Amyloid Beta and Tau as Alzheimer’s Disease Blood Biomarkers: Promise From New Technologies

Alzheimer’s disease (AD) core pathological components, amyloid beta (Aβ) peptide 42 (Aβ42), Aβ40, tau, and tau phosphorylated at threonine-181 (Thr181P), have been targets for biomarker development for two decades [1‐5]. The underlying rationale of using these molecules as biomarkers of AD is that definitive diagnosis of this condition relies on confirmation from neuropathological hallmarks containing these components, and the ability of tracking or measuring these components in brains or biofluids of living subjects could provide evidence of ongoing pathophysiology. There has been tremendous progress made towards achieving this goal. Ligands have been developed for visualizing amyloid or tau pathologies in the brain by positron emission tomography (PET) [6, 7]. The utility of amyloid imaging and cerebrospinal fluid (CSF) biomarkers of Aβ and tau for clinical diagnosis has been validated in large cohorts and population studies, as well as in neuropathologically confirmed cases [8‐15]. The use of the ratios of CSF Aβ42 to tau or Thr181P has been established as a means for identifying clinically diagnosed AD [2, 9, 16‐22]. The inclusion of AD core biomarkers in the criteria for the diagnosis of probable and possible AD has been recommended by the National Institute on Aging–Alzheimer’s Association workgroups [23‐26]. The strategy of combining amyloid visualization by PET and CSF core markers has been demonstrated to improve the accuracy of predicting the pre-clinical stage of AD [25, 26]. As PET primarily visualizes fibrillar amyloid deposits, CSF Aβ measures could be more sensitive in detecting changes in Aβ levels at the pre-clinical stage or even earlier [27‐29]. By contrast, the development of assays for use in measuring AD core pathological molecules in blood as disease biomarkers has fallen behind [30, 31]. Several factors could have hampered the development. One of the major challenges in developing AD core pathological components such as blood biomarkers has been the lack of sensitive assays. Analyses using singulex or multiplex enzyme-linked immunosorbent assay (ELISA) platforms for determining the levels of Aβ and tau in plasma or serum have led to conflicting findings—with levels either unchanged, decreased, or increased from the normal controls—suggesting limitations due to the assay platforms [32‐41].

This article is based on previously conducted studies and does not involve any new studies of human or animal subjects performed by any of the authors.

The discrepancy in the levels of AD core markers in blood samples could be due to a wide range of causes: biological nature, assay sensitivity, platform, sample processing, storage condition, clinical criteria, and/or demographic features of the participants, as discussed in a previous review [42]. From a biological point of view, the concentrations of Aβ and tau in the circulation are much lower than those in the CSF because these molecules present in the brain do not directly enter the circulation due to the presence of the blood–brain barrier (BBB). Some Aβ molecules are cleared at the BBB through receptor-mediated mechanisms, but others are cleared from the CSF through the lymphatic drainage system [43‐45].

The assay platforms for measuring CSF Aβ, tau, and Thr181P, namely traditional singulex ELISA and luminex-based multiplex ELISA, are well-established [19, 46‐48]. It has been consistently shown that compared to normal controls Aβ42 levels in the CSF are lower in patients with AD, while both tau and phosphorylated tau (p-tau) levels are higher in patients with AD (for example, see [48]). Combining CSF tau, p-tau, or Aβ40 levels as ratios with Aβ42 is more effective than standalone markers in predicting brain Aβ deposition detected by PET imaging in patients with AD or in the preclinical stage of AD [49‐51].

Figueski et al. measured Aβ42 and Aβ40 levels in plasma samples using similar ELISA platforms and found that the plasma levels were one-fifth to one-tenth the levels found in the CSF [33]. Using an ultra-sensitivity platform, Janelidze et al. found even wider differences in detection levels between plasma and CSF samples [36]. Plasma Aβ levels tend to be near the lower limits of detection of current ELISA assays and close to the low-end of the linear range of a calibration curve. Under these conditions, the ELISA assays lose their sensitivity for detecting narrow differences between biological samples. The performance of immunoassays also depends on the epitopes and affinity of the antibodies selected for capturing antigen contained in the samples. Moreover, plasma and serum contain high concentrations of albumin and immunoglobulins, which are known Aβ-binding proteins that can interfere with accurate detection of the free forms of Aβ [52, 53]. Endogenous immunoglobulins, autoantibodies, and heterophilic antibodies can also interfere with the performance of ELISAs. This has been discussed in depth previously [54].

