Plasma Levels of Phosphorylated Tau 181 Are Associated With Cerebral Metabolic Dysfunction in Cognitively Impaired Individuals

Alzheimer’s disease (AD) biomarkers are primarily evaluated through MRI, PET, and CSF methods in order to diagnose and monitor disease. Recently, advances in the assessment of blood-based biomarkers have shown promise for simple, inexpensive, accessible, and minimally invasive tools with diagnostic and prognostic value for AD. Most recently, plasma phosphorylated tau181 (p-tau181) has shown excellent performance. The relationship between plasma p-tau181 and cerebral metabolic dysfunction assessed by [ 18 F]FDG PET in AD is still unknown. This study was performed on a total of 892 individuals (297 cognitively unimpaired; 595 cognitively impaired) from the ADNI cohort. Plasma p-tau181 was assessed using single molecular array (Simoa) technology and metabolic dysfunction was indexed by [ 18 F]FDG PET. Cross-sectional associations between plasma and CSF p-tau181 and [ 18 F]FDG were assessed using voxelwise linear regression models, with individuals stratied by diagnostic group and by Aβ status. Associations between baseline plasma p-tau181 and longitudinal rate of brain metabolic decline were also assessed in a subset (n=389) of individuals using correlations and voxelwise regression models. levels and annual rate of change in global [18F]FDG SUVR in in the longitudinal dataset stratied by cognitive status. Baseline plasma p-tau181 and change in global [18F]FDG SUVR were not signicantly correlated in CU individuals (r=0.062, p=0.472), but were signicantly negatively correlated in CI individuals (r=-0.163, p=0.0096). (C) Voxelwise linear regression models were used to investigate associations between baseline log-transformed plasma p-tau181 and annual rate of change in [18F]FDG SUVR in CI individuals in the longitudinal dataset, adjusting for age and sex. Signicant associations were found in the medial and lateral temporal lobes (peak t-values of -5.01). Voxelwise results were corrected for multiple comparisons.

changes precede the appearance of clinical symptoms by many years [3], these pathologies may play an important role in both research and clinical trials for the screening, diagnosis, and progression monitoring of AD [4].
Currently, these biomarkers, i.e. Aβ, tau, glucose metabolism, and brain atrophy, are primarily assessed through positron emission tomography (PET), magnetic resonance imaging (MRI), and cerebrospinal uid (CSF) measures [3,5]. However the excessive cost, relative invasiveness, and time-consuming nature [6] of these methods obstruct their use in clinical practice. As such, given the need for more accessible AD biomarkers, blood-based biomarkers, such as measures of phosphorylated tau, Aβ42/40 ratio, and neuro lament light protein, constitute a viable promise and warrant thorough investigation with regards to their speci city to AD [7].
Phosphorylated tau is the principal component of neuro brillary tangles and dystrophic neurites in AD.
Tau protein phosphorylated at threonine-181 (p-tau181) has been examined in CSF [8], and it has been demonstrated that p-tau181 is highly speci c for AD-related tau aggregation [2]. Importantly, recent technological advancements have led to ultrasensitive assays of p-tau181 in blood samples (i.e. plasma and serum) using ultrasensitive immunoassays [9][10][11][12][13] and mass spectrometry methods [14]. Plasma p-tau181 levels have been shown to be strongly associated with brain tau pathology, signi cantly elevated in AD, and differentiate AD from other neurodegenerative diseases [9][10][11][12][13]. However, to date the associations between plasma p-tau181 and AD-related brain metabolic dysfunction, a well-recognized pathophysiological process underlying AD, remains unknown.
In order to address this knowledge gap, the current study was designed to measure plasma p-tau181 levels and brain glucose metabolism as assessed by [ 18 F]FDG PET in participants of the Alzheimer's Disease Neuroimaging Initiative (ADNI). The goal of the study is to examine (1) how the plasma biomarker compares to the CSF biomarker in terms of its association to [ 18 F]FDG PET cross-sectionally, and (2) how baseline levels of plasma p-tau181 relate to longitudinal trajectories of brain metabolic decline. We hypothesize that plasma p-tau181 performs similarly to CSF p-tau181 with regards to its relationship to brain metabolic dysfunction and that baseline plasma p-tau181 is able to predict reduction of brain metabolism over time.

