Cerebral accumulation of amyloid β protein (Aβ) is characteristic of Alzheimer disease (AD). Aβ can be detected in cerebrospinal fluid and in plasma. Although plasma Aβ has been proposed as a marker of risk of AD, it is unknown how plasma levels relate to neuropathologic levels. We compared plasma levels of Aβ40 and Aβ42 obtained during life with biochemical and pathologic levels in frontal and temporal neocortex in 25 individuals (17 AD, 3 control, and 5 non-AD dementia) who died a median of 1 year after blood collection. Plasma levels of Aβ40 and Aβ42 were not associated with any of the brain measures, even after adjusting for age and interval between plasma collection and death. The APOE ε4 allele may modify the relationship between plasma Aβ42 and formic acid-extractable Aβ42, with an inverse correlation in APOE ε4 carriers and a positive correlation in those lacking APOE ε4. We conclude that plasma levels of Aβ40 and Aβ42 are not robust correlates of histologic or biochemically assessed amyloid burdens in brain, although the influence of the APOE genotype should be further explored.
Cerebral accumulation of amyloid β protein (Aβ) is a characteristic pathologic feature of Alzheimer disease (AD). Aβ is produced by sequential β- and γ-secretase processing of the amyloid precursor protein, resulting primarily in Aβ species ending at positions Val-40 (Aβ40) or Ala-42 (Aβ42). Aβ40 is the principal component of vascular amyloid, whereas Aβ42 aggregates more readily and is the initial Aβ species deposited in brain parenchyma (1-4). Biochemically, Aβ40 and Aβ42 segregate into biochemical compartments (intracellular, membrane associated, extracellular soluble, and insoluble) defined by sequential extraction procedures (5,6). All biochemical and neuropathologic forms of Aβ are elevated in AD brain (6). Aβ accumulation is initiated years before the onset of symptoms, which begin after a threshold of neuronal loss, synaptic loss, and neurofibrillary tangle formation. Therefore, markers of amyloid accumulation in the brain may be useful as antecedent or clinical diagnostic biomarkers of AD.
The development of sensitive ELISA-based assays has allowed detection of Aβ in cerebrospinal fluid (CSF) and plasma. CSF Aβ42 levels are consistently lower in AD, a finding attributed to depletion of the monomeric protein from the CSF during incorporation into oligomeric soluble and insoluble forms in the brain (7). In part because of its availability, plasma Aβ has been studied as a marker of AD susceptibility, diagnosis, and treatment effects. Plasma total Aβ or Aβ42 is increased in cases of familial AD with presenilin or amyloid precursor protein mutations and in trisomy 21 (8-10). Plasma Aβ levels were not consistently related to diagnosis in clinic-based cross-sectional studies of the typical late-onset form of AD (8,9,11-15). Although these studies have examined the clinical, demographic, biochemical, and genetic factors that influence plasma Aβ levels, no studies have investigated the fundamental question of how plasma levels relate to neuropathologic evidence of AD. To further elucidate the role of plasma Aβ as a potential biomarker for AD, we compared plasma Aβ levels during life with measures of cortical biochemical and neuropathologic Aβ accumulation in study patients who subsequently came to autopsy.
Materials and Methods
The study participants consisted of patients followed in the Massachusetts Alzheimer Disease Research Center from 1991 to 2005 who donated plasma samples for clinical studies and subsequently came to autopsy. As part of institutional review board-approved protocols, informed consent for plasma collection was obtained from the patients and caregivers by a staff physician; autopsy consent was obtained from the next of kin. Clinical diagnoses at the time of plasma collection were based on established criteria for mild cognitive impairment (16), AD (17), progressive supranuclear palsy (PSP) (18), and primary progressive aphasia (19).
At autopsy, coronal slices from one hemisphere were fresh-frozen between dry ice-cooled aluminum plates, and the opposite hemisphere was fixed in 10% buffered formalin with subsequent processing and paraffin embedding. Neuropathologic diagnoses of AD (20-22), PSP (23), corticobasal degeneration (CBD) (24), and amyotrophic lateral sclerosis were made according to established diagnostic criteria.
