Changes of plasma Aβ and t-tau along with 18F-florbetapir PET in a cohort of cognitive decline


 Background

Blood based biomarkers for Alzheimer’s disease (AD) are becoming increasingly promising. Although plasma amyloid-β (Aβ) and total tau (t-tau) showed a potential ability in identifying cerebral amyloid deposition and AD. Comparisons of these plasma biomarkers along with Aβ-PET in a cohort of normal controls (NC), subjective cognitive decline (SCD), objectively-defined subtle cognitive decline (Obj-SCD), mild cognitive impairment (MCI) and AD remained lacking.
Methods

A total of 407 individuals aged from 40 to 90 years old were recruited, including 76 of NC, 77 of SCD, 61 of obj-SCD, 92 of MCI and 101 of AD. Plasma Aβ40, Aβ42 and t-tau were examined via single-molecule array (Simoa) immunoassay. A subset of 132 individuals underwent cerebral amyloid scans with 18F-florbetapir PET. Comparisons of plasma t-tau, Aβ40, Aβ42 and Aβ42/Aβ40 ratio were conducted between different diagnostic groups and cerebral Aβ burdens. Pearson correlation analysis was used to evaluate the correlation between Aβ42, Aβ42/Aβ40 ratio and 18F-florbetapir SUVR. Receiver operating characteristic (ROC) curve analyses were carried out to evaluate the capacity of plasma biomarkers in identifying high brain Aβ burden and diagnosis of AD.
Results

Plasma Aβ42 was significantly higher in SCD and obj-SCD than NC, MCI and AD. Plasma Aβ40 was significantly higher in SCD and obj-SCD than NC and AD. The lowest levels of plasma Aβ42 and Aβ42/Aβ40 ratio were found in AD. No significant difference of plasma t-tau was found among groups. Plasma Aβ42 and Aβ42/Aβ40 ratio were inversely correlated with 18F-florbetapir SUVR (r=-0.272, P = 0.003; r=-0.211, P = 0.021 respectively). Aβ42/Aβ40 ratio performed well in predicting high brain Aβ burden (area under the curve, AUC = 0.762). Plasma Aβ42 and Aβ42/Aβ40 ratio had acceptable diagnostic accuracy for AD (AUC = 0.714 and 0.706 respectively), even in Aβ-PET (+) individuals (AUC = 0.728 and 0.808 respectively).
Conclusions

Plasma Aβ40 and Aβ42 measured by Simoa immunoassay showed a significantly bidirectional trend of initially increasing from NC to SCD and obj-SCD, and then declining to MCI and AD. In addition, plasma Aβ was significantly correlated with 18F-florbetapir PET SUVR and showed potential value in predicting cerebral Aβ deposition and risk of AD.

dementia, but not took into account the earlier status such as subjective cognitive decline (SCD) and objectively-de ned subtle cognitive decline (Obj-SCD), which may lead to not accurately re ect the evolution of plasma biomarkers across a complete spectrum of cognitive decline.
The aims of this study were to examine plasma Aβ42, Aβ40 and t-tau with Simoa immunoassay in a cohort of cognitive decline, explore their evolution among normal controls (NC), SCD, Obj-SCD, MCI and AD, and investigate their abilities in identifying cerebral amyloid deposition and diagnosis of AD.

Study participants
In this monocentric retrospective cohort study, a total of 407 individuals were enrolled from Sixth People's Hospital, Shanghai, China, from January 2019 to January 2021. Participants were aged 40 to 90 years, educated more than one year. Individuals with a history of signi cant neurologic disease, psychiatric disorders, alcoholism, drug abuse and head trauma were excluded. Routine laboratory tests and cranial MRI scanning were carried out to preclude relevant diseases which may be adversely affecting cognitive function, such as abnormalities in folic acid, vitamin B12, thyroid function, cerebral infarction, subdural hematomas, hydrocephalus, intracranial tumors and infections. All the participants underwent a battery of standardized neuropsychological tests and examined plasma Aβ42, Aβ40 and t-tau via Simoa immunoassay. A total of 132 participants underwent 18F-orbetapir PET scan within 3 months after blood sampling. Written informed consent was obtained from all the participants or their caregivers. The ethics committee of Shanghai Jiao Tong University A liated Sixth People's Hospital approved this study.