Study participants
The current study was based on data from the ADNI database. ADNI is a multicentre study launched in 2003 as a public-private partnership, led by Principal Investigator Michael W. Weiner, MD. ADNI's primary goal is to test whether the combination of neuroimaging and biochemical biomarkers and clinical and neuropsychological assessments can be used for early detection and monitoring of AD dementia [15]. The ADNI study was approved by local Institutional Review Boards of all of the participating institutions, and informed written consent was provided by enrolled participants at each site. Full information regarding the ADNI inclusion/exclusion criteria is described elsewhere [16]. ADNI is a prospective cohort study that continues to recruit participants; this study was based on participants with available plasma p-tau181 data (data downloaded in June 2020).
The study population was classi ed into two diagnostic groups: cognitively unimpaired (CU) and cognitively impaired (CI) individuals. The CU classi cation was based on a CDR of 0; participants who had no cognitive dysfunction but reported subjective cognitive decline were analyzed together with CU, as per the National Institute of Aging-Alzheimer's Association's biological AD research framework [2]. The CI group consisted of individuals that were clinically de ned as having MCI or AD dementia. MCI and AD dementia classi cation followed the criteria described elsewhere. [15,17]  Plasma p-tau181 measurement Blood samples were collected, shipped, and stored as described by the ADNI Biomarker Core Laboratory [18]. Plasma p-tau181 was analyzed with the Single Molecule Array (Simoa) technique, using a clinically validated in-house assay described previously [9]. Plasma p-tau181 was measured on Simoa HD-X instruments (Quanterix, Billerica, MA, USA) in April 2020 at the Clinical Neurochemistry Laboratory, University of Gothenburg, Mölndal, Sweden. Plasma p-tau181 data was collected over 47 analytical runs. Assay precision was assessed by measuring two different quality control samples at the start and end of each run, resulting in within-run and between-run coe cients of variation of 3.3%-11.6% and 6.4%-12.7%, respectively. Out of 3762 ADNI samples, four were removed due to inadequate volumes. The remaining 3758 all measured above the assay's lower limit of detection (0.25 pg/ml), with only six below the lower limit of quanti cation (1.0 pg/ml). Plasma p-tau181 measurements were downloaded from the ADNI database (accessed 2020-06-20).
CSF p-tau181 measurement CSF samples were collected by lumbar puncture, shipped, and stored as described by the ADNI Biomarker Core Laboratory [18]. CSF concentrations of p-tau181 were quanti ed using fully automated Elecsys immunoassays (Roche Diagnostics) at the ADNI Biomarker Laboratory at the University of Pennsylvania. The lower and upper technical limits for CSF p-tau181 were 8 and 120 pg/mL. Procedures have been described in detail previously [19,20].

MRI acquisition and processing
Pre-processed 3T MRI T1-weighted magnetization-prepared rapid acquisition gradient echo images were downloaded from the ADNI database; full information regarding ADNI acquisition and pre-processing protocols of MRI data can be found elsewhere [21,22]. Images underwent linear and non-linear registration to the ADNI template space, and all images were visually inspected to ensure proper alignment to the ADNI template.

PET acquisition and processing
Pre-processed [ 18 F]FDG and [ 18 F]Florbetapir PET images were downloaded from the ADNI database; full information regarding ADNI acquisition and pre-processing protocols of PET data can be found elsewhere [23]. Images underwent spatial normalization to the ADNI standardized space using the automatic registration of PET images to their corresponding T1-weighted image space as well as the linear and non-linear transformations from the T1-weighted image space to the ADNI template space. PET images were spatially smoothed to achieve a nal resolution of 8 mm full width at half maximum (FWHM) and were visually inspected to ensure proper alignment to the ADNI template.
[ 18 F]FDG and [ 18 F]Florbetapir standardized uptake value ratio (SUVR) maps were generated using the pons and the full cerebellum as the reference region, respectively. For each participant, a global [ 18 F]FDG SUVR value was estimated by averaging the SUVR from the angular gyrus, posterior cingulate, and inferior temporal cortices [24]. A global [ 18 F]Florbetapir SUVR value was similarly estimated using the precuneus, prefrontal, orbitofrontal, parietal, temporal, anterior, and posterior cingulate cortices [24].

