Synergistic Association between Plasma Aβ1-42 and p-tau in Alzheimer's Disease but Not in Parkinson's Disease or Frontotemporal Dementia.

Beta-amyloid (Aβ1-42) triggers the phosphorylation of tau protein in Alzheimer's disease (AD), but the relationship between phosphorylated tau (p-tau) and Aβ1-42 in the blood is not elucidated. We investigated the association in individuals with AD (n = 62, including amnesic mild cognitive impairment and dementia), Parkinson's disease (n = 30), frontotemporal dementia (n = 25), and cognitively unimpaired controls (n = 41) using immunomagnetic reduction assays to measure plasma Aβ1-42 and p-tau181 concentrations. Correlation and regression analyses were performed to examine the relation between plasma levels, demographic factors, and clinical severity. Both plasma Aβ1-42 and p-tau concentrations were significantly higher in AD and frontotemporal dementia than in the controls and Parkinson's disease. A significant positive association was found between plasma p-tau and Aβ1-42 in controls (r = 0.579, P < 0.001) and AD (r = 0.699, P < 0.001) but not in frontotemporal dementia or Parkinson's disease. Plasma p-tau was significantly associated with clinical severity in the AD in terms of scores of clinical dementia rating (r = 0.288, P = 0.025) and mini-mental state examination (r = -0.253, P = 0.049). Regression analysis showed that plasma Aβ1-42 levels explain approximately 47.7% of the plasma p-tau levels in the AD after controlling age, gender, and clinical severity. While in non-AD participants, the clinical dementia rating explained about 47.5% of the plasma p-tau levels. The disease-specific association between plasma Aβ1-42 and p-tau levels in AD implies a possible synergic effect in mechanisms involving these two pathological proteins' genesis.

Introduction

The key pathological features of Alzheimer’s disease (AD) include extracellular accumulation of amyloid β (Aβ) peptides and intracellular hyperphosphorylation of tau protein and consequent neurofibrillary tangle (NFT) formation.[1,2] The amyloid cascade hypothesis has dominated research and therapeutic drug development of AD since it was proposed three decades ago;[3] this hypothesis states that the imbalance of Aβ production and clearance is an early or initiating event.[4−6] However, recent failures of clinical trials of disease-modifying therapies targeting amyloid deposition and formation have not supported the amyloid cascade hypothesis. On the other hand, although the amyloid cascade hypothesis also endorses the hyperphosphorylation of intracellular tau protein to be triggered by soluble Aβ oligomers in the interstitial fluid, its mechanism has not been fully elucidated.[4−6] Glycogen synthase kinase-3β has been proposed as a mediator of tau phosphorylation in neurons;[7] insulin-like growth factor-binding protein 3 may also be an essential tau kinase.[8] The coexistence of Aβ and phosphorylated tau (p-tau) was demonstrated pathologically in human post-mortem studies[8] and AD animal models.[4−6] Biomarker studies in cerebrospinal fluid (CSF) of individuals with AD show concomitant reduction of Aβ1–42 and an increase in p-tau levels,[9,10] which was suggested to reflect the deposition of Aβ into plaques and the increased phosphorylation state of tau protein.[11] However, the interrelation or effect size of the two abnormal proteins remains to be explored.

With the developments of ultrasensitive technologies in the past decade, it has been shown that Aβ1–42 and p-tau181 are present at low concentrations in human plasma.[12−14] These technological advances enable the investigation of the interrelation between p-tau181 and Aβ1–42 in human plasma, which may help to elucidate the pathogenic mechanism of proteinopathies. For instance, in individuals with prodromal or clinical AD, extracellular Aβ deposition precedes and precipitates aggregation of pathological intracellular p-tau protein. This report aims to use immunomagnetic reduction (IMR) to measure Aβ1–42 and p-tau181 in cognitively unimpaired individuals and participants with mild cognitive impairment (MCI) due to AD and AD dementia (ADD). In addition to AD, individuals with Parkinson’s disease (PD) and frontotemporal dementia (FTD) were enrolled as positive controls for assessing these two plasma protein biomarkers’ disease-specific association.