Cognitive groups
The clinical diagnosis of AD was made by experienced neurologists according to the National Institute on Aging and Alzheimer's Association (NIA-AA) criteria for probable AD dementia(28). In the non-dementia participants, MCI and Obj-SCD was classi ed according to the criteria proposed by Jak and Bondi (29). Diagnosis of MCI was given if the participant met one of the following criteria: (1) impaired scores (de ned as >1 standard deviation (SD) below the age-corrected normative mean) on two of the six neuropsychological indexes in the same cognitive domain; (2) impaired scores in each of the three cognitive domains; (3) FAQ score≥9. Participants were classi ed as Obj-SCD according to one of the following criteria: (1) impaired scores on two of the six neuropsychological indexes in the different cognitive domains; (2) impaired scores on one of the six neuropsychological indexes; (3) FAQ score of 6-8. Participants with self-reported experiences of cognitive decline but not met the criterion for MCI and Obj-SCD were classi ed as SCD according to the conceptual framework proposed by the Subjective Cognitive Decline Initiative (SCD-I) working group (30). Normal controls were recruited from caregivers or through advertisements, with no cognitive complaint and impairment determined by neuropsychological tests.
Blood processing and measurements of plasma Aβ42, Aβ40, t-tau Blood samples were centrifuged at 500 g for 5 minutes at 4 ºC to collect plasma. Then plasma was immediately aliquoted into ultra-low adsorption tubes (AXYGEN MCT-150-L-C) on ice and stored at -80 ºC refrigerator. Before the test, plasma samples were transferred from the refrigerator to ice plate for 30 minutes, and then centrifuged at 10,000 g for 5 min at 4 ºC. Total tau measurement was performed using Neurology 4-Plex A kit (Quanterix 102153), Aβ42 and Aβ40 were performed using Neurology 3-Plex A assay (Quanterix 101995) on the Simoa HD-1 platform (Quanterix) (31). Reagent pretreatment and sample loading were carried out according to the instructions of the manufacture. In brief, 350μL pre-diluted calibrator reagent (0−100pg/mL for t-tau, 0−50 ng/mL for Aβ42 and 0−150 ng/mL for Aβ40) and 320μL plasma sample (1:4 diluted) were loaded on the plate. Concentrations of each marker (pg/mL) were calculated from the calibration curve. All the biomarker measurements were performed by laboratory technicians who were blinded to the clinical data.
18F-orbetapir PET acquisition and analysis Amyloid PET images were obtained from a PET/CT system (Biograph mCT Flow PET/CT, Siemens, Erlangen, Germany) at the PET center of Huashan hospital, Fudan University. Cerebral amyloid PET scans were carried out 50min after the intravenous injection of 7.4 MBq/kg(0.2mCi/kg) orbetapir and lasted for 20min. PET images were reconstructed using ltered back projection algorithm with corrections for decay, normalization, dead time, photon attenuation, scatter and random coincidences. PET images were coregistered to the individual structural MRI and partial volume error correction (PVC) was performed using the Muller-Gartner methods (32,33), followed by spatially normalized in the Montreal Neurological Institute (MNI) template and Gaussian smoothing. Standard uptake value ratios (SUVR) were calculated for each cortical region of interest (ROI) relative to cerebellar crus. Global SUVR scores were calculated by weighted averaging of these ROIs. Subjects were diagnosis as 18F-orbetapir PET positive according to the visual interpretation by 3 physicians independently as previously reported.

Statistical Analyses
Quantitative variables were expressed as mean and standard deviation (SD). Categorical variables were expressed as percentage (%).
Differences between the groups were assessed by one-way analysis of variance (ANOVA) or Chi-squared analysis according to the characteristics of data. Post hoc pairwise comparisons of plasma biomarkers between groups were assessed using the least signi cant difference (LSD) test. Pearson correlation analysis was used to evaluate the relationships between plasma biomarkers and values of 18Forbetapir PET SUVR. Receiver operating characteristic (ROC) curve analyses were carried out to evaluate the capacity of plasma biomarkers in discriminating subjects with high composite 18F-orbetapir PET SUVR and subjects diagnosed with AD. All hypothesis testing was twosided, and the level of signi cance was set at α= 0.05. Statistical analyses were conducted using IBM SPSS Statistics 23.0 and MedCalc 19.1. A graphics package (GraphPad Prism, version 8.0) was used to create gures.

