11/36/5	0/20/2	0.034
Age at time of scan (years, Mean±SD)	74.06±9.09	75.35±7.93	77.14±6.17	0.428
Sex (Male/Female)	15/7	35/17	14/8	0.940
Education (years, Mean±SD)	15.73±3.04	16.31±2.65	15.50±3.32	0.492
Handedness (Right/Left)	20/2	48/4	17/5	0.165
APOE ε4 status (ε4−/ε4+)	8/14	24/28	16/6	0.039
Average regional [11C]PiB uptakea (Mean±SD)	2.01±0.31	1.81±0.44	1.56±0.34	0.001
Plasma Aβ1-40 (pg/mL, Mean±SD)	160.99±47.89	171.56±48.60	168.75±36.57	0.666
Plasma Aβ1-42 (pg/mL, Mean±SD)	36.05±9.19	40.81±12.48	41.93±9.06	0.160
Plasma Aβ1-40/Aβ1-42 (Mean±SD)	4.53±1.20	4.38±1.19	4.11±0.81	0.440

Abbreviations: AD, Alzheimer’s disease; MCI, mild cognitive impairment; HC, healthy control.

^*For categorical variables, Pearson chi-square was used to compute the p value. For continuous variables, one-way analysis of variance was used to compute the p value.

^aAverage regional [¹¹C]PiB uptake is the average of [¹¹C]PiB uptake values from four brain regions: anterior cingulate, frontal cortex, parietal cortex and precuneus, normalized to cerebellum.

2.2.1. [¹¹C]PiB PET image data

For all participants, PET data consisted of each participant’s initial [¹¹C]PiB scan in the ADNI longitudinal imaging protocol. Initial scans were acquired at either the participant’s baseline visit, 12-month visit, or 24-month visit (Table 1). Methods for the acquisition and processing of [¹¹C]PiB PET scans for the ADNI sample have been described elsewhere [31, 32]. The PET data used in the present study were what was available as of October 2010. The ADNI database includes normalized whole-brain [¹¹C]PiB images, as well as normalized regional [¹¹C]PiB average uptake values extracted from anatomically-defined regions of interest (ROIs). Both pre-existing ROI data and whole-brain [¹¹C]PiB images were downloaded and analyzed in the present study. Regional [¹¹C]PiB standardized uptake value ratios (SUVR) from four ROIs (anterior cingulate, frontal cortex, parietal cortex and precuneus) were averaged and used as a quantitative phenotype, which will be referred to as “average regional [¹¹C]PiB uptake”. This metric has been previously used to classify participants as positive or negative for amyloid deposition [32].

The whole-brain [¹¹C]PiB PET images evaluated in the present study were preprocessed (“PIB Coreg, Avg, Std Img and Vox Size, Uniform Resolution”), as has been previously described [32]. Briefly, the images were set to a standard orientation and voxel size, intensity normalized using a cerebellar grey matter (GM) ROI, and smoothed to a common resolution of 8 mm full-width at half maximum. These pre-processed scans were downloaded in Neuroimaging Informatics Technology Initiative (NIfTI) format from the ADNI scan repository (http://adni.loni.ucla.edu) and processed further using Statistical Parametric Mapping [33] version 5 (SPM5) (http://www.fil.ion.ucl.ac.uk/spm/) implemented via MATLAB v7.1.0 (MathWorks, Natick, MA, USA) [34]. Specifically, for all participants, 1.5 Tesla T1-weighted 3D magnetization prepared rapid acquisition gradient echo (MP-RAGE) MRI scans [35] acquired at the same time point as the [¹¹C]PiB scans were also downloaded from the ADNI site (http://adni.loni.ucla.edu). The pre-processed [¹¹C]PiB PET image of each participant was co-registered to their corresponding MRI scan. PET and MRI data were then spatially normalized to Montreal Neurological Institute (MNI) space using transformation parameters estimated from the SPM segmentation algorithm [36]. These spatially-normalized PET images were used for the whole-brain voxel-wise analysis.