Statistical analyses
All nonimaging statistical analyses were performed using R v4.0.0. Voxelwise imaging statistical analyses were executed using the VoxelStats toolbox [26] in MATLAB version 9.4. Subjects were considered outliers if their baseline plasma p-tau181 value was three standard deviations above the population mean, and their data were excluded. Comparing demographic and clinical characteristics between diagnostic groups was done using χ 2 test with continuity correction for categorical variables, Mann-Whitney U test for non-normal continuous variables, and one-way ANOVA for normal continuous variables. Correlations between plasma p-tau181 levels and demographic and clinical characteristics used Pearson's correlation coe cient (r). All p values were two-tailed and p values <0.05 were considered signi cant.
Cross-sectional data were evaluated with correlations between CSF and plasma p-tau181 concentrations using Pearson's correlation coe cient, with subjects strati ed by diagnostic group and Aβ status.
Voxelwise linear regression models tested the cross-sectional associations between [ 18 F]FDG PET uptake and both CSF and plasma p-tau181 concentrations, adjusting for age and sex, in diagnostic groups (with and without Aβ status strati cation).
Longitudinal analyses investigated the associations between baseline plasma p-tau181 levels and longitudinal metabolic decline. Annual rates of change were calculated both for global [ 18 F]FDG SUVR and voxelwise for [ 18 F]FDG images by subtracting the baseline value from the follow-up value and normalizing by time difference between time points, in years. Correlations and voxelwise linear regression models then tested the associations between annual rate of change in metabolic decline (using [ 18 F]FDG SUVR and images, respectively) and baseline concentration of plasma p-tau181 and, adjusting for age and sex. Log-transformation of CSF and plasma p-tau181 measurements in pg/mL was used in all voxelwise analyses in order to reduce the skew of the distribution. Random eld theory with a cluster threshold of p < 0.001 was used to correct voxelwise analyses for multiple comparisons [27].

Discussion
In this study, performed in 892 participants from the ADNI cohort, we provide evidence for associations between plasma measures of p-tau181 and brain metabolism as assessed by [ 18 F]FDG PET. To our knowledge, this is the rst investigation of the cross-sectional and longitudinal associations between plasma p-tau181 and cerebral hypometabolism. Our main ndings were that cross-sectionally, plasma p-tau181 is associated with the metabolic signatures of AD. Moreover, in cognitively impaired individuals, levels of plasma p-tau181 at baseline were associated with rates of metabolic decline in the medial and lateral temporal lobes. Taken together, our study suggests that plasma p-tau181 may provide a costeffective and minimally invasive method to assess existing disease pathophysiology highly associated with metabolic dysfunction.
We found that plasma p-tau181 was higher in males and in APOE ε4 carriers, which to our knowledge is a nding that has not been yet described. We also observed plasma p-tau181 to be signi cantly associated with older age, fewer years of education, elevated global cortical composite measure of Aβ-PET, and worse performance on cognitive scores, which, with the exception of education, concur with earlier studies on plasma p-tau181 [9][10][11][12]28]. As previously described, in our cross-sectional analyses, plasma p-tau181 was correlated with CSF p-tau181. [9][10][11] Moreover, in agreement with previous research, plasma levels of p-tau181 in our sample were signi cantly elevated in cognitively impaired individuals, as well as in Aβ+ individuals independent of their cognitive status. [9][10][11][12] Our cross-sectional data indicated that p-tau181 was associated with neuronal injury in AD. Plasma p-tau181 levels and metabolic dysfunction were associated in the inferior temporal, posterior cingulate, precuneus, and orbitofrontal cortices in CI individuals. Interestingly, one can speculate that the small clusters in the anterior corpus callosum present in CU individuals may indicate a link between white matter energetic abnormalities in early states of the disease. [29] Importantly, in both groups, higher plasma p-tau181 levels were linked with Aβ status. Furthermore, baseline plasma and CSF p-tau181 had highly similar associations with [18F]FDG PET, with higher correlations in individuals on the Alzheimer continuum as well as in similar brain regions. This indicates that glucose metabolism associates with abnormal tau phosphorylation at threonine-181 measured in either blood or CSF.
The link between plasma p-tau181 and neuronal dysfunction is further corroborated in our longitudinal analyses. We found that baseline concentrations of plasma p-tau181 were signi cantly associated with annual rate of metabolic decline assessed by decrease in global [ 18 F]FDG SUVR, only within CI individuals. Similarly, voxelwise analysis conducted in the CI group revealed that plasma p-tau181 was associated with annual rate of change in [ 18 F]FDG uptake in the medial and lateral lobes. Together, these results support the concept that elevated plasma p-tau181 implies the presence of amyloidosis and neurodegeneration.
In our results, the topography of hypometabolism was consistent with brain regions that are known to be affected in AD. Speci cally, metabolic dysfunction in the posterior cingulate gyrus, precuneus, and medial and lateral temporal lobes are commonly observed in amnestic MCI and AD dementia [30][31][32]. Moreover, the posterior cingulate gyrus, precuneus, and medial and lateral temporal lobes are brain regions that are affected by signi cant tau aggregation in AD [33,34]. Metabolic dysfunction in these regions is further associated with cognitive decline as well as increased risk of progression to dementia [35]. Because tau aggregation as measured by PET [36] and by CSF [37] is tightly associated with brain metabolism, the results of our study suggest that plasma p-tau181 can serve as a less invasive and more accessible measure of AD-related cerebral metabolic dysfunction.
Neurodegeneration biomarkers in isolation are neither sensitive nor speci c to AD [2]. In both crosssectional and longitudinal analyses, we found little associations between plasma p-tau181 and [ 18 F]FDG PET in CU individuals. This nding is consistent with the observation that metabolic dysfunction is tightly related to cognitive decline [38] and thus signi cant metabolic decline is more commonly observed in individuals with cognitive impairment. Furthermore, we conducted strati ed analyses of relationships between plasma p-tau181 and [ 18 F]FDG PET in Aβ+ and Aβ-individuals. In individuals on the AD continuum (Aβ+), we observed signi cantly more pronounced cross-sectional associations between plasma p-tau181 and [ 18 F]FDG PET in regions vulnerable to hypometabolism in AD. However, in individuals who were not on the AD continuum (Aβ-), we did not observe associations between plasma measures of p-tau181 and brain metabolic dysfunction. This was observed for both CU and CI Aβindividuals. These results are consistent with accepted disease models in which (detectable) Aβ aggregation occurs upstream of (detectable) tau aggregation [3,39]. Taken together, these results suggest that the speci city of plasma p-tau181 for AD-type pathology [9,10,40] provides important information about the etiology of neurodegeneration and corresponding cognitive decline.
Associations between plasma measures of p-tau181 and [ 18 F]FDG PET have important clinical implications. [ 18 F]FDG PET is a commonly employed test in the differential diagnosis of individuals with cognitive impairment [41]. Brain metabolism, as indexed with [ 18 F]FDG PET, correlates with cognitive function [38] and reduced brain metabolism constitutes an important risk for clinical progression to dementia [42]. [ 18 F]FDG PET abnormalities are also observed before MRI atrophy [43], suggesting that [ 18 F]FDG PET is a sensitive marker of neurodegeneration. Therefore, plasma measures of p-tau181 have potential as a simple tool for the diagnosis and monitoring of AD, as well as for the screening of individuals for disease-modifying clinical trials.
The validity of our results is potentially in uenced by methodological limitations. First, we only included ADNI participants with a 24-month follow-up [ 18 F]FDG PET relative to plasma p-tau181 assessment as imaging data was not consistent at later time points (i.e. 36, 48 months). As a consequence, we may have observed stronger and more compelling associations between plasma p-tau181 and [ 18 F]FDG PET, as accepted biomarker models of AD demonstrate that tau accumulation occurs upstream of metabolic dysfunction and neurodegeneration [3]. Moreover, the ADNI cohort, from which all subjects in this study were selected, does not encompass individuals with neurodegenerative or tau-related diseases other than AD. Therefore, it is not known how plasma p-tau181 may perform in predicting current and future metabolic dysfunction in other neurodegenerative diseases. Further studies should conduct similar analyses in other more varied observational cohorts, as well as track brain metabolism through [ 18 F]FDG PET over a longer time frame.

