Plasma P‐tau217, GFAP, and NfL as biomarkers for Alzheimer's disease: role in disease stratification, pathological progression, and cognitive decline

Participants for our study were sourced from the Chinese Preclinical Alzheimer's Disease Study (C‐PAS) research cohort.19 A total of 1275 participants underwent both 3.0T brain magnetic resonance imaging (MRI) and ¹⁸F‐florbetapir (AV45) PET scans, along with concurrent blood sampling for the measurement of plasma p‐tau217, GFAP, and NfL, were included in the present analysis. A subset of 584 participants underwent MK6240‐PET scans. Written informed consent was obtained from all participants or their caregivers. The ethics committee of Shanghai Sixth People's Hospital, affiliated with Shanghai Jiao Tong University School of Medicine, reviewed and approved this study.

Global cognitive performance and functional status were assessed via the Montreal Cognitive Assessment‐Basic (MoCA‐B) and the Activities of Daily Living (ADL), respectively.20, 21 A comprehensive battery of standardized neuropsychological tests was administered to derive composite scores across five cognitive domains, as previously documented.22 In brief, memory is calculated as the average of the z‐scores for Auditory Verbal Learning Test delayed recall and Brief Visuospatial Memory Test‐Revised delayed recall. Language is calculated as the average of the z‐scores for Boston Naming Test and Animal Verbal Fluency Test. Attention is calculated as the average of the z‐scores for Symbol Digit Modalities Test and Digit Span Test. Visuospatial is calculated as the average of the z‐scores for silhouette test and judgment of line orientation. Execution is calculated as the average of the z‐scores for Shape Trail Test Parts A and B. The diagnosis of mild cognitive impairment (MCI) was established based on the actuarial neuropsychological criteria proposed by Jak and Bondi.23 The diagnosis of probable AD dementia was determined according to the criteria set forth by the National Institute on Aging‐Alzheimer's Association (NIA‐AA) in 2011.24 Participants who performed within the normal range and did not meet the criteria for MCI or dementia were classified as cognitively normal (CN).

Brain MRI images were obtained utilizing a 3.0T MRI scanner (SIEMENS MAGNETOM Prisma 3.0T, Siemens, Erlangen, Germany) within the Department of Radiology at Shanghai Sixth People's Hospital. The volumes of the left and right hippocampus (HV‐L, HV‐R) were determined by dividing the total hippocampal volume by the estimated total intracranial volume and subsequently multiplying by 1000. For details regarding the acquisition and preprocessing of the MRI images, see. Supporting Information

Both AV45‐PET and MK6240‐PET scans were conducted utilizing a PET/computed tomography (CT) system (Biograph mCT Flow, Siemens, Erlangen, Germany) at the Department of Nuclear Medicine & PET Center, Huashan Hospital, Fudan University. Subjects were classified as Aβ positive (A+) and negative (A−) through visual interpretation of AV45‐PET scans, following established visual rating guidelines as previously documented.25 The visual rating of MK6240‐PET images adhered to the guidelines for tau PET interpretation.26 Based on the Braak staging system,27 subjects were categorized into tau negative (T−) and tau positive (T+) stages, which include Braak I–II (entorhinal cortex and hippocampus), Braak III–IV (limbic and temporal neocortex), and Braak V–VI (neocortical association areas). Consequently, Aβ/tau‐PET images were classified into A−T−, A−T+, A+T−, A+T+_{Braak I–II}, A+T+_{Braak III–IV}, and A+T+_{Braak V–VI}. Brain Aβ burden was quantified using standard uptake value ratios (SUVRs) for the global cortical region, determined through a weighted average of the region of interest (ROI) relative to the cerebellar crus. In the context of MK6240 PET imaging, SUVR was computed using the inferior cerebellar gray matter as the reference region. Both global MK6240‐SUVR and three composite regions corresponding to Braak I–II, Braak III–IV, and Braak V–VI stages were calculated independently. Comprehensive details regarding image acquisition, data preprocessing, visual interpretation, and SUVR calculation for both AV45‐PET and MK6240‐PET are provided in the Methods section of the Supporting Information.