Results and Discussion

The enrolled subjects’ demographic information is listed in Table with the information on the AD spectrum subgroups, including MCI due to AD and ADD, and PD spectrum subgroups, including PD with normal cognition (PD-NC) and PD with impaired cognition (PD-IC). There were 41 individuals (65.1 ± 6.8 years, 75.6% women) in the control group. There were 62 (74.4 ± 7.8 years, 72.6% women) in the AD spectrum group, 36 with MCI due to AD, and 26 with ADD. There were 30 patients in the PD spectrum group (67.8 ± 11.6 years, 40% women), 17 with PD-NC, and 13 with PD-IC. There were 25 patients in the FTD (age: 63.8 ± 7.4 years, 72% women). Individuals in the FTD group were further classified according to their clinical features, including 12 with semantic-type primary progressive aphasia (PPA), 3 with nonfluent PPA, 5 with behavior-variant FTD, 2 with behavior-variant FTD-amyotrophic lateral sclerosis, 2 with FTD-progressive supranuclear palsy, and, finally, 1 with FTD corticobasal degeneration. One subject with logopenic variant PPA was included in the AD spectrum group for analysis. Around 90% of the subjects with logopenic variant PPA showed amyloid positivity in brain image, and about 80% showed AD pathology.[15,16] Individuals in the AD spectrum group were significantly older than those in the other groups (all P < 0.005); therefore, subsequent statistical analyses were performed by adjusting for the age effect. Significant between-group differences in gender distribution (Pearson’s χ = 17.327, P = 0.004) and clinical severity in terms of global clinical dementia rating (CDR) scores (Pearson’s χ = 259.678, P < 0.001) were also observed. We used regression analyses to examine gender and clinical severity effects.

Table 1

Demographic, Clinical Information, and Measured Levels of Plasma Aβ1-42 and p-tau181 for Enrolled Subjectsa

		AD spectrum			PD spectrum
group	CON	aMCI	ADD	combined	PD-NC	PD-IC	combined	FTD
N (female%)	41 (75.6%)	36 (83.3%)	26 (57.7%)	62 (72.6%)	17 (35.3%)	13 (46.2%)	30 (40%)	25 (72%)
age (yr.)	65.1 ± 6.8	72.7 ± 7.8	76.7 ± 7.5	74.4 ± 7.8	66.3 ± 13.1	69.8 ± 9.5	67.8 ± 11.6	63.8 ± 7.4
MMSE	29.2 ± 0.7	25.9 ± 2.6	19.0 ± 4.0	23.0 ± 4.7	29.5 ± 0.9	27.1 ± 1.0	28.5 ± 1.5	18.4 ± 9.1
CDR (0/0.5/1/2/3)	41/0/0/0/0	0/36/0/0/0	0/6/17/3/0	0/42/17/3/0	17/0/0/0/0	0/13/0/0/0	17/13/0/0/0	0/8/11/3/3
Aβ1–42(pg/mL)	15.6 ± 2.3	18.2 ± 1.9	20.4 ± 3.5	19.2 ± 2.9	15.4 ± 2.4	16.4 ± 3.6	15.9 ± 2.9	18.3 ± 2.7
p-taul81(pg/mL)	2.6 ± 1.2	4.3 ± 1.5	6.6 ± 2.6	5.3 ± 2.4	3.9 ± 1.0	3.3 ± 1.1	3.6 ± 1.1	6.6 ± 1.3

N: number of subjects; CON: control group; aMCI: amnestic mild cognitive impairment due to AD; ADD: Alzheimerʼs disease dementia; AD spectrum: aMCI plus AD; PD: Parkinsonʼs disease; PD-NC: Parkinsonʼs disease with normal cognition; PD-IC: Parkinsonʼs disease with impaired cognition including Parkinsonʼs disease dementia and Parkinsonʼs disease mild cognitive impairment; PD spectrum: PD-NC plus PD-IC; FTD: frontotemporal dementia; MMSE: mini-mental state examination; CDR: clinical dementia rating.

The measured levels of plasma Aβ1–42 in the four groups are plotted in Figure a. Group differences in plasma Aβ1–42 and p-tau181 were compared using multivariate analysis of covariance (MANCOVA), controlling for the age effect. The controls had the mean plasma Aβ1–42 level of 15.6 ± 2.3 pg/mL. The mean level of the MCI group was 18.2 ± 1.9 pg/mL, and the ADD was 20.6 ± 3.5 pg/mL for plasma Aβ1–42. The MCI group had higher levels of plasma Aβ1–42 than the controls (P < 0.001); the levels of plasma Aβ1–42 in the ADD group were higher than those in the MCI group (P = 0.007).