Conclusion
Our study provides evidence for associations between plasma measures of p-tau181 and brain metabolic dysfunction as measured by [ 18 F]FDG PET. Subgroup analyses revealed more widespread associations in CI individuals as compared to CU individuals. Moreover, extensive associations were observed in Aβ+ individuals, whereas no associations between plasma p-tau181 and [ 18 F]FDG PET were observed in ABindividuals. Finally, baseline levels of plasma p-tau181 were associated with rates of metabolic decline in CI individuals. Together, our results suggest that plasma p-tau181 provides interrelated information to

Declarations
Ethics approval and consent to participate The ADNI study was approved by local Institutional Review Boards of all of the participating institutions.
Informed written consent was provided by enrolled participants at each site.

Consent for publication
Not applicable.

Availability of data and materials
The dataset supporting the conclusions of this article is available on the ADNI site, at http:// http://adni.loni.usc.edu/data-samples/access-data/. and MS contributed to data analysis; FZL, ALB, JT, TAP, CT, and PR-N wrote the original manuscript draft. All authors reviewed, edited and approved the nal manuscript for submission. Cross-sectional associations between plasma and CSF p-tau181 and between p-tau181 and [18F]FDG SUVR Description: (A) CSF p-tau181 levels in pg/mL were compared in individuals in the cross-sectional dataset, strati ed by both their cognitive status (cognitively unimpaired (CU) or impaired (CI)) and Aβ status (+ or -), using Mann-Whitney U test. Signi cant differences in CSF p-tau181 were found between the CU-and CU+ groups (p<0.001), the CU+ and CI+ groups (p<0.001), and the CI-and CI+ groups  and Aβ status (+ or -), using Mann-Whitney U test. Signi cant differences in baseline plasma p-tau181 levels were found between the CU-and CU+ groups (p<0.001), and the CI-and CI+ groups (p<0.001). (B) Pearson's correlation coe cient (r) was computed for associations between baseline plasma p-tau181