Plasma samples were obtained from ethylenediaminetetraacetic acid (EDTA)‐treated blood and subsequently subjected to centrifugation, aliquoting, and storage at −80°C. The concentrations of plasma biomarkers, including p‐tau217, GFAP, and NfL, were quantified using the Light‐initiated Chemiluminescent Assay (LiCA) on the Chemclin LiCA 800 automated immunoassay analyzer (Chemclin Diagnostics, Beijing, China). Further description of assays is provided in the Methods in the. Supporting Information

Categorical variables were represented as frequencies (percentages) and analyzed utilizing the chi‐squared test. Continuous variables were presented as means ± standard deviation (SD), with multigroup comparisons conducted using analysis of variance (ANOVA) followed by post hoc Bonferroni correction. Segmental linear regression analysis was further conducted to compare the variation characteristics of different plasma biomarkers in relation to changes in brain Aβ and tau burden. Prior to analysis, plasma biomarkers were standardized using z‐scores. The area under the receiver operating characteristic (ROC) curve (AUROC) was employed to assess the efficacy of plasma biomarkers in differentiating various AT statuses. Cutpoints were determined according to the Youden index of ROC curves. A general linear model (GLM) adjusted for age and gender was utilized to evaluate the associations between Aβ and tau burden in different statuses of astrocyte activation. Linear regression analyses adjusted for age, sex, and years of education were conducted to examine the predictive effects of plasma p‐tau217 and NfL on hippocampal volume and cognitive performances. Plasma p‐tau217 and NfL were z‐scored before entering the analysis. Statistical significance was determined using a two‐sided p value threshold of <0.05. Data analyses were executed using IBM SPSS Statistics version 26 and GraphPad Prism version 9.0 (GraphPad Software, Inc.).

In this study, comprising 1275 participants, 462 individuals were classified as CN, 539 as having MCI, and 274 diagnosed with probable AD dementia. As indicated in Table 1, there was a significant rise in plasma levels of GFAP, p‐tau217, NfL, apolipoprotein E (APOE) ε4 prevalence, AV45‐PET positivity, and global AV45‐SUVR from the CN group to the MCI group and then to the dementia group, while hippocampal volumes significantly decreased. Among participants who underwent MK6240‐PET scanning, the prevalence of individuals at Braak V–VI stage showed a significant increase from the CN group to the MCI group and then to the dementia group. However, no significant difference was observed in the prevalence of individuals at Braak I–II and III–IV stages across the different cognitive stages. MK6240‐SUVR in Braak I–II regions showed a significant increase across cognitive stages, whereas significant increases in the global cortex and Braak III–IV and V–VI regions were observed exclusively in individuals with dementia. The global and regional MK6240‐SUVRs for subgroups categorized by the final Braak stages are illustrated in Figure S1.

Segmental linear regression analysis was employed to investigate the progression of plasma biomarkers in relation to increasing amyloid/tau burdens. Breakpoints indicating significant changes were identified using the least‐squares method, which minimized the residual sum of squares. As shown in Figure 1A, plasma p‐tau217 did not exhibit significant correlation with AV45‐SUVR levels below the breakpoint but increased gradually once the breakpoint was exceeded (Slope 2 = 0.768). Plasma GFAP initially demonstrated a transient decrease prior to the breakpoint, followed by a consistent increase (Slope 1 = −0.201, Slope 2 = 0.519). In contrast, plasma NfL levels showed a significant yet weak increase before the breakpoint (Slope 1 = 0.142) and did not exhibit a progressive increase with further elevations in AV45‐SUVR. Figure 1B–E illustrates the trends in plasma biomarkers as MK6240‐SUVR levels rise. Plasma p‐tau217 levels exhibited a progressive rise with increasing global MK6240‐SUVR (Slope 1 = 0.610), until a high SUVR threshold was reached, beyond which the rising trend ceased to be significant. Plasma GFAP levels showed a progressive rise with increasing global MK6240‐SUVR (Slope 1 = 0.592), followed by a plateau beyond a breakpoint, which was identified at a lower SUVR value compared to that of p‐tau217. In the Braak regions, plasma GFAP displayed similar trends, while plasma p‐tau217 demonstrated distinct patterns in the Braak I–II and Braak III–IV regions. Specifically, plasma p‐tau217 demonstrated a progressive increase with MK6240‐SUVR before reaching the breakpoints (Slope 1 = 0.531 and 0.577, respectively) and continued to show a relatively modest yet statistically significant upward trend after the breakpoints (Slope 2 = 0.190 and 0.175, respectively). Regarding plasma NfL, aside from a slight increase observed before the breakpoints, no elevation in NfL levels was associated with increasing MK6240‐SUVR. Figure S2 shows a scatterplot of plasma biomarkers with AV45‐SUVR and MK6240‐SUVR.

The differences in plasma biomarkers across the different A/T statuses defined by Aβ/tau‐PET are shown in Figure 2A–C. Plasma p‐tau217 and GFAP levels were elevated in all A+ groups relative to the A−T− and A−T+ groups, with the exception of GFAP levels, which showed no significant difference between the A+T− and A−T+ groups. Moreover, plasma p‐tau217 and GFAP concentrations were significantly higher in the A+T+_{Braak III–IV} and A+T+_{Braak V–VI} groups compared to the A+T− and A+T+_{Braak I–II} groups. No significant differences were noted between the A−T− and A−T+ groups or between the A+T− and A+T+_{Braak I–II} groups. Additionally, plasma p‐tau217 levels were elevated in the A+T+_{Braak V–VI} group compared to the A+T+_{Braak III–IV} group. Compared to the A−T− group, plasma NfL showed a slight increase in the A+T− group, and the elevation in NfL levels in the A+T+_{Braak III–IV} and A+T+_{Braak V–VI} groups was not as pronounced as the increases observed in p‐tau217 and GFAP.