Five-micrometer paraffin sections were deparaffinized in xylenes, rehydrated with graded alcohols, pretreated with 70% formic acid (5 minutes, room temperature), and blocked in 3% nonfat milk in Tris-buffered saline (TBS) (1 hour). Sections were incubated overnight at 4°C with mouse anti-Aβ40 BA27 (1:1,000) (monoclonal anti-Aβ40; Takeda, Osaka, Japan) or mouse anti-Aβ42 BC05 1 (1:40,000) (monoclonal anti-Aβ42; Takeda) in 1.5% normal goat serum in TBS. Sections were washed in TBS (3 × 5 minutes), incubated with biotinylated anti-mouse IgG 1:200 (1 hour, room temperature), and treated with preformed avidin/biotin conjugated enzyme complex (ABC; Vector Laboratories, Burlingame, CA). The staining was visualized with diaminobenzidine (Vector Laboratories). Nuclei were counterstained with hematoxylin. Sections were cleared in alcohols and xylenes and cover-slipped with Permount mounting media.
Amyloid Burden Quantitation
To calculate the percentage of cortical area covered by Aβ40 and Aβ42 immunostaining (% Aβ40 burden and % Aβ42 burden, respectively), stained sections were placed under a light microscope (Leica Microsystems, Fairfax, VA) equipped with a digital camera linked to a computer running Bioquant software (R and M Biometrics, Nashville, TN). An 800-μm-wide rectangular area spanning all cortical layers in the mid-superior temporal sulcus and the frontal neocortex region were delineated as described previously (25). At least 10 counting frames of 500 × 500 μm were systematically sampled over the delineated areas. The areas of all immunostained plaques were determined under 10x objective. The sum total of all plaque areas was divided by the total area of the counting frames to obtain the percentage of cortex covered by Aβ40 and Aβ42 immunoreactivity, respectively.
For biochemical analysis of Aβ, approximately 1-cm strips of cortex from frozen temporal (Brodmann areas 20, 21, and 22) and frontal neocortex (Brodmann areas 9 and 10) was carefully dissected from the underlying white matter. The tissue was homogenized in 14 μL/mg wet weight Tris buffer, pH 7.2 (50 mmol/L Tris, 200 mmol/L NaCl2, 2 mmol/L EDTA, 0.1% Triton X-100, and 2% protease-free bovine serum albumin [BSA]) with protease inhibitors (Complete; Roche, Indianapolis, IN). After centrifugation (15,000 rpm, 21,000 × g, 4°C, 5 minutes), the supernatant was collected (Tris-soluble Aβ fraction). The pellet was homogenized in 8 μL/mg wet weight 70% formic acid (FA) and recentrifuged (22,000 rpm, 44,000 × g, 4°C, 5 minutes), and the resulting FA-extracted supernatant was neutralized with 1 mmol/L Tris buffer (pH 11.0) (FA-extracted Aβ fraction) (5,6).
Blood was collected in K-EDTA tubes, spun, and stored in 1-mL aliquots at −80°C. For ELISA, plasma samples were precleared with IgG1 κ (Sigma, St. Louis, MO) cross-linked Sepharose beads to block cross-reaction of unidentified components of human plasma with the Aβ ELISA (9,14). Three hundred microliters of plasma was diluted with 525 μL of buffer C (20 mmol/L PO4, 400 mmol/L NaCl, 2 mmol/L EDTA, 0.2% BSA, 10% BlockAce [Serotec, Oxford, UK], and 0.076% 3-[(3-cholamidopropyl)dimethylammonio]propanesulfonate [CHAPS], pH 7.2) and 75 μL of CNBr-activated Sepharose beads (Pharmacia, Peapack, NJ) covalently crosslinked to nonspecific IgG1 κ. After incubation for 2 hours at 4°C, the beads were removed by centrifugation.