Figure 1

Concentrations of (a) plasma Aβ1–42 and (b) plasma p-tau181 for subjects in the controls (CON), AD spectrum (AD), PD spectrum (PD), and FTD spectrum (FTD). The black triangles denote MCI due to AD, and gray triangles represent ADD. The black rectangles are for PD-NC, and the gray rectangles indicate PD-IC (PDMCI and PDD). The symbols * indicate MANCOVA p = 0.01, ** indicate p = 0.001, and *** indicate p < 0.001 with Bonferroni correction for confidence intervals.

In the PD group, both PD-NC (15.4 ± 2.4 pg/mL) and PD-IC (16.4 ± 3.6 pg/mL) did not show significantly different between-group levels of plasma Aβ1–42 from each other or the controls (P > 0.5). However, the FTD group had a mean level of plasma Aβ1–42 of 18.3 ± 2.7 pg/mL, which was significantly higher than the controls (P = 0.002). We removed one outlier by the 3-standard deviation rule in the FTD group due to a plasma level of Aβ1–42 28.9 before further correlation and regression analyses.

Figure b shows plasma p-tau181 concentrations for all four groups. The controls had a mean level of 2.6 ± 1.2 pg/mL for plasma p-tau181. The MCI and ADD groups had mean levels of 4.3 ± 1.5 and 6.6 ± 2.6 pg/mL, respectively. The concentrations of plasma p-tau181 in both aMCI and mild AD were found to be significantly higher than the controls (P < 0.001); similarly, the concentrations of plasma p-tau181 in the ADD group were significantly higher than those in the MCI group (P < 0.001). Compared to the controls, the concentrations of plasma p-tau181 in the PD-NC and PDD groups were not significantly different (PD-NC: 3.56 ± 1.16 pg/mL; PDD: 3.70 ± 0.96 pg/mL, both P > 0.05). The mean plasma p-tau181 concentration in the FTD group (6.67 ± 1.34 pg/mL) was significantly higher than the controls (P < 0.001).

In addition to the comparison with the controls, we also examined the between-group differences of various neurodegenerative diseases. For plasma Aβ1–42, there were significant between-group differences between AD and PD spectrum (p < 0.001) and PD and FTD (p = 0.01). For plasma p-tau, there were significant between-group differences between AD and FTD and between PD and FTD (both p = 0.01) and AD and PD (p < 0.001). However, for clinical classification of various neurodegenerative diseases, multiplex plasma biomarkers with the help of a machine-learning model are helpful.[17]

We further examined the subgroup difference in various FTD subtypes but found no significant difference, which was probably due to small sample size (both ANCOVA p > 0.4). However, some premature yet interesting findings could be presented. The highest plasma p-tau was in the PSP subgroup (7.6 ± 0.8 pg/mL, n = 2) and the lowest in the ALS subgroup (5.2 ± 0.5 pg/mL, n = 2). On the other hand the highest plasma Aβ1–42 was in the behavioral variant (20.5 ± 4.9 pg/mL, n = 5) and the lowest in the ALS subgrouop (16.4 ± 1.3 pg/mL, n = 2). Further exploration with a large sample size should be warranted.

The relationships between p-tau181 and Aβ1–42 concentrations in plasma for various types of dementia are plotted in Figure . A partial correlation controlling the age effect was performed to examine the interrelation between plasma Aβ1–42 and p-tau181 concentrations. The correlation coefficients between plasma p-tau181 and Aβ1–42 concentrations reached significant levels for the controls (r = 0.579, P < 0.001) and for the AD spectrum (r = 0.699, P < 0.001) but not for the PD spectrum (r = 0.208, P = 0.280) or for FTD (r = 0.052, P = 0.811).

Figure 2

Concentration relationships displayed by ordinary least-squares regression between plasma p-tau and Aβ1–42 in the control (green), AD spectrum (black), PD spectrum (orange), and FTD spectrum (blue). AD denotes AD spectrum, including aMCI due to AD and ADD groups. PD spectrum includes PD-NC and PD-IC groups. Only the control and AD spectrum reach significant positive associations.