Demographic and Clinical Characteristics
The demographic and clinical characteristics in the groups of NC, SCD, Obj-SCD, MCI and AD are shown in Table 1. Compared to other groups, subjects with AD were signi cantly older and had a lower education level. There was no signi cant difference of gender, diabetes and hypertension prevalence among the groups. Performances on all the neuropsychological tests gradually got worse along with the aggravation of cognitive impairment. Of the 132 subjects who underwent cerebral amyloid scans, the Aβ PET positive rates for each group were 7.6% (2/26, NC), 13.7% (4/29, SCD), 36.0% (9/25, Obj-SCD), 48.5% (17/35, MCI) and 94.1% (16/17, AD) respectively.
Plasma t-tau, Aβ40, Aβ42 and Aβ42/Aβ40 among diagnostic groups The plasma levels of t-tau, Aβ40, Aβ42 and the ratio of Aβ42/Aβ40 in different diagnostic groups were shown in Table 2 and Figure 1. There were signi cant differences of plasma Aβ40, Aβ42 and Aβ42/Aβ40 ratio among the groups (P=0.021, P<0.001, P=0.007, respectively), though no signi cant difference was found for plasma t-tau (P=0.762). Post hoc analyses indicated that plasma Aβ40 was signi cantly higher in SCD and Obj-SCD than in NC (P= 0.0101 and 0.0105) and AD (P= 0.0326 and 0.0313), while no signi cant difference was found between NC, MCI and AD. Plasma Aβ42 was signi cantly higher in SCD and Obj-SCD than in NC (P=0.0006 and 0.0058), MCI (P=0.0042 and 0.0262) and AD (both P<0.0001), while no signi cant difference was found between NC and MCI. Signi cantly lower plasma Aβ42 and Aβ42/Aβ40 ratio was observed in AD than all other groups, while no signi cant difference of Aβ42/Aβ40 ratio was found among NC, SCD, Obj-SCD and MCI.

Associations between plasma biomarkers and amyloid PET
Associations between plasma biomarkers and SUVR values of amyloid PET were rst evaluated in all the subjects who underwent 18Forbetapir PET scan. As a result, though no signi cant correlation was found between plasma t-tau, Aβ40 and 18F-orbetapir SUVR, plasma Aβ42 and Aβ42/Aβ40 ratio were signi cantly associated with 18F-orbetapir SUVR (r=-0.272, P=0.003; r=-0.211, P=0.021 respectively) (  Table 3, plasma Aβ42 and Aβ42/Aβ40 ratio were signi cantly lower in the subgroup of high SUVR than low and moderate SUVR (All P<0.05). Although not statistically signi cant, plasma Aβ40 and Aβ42 were both showed an initially increasing trend from low SUVR subgroup to moderate SUVR subgroup, and then decreased in high SUVR subgroup.
Plasma biomarkers as predictors of high brain Aβ burden and AD diagnosis In a subset of individuals who underwent both amyloid PET and plasma biomarkers, ROC Analyses were carried out to evaluate the performances of plasma t-tau, Aβ40, Aβ42 and Aβ42/Ab40 ratio in predicting high brain Aβ burden (18F-orbetapir SUVR >1.35). As shown in Figure 3. A, the highest value of area under the ROC curve (AUC) was seen with Aβ42/Aβ40 ratio (AUC=0.762), followed by plasma Aβ42 (AUC=0.677), though relatively low values were observed with plasma t-tau and Aβ40 (AUC=0.555 and 0.529 respectively).
Comparative diagnostic performances of plasma biomarkers in identifying AD were carried out in all participants regardless of the results of cerebral Aβ status rst (Figure 3.B). ROC curves demonstrated the diagnostic accuracy of plasma Aβ42 (AUC=0.714) and Aβ42/Aβ40 ratio (AUC=0.706) in discriminating AD from non-AD, though plasma t-tau and Aβ40 showed relatively poor performances (AUC=0.613 and 0.563 respectively). Furthermore, in subjects with 18F-orbetapir PET positive, both plasma Aβ42 (AUC=0.728) and Aβ42/Aβ40 ratio (AUC=0.808) showed even higher diagnostic accuracy in identifying AD, though no signi cant difference was found when compared to their performances in all subjects (Figure 3.C).