Homogenates and plasma samples (100 μL, in triplicate) were incubated in BNT77 (Takeda, Osaka, Japan)-coated wells containing 50 μL of buffer C for plasma and 25 μL of buffer C for brain homogenate (20 mmol/L PO4, 400 mmol/L NaCl, 2 mmol/L EDTA, 0.2% BSA, 10% BlockAce, and 0.076% CHAPS, pH 7.2) overnight at 4°C. The plates were washed four times with PBS, then reacted with horseradish peroxidase-conjugated detector antibodies (BA27 IgG2 mouse anti-Aβ40 1:1000 for homogenate, 1:5000 for plasma; BC05 IgG1 mouse anti-Aβ42 1:10,000-0.5 μg/mL; Takeda) in 75 μL of buffer D (20 mmol/L PO4, 400 mmol/L NaCl, 2mmol/L EDTA, and 1% BSA, pH 7.0) for 3 h at room temperature. After 6 washes with PBS, horseradish peroxidase enzyme activity was measured using QuantaBlu Fluorogenic Substrate (Pierce, Rockford, IL) on a Wallac Victor V2 plate reader using a 320-nm excitation filter and 400-nm emission filter. Each plate contained known concentrations of human synthetic Aβ 1-40 and Aβ 1-42 (Bachem, King of Prussia, PA) in Tris or neutralized FA buffer to construct a log-log standard curve. These ELISAs can detect N-terminally truncated BACE-cleaved Aβ species (Aβ 11-40/42) as well as full-length Aβ (Aβ 1-40/42) but not α-secretase cleaved products (p3-Aβ 17-40/42) (26).
AD and non-AD cases were compared by the nonparametric Wilcoxon rank-sum test. Spearman correlations between each biochemical and neuropathologic measure were evaluated with plasma Aβ measures. Linear regression examined the association of the plasma Aβ measures with the biochemical and neuropathologic measures, adjusting for age, diagnosis, and time from plasma collection to autopsy.
Matched plasma and brain samples were obtained from 25 research subjects. Plasma was collected from 24 days to 3.5 years before death (median: 1 year). Seventeen patients had a neuropathologic diagnosis of AD; the remainder had neuropathologic diagnoses of PSP (2 patients), CBD (1 patient), hippocampal sclerosis (1 patient), frontotemporal dementia (FTD) (1 patient), and no diagnostic abnormality (3 patients). The clinical diagnosis at the last visit was consistent with the neuropathologic diagnosis, except for 2 individuals with a clinical diagnosis of AD at the last visit who had a neuropathologic diagnosis of FTD and hippocampal sclerosis, respectively. Two subjects had mild dementia/minimal cognitive impairment; of these, both met Consortium to Establish a Registry for Alzheimer Disease neuropathologic criteria for "possible AD,” with Braak and Braak stages of I/VI and II/VI, respectively. In addition, an individual with a clinical diagnosis of primary progressive aphasia had CBD, and an individual with a clinical diagnosis of Pick disease had AD.
There was no significant difference between the AD and non-AD groups in age at blood draw (p = 0.22), age at death (p = 0.21), or time from the blood draw to death (p = 0.41). An APOE ε4 allele was present in 53% of the AD patients and 25% of the non-AD patients (Table 1).
Plasma Aβ40 levels were lower in AD than in non-AD cases (p = 0.05), whereas Aβ42 levels were similar. We surveyed a range of biochemical and pathologic Aβ measures in brain for correlation with plasma levels. These included Aβ40 and Aβ42 levels in soluble (Tris) and insoluble (FA) fractions of temporal and frontal neocortex and the deposited Aβ40 and Aβ42 burden in these same brain regions. All Aβ measures were elevated in the AD versus non-AD neocortex (p < 0.05, except p = 0.07 for temporal soluble [Tris] Aβ40) (Fig. 1). As expected, Aβ42 burden tended to exceed Aβ40 burden; Aβ40 preferentially localized to plaque cores and amyloid angiopathy (Fig. 2).
