Plasma amyloid-beta oligomer is related to subjective cognitive decline and brain amyloid status

We recruited participants who had concerns about their cognitive decline through advertisements from the community. Recruited participants underwent assessment for eligibility in a university-affiliated general hospital from September 2020 to April 2021. The inclusion criteria included age from 60 to 79 years, literacy, and preserved hearing and vision sufficient to perform neuropsychological tests. The exclusion criteria were any history of major neurological or psychiatric diseases (such as Parkinson’s disease, stroke, schizophrenia, or bipolar disorder), alcohol/substance use disorder within 3 years before enrollment, and intracranial hemorrhage confirmed by magnetic resonance imaging. We also excluded participants with any history of dementia or those considered to have AD at the time of screening according to the National Institute on Aging-Alzheimer’s Association (NIA-AA) criteria [17]. All participants completed self-report scales and underwent neuropsychological tests, clinical interviews, neuroimaging of Aβ by PET, and blood sampling. After assessing the initial 164 participants, we excluded 24 participants who were determined to have MCI by the NIA-AA criteria [18] so that the remaining 140 had normal objective cognition. We also excluded 14 participants who refused blood sampling after the initial assessment with written consent. The remaining 126 participants with normal objective cognitive function (SCD or normal cognition) were finally analyzed. All patients were enrolled after providing informed consent under the approval of the institutional review board.

The Seoul Neuropsychological Screening Battery-Core (SNSB-C) was used to assess objective cognitive function. To present global cognitive function, we used the composite score of the SNSB-C [19], calculated from ten subtests for five cognitive domains as followings: (i) Digit Span Test for attention; (ii) short form of Korean-Boston Naming Test for language; (iii) Rey Complex Figure Test for visuospatial function; (iv) Seoul Verbal Learning Test-Elderly’s version: Immediate recall, Delayed recall, and Recognition for memory; and (v) short form of the Korean-Color Word Stroop Test, Controlled Oral Word Association Test, Korean-Trail Making Test-Elderly’s version B, and Digit Symbol Coding for frontal/executive function. The SNSB-C composite scores were transformed into Z-scores that were corrected for age, sex, and level of education.

Among our participants with normal objective cognition (SCD or normal cognition), we did not dichotomize participants as those with SCD and those without (normal cognition). Instead, we quantified the severity of the self-perceived cognitive decline in all participants by two widely used self-report scales: the Subjective Cognitive Decline Questionnaire (SCDQ) [20] and the Memory Age-associated Complaint Questionnaire (MACQ) [21]. SCDQ consists of 24 yes-or-no questions assessing the difficulty of performing activities requiring cognitive function in the past 2 years. The total SCDQ score ranges from 0 to 24, with a higher score indicating more self-perceived cognitive decline. The MACQ was developed to measure individuals’ subjective age-related memory decline compared with memory at their younger ages. The MACQ is comprised of six items that are rated from 7 to 35, with a higher score indicating more subjective memory decline.

The plasma AβO level was measured by Multimer Detection System (MDS) [22, 23] using the inBlood™ AβO Test commercial kit (Peoplebio Inc., Gyeonggi-do, Korea). MDS is a modified sandwich enzyme-linked immunosorbent assay (ELISA) which exclusively detects oligomers or multimers by epitope-overlapping antibodies targeting a unique epitope in the Aβ monomer. Details of the quantification technique are available in previously published papers [13, 24]. In brief, MDS uses two types of antibodies to detect AβO: the mouse monoclonal antibody 6E10 (BioLegend, San Diego, CA, USA) as the capturing antibody and WO2-HRP antibody (Absolute Antibody Ltd., Oxford, UK) as the detection antibody. Capturing antibodies pre-coated in well-plate react with and capture the epitope of Aβ (the N-terminus 3–8). Then, detection antibodies also react with and bound to the epitope of Aβ (the N-terminus 4–10). Since the epitopes of these antibodies overlap at the N-terminus 4–8 of Aβ, AβO with multiple epitopes could react with both capturing and detection antibodies. However, Aβ monomer with one epitope could only react with capturing antibodies and not be detected by detection antibodies. After the antibody–antigen reaction, a multispectrophotometer quantified the luminescence signal of the plate, which allowed the calculation of relative luminescence units. Based on the relative luminescence units compared with a standard curve, we determined the concentration of AβO, expressed in ng/mL. To validate the plasma AβO concentration, we measured the AβO level in another sample from the same participant. In this additional sample, the AβO level was expressed as a ratio of the concentration to the mean value of two internal standards, as previously described [12]. The plasma AβO level was repeatedly measured for each participant at 6-month intervals. The mean value of two successive measurements was defined as the participant’s l plasma AβO level.

