Brain to Plasma Amyloid-β Efflux: a Measure of Brain Amyloid Burden in a Mouse Model of Alzheimer's Disease

See allHide authors and affiliations

Science  22 Mar 2002:
Vol. 295, Issue 5563, pp. 2264-2267
DOI: 10.1126/science.1067568


The deposition of amyloid-β (Aβ) peptides into amyloid plaques precedes the cognitive dysfunction of Alzheimer's disease (AD) by years. Biomarkers indicative of brain amyloid burden could be useful for identifying individuals at high risk for developing AD. As in AD in humans, baseline plasma Aβ levels in a transgenic mouse model of AD did not correlate with brain amyloid burden. However, after peripheral administration of a monoclonal antibody to Aβ (m266), we observed a rapid increase in plasma Aβ and the magnitude of this increase was highly correlated with amyloid burden in the hippocampus and cortex. This method may be useful for quantifying brain amyloid burden in patients at risk for or those who have been diagnosed with AD.

Abundant evidence suggests that a key event in the pathogenesis of AD is the conversion of Aβ peptides from soluble to insoluble forms in the brain (1). This process is among the earliest pathological changes that characterizes AD, and is estimated to occur ∼10 to 20 years before the appearance of the earliest cognitive changes of the disease (2, 3). Whereas individuals with pre-clinical AD (i.e., cognitively normal individuals with plaque and tangle densities similar to those with AD) have no measurable neuronal loss in affected brain regions, individuals with even very mild cognitive impairment indicative of clinical AD have plaques and tangles and also have already lost a significant number of neurons (4, 5). Identification of biomarkers predictive of the presence and magnitude of amyloid plaque burden may be useful in the diagnosis of both pre-clinical and clinical AD. In the case of pre-clinical AD, these biomarkers might allow for detection of individuals who may benefit from preventative therapies even before the development of neuronal loss and cognitive impairment. Further, such biomarkers may be of help in distinguishing patients with mild cognitive impairment (MCI) due to AD and those with other underlying disease processes.

The Aβ peptides are predominantly 40 to 42 amino acids in length and are synthesized as soluble proteolytic products of the amyloid precursor protein (APP), a large integral membrane protein expressed at high levels in the brain (1). Recent data suggest that after Aβ synthesis in the central nervous system (CNS), the peptide(s) can be locally metabolized or cleared into the plasma (6–9). With the use of a tg mouse model of AD amyloidosis, PDAPP (APPV717F) mice, in which a humanAPP transgene with a familial AD mutation is expressed (10), we have recently demonstrated that peripheral administration of a monoclonal antibody to Aβ (m266) results in a rapid and massive increase of CNS-derived Aβ in the plasma (9). Our data suggested that m266 was both decreasing plasma Aβ degradation as well as facilitating Aβ efflux from CNS to plasma. The presence and amount of insoluble Aβ deposited in plaques in the brain appears to influence the metabolism and clearance of soluble Aβ. For example, as plaque deposition progresses, the amount of insoluble Aβ in brain lysates increases markedly (11). Though the amount of soluble Aβ in brain lysates does not increase to the same extent, it does increase (12). Thus, we reasoned that the accumulation of Aβ in plasma after the administration of m266 might be a measure of (or in some way reflect the amount of) Aβ burden in the brain.

We assessed a large cohort (n = 49) of PDAPP mice homozygous (+/+) for the APPV717F transgene, all of which were 12 to 13 months of age. These mice begin to exhibit Aβ and amyloid deposition in the brain at ∼3 to 6 months of age. Though all PDAPP+/+ mice ultimately develop deposits of Aβ in the form of diffuse and neuritic (thioflavine-S-positive, amyloid) plaques after 6 months, there is a rather large variability in the degree of amyloid pathology in individual animals of the same age (9,13, 14). This inherent variability in amyloid burden in our age-matched cohort allowed us to compare plasma Aβ levels with brain Aβ burden. Five minutes before administering m266 to the mice, we obtained a baseline plasma sample from each animal. Each mouse then received an intravenous (i.v.) injection of m266 (500 μg). Plasma samples from each animal were then obtained 5 min and 1, 3, 6, and 24 hours later. In each plasma sample, levels of Aβ40 and Aβ42 were assessed by enzyme-linked immunosorbent assay (ELISA) as previously described (9). Mice were killed after 24 hours. One hemisphere was assessed with the use of quantitative Aβ-immunofluorescent and thioflavine-S (amyloid) staining to determine the area of the hippocampus or cingulate cortex occupied by Aβ and amyloid, respectively (% Aβ or amyloid load), and regions from the other hemisphere were assessed for Aβ by ELISA as described (9,14, 15). Neuropathological assessment of Aβ and amyloid load was carried out by investigators blind to the plasma Aβ levels.