Correlation Between Plasma Aβ and Brain Aβ
We determined the Spearman correlation between the plasma and brain Aβ measures among all the cases (Table 2) and among the AD cases alone (Table 3). Among all cases, the neuropathologic and brain biochemical measures of Aβ40 and of Aβ42 were intercorrelated. Plasma levels of Aβ40 and Aβ42 did not correlate with any of the brain measures (Table 2, Fig. 3).
Among the AD cases alone, the intercorrelation among the brain measures was weaker. Plasma Aβ42 was inversely correlated with frontal Aβ42 burden (p = 0.03) and frontal soluble (Tris) Aβ42, but not any other brain measures (Table 3, Fig. 4).
Regression analyses were also adjusted for age, interval between plasma draw and death, presence of cerebral amyloid angiopathy, and duration of illness. Even with these adjustments, no consistent correlations of plasma levels of Aβ40 and Aβ42 were found with the brain measures across the full sample. We were also interested in whether the APOE ε4 allele influenced the relation between plasma and brain Aβ. Although the individual correlations were not statistically significant, plasma Aβ42 tended to be inversely correlated with biochemical Aβ42 measures in the brain in APOE ε4 carriers (Spearman r = −0.12 to −0.53, n = 11), but positively correlated with biochemical brain Aβ42 measure in noncarriers (r = +0.26 to +0.53, n = 13) (Fig. 5). In a linear regression model predicting plasma Aβ42, a test for interaction between FA temporal Aβ42 and APOE ε4 was significant (p < 0.05), adjusting for diagnosis, and an interaction between FA Aβ42 and APOE ε;4 was marginal (p = 0.07). This suggests that the APOE ε4 allele may modify the relationship between plasma and brain Aβ42 measures.
To assess the relevance of plasma Aβ levels as a reflection of pathologic changes in the AD brain, we compared plasma Aβ levels during life to postmortem brain Aβ measures in well-characterized and diverse neuropathologic specimens. As expected, biochemical and neuropathologic measures of Aβ40 and Aβ42 were elevated in AD brains relative to non-AD brains. However, across all cases, plasma Aβ levels were not associated with brain levels.
Plasma Aβ levels are probably modulated by peripheral and brain metabolism and clearance as well as by transport across brain, CSF, and vascular compartments. Although all forms of brain Aβ are elevated in AD, the weak correlations of the various brain Aβ measures in AD suggest that these may reflect distinct biochemical and morphologic pools of Aβ (6). Furthermore, plasma Aβ levels and CSF Aβ levels are not correlated in AD (27). It is not unexpected, therefore, that plasma Aβ and brain Aβ levels are not strongly correlated.
The results of our study should be interpreted in the context of the patient population, sample size, and study design. The number of patients in the study was small, with an approximately 80% power to detect a correlation of 0.54 or greater at α = 0.05. However, correlations were detected among the brain Aβ measures in the entire cohort. Although plasma was obtained from 1 to 3 years before death, plasma measures of Aβ tended to be relatively stable (28). Our finding that the APOE genotype might alter the relationship between plasma and brain Aβ level is based on a small population and should be extended in further studies. If confirmed, it suggests that interpretation of plasma biomarkers would need to be adapted in light of other genetic factors.
There are several implications of these results to the use of plasma Aβ as a biomarker of AD. Plasma Aβ cannot reliably distinguish cases with prominent brain Aβ deposition (AD) from those without Aβ deposition (non-AD), limiting its usefulness as a diagnostic or predictive biomarker. The interesting inverse relation of plasma Aβ42 with some brain Aβ42 measures within the AD cases may be a peripheral reflection of low CSF Aβ42 in AD. Finally, the pathology of AD reflects the end stage of a pathophysiologic process beginning years before the onset of symptoms. If plasma Aβ is a risk factor for AD as is suggested in two longitudinal cohort studies (28), the relevant Aβ levels may be those seen 5 to 20 years before death (29).