Plasma amyloid beta peptides: an Alzheimer’s conundrum or a more accessible Alzheimer’s biomarker?

KEYWORDS:

• Alzheimer’s disease
• amyloid beta
• plasma
• biomarkers
• oligomers
• sink model
• amyloid fibrils
• tau
• amyloid cascade hypothesis
• dementia

The vexed question of whether Aβ peptides are diagnostic of Alzheimer’s disease (AD) continues to pepper the Alzheimer’s literature. From a mechanistic viewpoint, these peptides are hallmark features of the AD pathomechanism, being major constituents of both plaques and cerebral amyloid angiopathy (CAA) of the microvasculature, features on which the very diagnosis of AD is predicated. Conversely, overall plaque deposition in the AD brain is poorly correlated with clinical severity of AD, plaque load is common in the normal elderly [1], and its relationship to disease mechanisms still contentious. Ambiguities on the role of Aβ in AD pathogenesis have lead some to question the veracity of the amyloid cascade hypothesis, in particular, reliance on this model as a sole explanation for AD [2]. However, most researchers accept that Aβ is necessary, if not sufficient, for AD causation [3,4] and that plaque load and/or abnormal Aβ levels even in the cognitively normal are adverse features and indicative of pathology. This current paradigm is supported by strong evidence that Aβ acts as a trigger for downstream events, which may include (i) dysregulation of tau [4], (ii) ‘prion-like’ effects spreading to/seeding other neurotoxic AD associated proteins, such as tau [3,4], (iii) N-methyl-d-aspartate (NMDA) and calcium ion driven excitotoxicity [5], (iv) alteration of proteostatic mechanisms [6], (v) inflammatory processes and oxidative stress [7], (vi) compromised metabolic and mitochondrial functions [8], and (vii) disruption of autophagy [9]. The full spectrum of mechanistic links between Aβ dysregulation and AD pathophysiology is still incompletely understood, but Aβ remains a central and inevitable feature of AD, and arguably a pivot-point on which the balance can sway to health or disease. The complexity of AD clearly makes additional lines of investigation necessary, not least the factors which may cause initial dysregulation of Aβ, but an integrated mechanistic understanding of AD cannot avoid the amyloid cascade concept.

Much of the current mechanistic insight and temporal sequencing of molecular events in AD has been provided by cellular and animal models, but in the end, there is no substitute for human studies of AD. A variety of biomarker approaches in humans now utilize Aβ measures in vivo, so that quantifying the brain Aβ load is no longer a post-mortem only option. These measures include positron emission tomography assay of amyloid deposition in the brain, as well as soluble forms such as cerebrospinal fluid (CSF) Aβ1–40, Aβ1–42 monomers, the Aβ1–42/Aβ1–40 ratio, and/or oligomers of Aβ. A growing body of evidence shows that changes to these various Aβ forms occur at preclinical stages of late onset AD [10]; the soluble multimers/oligomers are the most toxic forms [5]; they are early features of familial AD [11]; their accumulation/aggregation temporally precedes tau dysregulation [4]; and, tau is neurotoxic but in the absence of Aβ aggregation does not lead to AD [3]. Therefore, Aβ measurement will no doubt continue to be of interest as both a research and a clinical tool.

The current popular view of the Aβ lifecycle is that the shorter (Aβ1–40) version has normal functions (though potentially contributing to CAA and vascular impairment), whereas the longer peptide variants (Aβ1–42/43) prevalent in AD are the more amyloidogenic. These have a greater tendency to form fibrillar structures which are not only likely to be toxic in their own right [12], but also lead to formation of soluble multimers and oligomers with well-established evidence for neurotoxicity [13]. The end point is insoluble plaque, which may be a relatively innocuous ‘gravestone’ with relatively low toxicity, possibly acting as a sink for the more toxic soluble variants, since many cognitively healthy older individuals carry substantial plaque load [2]. A major question relates to which of these multiple Aβ variants should we take as being representative of disease diagnosis, prognosis, or predisposition, and will a different Aβ measure serve best for each purpose. Major confounders include the dynamic nature of Aβ over time as it follows the sink model trajectory, the considerable overlap with normal aging [14], and its specificity for Alzheimer’s dementia, since changing Aβ levels are reported in other neurodegenerative disorders and also in normal aging [15].