For the correlation of plasma biomarkers to clinical severity, the only significant findings were between plasma p-tau levels and both global CDR scores (r = 0.288, P = 0.025) and mini-mental state examination (MMSE) scores (r = −0.253, P = 0.049) but not between plasma Aβ1–42 and global CDR scores (r = 0.173, P = 0.184) or MMSE scores (r = −0.246, P = 0.056) in the AD spectrum group.

We further performed group-wise (in terms of the spectrum) regression analyses using plasma p-tau levels as the dependent variable. The results showed that plasma Aβ1–42 levels indicated an r square (R2) change of 0.324 (P < 0.001) in the controls, excluding age, gender, MMSE, and CDR. In the AD spectrum, plasma Aβ1–42 levels indicated an R2 change of 0.477, while gender indicated an additional R2 change of 0.064, and CDR further indicated an R2 change of 0.031, excluding age and MMSE (Table ). No significant effect of the independent variables could be found in the PD spectrum or FTD group.

Table 2

Regression Analyses Using Plasma p-tau Levels as Dependent Variable and Plasma Aβ1-42 Levels and Demographic and Clinical Information as Independent Variablesa

group-wise	R2/R2 change	F change	P-value	excluding variables
Control (n = 41)
model I: Aβ1–42	0.324/0.324	18.659	<0.001	age, gender, MMSE
AD Spectrum (n = 62)
model I: Aβ1–42	0.477/0.477	54.676	<0.001	Age, gender, CDR, MMSE
model II: Aβ1–42, gender	0.540/0.064	8.155	0.006	age, CDR, MMSE
model III: Aβ1–42, gender, CDR	0.572/0.031	4.254	0.044	age, MMSE
All Excluding AD Spectrum (n = 96)
model I: CDR	0.473/0.473	84.333	<0.001	age, gender, Aβ1–42, MMSE
model II: CDR, Aβ1–42	0.095/0.558	20.344	<0.001	age, gender, MMSE

MMSE: mini-mental state examination; CDR: clinical dementia rating; no significant results found in all independent variables of both PD spectrum and FTD groups.

Finally, we also performed a regression analysis with non-AD participants, all subjects excluding the AD spectrum. The result showed that p-tau was mainly explained by CDR with an R2 change of 0.477, excluding age, gender, MMSE, and Aβ1–42 levels, and Aβ1–42 levels further added an R2 change of 0.095 excluding age, gender, and MMSE (Table ).

In tau positron-emission tomography (PET) studies, tracer retention increases modestly with age throughout the brain in cognitively unimpaired individuals, while elevated tau is seen more often when amyloid brain accumulation is present.[18] In cognitively unimpaired individuals, total tau concentrations in CSF increase with age, and CSF p-tau concentrations increase with age in ApoE4 carriers.[19] In our previous study, age explained approximately 13% of the variance in plasma total tau levels in a group of 126 cognitively unimpaired individuals aged 45–95 years.[20]

On the other hand, the age-related deposition of brain amyloid protein in cognitively unimpaired individuals is less well studied, and most relevant studies have reported that approximately 15–30% of cognitively unimpaired aging adults showed positive amyloid deposition, with ApoE4 carriers having a higher risk.[21,22] The Australian Imaging, Biomarkers and Lifestyle study of aging showed that amyloid burden increased with age most strongly in ApoE4 carriers.[23] However, the impact of ApoE4 on cognitive function has not been determined.[24] For plasma biomarkers, our previous studies showed that ApoE4 carriers have higher plasma Aβ1–42 /Aβ1–4o ratios[12] and have higher plasma total tau levels than their noncarrier counterparts.[25]

In contrast to CSF tau levels, the CSF-Aβ1–42 or Aβ1–42/Aβ1–40 ratios did not correlate with age.[19] In our previous study, the plasma Aβ1–42 levels showed a modest but significant negative relation with age (r = −0.126, p = 0.0128) in a population of 391 cognitively unimpaired adults aged 23–91 years.[26]

According to the results in Figure a, subjects in the AD spectrum showed the most noticeable difference in plasma Aβ1–42 levels compared with the other diagnostic groups. This finding is in keeping with the results of our previous studies of plasma Aβ1–42 using IMR measurements,[12,27] which was found to be positively correlated with brain amyloid deposition in terms of the Aβ1–42/Aβ1–40 ratio[28] but moderately negatively correlated with CSF Aβ1–42 levels.[13]