The brain Aβ deposition was visualized by amyloid PET using tracer [¹⁸F] flutemetamol and quantified as the standardized uptake value ratio (SUVR) of each cortical region. The details are previously described [25, 26]. Briefly, PET scans were rigidly coregistered to the corresponding structural MRI scans, corrected for partial volume effects. Regional PET uptake values were sampled from 82 brain regions defined in the Desikan-Killiany atlas [27], covering the whole cerebral cortex. The atlas labels multiplied with a binary gray matter mask of the reference template threshold at 50% gray matter probability and were propagated to the participant’s native space using nonlinear image registration. Regional PET uptake means were converted to SUVR by scaling to the mean uptake of the whole cerebellum. The composite SUVR for each participant was derived by calculating the average SUVR values for the bilateral frontal, lateral temporal, parietal, cuneus, and anterior and posterior cingulate cortices [25, 26]. According to a previous study of flutemetamol PET, participants with SUVR ≥ 1.23 were considered as those with abnormal Aβ deposition (flutemetamol PET [+]) [28].

Participants were asked for their medical, smoking, and alcohol drinking history. According to the National Institute on Alcohol Abuse and Alcoholism, heavy alcohol drinking is defined as > 14 standard drinks/week for men and > 7 standard drinks/week for women [29]. The Frail scale (FRAIL) [30, 31] was used to assess frailty, which is known to be related to SCD [1, 32]. Participants with a FRAIL score ≥ 3 were regarded as those with frailty [1, 31]. Depression is associated with a high risk of SCD [32]; therefore, the degree of depression was evaluated using the Geriatric Depression Scale (GDS) [33, 34], in which a high score indicates more depressive symptoms. GDS score ranges from 0 to 30, with 17 as a validated cutoff for major depressive disorder in Korea [34]. We also assessed the presence of apolipoprotein E ε4 allele (APOE4), determined by polymerase chain reaction, and the Clinical Dementia Rating–Sum of Boxes (CDR-SB).

Using binary logistic regression models, we explored whether plasma AβO levels were associated with the presence of amyloid PET positivity in our study sample of normal objective cognition. In these models, the outcome variable was flutemetamol PET positivity, and the independent variable was plasma AβO level, corrected for the same covariates of multiple regression models. Flutemetamol PET (+) was determined by SUVR ≥ 1.23 which is previously described [28]. To assess the capacity of plasma AβO level as a surrogate marker of flutemetamol PET, receiver operative characteristic (ROC) analyses were used. The area under the curve (AUC), sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and optimal cutoff value by the Youden index were calculated.

All statistical analyses were performed using R version 4.1.1 and RStudio version 1.4.1106. Missing data were addressed by listwise deletion. Significance was set at alpha = 0.05.

Overall, the mean age of participants was 73.3 years, with females being predominant (69.0%). The mean SNSB-C Z-score of participants was − 0.007. The number of participants with positive flutemetamol PET scan (SUVR ≥ 1.23) was 13 (10.3%, Table 1).

Multiple linear regression analyses revealed that, in overall participants of normal objective cognition, both higher SCDQ and MACQ scores were significantly associated with the higher plasma AβO levels presented as ratios (Fig. 1A, B). Similarly, these associations were also significant in samples presented as concentrations (ng/mL, Fig. 1C, D). Contrary to plasma AβO levels, brain Aβ deposition, represented as SUVR, was not associated with SCDQ and MACQ scores (Fig. 2A, B).

After correction for the same covariates in multiple linear regression models, binary logistic regression models found a significant association between the presence of flutemetamol PET (+) and plasma AβO levels presented as both ratios (adjusted OR 2.04) and concentrations (ng/mL; adjusted OR 1.58, Table 2). This result means that a 0.1 unit increase in plasma AβO level raised the likelihood of flutemetamol PET (+) in 2.04 (ratio) or 1.58 times (concentration [ng/mL]). In ROC analyses, plasma AβO level presented as a ratio of optimal cutoff 1.047 could discriminate participants with flutemetamol PET (+) with AUC 0.694, sensitivity 69.2%, specificity 73.0%, PPV 23.05%, and NPV 95.29% (Fig. 3A, red line). As presented in concentration (ng/mL), a plasma AβO level of optimal cutoff 0.893 ng/mL could identify flutemetamol PET (+) with AUC 0.662, sensitivity 46.2%, specificity 87.4%, PPV 30.0%, and NPV 93.3% (Fig. 3A, blue line). When combined with APOE4 and plasma AβO level, AUC increased to 0.789 (plasma AβO as ratio) or 0.783 (plasma AβO as concentration [ng/mL]; Fig. 3B).