2.2.2. Plasma Aβ data

Plasma Aβ_1-40 and Aβ_1-42 levels for participants with [¹¹C]PiB PET scans from the same time point that the PET data were acquired were obtained from the ADNI database. The methods for the collection, measurement and quality control (QC) of plasma samples have been previously described [22, 37].

2.2.3. APOE ε4 genotyping

The APOE ε4 status of all participants was determined by two single nucleotide polymorphisms (rs429358 and rs7412), as previously described [38]. Participants were classified as APOE ε4− (absence of the ε4 allele), or APOE ε4+ (presence of the ε4 allele).

2.3. Statistical analyses

The influence of APOE ε4 status on the association between plasma Aβ and average regional [¹¹C]PiB uptake was assessed in R version 2.10.0 [39] and SAS 9.3 (SAS Institute Inc., Cary, NC, USA) using the following regression models:

1. Average regional [¹¹C]PiB uptake=Plasma Aβ_1-40+APOE ε4 status+(Plasma Aβ_1-40*APOE ε4 status)

2. Average regional [¹¹C]PiB uptake=Plasma Aβ_1-42+APOE ε4 status+(Plasma Aβ_1-42*APOE ε4 status)

3. Average regional [¹¹C]PiB uptake=Plasma Aβ_1-40/Aβ_1-42+APOE ε4 status+(Plasma Aβ_1-40/Aβ_1-42*APOE ε4 status)

Models with significant interactions between plasma Aβ and APOE ε4 status on average regional [¹¹C]PiB uptake were identified. For these models, the variance in average regional [¹¹C]PiB uptake explained by different terms in the model was determined using the following regression models:

1. Variance in average regional [¹¹C]PiB uptake explained by plasma Aβ term alone:

   Average regional [¹¹C]PiB uptake=Plasma Aβ

2. Variance in average regional [¹¹C]PiB uptake explained by APOE ε4 status term alone:

   Average regional [¹¹C]PiB uptake=APOE ε4 status

3. Variance in average regional [¹¹C]PiB uptake explained by plasma Aβ and APOE ε4 status terms together:

   Average regional [¹¹C]PiB uptake=Plasma Aβ+APOE ε4 status

4. Variance in average regional [¹¹C]PiB uptake explained by plasma Aβ, APOE ε4 status, and (plasma Aβ*APOE ε4 status) terms together:

   Average regional [¹¹C]PiB uptake=Plasma Aβ+APOE ε4 status+(Plasma Aβ*APOE ε4 status)

The influence of APOE ε4 status on the association between plasma Aβ and [¹¹C]PiB uptake was further assessed in whole-brain voxel-wise analyses in SPM5 using the following regression models:

1. Voxel [¹¹C]PiB uptake=Plasma Aβ_1-40+APOE ε4 status+(Plasma Aβ_1-40*APOE ε4 status)

2. Voxel [¹¹C]PiB uptake=Plasma Aβ_1-42+APOE ε4 status+(Plasma Aβ_1-42*APOE ε4 status)

3. Voxel [¹¹C]PiB uptake=Plasma Aβ_1-40/Aβ_1-42+APOE ε4 status+(Plasma Aβ_1-40/Aβ_1-42*APOE ε4 status)

In the voxel-wise analyses, an explicit GM mask was used to restrict analyses to GM regions. Significant interactions were determined using a voxel-level threshold of p<0.005 (uncorrected) and cluster-level threshold of k≥200 contiguous voxels to achieve cluster-level uncorrected p<0.05. Clusters identified in the left or right cerebellum were not considered as the [¹¹C]PiB PET images had been intensity normalized using a cerebellar GM region of interest. Voxels at which significant relationships existed were displayed on a three-dimensional rendered brain. The MNI coordinates of voxels that were peak maxima and local maxima (voxels>4 mm apart) in each cluster were converted to Talairach coordinates, and queried in Talairach Client v2.4.2 [40, 41] software to determine the associated anatomic labels. Random field theory corrected p values (p_corr) were used to identify significant clusters. Mean [¹¹C]PiB uptake from each significant cluster was extracted for all participants and their distribution in APOE ε4− and APOE ε4+ participants was further examined in R version 2.10.0 and SAS 9.3. We then further evaluated the relationship of mean [¹¹C]PiB uptake from the biggest and most significant cluster and plasma Aβ levels. The variance in mean [¹¹C]PiB uptake extracted from the significant cluster explained by different terms in the model was determined using the following regression models:

1. Variance in mean [¹¹C]PiB uptake from the significant cluster explained by plasma Aβ term alone:

   Mean [¹¹C]PiB uptake from the significant cluster=Plasma Aβ

2. Variance in mean [¹¹C]PiB uptake from the significant cluster explained by APOE ε4 status term alone:

   Mean [¹¹C]PiB uptake from the significant cluster=APOE ε4 status

3. Variance in mean [¹¹C]PiB uptake from the significant cluster explained by plasma Aβ and APOE ε4 status terms together:

   Mean [¹¹C]PiB uptake from the significant cluster=Plasma Aβ+APOE ε4 status

4. Variance in mean [¹¹C]PiB uptake from the significant cluster explained by plasma Aβ, APOE ε4 status, and (plasma Aβ*APOE ε4 status) terms together:

   Mean [¹¹C]PiB uptake from the significant cluster=Plasma Aβ+APOE ε4 status+(Plasma Aβ*APOE ε4 status)

3. Results

We first investigated the influence of APOE ε4 status on the association between plasma Aβ and average regional [¹¹C]PiB uptake. No significant interactions between plasma Aβ_1-40 and APOE ε4 status or between plasma Aβ_1-42 and APOE ε4 status were observed on average regional [¹¹C]PiB uptake. However, a significant interaction between plasma Aβ_1-40/Aβ_1-42 and APOE ε4 status (p=0.025) on average regional [¹¹C]PiB uptake was observed. Inclusion of age at time of scan and gender as covariates did not alter this finding. APOE ε4 genotype status (ε4− or ε4+) conferred different patterns of association between plasma Aβ_1-40/Aβ_1-42 and average regional [¹¹C]PiB uptake. Specifically, in the APOE ε4− participants, there was a positive relationship between plasma Aβ_1-40/Aβ_1-42 and average regional [¹¹C]PiB uptake (Slope=0.162; p=0.008; r²=0.141) (Fig. 1A). However, this relationship did not exist in the APOE ε4+ participants (Slope=−0.005; p=0.901; r²<0.001) (Fig. 1B). The APOE ε4 status and plasma Aβ_1-40/Aβ_1-42 terms explained 17% and 6% of variation in average regional [¹¹C]PiB uptake, respectively. The two terms together explained 19% of variation in average regional [¹¹C]PiB uptake. Inclusion of the (plasma Aβ_1-40/Aβ_1-42*APOE ε4 status) interaction term in the model increased the explained variance in average regional [¹¹C]PiB uptake to 24%.

Fig. 1.

Scatter plots of plasma Aβ_1-40/Aβ_1-42 versus average regional [¹¹C]PiB uptake from the (Average regional [¹¹C]PiB uptake=Plasma Aβ_1-40/Aβ_1-42+APOE ε4 status+(Plasma Aβ_1-40/Aβ_1-42*APOE ε4 status) model (A and B), and plasma Aβ_1-40/Aβ_1-42 versus mean [¹¹C]PiB uptake from the cluster identified in the (Voxel [¹¹C]PiB uptake=Plasma Aβ_1-40/Aβ_1-42+APOE ε4 status+(Plasma Aβ_1-40/Aβ_1-42*APOE ε4 status)) model (C and D).