In addition to the above-mentioned factors, improper sample handling can affect assay accuracy. Most of the lessons in this regard were learned during the process of establishing these AD core markers for CSF. As plasma contains much higher concentrations of proteins, including degradative enzymes, as well as more complex components than CSF, additional problems of these sorts are to be expected. These include the need for rigorous procedures of sample collection, volume of sample aliquots, type of tubes for storing aliquots, number of freeze–thaw cycles, type of calibration proteins, batch-to-batch reagent variability, and site-to-site operations [55].

To overcome the challenges of detection encountered using traditional ELISA platforms, new approaches and technologies are emerging with the potential to provide superior sensitivity and specificity for measuring Aβ and tau in blood samples [56‐58]. For example, a new approach that used immunoprecipitation to pull down various Aβ fragments in plasma samples followed by mass-spectrophotometry analysis has led to discovery of a new marker, APP669–711, whose ratio to Aβ1-42 demonstrated 93% sensitivity and 96% specificity to discriminate Pittsburgh compound B (PIB)-positive subjects from PIB-negative subjects [58]. New immunoaffinity-based assays have also been applied to AD core marker analysis in biofluids, including immunomagnetic reduction (IMR) and single molecule array (SIMOA) for the analysis of plasma samples and the assay by Meso Scale Discovery (Rockville, MD) for CSF samples [59]. Here, we limit the scope of this review to the IMR and SIMOA assays, which have been used to quantify plasma Aβ and tau in studies involving medium to large numbers of subjects.

The IMR technology was developed by MagQu Company, Ltd. (New Taipei City, Taiwan), and the SIMOA technology was developed by Quanterix (Lexington, MA). Quantification with both platforms is based on immunoreactivity between specific antibodies and analytes or protein standards. However, the principle and design of the two detection systems are quite different. IMR technology detects alternating-current magnetic susceptibility by a superconducting quantum interference device (SQUID), while SIMOA technology detects the presence of antigen by fluorescence imaging of single enzyme-labeled immunocomplexes reacting with the fluorogenic substrate, resorufin β-d-galactopyranoside [57, 60]. The technical features of the two platforms are summarized in Table 1.

The IMR and SIMOA assays were developed to increase the detection sensitivity of immunoassays, and they have been used to analyze plasma levels of Aβ and tau in human subject studies.

Studies using IMR assays have mainly been conducted in Taiwanese cohorts [57, 61, 64‐66], and studies involving other ethnic groups are ongoing (personal communication by authors). The IMR assays performed to date in Taiwanese subjects have revealed elevated Aβ42 levels, reduced or no change in Aβ40 levels, and increased tau levels in patients with AD when compared to normal controls [61, 64‐66]. Receiver operating characteristics curve analyses showed 96% sensitivity and 97% specificity for distinguishing healthy controls from a heterogeneous group of study subjects consisting of those with mild cognitive impairment (MCI) due to AD and those with mild to severe AD (Clinical Dementia Rating scores 0.5–3), whereas an 80% sensitivity and 82% specificity was obtained for discriminating patients with AD from those with MCI [61]. When amyloid PET imaging was used to stratify the study subjects, there was 84% sensitivity and 100% specificity to predict the results of amyloid detected by PET when the ratio of Aβ42 to Aβ40 was used [66]. The excellent AD diagnostic performance indicated by sensitivity and specificity in these studies using the product of Aβ42 and tau has yet to be compared in independent studies from other sites.