Given that the primary beneficiaries of current anti‐Aβ AD‐modifying therapies are likely to be the A+T− and A+T+_{Braak I–II} groups, we assessed the efficacy of plasma biomarkers in distinguishing these populations. ROC analyses indicated that plasma p‐tau217 exhibited the highest discriminative capacity among the various AT groups (Figure 2D–F and Table S1). Within the overall population, plasma p‐tau217 demonstrated a robust ability to exclude both the A−T− and A−T+ groups (area under the curve, AUC = 0.931), while also effectively identifying the A+T+_{Braak III–IV} and A+T+_{Braak V–VI} groups (AUC = 0.959). However, because the plasma p‐tau217 levels in individuals with A+T− and A+T+_{Braak I–II} largely overlapped with those in either the A−T− and A−T+ groups or the A+T+_{Braak III–IV} and A+T+_{Braak V–VI} groups (Figure 2G), it could not effectively distinguish individuals with A+T− and A+T+_{Braak I–II} in all the participants. Nevertheless, plasma P‐tau217 retained a high ability to differentiate the A+T− and A+T+_{Braak I–II} groups within the Aβ‐PET‐positive population (AUC = 0.876).

Plasma GFAP elevation serves as an indicator of astrocyte activation. Within the comprehensive C‐PAS cohort, astrocyte activation is defined by the 95th‐percentile level of plasma GFAP in CN young controls who are Aβ− (annotated as GFAP+). After adjusting for age and gender, there was a significant increase in global MK6240‐SUVR corresponding to the rise in AV45‐SUVR within the GFAP+ population (Figure 3A; β = 1.023, p < 0.001). Conversely, in the GFAP− population, the increase in AV45‐SUVR did not significantly impact global MK6240‐SUVR (β = 0.087, p = 0.211). A similar pattern was found in the distinct Braak regions for both GFAP+ and GFAP− populations (Figure 3B–D). Additionally, the similar impact of AV45‐SUVR on plasma p‐tau217 levels was also observed in the GFAP− and GFAP+ populations (Figure 3E). Figure 3F illustrates the variations in MK6240‐SUVR across the Aβ and GFAP subgroups in the global cortex and distinct Braak regions, revealing a significant increase in MK6240‐SUVR exclusively in the cohort positive for both Aβ and GFAP. Furthermore, Figure 3G indicates that tau pathology beyond Braak stage III was infrequent in populations negative for both Aβ and GFAP or when only one marker was positive. In contrast, this was more prevalent in the population with concurrent Aβ+ and GFAP+.

The associations of plasma p‐tau217 and NfL on hippocampal volume, global cognitive decline, and performances across diverse cognitive domains were assessed via linear regression analyses, with age, gender, and years of education included as covariates (Figure 4 and Table S2). Z‐scored plasma p‐tau217 and NfL were incorporated into the models concurrently, and no collinearity between variables was observed (VIF range, 1.066 to 1.361). Among all participants, elevated plasma p‐tau217 levels demonstrated a significant correlation with reduced hippocampal volumes, whereas the association between increased plasma NfL levels and hippocampal volume reduction was only marginally significant. When stratified by Aβ status, the significant relationship between elevated plasma p‐tau217 levels and hippocampal volume reduction was evident exclusively in the Aβ+ group, with no such association observed in the Aβ− group. Conversely, increased plasma NfL levels were associated with hippocampal volume reduction solely in the Aβ− group, with no significant association found in the Aβ+ group. In relation to cognitive performance, increased plasma p‐tau217 levels were linked to declines in overall cognitive function and each of the cognitive domains. However, within subgroups categorized by Aβ status, the association between elevated plasma p‐tau217 and cognitive decline was predominantly observed in the Aβ+ group, where it demonstrated significant correlations with declines in overall cognitive function, as well as in memory, language, and attention. Conversely, elevated plasma NfL was primarily associated with declines in overall cognitive function and language, visuospatial, and executive abilities, with these associations evident in both Aβ− and Aβ+ groups. Importantly, within the Aβ+ group, elevated plasma p‐tau217, as opposed to plasma NfL, was significantly associated with memory decline, whereas plasma NfL, rather than p‐tau217, was significantly associated with declines in executive function.