Baseline levels of plasma Aβ40, Aβ42, and the calculated Aβ40/42 ratio before the administration of m266 did not correlate with the amount of Aβ or amyloid deposition present in the brain (Table 1,Fig. 1). This finding is similar to that previously reported in humans where plasma Aβ has been shown not to be a useful biomarker in distinguishing AD patients from age-matched controls (16). As we have previously reported, after parenteral (i.v.) administration of m266, there was a rapid and marked increase in plasma Aβ40 and Aβ42. We also observed highly significant correlations between levels of plasma Aβ (Aβ40, Aβ42, Aβ40/42 ratio) and both Aβ and amyloid burden in the hippocampus (Fig. 1, Table 1) and in the cingulate cortex. In addition, highly significant correlations were observed when comparing the total amount of plasma Aβ40 and Aβ42 accumulated over 24 hours (area under the curve, AUC) and both Aβ and amyloid burden (Table 1). Highly significant correlations were obtained when comparing plasma Aβ levels after m266 administration to brain Aβ levels measured by ELISA. We next grouped mice according to their Aβ burden and compared plasma Aβ levels among those with the lowest, middle two, and highest quartiles of Aβ burden (Fig. 2A). There was no overlap in plasma Aβ40 levels assessed at 24 hours (or in AUC for plasma Aβ40) after m266 administration between mice in the lowest quartile (0 to 1.4%) versus those in the highest quartile (18.2 to 34.5%) of brain Aβ load (Fig. 2, A, C, and D). By contrast, plasma Aβ levels measured just before m266 administration failed to differentiate these two groups of mice (Fig. 2B).

Figure 1

Plasma levels of Aβ40 following m266 administration are highly correlated with Aβ and amyloid burden in hippocampus. Just before (pre-bleed) and after the i.v. administration of m266 (500 μg), plasma samples were collected at various times (1, 3, 6, and 24 hours). Plasma Aβ40 and Aβ42were measured by ELISA as previously described (9). Aβ and amyloid load were quantitated as described (14). Before m266 administration, there was no correlation between the plasma levels of Aβ40 (or Aβ42) (Table 1) and % Aβ or amyloid load in the hippocampus. In contrast, 24 hours after i.v. administration of m266, there were highly significant correlations (see Table 1) when comparing the plasma levels of Aβ40 and the Aβ40 accumulation over 24 hours (AUC) to Aβ and amyloid load in the hippocampus.

Figure 2

PDAPP +/+ mice (12 to 13 months of age) with varying amounts of Aβ deposition can be differentiated by plasma Aβ measurements after the parenteral (i.v.) administration of m266. (A) Mice were divided into quartiles based on Aβ load in the hippocampus: the lowest quartile of hippocampal Aβ load (low, 0 to 1.4% Aβ load), the middle two quartiles of Aβ load (medium, 1.4 to 18.2% Aβ load), and the highest quartile of Aβ load (high, 18.2 to 34.5%). Aβ-immunofluorescent staining (top panels) and thioflavine-S (amyloid) staining (bottom panels) reveals the range in the amount of deposition between the groups. Representative images from a mouse in the lowest quartile, middle two quartiles, and highest quartile of pathology are shown. (B) Baseline plasma Aβ40 levels, before the injection of m266, did not differentiate the groups. However, after injection of m266, plasma Aβ40 levels after 24 hours (C) and cumulative plasma Aβ40 levels (AUC) (D) that reflect the amount of Aβ accumulating in plasma over 24 hours were all highly correlated to the amount of Aβ deposition in the hippocampus. There is no overlap between plasma Aβ levels in the low versus the high Aβ load groups after m266 administration (C) and (D).