To further examine the spatial extent of the potential influence of the APOE ε4 status on the association of plasma Aβ and [¹¹C]PiB uptake, we performed whole-brain voxel-wise regression analyses. APOE genotype did not significantly affect the positive or negative associations between plasma Aβ_1-40 and [¹¹C]PiB uptake and between plasma Aβ_1-42 and [¹¹C]PiB uptake in cerebral regions. However, APOE genotype significantly altered the negative correlation of plasma Aβ_1-40/Aβ_1-42 and [¹¹C]PiB uptake in a significant cluster in the left inferior frontal gyrus (MNI peak coordinates: x=−40, y=18, z=−6; k=6152 voxels; cluster-level p_corr<0.001) (Fig. 2 and Table 2). Similar to the data from the average regional [¹¹C]PiB uptake analysis, there was a positive relationship between plasma Aβ_1-40/Aβ_1-42 and mean [¹¹C]PiB uptake from the inferior frontal gyral cluster in the APOE ε4− participants (Slope=0.250; p=0.001; r²=0.213) (Fig. 1C), but not in the APOE ε4+ participants (Slope=−0.050; p=0.404; r²=0.015) (Fig. 1D). Inclusion of age at time of scan and gender as covariates again did not alter this finding. The APOE ε4 status and plasma Aβ_1-40/Aβ_1-42 terms by themselves explained 13% and 4% of variation in mean [¹¹C]PiB uptake from the significant cluster, respectively. The two terms together explained 14% of variation in mean [¹¹C]PiB uptake from the significant cluster. Inclusion of the (plasma Aβ_1-40/Aβ_1-42*APOE ε4 status) interaction term along with the two terms in the model increased the variance explained in mean [¹¹C]PiB uptake from the significant cluster to 23%. No clusters were identified when considering the positive correlation of plasma Aβ_1-40/Aβ_1-42 and [¹¹C]PiB uptake.

Fig. 2.

Brain regions identified in the (Voxel [¹¹C]PiB uptake=Plasma Aβ_1-40/Aβ_1-42+APOE ε4 status+(Plasma Aβ_1-40/Aβ_1-42*APOE ε4 status)) model (voxel-level threshold of p<0.005 (uncorrected), cluster size≥200 voxels). The red-to-yellow scale indicates increasing statistical significance of association.

Table 2.

Brain regions identified in the (Voxel [¹¹C]PiB uptake=Plasma Aβ_1-40/Aβ_1-42+APOE ε4 status+(Plasma Aβ_1-40/Aβ_1-42*APOE ε4 status) model (voxel-level threshold of p<0.005 (uncorrected), cluster size≥200 voxels)

Region	Broadmann Area	Peak value coordinates (mm)			Voxel-level		Cluster-level
		x	y	z	T value	pFWE-corr	k	puncorr	pcorr
Left inferior frontal gyrus	BA47	−40	18	−6	3.92	0.713	6152	<0.001	<0.001
Left superior temporal gyrus	BA22	−46	6	−4	3.85	0.774
Left middle frontal gyrus	BA6	−30	12	60	3.81	0.816
Left superior frontal gyrus	BA8	−28	20	56	3.76	0.855
Left precentral gyrus	BA4	−34	−20	52	3.68	0.910
Left insula	BA13	−42	−8	−8	3.58	0.955
Left middle frontal gyrus	BA8	−24	24	50	3.43	0.990
Left middle frontal gyrus	BA10	−36	42	14	3.40	0.993
Left middle temporal gyrus	BA21	−58	−10	−8	3.35	0.996
Left superior frontal gyrus	BA9	−18	48	34	3.29	0.998
Left inferior frontal gyrus	BA45	−50	20	20	3.24	0.999
Left anterior cingulate	BA32	0	44	12	3.23	0.999
Left superior frontal gyrus	BA10	−26	42	30	3.19	1.000
Left medial frontal gyrus	BA10	−2	58	−4	3.17	1.000
Left inferior parietal lobule	BA40	−54	−22	26	3.16	1.000
Left transverse temporal gyrus	BA41	−48	−26	10	3.16	1.000
Left postcentral gyrus	BA40	−48	−34	54	3.73	0.878	273	0.023	0.448
Left inferior parietal lobule	BA40	−36	−52	56	3.54	0.968
Left inferior temporal gyrus	BA20	−64	−8	−26	3.32	0.998	231	0.035	0.586
Left fusiform gyrus	BA20	−62	−4	−28	3.20	1.000
Left middle temporal gyrus	BA21	−54	0	−18	2.84	1.000
Right inferior frontal gyrus	BA47	42	14	−14	3.32	0.998	262	0.026	0.482
Right superior temporal gyrus	BA22	46	0	−4	3.16	1.000
Right sub-gyral	BA13	44	2	−8	3.08	1.000
Right insula	BA13	42	−6	4	2.92	1.000