CSF contains relatively abundant levels of soluble Aβ and there is good consensus in the literature that higher levels of Aβ1–42 are observed at preclinical stages of AD and herald AD onset, with a longitudinal trajectory of declining levels consistent with the sink model of aggregation to plaque [16]. Furthermore, tau and ptau are relatively easily assayed in CSF, and if we accept the ‘trigger and bullet’ concept of ‘Aβ and tau,’ there is merit in monitoring both [4]. However, routine CSF sampling is unlikely to gain universal acceptance, particularly for repeated measurement, tracking effectiveness of disease modifying strategies or basic research projects. Plasma, which is an excellent and routinely used resource for clinical assessment of health status, contains measurable quantities of Aβ and continues to appear in the research literature as the basis for measures of disease status and response to treatment [14]. However, it is also a controversial sample, fraught with questions about assay techniques, non-central nervous system (CNS) sources, considerable dissonance across laboratories in measures used, and poor replication of results. Before disregarding the usefulness of the plasma measure altogether, it is worth considering whether this conflicted state is a reflection of biological reality or a consequence of technical limitations, or both. If the biology of Aβ is multifaceted, then our measures or approaches also need to be, and perhaps deposited, soluble monomeric, multimeric, and oligomeric versions must all be assessed to provide a full picture of normal versus pathological state and prognosis [10]. The current paradigm of Aβ (particularly Aβ1–42) as an AD trigger suggests that Aβ measures should be evaluated early and prior to onset of clinical/neurological sequelae. Recent studies of plasma Aβ levels have shown that they recapitulate the same sink-model trends observed in CSF (albeit with considerably lower Aβ levels in plasma): higher than normal Aβ1–42 levels at preclinical stages, which drop longitudinally with disease onset and progression. This paradigm is supported by the following observations: (i) higher levels in early onset, genetically associated AD, such as in Down’s syndrome and autosomal dominant AD (ADAD), and even as early as children with ADAD [11]; (ii) higher levels in animal models prior to disease onset [17]; (iii) associations with cognitive change in normal individuals [14]; and (iv) association between plasma and CSF levels [18], in longitudinal studies being predictive of progression to AD [19].

For plasma Aβ, some empirical questions that still need to be resolved are as follows: (i) How much is derived from the CNS versus peripheral sources? (ii) Can we reliably measure the toxic variants (i.e. oligomers) in plasma? (iii) In what forms does it travel in plasma (i.e. free Aβ, protein-bound Aβ, bound to other plasma components, such as: high, low, and very low density lipoprotein (HDL, LDL, VLDL, respectively), chylomicrons, exosomes, platelets, and blood cells)? (iv) Does this partitioning vary during disease? (v) What is the normal function of Aβ, beyond its role in disease? (vi) What molecular processes during aging predispose to dysregulation of Aβ, in the absence of genetic mutations to amyloid precursor protein or the presenilins? Furthermore, common covariates used in AD research of Aβ include age, gender, and apolipoprotein E allele, but a variety of other possible confounders are often not considered or controlled for, such as medications, supplements, cholesterol, HDL, physical activity, vitamins, homocysteine, depression, renal function, glucose status, and body mass index. What effect these variables have on plasma Aβ levels warrants further exploration?

These questions and dilemmas notwithstanding, plasma Aβ remains an important research tool, and improvement/development of techniques, such as plasma oligomer measures, standardization of plasma handling, and assay techniques across laboratories, use of well characterized, highly specific and sensitive approaches, together with use of large well-researched aging cohorts, which include multiple measures of Aβ load such as amyloid imaging and CSF as well as plasma estimation, will continue to provide insight into disease mechanisms. From recent studies, a picture is emerging of a dynamic and variable plasma Aβ trajectory, with lower levels associated cross-sectionally with diagnosed disease or impaired cognition, and higher basal levels predictive of future cognitive decline or conversion to MCI or AD [14]. Population-based cohorts of older age participants (>60 years), such as the Framingham study, the 3-city Dijon study, the Australian Imaging Biomarkers and Lifestyle (AIBL) study of Ageing and the Alzheimer’s Disease Neuroimaging Initiative (ADNI) (2189, 1530, 1112, and 715 subjects, respectively), are all now pointing to the value of plasma Aβ measures as predictors/indicators of pathology or cognitive change [20–24] and are of value in research studies to understand mechanisms or evaluate response to treatments.

Plasma Aβ is not yet ready for the clinic, and it is unlikely that single time-point measures will on their own be diagnostic of late-onset AD, largely due to the considerable overlap with, and range width, of the normal aging population. Repeat analyses to track longitudinal change may yield a better return in the clinic, and plasma is an ideal sample for this purpose. Cumulative data suggest that plasma Aβ measures are indeed meaningful of CNS events, and part of its appeal is that so far it is the only biomarker in plasma that is also a hallmark of AD pathology. Robust assays for tau and ptau in plasma would also be useful, though plasma levels of these proteins are vanishingly low and will require much more sensitive approaches. Furthermore, proteomics, metabolomics, and lipidomics techniques are identifying a variety of molecules associated with cognitive change, aging, and dementia. In particular, markers of inflammation and lipid/cholesterol transport and metabolism have frequently been investigated as biomarkers, and may give a fuller picture of the early events which lead to AD. In the hunt for the elusive AD plasma biomarker, perhaps we need to consider Aβ assays as just one, albeit important, arrow to the quiver.

Declaration of interest

The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

Additional information

Funding

This paper has been supported by the Australian National Health and Medical Research Council Program Grant (NHMRC) [grant number APP1093083], The Australian Research Council Discovery Project Grant [grant number DP120102078] and the Rebecca L. Cooper Medical Research Foundation.