The SIMOA platform has been shown to detect increases in tau level in patients with AD from tau levels in patients with MCI and normal controls, and for Aβ and for tau in studies of AD [36, 39, 56, 67‐69]. The results of plasma tau studies have shown significant increases in tau in patients with AD, and increases or no changes in those with MCI compared with normal controls, while higher plasma tau levels have also been associated with reduced memory performance [39, 68]. The authors of these studies drew the same conclusion that due to substantial overlap between clinically diagnosed groups plasma tau concentration cannot be used as a prognostic or diagnostic marker. When plasma Aβ42 and Aβ40 levels were assayed by SIMOA, significant decreases were detected between patients with AD and those with MCI, between those with AD and subjects with subjective cognitive decline (SCD), and between those with AD and normal controls [36]. The detection levels of plasma Aβ1–42 [mean ± standard deviation (SD); normal controls: 19.6 ± 5.2 pg/ml; AD: 13.2 ± 7.3 pg/ml) were less than 10% of those of plasma Aβ1–40 (mean ± SD; normal controls: 276.7 ± 66.1 pg/ml; AD: 244.3 ± 105.8 pg/ml). In this study, the CSF Aβ levels were not assayed by the SIMOA assay, but by the Euroimmun immunoassay (EUROIMMUN AG, Lübeck, Germany) [49, 51]. CSF Aβ1–42 levels in patients with AD were significantly decreased compared with other groups (P < 0.0001), while Aβ1–40 levels in patients with AD were only significantly lower than those of SCD subjects (P = 0.003). The detected values of CSF Aβ42 were 554.0 ± 195.1 pg/ml in normal controls and 289.5 ± 103.8 pg/ml in patients with AD (P < 0.0001), whereas the values of Aβ40 were 4688.5 ± 1650.0 pg/ml in normal controls and 4387.2 ± 1761.6 pg/ml in patients with AD (no statistical significance). Although different platforms were used to assay CSF and plasma samples, the results showed that CSF and plasma Aβ42 and Aβ40 levels were significantly positively correlated (Pearson’s correlation analyses in all participants: r = 0.274, P < 0.001 for Aβ42; r = 0.136, P = 0.001 for Aβ40).

Strategies for the development and utility of blood-based biomarkers for AD have been discussed in detail recently in several review articles [30, 70‐72]. In this article, we focused on two new ultra-sensitive immunoaffinity-based technologies that offer promise for establishing Aβ and tau as blood biomarkers for AD. Currently, these two platforms are uniquely situated for further assessment, especially in large population studies. However, the advance could be limited by the cost of the instruments, the lack of high-throughput capacity, and single suppliers of assay reagents. The availability of a throughput automated instrument, such as the SIMOA HD-1 analyzer, will certainly appeal to pharmaceutical companies when considering biomarkers to assess the progress of clinical trials in large numbers of subjects, in which plasma Aβ and tau measurements might have potential utilities. Nevertheless, recent studies on plasma tau have not confirmed its feasibility as a diagnostic or prognostic biomarker due to large overlap between AD and MCI, and between AD and normal controls, regardless of the presence of differences in disease-associated expression. The SIMOA assay for plasma Aβ had shown preliminary potential for utility of diagnosis. In these regards, the IMR technology seems to make more progress, evident from a series of cohort studies that showed good sensitivity and specificity and promising correlations with PET imaging of amyloid and tau. However, the IMR technology will need to be assessed vigorously in cohorts of different ethnicity, and in longitudinal study of those subjects stratified by amyloid or tau imaging, or by CSF Aβ and tau profiles. Although both platforms are consistent in showing increases in plasma tau levels in patients with AD, the Aβ42 findings were opposite. It has been cautioned that comparing findings between different platforms could be problematic [73, 74]. However, future studies are needed to replicate the differences in findings between platforms before the issue of whether plasma Aβ42 levels are increased or decreased in AD can be resolved.

In summary, ultrasensitive platforms are necessary for establishing whether plasma AD core markers can be valid blood-based biomarkers. As pointed out recently by O’Bryant and colleagues, significant breakthrough in establishing blood-based biomarkers could be achieved when the context of use can be defined at the beginning of biomarker development and if approaches from academic research and industry can be integrated during the process [71]. Preliminary assessment of published findings support that both IMR and SIMOA technologies warrant multicenter cross-validation study.