Abbreviations: p_FWE-corr, Voxel-level p value after family-wise error correction; k, number of voxels in cluster; p_uncorr, Cluster-level uncorrected p value; p_corr, Cluster-level p value after random field theory correction.

4. Discussion

The investigation of Aβ species in plasma offers advantages over conventional methods for measuring Aβ levels in the brain and CSF. Obtaining and analyzing plasma samples is relatively inexpensive, minimally invasive, and can easily be performed at multiple time points. Therefore, a plasma-based biomarker for early detection and diagnosis of AD would be ideal. In the ADNI cohort, a strong association has been observed between CSF Aβ_1-42 and fibrillar brain Aβ (indexed with [¹¹C]PiB) [31], while a weak but significant association has been observed between CSF Aβ_1-42 and soluble plasma Aβ_1-42 [37]. However, the association between soluble plasma Aβ and fibrillar brain Aβ is unclear.

In the present report, we investigated the relationship of soluble plasma Aβ (Aβ_1-40, Aβ_1-42 and Aβ_1-40/Aβ_1-42), APOE ε4 status, and fibrillar brain Aβ (indexed with [¹¹C]PiB) in the ADNI cohort. In two types of analytic approaches, APOE ε4 genotype status had a significant effect on the relationship between plasma Aβ_1-40/Aβ_1-42 and [¹¹C]PiB uptake. Specifically, a positive relationship between plasma Aβ_1-40/Aβ_1-42 and [¹¹C]PiB signal was observed in APOE ε4− participants, but not in APOE ε4+ participants (Fig. 1). This finding may reflect a stronger relationship between plasma Aβ and accumulation of fibrillar amyloid in the brain in individuals at an earlier and/or less severe disease state (e.g. APOE ε4−). A recent study suggested that plasma Aβ levels in cognitively stable individuals tend to increase slightly with age [15]. Cognitively normal individuals with higher plasma Aβ levels are thought to be at an increased risk of progression to AD. Plasma Aβ levels in individuals who go on to develop AD tend to be elevated in the pre-dementia stage, reach a peak, and then fall prior to developing clinical AD symptoms. Increasing brain amyloid deposits in the later stages of disease may perhaps reduce interstitial Aβ in the brain and CSF, confounding the relationship between plasma Aβ and brain amyloid.

The molecular mechanism by which the APOE ε4 allele leads to increased risk for AD is unclear. The APOE gene codes for the apoE protein, which is essential for maintaining BBB integrity [23]. Furthermore, the various apoE protein isoforms are thought to differentially clear Aβ from the brain into the plasma across the BBB. In a recent study, the authors found that mice expressing the human apoE4 protein had greater Aβ concentrations in the interstitial fluid of the brain and hippocampus and showed reduced Aβ clearance from the interstitial fluid of the brain when compared to mice expressing the human apoE2 or apoE3 proteins [24]. However, the authors did not find differences in Aβ synthesis or amyloidogenic processing between mice expressing the different apoE protein isoforms. In another study conducted in APOE ε2, ε3, and ε4 knock-in and APOE knock-out (KO) mice injected with lipidated recombinant apoE2, E3, and E4 proteins, the authors found a difference in peripheral Aβ clearance from the plasma by APOE genotype [25]. The results suggested that APOE ε4 gene expression results in a protein (apoE4) with lowered efficiency of peripheral Aβ clearance from the plasma. Tight junction integrity in the BBB is also regulated by the apoE protein in an isoform-dependent manner, with impaired tight junction integrity and increased BBB permeability observed in mice with the apoE4 isoform compared to mice with expression of the apoE3 isoform [26]. Thus, reduced Aβ clearance by apoE4 protein may lead to an impaired BBB, in turn, affecting Aβ levels in the brain and plasma and the relationship between these compartments.