In Figure b, both the AD spectrum and FTD groups showed increased levels of p-tau in plasma. AD pathology is a dual proteinopathy characterized by the coexistence of extracellular aggregates of mainly Aβ1–42 forming neuritic Aβ plaques and intracellular aggregates of p-tau forming NFT.[29] Despite the potential synergistic relationship between these two proteinopathies, the observation of tau pathology early in the disease course[30] and relatively good association with clinical severity suggest that AD is a dual proteinopathy consisting of both Aβ1–42 and p-tau. The latter finding was supported by this study that only plasma p-tau levels but not Aβ1–42 were associated with clinical severity in the AD spectrum groups, compatible with a recent study which also reported that plasma p-tau181 was associated with clinical severity and tau-PET.[14]

On the other hand, FTD consists of a spectrum of clinical syndromes associated with several underlying neurodegenerative diseases characterized by frontotemporal lobar degeneration (FTLD).[31] From a pathological point of view, most (90–95%) FTLDs are caused by intracellular aggregates of p-tau or TAR DNA-binding protein 43 (TDP-43).[32,33] Mixed FTLD pathologies or unclassifiable tauopathies are also not infrequently observed.

A recent study assessing Aβ in 98 individuals with pathologically confirmed frontotemporal dementia syndromes showed that in individuals with various types of frontotemporal dementias, 8%–29% of individuals showed Aβ deposition in the frontotemporal cortexes, and the prevalence increased to 29%–50% if the basal ganglia or substantia nigra were included.[34] Amyloid molecular imaging studies using Pittsburgh Compound B (PiB)-PET scans also revealed amyloid depositions in cortexes and subcortical areas in individuals with FTD.[34,35] The coexistence of Aβ and p-tau or other proteinopathies in FTD syndrome, such as TDP-43, fused in sarcoma can be expected; thus, it is unsurprising to observe elevated plasma Aβ1–42 levels in individuals with FTD. The mean level of plasma Aβ1–42 is approximately at the level of MCI due to the AD in this study.

The main pathological hallmark for PD is Lewy bodies, which mainly consist of α-synuclein. Although amyloid molecular imaging studies showed amyloid deposition in individuals with α-synucleinopathy, those with amyloid positivity mostly have dementia with Lewy bodies (DLB) and PDD. In a meta-analysis study including 233 individuals with DLB, PD, and PD-MCI receiving PiB-PET scans, the prevalence rates were 0.68 (95% CI 0.55–0.82) in the DLB group, 0.34 (95% CI 0.13–0.56) in the PDD group, and 0.05 (95% CI −0.07–0.17) in the PD-MCI group. In our PD-IC group (combined PD-MCI and PDD), most of the cases were PD-MCI (with a mean MMSE score of 27.1 ± 0.1), and the mean plasma level was 3.6 ± 1.1 pg/mL, which, although elevated, did not reach significant levels compared to the controls.

Although we observed elevation of both levels of plasma Aβ1–42 and p-tau181 concentrations in both the AD spectrum and FTD groups, their interrelations are notably different and might have different implications. In Figure , a significant linear correlation between p-tau181 and Aβ1–42 in plasma was observed in the AD spectrum group (r = 0.699, P < 0.001) but not in the FTD group (r = 0.052, P = 0.811). Groupwise regression analysis showed that in the AD spectrum, plasma Aβ1–42 levels explained approximately 47.7% (P < 0.001) of the plasma p-tau levels, followed by gender (6.4%, P = 0.006) and severity in terms of CDR (3.1%, P = 0.044). Amyloid hypothesis obtained supporting mechanistic evidence, at least in parts, from those observations that soluble oligomers of Aβ1–42 decrease synapse number, inhibit long-term potentiation, and enhance long-term depression hippocampal neurons. The Aβ1–42 oligomers also increase abnormal phosphorylation of tau, driving vicious cycles leading to AD pathology.[36] The amyloid hypothesis finds its support mainly from early onset AD when dominant mutations involving amyloid-β-related pathogenesis. However, recent evidence from observation of aging human brains and an animal model of late-onset AD, aging rhesus macaque,[37] leads to a hypothesis that tau pathology is probably an initiating factor for sporadic late-onset AD.[38] Pathological tau may also drive Aβ cleavage, consequently increasing Aβ production by p-tau in the microtubules trapping amyloid precursor protein-containing endosomes in dendrites, propelling a vicious cycle of tau and amyloid pathology over a long lifetime.[38]

Regarding the positive association between plasma Aβ1–42 and p-tau181 concentrations in the controls, we proposed that some individuals of the controls might be at the preclinical stage of AD, the assumption is reasonable considering the high prevalence of AD in the general population, especially in individuals with subjective memory decline and in ApoE4 carriers, or it also reflects a brain aging process with both proteinopathies.[18,23,26] These findings warrant further exploration with a longitudinal study.