Discussion
The primary objective of this study was to compare plasma Aβ40, Aβ42 and t-tau examined by Simoa immunoassay in a continuum of cognitive decline including NC, SCD, Obj-SCD, MCI and AD. Consistent with the previous reports(6, 11, 12), concentration of plasma Aβ42 and ratio of Aβ42/Aβ40 were signi cantly lower in AD patients, while plasma Aβ40 was less distinct. Plasma t-tau showed no signi cant difference among the diagnostic groups, which was in agreement with the ndings in the cohort of BioFINDER (34). It is worth noting that levels of plasma Aβ40 and Aβ42 were obviously higher in our groups of SCD and Obj-SCD, and showed a signi cant trend of initially increasing from NC to SCD and Obj-SCD, and then decreasing from SCD and Obj-SCD to MCI and AD. To a great extent, this change was similar to the ndings that levels of plasma Aβ isoforms manifested as a bidirectional character over the course of cognitive decline, which was more likely to increase in the early stages of cognitive impairment and decrease prior to clinical AD onset(16-18). As in the Australian Imaging, Biomarkers and Lifestyle (AIBL) study of aging, plasma Aβ40 and Aβ42 levels measured by double sandwich ELISA tended to be higher for MCI than both NC and AD, though were statistically less signi cant (16). However, with more detailed grouping criteria for cognitive decline, our results indicated that plasma Aβ42 and Aβ40 were signi cantly higher in SCD and Obj-SCD but not MCI. This may be partly due to the diagnosis of MCI in this study were based on an actuarial method with comprehensive neuropsychological tests, which may lead to less false positive and greater percentage of progression to dementia compared to conventional MCI diagnostic criteria (29). Compared to the marginal signi cance in AIBL, signi cant differences of plasma Aβ40 and Aβ42 between diagnostic groups in this study may also attribute to the ultrasensitive performance of SIMOA platform we used. On the other hand, given the bidirectional character of plasma Aβ40 and Aβ42 during the progression of cognitive decline in our study, no signi cant difference of plasma Aβ40 and Aβ42 between the groups of NC and MCI was not surprising, as observed in this study and previous report (35). In contrast, this study showed no biphasic evolution of Aβ42/Aβ40 ratio across the cognitive decline but signi cantly lower Aβ42/Aβ40 ratio in AD than all other groups. At this point, though the relative ratio of Aβ42 and Aβ40 may normalize the pre-analytical variability and eliminate the inter-individual differences for total Aβ concentrations(36, 37), as plasma Aβ42 and Aβ40 showed synchronous changes in SCD and Obj-SCD, the change of Aβ42/Aβ40 ratio would inevitably be weakened.
Previous studies demonstrated that plasma amyloid-β examined using the Simoa immunoassay could be useful as a potential surrogate for brain Aβ pathology, though the performances were not su cient and discrepancies remained (6,7,9,38,39). In the present study, relationships between plasma amyloid-β and 18F-orbetapir SUVR were assessed not only in all the subjects with Aβ-PET scan, but also in subjects with different Aβ-PET status. As a result, both plasma Aβ42 and Aβ42/Aβ40 ratio were signi cantly corelated with 18F-orbetapir SUVR in all the subjects, though the correlation became weak in the subgroup of Aβ-PET-positive, and no signi cant relationship was found in the subgroup of Aβ-PET-negative. We therefore speculate that signi cant correlation between plasma amyloid-β and 18F-orbetapir SUVR in all the subjects may due to their marked alteration between the subgroups of cerebral Aβ positive and negative. In view of the bidirectional character of amyloid-β with cognitive decline in our study, we further compared the levels of plasma biomarkers between different cerebral Aβ burden subgroups but not the simple status of positive or negative via visual interpretation of Aβ-PET. As a result, similar to their bidirectional character in the progression of cognitive decline, the levels of plasma Aβ42 and Aβ40 showed an increasing trend from low SUVR to moderate SUVR and decreasing from moderate SUVR to high SUVR, though only the decreasing of plasma Aβ42 in the subgroup of high 18F-orbetapir SUVR had statistical signi cance. Taken together, as transportation of Aβ across blood-brain barrier (BBB) plays a major role in Aβ clearance (40), the bidirectional trend of plasma amyloid-β along with cognitive decline and cerebral Aβ burden may be associated with the alteration of equilibrium between amyloid-β production and clearance in AD continuum. A higher level of amyloid-β in peripheral blood may be attributed to the compensatory increased transportation of amyloid-β across blood-brain-barrier. At this point, though it is typically considered that decreased Aβ clearance contribute to the predominant pathogenesis of late-onset AD (LOAD) (41), the elevated plasma Aβ40 and Aβ42 in SCD and Obj-SCD, or not decreased plasma Aβ40 and Aβ42 in the group of moderate Aβ-PET SUVR indicated that amyloid-β clearance from brain by the BBB was not initially impaired, at least in preclinical AD. However, plasma amyloid-β showed signi cantly decreased with disease progression, consistent with the change in CSF. This may, at least in part, be due to the dysfunction of clearance systems for removing Aβ from brain, such as BBB clearance, Interstitial uid bulk-ow clearance and CSF absorption clearance (40). In addition, this can also be explained by the acceleration of Aβ aggregation in brain (42), rather than an intrinsic defect of clearance system. On the other hand, whether the initially elevated level of plasma Aβ in SCD and Obj-SCD represents more soluble forms of amyloid-β in brain and be a potential window period for anti-Aβ immunotherapy still need more studies to con rm. Either way, the ratio of Aβ42/Aβ40 showed a gradually decreasing but not bidirectional trend with the increasing of SUVR. This may contribute to the fact that Aβ42 is more likely to form hard-to-clear aggregates than Aβ40 in the progression of cerebral amyloid deposition according to their structural characteristics (43).
Owning to the invasive procedure of CSF and high cost of PET image, blood-based assessments with comparable accuracy in predicting cerebral pathology of AD are urgently needed, especially in population screening. In our current study, a lower ratio of plasma Aβ42/Aβ40 demonstrated acceptable value (AUC = 0.762) in identifying high brain Aβ burden. This was similar to the study in which amyloid-PET positivity was de ned by composite SUVR (AUC = 0.79)(38), and higher than the study in which amyloid-PET positivity was de ned by visually read (AUC = 0.68) (7). In any case, with the ultrasensitive technology of Simoa, plasma amyloid-β may be a potential surrogate in predicting cerebral Aβ deposition and could undoubtedly serve as a prescreening method to reduce the need of invasive or expensive methods such as lumbar puncture or PET scanning. In our whole cohort regardless the underlying Aβ pathology, both plasma Aβ42 and Aβ42/Aβ40 ratio showed potential ability in discriminating the diagnosis of AD from non-AD. Furthermore, in the subset of AD continuum with Aβ positive determined by 18F-orbetapir PET, both plasma Aβ42 and Aβ42/Aβ40 ratio showed even higher diagnostic accuracy in identifying AD. Thus, it can be seen that, even Aβ deposition reaches the threshold of positivity in PET imaging, plasma amyloid-β may continue to change along with cerebral Aβ deposition during the natural course of AD.
Several limitations should be noted in this study. First, only parts of our participants underwent 18F-orbetapir PET scan, which may have led to the non-signi cant differences of plasma amyloid-β between the groups of low and moderate 18F-orbetapir SUVR. By expanding the sample size of our cohort, we would like to re-assess the changes of plasma amyloid-β between different cerebral Aβ burdens or different distributions of Aβ in brain regions. In addition, the NC subgroup had relatively lower prevalence of Aβ-PET positivity in our study than previous reports, though this could contribute to the fact that SCD and Obj-SCD may be regarded as normal controls in previous studies. Second, as another primary neuropathological hallmarks of AD, phosphorylated tau in blood was not measured in this study. Plasma p-tau181 examination and cerebral tau PET scan will be carried out in our future studies to further increase our understanding of cerebral and peripheral biomarkers of AD. Third, designed as a cross sectional study, longitudinal data are needed to con rm the evolution of plasma biomarkers along with cognitive decline within our ndings.