Table 1

Plasma Aβ correlations with Aβ load and fibrillar amyloid in hippocampus with AD-like pathology. The Pearson correlation coefficient (Pearson r) and significance (P value) were determined between plasma Aβ values (before and after injection of m266) and hippocampal Aβ or amyloid load with the use of GraphPad Prism software (v. 3.00 for Windows, San Diego, CA). Aβ load was defined as the % area of the hippocampus covered by Aβ-immunoreactive deposits. Amyloid load was defined as the % area of the hippocampus covered by thioflavine-S positive deposits. Correlations were also determined between the plasma Aβ accumulation over 24 hours (AUC) and hippocampal Aβ load or amyloid load.

View this table:

To further explore whether measuring plasma Aβ levels after m266 administration could potentially be used to develop a diagnostic test to estimate brain Aβ or amyloid burden, we used recursive-partitioning, an exploratory statistical technique for revealing rule-based relations in data (17, 18). This technique generates a set of hierarchical “rules” which can be graphically displayed as a tree dendogram (decision tree). To develop predictive models, we examined all plasma Aβ values after m266 administration in relation to Aβ burden in the hippocampus. Using diagnostic rules based on only one plasma Aβ measurement [Aβ40 at 24 hours (19)] or on multiple measurements (19), several models were generated that accurately predicted the levels of hippocampal Aβ burden (19). We next developed a rule-based “diagnostic” procedure whereby, after the administration of an antibody to Aβ, the presence of “high” or “low” plaque burden could be predicted with acceptable false-negative and false-positive rates for such a test. Using a decision tree, one model allowed us to discriminate between animals with greater than the 50% percentile of Aβ burden in the cohort (n = 49) from those with less than that percentile, with a sensitivity of 96% and a specificity of 84%. The recursive-partitioning models discussed were intentionally forced to be very conservative, which suggests that more complex models with even higher predictive power can be developed for eventual use in humans.

Here, we found no correlation between baseline levels of plasma Aβ and brain Aβ burden in a large cohort of 12-to 13-month-old PDAPP mice. In contrast, after parenteral administration of m266, we observed rapid and large increases in plasma Aβ40, Aβ42, and the Aβ40/42 ratio that were highly correlated with the amount of Aβ deposition and amyloid burden in brain. There was no overlap between plasma Aβ40 levels at 24 hours after m266 administration in mice from the lowest and highest quartiles of Aβ burden quantitated in the hippocampus. Our findings have obvious implications for quantifying Aβ and/or amyloid burden in humans. However, whether such a “diagnostic” test in humans would yield similar results is unclear. Though there is a wide range of Aβ and amyloid deposition that occurs with age in both humans and PDAPP mice, all PDAPP mice overproduce human Aβ and, unlike humans, all mice will eventually develop Aβ and amyloid deposition in the brain. In quantitative terms, this contrasts with what is observed in the aging human brain. Whereas cortical amyloid plaque burden in humans with pre-clinical and clinical AD are similar to each other and to that observed in the PDAPP mice we studied with high Aβ burden, studies have shown that most cognitively normal elderly humans (∼70% by age 75) have either no or only very small amounts of cortical Aβ deposition (3, 20). The latter human group would be analogous to the mice in our study with little to no Aβ deposition (lowest quartile). This dichotomy in amyloid plaque burden observed in the aging human brain suggests, therefore, that measuring plasma Aβ after administration of antibody to Aβ may be able to clearly distinguish such individuals. Thus, the use of a monoclonal antibody with characteristics similar to m266 but developed for humans may provide a means to develop a facile diagnostic test to quantify amyloid burden in persons with pre-clinical AD, as well as to assist in the differential diagnosis of clinical AD. Such a test may also have utility for monitoring the response to anti-amyloid therapy.

The highly significant correlations between plasma Aβ and both brain Aβ and amyloid burden strongly suggest that the presence of m266 in the peripheral circulation directly facilitated net Aβ efflux from the brain, acting as a “peripheral sink.” Further supporting this model is that significant correlations were observed within 5 min after peripheral injection of m266. By increasing Aβ efflux from brain, it appears that the presence of m266 in plasma can also reveal quantitative differences in brain Aβ deposition, presumably by facilitating efflux of soluble Aβ from brain. Taken together, our data suggest that brain Aβ clearance is a dynamic process and that modifying this process may be useful in both diagnosing and treating AD.

  • * These authors contributed equally to this work.

  • To whom correspondence should be addressed. E-mail: holtzman{at} (D.M.H.) or Paul_Steven_M{at}


Stay Connected to Science

Navigate This Article