There are two limitations to this study. First, most participants did not have molecular imaging, such as PiB PET or tau PET, to support their diagnoses. However, with a careful clinical diagnosis and structural brain imaging (mostly magnetic resonance imaging), we can still achieve a degree of clinical confidence.

Second, tau has over 40 sites for phosphorylation, but we measured only plasma p-tau181 in this study. Although CSF p-tau181 was considered capable of discriminating between AD and other neurodegenerative diseases, p-tau181 changes not entirely specific to AD. CSF p-tau181 elevations could be found in FTLD,[39] especially late-onset types.[40] Recent studies showed that p-tau217 outperforms p-tau181 in the differential diagnosis between AD and non-AD dementia.[41] In the future, the inclusion of the plasma TDP-43, α-synuclein, and p-tau217 is warranted.

In conclusion, by using immunomagnetic reduction to assay plasma Aβ1–42 and p-tau181 concentrations, we observed a positive correlation between Aβ1–42 and p-tau181 in plasma in AD but not in PD or FTD. The disease-specific association between plasma Aβ1–42 and p-tau levels in AD warrants further exploration of a possible bidirectional synergic effect involving the genesis of these two pathological proteins.

Summary

Alzheimer’s disease’s (AD) key pathological features include extracellular aggregates of amyloid-beta (Aβ) forming neuritic plaques and intracellular aggregates of hyperphosphorylated tau (p-tau) forming neural fibrillary tangles. Aβ triggers the hyperphosphorylation of tau in AD, but the relationship between blood p-tau and Aβ is not elucidated. We investigated the relation in AD, Parkinson’s disease, frontotemporal dementia, and controls using immunomagnetic reduction assays to measure plasma Aβ and p-tau. We found a positive association between plasma p-tau and Aβ in controls and AD but not in frontotemporal dementia or Parkinson’s disease. Plasma p-tau but not Aβ was associated with clinical severity in the AD. Regression analysis showed that plasma Aβ explains approximately 47.7% of the plasma p-tau levels in the AD. In comparison, CDR explains about 47.5% of plasma p-tau in the non-AD participants. The disease-specific association in AD implies a possible synergic effect involving the genesis of these two pathological proteins.

Methods

Recruitment of Subjects

The 158 subjects were enrolled at National Taiwan University Hospital (NTUH), Taiwan, Triservice General Hospital (TSGH), Taiwan, and Sahlgrenska University Hospital (SUH), Sweden. Individuals with either MCI due to AD or ADD were diagnosed according to Naitonal Institute on Aging and Alzheimer’s Association (NIA-AA) diagnostic guidelines.[42,43] Individuals with Parkinson’s disease (PD) were diagnosed using the United Kingdom PD Society Brain Bank clinical diagnostic criteria.[44] PD with dementia (PDD) or mild cognitive impairment (PD-MCI) was further diagnosed according to the diagnostic guidelines suggested by the Movement Disorder Society Task Force to separate them from PD-NC.[45,46] We used MMSE scores of ≤25 as the cutoff value for significant cognitive dysfunction in PDD as well as impairment of instrumental activities of daily living (IADL) and MMSE scores of 26–28 with normal informant-based IADL for PD-MCI.[47] The two groups, PDD and PD-MCI, were combined to constitute PD-IC for further analysis. Individuals with FTD were enrolled and classified according to two diagnostic criteria consensus.[48,49] In SUH, AD patients were further confirmed by CSF Aβ1–42 ≤ 530 pg/mL and total tau protein ≥350 pg/mL.[50] The SUH cohort recruited 29 participants, including 14 ADD patients and 15 control subjects (Supplementary Table 1). For both the controls and ADD groups, there is no significant difference in both plasma levels of Aβ1–42 and p-tau between the Taiwanese and Swedish cohorts (Supplementary Table 2).