Conclusions
The present study demonstrated that plasma Aβ42 and Aβ42/Aβ40 ratio were signi cantly lower in AD than all other groups. In addition, plasma amyloid-β was signi cantly correlated with 18F-orbetapir PET SUVR and showed potential value in predicting cerebral Aβ deposition and diagnosis of AD. Furthermore, plasma Aβ40 and Aβ42 measured by Simoa immunoassay showed a signi cantly bidirectional trend of initially increasing from NC to SCD and Obj-SCD, and then declining to MCI and AD. This may contribute to the status that more soluble forms of amyloid-β in brain, or compensatory increased clearance of amyloid-β from the brain to periphery in preclinical AD. Our ndings encourage future longitudinal investigations on the changes of plasma Aβ along with cognitive decline and cerebral amyloid deposition, and to further explore the mechanisms of this bidirectional character of plasma Aβ with disease progression. The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests. Authors' contributions FFP analysed and interpreted the data, and was a major contributor in writing the manuscript. QH contributed to the PET experiments and drafted and revised the manuscript. YW performed the blood sample collection and processing. YFW had a major role in the acquisition of data. YHG had a major role in PET experiments and data analyses. FX assisted in the PET experiments and revised the manuscript. QHG designed and conceptualised the study, and revised the manuscript. All authors read and approved the nal version of the paper. Table 1. Demographics, neuropsychological tests and 18F-orbetapir PET imaging for NC, SCD, Obj-SCD, MCI and AD Index NC (n=76) SCD(n=77) Obj-SCD(n=61) MCI (n=92) AD n=101 F/x 2 (P value) Table 2. Plasma t-tau, Aβ40, Aβ42 and Aβ42/Aβ40 for NC, SCD, Obj-SCD, MCI and AD Plasma biomarkers NC (n=76) SCD(n=77) Obj-SCD(n=61) MCI (n=92) AD n=101 F (P value)