Binding of the apoE protein to Aβ may also influence Aβ clearance from the brain into the plasma. The apoE protein binds strongly to Aβ, but the binding characteristics of the apoE4 isoform are different than those of the apoE3 isoform [42, 43]. Oxidized apoE4 binds more rapidly to synthetic Aβ than oxidized apoE3. Binding by oxidized apoE4 was also more sensitive to pH changes than oxidized apoE3. In addition, the APOE ε4 allele is associated with increased vascular and plaque amyloid deposits [44]. APOE ε4 homozygotes have higher amyloid deposits in the vasculature and tissue compared to APOE ε3 homozygotes. APOE ε3/ε4 heterozygotes have intermediate amyloid deposits. In a recent study, free apoE protein was shown to facilitate a greater removal of Aβ from the brain into the periphery across the BBB compared with apoE protein bound to Aβ [45]. Furthermore, apoE isoform-specific differences were observed in Aβ transport. Specifically, Aβ bound to the apoE4 isoform had increased blood to brain transport when compared to Aβ bound to the apoE3 isoform. Similar to a previous study, the authors found that the apoE4 isoform had decreased Aβ clearance across the BBB in comparison to the apoE3 isoform. More recently, the authors in a different study observed that retinoid X receptor stimulation increased Aβ clearance across the BBB, an effect which was believe to be partially mediated by the apoE protein [46].

The apoE protein has been shown to be present in greater amounts in the AD brain relative to those of healthy elders. Furthermore, apoE undergoes significantly more cleavage in the AD brain than in HC, especially in APOE ε4 carriers [47]. The N-terminal domain of the apoE protein contains the major receptor binding region and the C-terminal domain contains the lipid binding region. The C-terminal domain of both the apoE3 and apoE4 isoforms has been shown to interact closely with Aβ [47]. Isolated C-terminal–truncated apoE4 protein fragments have been shown to be associated with Aβ plaques [47]. Finally, inefficient clearance of Aβ peptides produces neuronal and behavioral deficits in mice [48]. Thus, differential clearance of Aβ by apoE protein isoform, which is coded for by different APOE genotypes, may be a potential explanation for the APOE genotype effects observed in the present study. Further investigation of mechanistic explanations is warranted.

It is important to discuss the limitations of the present study. First, the sample size in the present study was relatively modest, as only 96 participants in the ADNI cohort had both [¹¹C]PiB PET scans and concomitant plasma measurements of Aβ_1-40 and Aβ_1-42. Our results suggest a complex relationship between plasma Aβ, brain Aβ, and APOE genotype, warranting further investigation in independent and larger samples. Second, a majority of the participants (n=52) in the present study had a diagnosis of MCI. Therefore, it is possible that the observed association may have been driven primarily by these participants. Due to the relatively small number of participants in the three diagnostic groups, we were unable to perform analyses within each diagnostic group to determine if there was an effect of diagnosis. Finally, genetic factors other than APOE may also play a role in modulating the association of plasma and brain Aβ. Investigation of the impact of other known and novel AD-associated genetic variants on the relationship between plasma and brain Aβ represents a possible future direction of this work.

5. Conclusions

In summary, we detected an association between soluble plasma Aβ and fibrillar brain Aβ that was modulated by APOE ε4 status. Replication and additional study in independent samples is needed to clarify the nature of this interaction, as well as to understand the underlying biological mechanisms. The present report suggests that plasma Aβ levels have great potential as an AD biomarker and underscores the importance of genetic variation in the interpretation of plasma Aβ levels.