Most participants received a CDR and an MMSE to evaluate their clinical severity.

All study participants or their primary caregivers provided informed consent before participation in this investigation. The study was approved by the Institutional Research Ethics Committee of NTUH, Institutional Review Board of TSGH, National Defense Medical Center, and Central Ethical Review Board, University of Gothenburg for SUH. The study was carried out following the Helsinki Declaration of 1975.

Preparation of Plasma

An EDTA blood collection tube was used for blood collection, followed by centrifugation with speeds ranging from 1500–2500 g for 15 min at room temperature. The plasma in the EDTA tube was transferred and aliquoted into 0.5 mL microcentrifuge tubes and stored at −80 °C until IMR measurements were performed. Plasma was frozen no later than 3 h after blood collection. Collected plasma samples were delivered to MagQu Co., Ltd., Taiwan, by a dry ice package for blindly assaying plasma Aβ1–42 and p-tau181. The protocols for plasma preparation were identical in both Taiwanese and Swedish cohorts.

Assay of Aβ1–42 and p-tau181

IMR was utilized to assay Aβ1–42 and p-tau181 in collected plasma samples. Briefly, IMR, using the reduction in the ac magnetic susceptibility, that is, the IMR signal, of the reagent after being mixed with a sample, was measured using a superconducting quantum interference device-based ac magnetic susceptometer (XacPro-S, MagQu).[12,51]

For assaying Aβ1–42, 60 μL of plasma was mixed with 60 μL of reagent (MF-AB2–0060, MagQu), which involves the use of an antibody (ab34376, Abcam) against Aβ1–42 immobilized on dextran-coated magnetic Fe3O4 nanoparticles (MF-0060-DEX, MagQu). The epitope of the Aβ1–42 antibody is amino acid region 37–42 aa. For assaying p-tau181, 40 μL of plasma was mixed with 80 μL of reagent (MF-PT1–0060, MagQu), involving the use of an antibody (MN1050, Thermo), which is the same as the commonly used AT270 antibody, against p-tau181 immobilized on dextran-coated magnetic Fe3O4 nanoparticles (MF-0060-DEX, MagQu), and an IMR analyzer (XacPro-S, MagQu) was used to analyze the concentrations of Aβ1–42 and p-tau181 in each plasma. Duplicate measurements were conducted for each biomarker of every sample. The reported concentration of Aβ1–42 or p-tau181 of a plasma sample was the averaged value of duplicate measurements. The variation in the duplicate measurements’ levels was lower than 20% for both Aβ1–42 and p-tau181.

Before performing the analysis of Aβ1–42 and p-tau181, two calibrators consisting of magnetic fluid (CA-DEX-0060, CA-DEX-0080, MagQu) were applied to calibrate the IMR analyzer readings. The details of the preparation of the magnetic fluid are described in our previous work. For every batch of analysis of Aβ1–42 and p-tau181, control solutions were used for quality control. The control solutions for Aβ1–42 are pure PBS solution (negative control) (CL-AB2-00T, MagQu) and 20 pg/mL Aβ1–42 solution (CL-AB2-020T, MagQu). The control solutions for p-tau181 were pure PBS solution (negative control) (CL-PT1-00T, MagQu), and 5 pg/mL p-tau181 solution (CL-PT1-005T, MagQu). The acceptable measured concentrations for the 20 pg/mL Aβ1–42 control solution are 17–23 pg/mL and 4.25–5.75 pg/mL for the 5 pg/mL p-tau181 control solution.

Statistical Analyses

Continuous variables for each measurement were presented as the mean ± standard deviation, and group differences were compared using analysis of variance for continuous variables and Pearson’s χ2 test for noncontinuous variables. Missing data were handled by using subgroup means. Group differences of plasma biomarkers were compared using MANCOVA, controlling for the age effect. Partial correlation controlling for the age effect was performed to examine the interrelation of plasma biomarkers or their relation to clinical data. Ordinary least square regression was also used to explore relationship of Aβ1–42 levels to plasma p-tau levels. We further performed a stepwise linear regression analysis to estimate the independent variables’ potential contribution to the dependent variables. We examined and removed possible outliers in further correlation and regression analyses. The statistical analyses were performed with IBM SPSS Statistics version 20 (Armonk, NY), and graphs were produced using GraphPad Prism version 8 (San Diego, California).