Correlations between plasma markers and brain Aβ deposition across the AD continuum: Evidence from SILCODE

Abstract

BACKGROUND

Previous studies have found that Alzheimer's disease (AD)‐related plasma markers are associated with amyloid beta (Aβ) deposition, but the change of this association in different Aβ pathological stages remains unclear.

METHODS

Data were obtained from the SILCODE. According to the standardized uptake value ratio (SUVR) and Aβ stage classification, correlation analysis was performed among plasma biomarkers, and voxel/SUVR values in the regions of interest (ROI) and clinical scale information, respectively. Mediation analysis was used to study the possible pathways.

RESULTS

The proportion of cognitively normal (CN) and subjective cognitive decline (SCD) was the highest in stages A0 to 1, while in stages A2 to 4, the proportion of mild cognitive impairment (MCI) and AD increased. Plasma phosphorylated tau (p‐tau)181 and glial fibrillary acidic protein (GFAP) levels were significantly lower in stage A0 compared to the later phases. Two pathways demonstrated fully mediated effects: positron emission tomography (PET) SUVR–plasma p‐tau181–Mini‐Mental State Examination (MMSE) and PET SUVR–plasma GFAP–MMSE.

DISCUSSION

This study demonstrated the role of plasma biomarkers in the early stage of AD, especially in SCD, from both the clinical diagnosis and Aβ stage dimensions.

Highlights

• Plasma ptau181 and GFAP level serve as indicators of early Alzheimer's disease and the pathologic Aβ staging classification.

• A possible ceiling effect of GFAP was observed in the mid‐to‐late stages of the AD course.

• This study confirms the role of AD plasma markers in promoting Aβ deposition at an early stage, particularly in females with subjective cognitive decline(SCD).

• The overlapping brain regions of plasma p‐tau181, GFAP, and neurofilament light for Aβ deposition in the brain in early AD were distributed across various regions, including the posterior cingulate gyrus, rectus gyrus, and inferior temporal gyrus.

1. INTRODUCTION

The abnormal accumulation of aggregated amyloid beta (Aβ) peptides and abnormally phosphorylated tau (p‐tau) as extracellular plaques and neuronal neurofibrillary tangles (NFTs), respectively, are fundamental hallmarks of Alzheimer's disease (AD) neuropathology and remain the definitive confirmation of the disease; therefore, the diagnosis cannot be established until autopsy. ¹ Nonetheless, visualization of Aβ ² and tau ³ aggregates by positron emission tomography (PET) greatly enhanced the characterization of AD in vivo, but the change of this association in different Aβ pathological stages remains unclear. Decreases in Aβ42 or Aβ42/Aβ40 and increases in total tau (t‐tau) and p‐tau represent the core changes in cerebrospinal fluid (CSF) biomarkers. ⁴ However, although both CSF and PET biomarkers have obvious practical and diagnostic value, they are unlikely to be widely used in primary care to assess AD or other neurodegenerative diseases due to invasiveness and expensive economic drawbacks, respectively. Furthermore, in light of recent clinical trial failures, there is a growing general need for confirmatory evidence of the underlying pathology as a basic selection criterion for AD treatment trials. Thus, the field would greatly benefit from an easier‐to‐implement and cost‐effective approach to initial patient or participant assessment that could meet clinical and drug development needs.

The development of plasma biomarkers has been the most important recent advance in the diagnosis of AD. ⁵ , ⁶ However, the dynamic changes in various blood biomarkers across the AD continuum remain to be explored. On the one hand, recruiting individuals from the preclinical to symptomatic stages of AD is particularly challenging. Several previous studies have been limited by the unavailability of CSF or PET data and have focused on the clinical spectrum of AD ⁷ , ⁸ or simply divided the AD continuum into A− and A+ stages. ⁹ On the other hand, many studies have included a single or very small number of markers and have not been graded according to the pathological degree of Aβ stage, which lacks an understanding of the entire AD continuum. In addition, these previous studies were mainly from Western countries, and the timing of the onset of changes in blood biomarkers, the characteristics of the changes, and whether these biomarkers reflect AD pathology in the Chinese population are unknown. Due to recent reports of potential ethnic differences in AD biomarkers, ¹⁰ , ¹¹ our data may provide new insights for studies in Chinese cohorts.

In this study, we aim to investigate the role of plasma in promoting Aβ deposition in the subjective cognitive decline (SCD) stage of AD according to the clinical spectrum and pathological Aβ stage classification, as well as which plasma biomarkers are still in effect in the middle and late stages of AD and whether there is a ceiling effect. Additionally, we also paid attention to the existence of plasma‐mediated effects under the two dimensions of clinical and Aβ pathologic staging, and focused on some influencing factors, especially sex, to study its possible influencing factors. Finally, two dimensions of AD were used to demonstrate the early role of plasma and its early impact on high‐risk populations.

2. METHODS

2.1. Participants

This study used data collected from Xuanwu Hospital in Beijing between July 2016 and December 2022, involving a sample size of 148 participants. These individuals underwent neuroimaging and neurological scales. Specifically, the dataset comprised 66 cognitively normal (CN) individuals, 59 with SCD, 12 with mild cognitive impairment (MCI), and 11 diagnosed with AD. All subjects gave written informed consent prior to enrollment. Subcategories of participants are described as follows. Diagnosing CN was based on excluding individuals with MCI ¹² and AD dementia. ¹³ The SCD group was defined as a persistent cognitive decline in self‐perception, compared to a previous normal state not associated with an acute event, that did not meet the criteria for MCI or AD dementia. ¹⁴

The imaging examinations included Aβ PET and T1 magnetic resonance imaging (MRI). Additionally, all participants underwent examinations covering 11 neurocognitive assessments, including Subjective Cognitive Decline Questionnaire 9 (SCD‐9), Hamilton Depression Rating Scale (HAMD), Hamilton Anxiety Rating Scale (HAMA), the Mini‐Mental State Examination (MMSE), long‐term delayed recall (N5), and recognition (N7), Shape Trail Test A (STT‐A), Shape Trail Test B (STT‐B), Verbal Fluency Task (VFT), Boston Naming Test (BNT), and basic version of Montreal Cognitive Assessment (MoCA‐B).

Exclusion criteria were current major psychiatric diagnoses (e.g., major depression or anxiety), other neurological disorders or conditions that can lead to cognitive decline (e.g., Parkinson's disease, encephalitis, or thyroid dysfunction) other than AD spectrum disorders, the inability to complete the study protocol, or the presence of contraindications to MRI.

RESEARCH IN CONTEXT

1. Systematic review: Literature reviews in PubMed and Google Scholar suggest that Alzheimer's disease (AD)‐related plasma markers are associated with amyloid beta (Aβ) deposition, but the change of this association in different Aβ pathological stages remains unclear. In this study, we aim to investigate the role of plasma in promoting Aβ deposition in the subjective cognitive decline (SCD) stage of AD according to the clinical spectrum and pathological Aβ stage classification, as well as which plasma biomarkers are still in effect in the middle and late stages of AD and whether there is a ceiling effect.

2. Interpretation: Our findings suggest a suggestive role of plasma phosphorylated tau (p‐tau)181 and glial fibrillary acidic protein (GFAP) in early AD, as evidenced by both the clinical spectrum of AD and the pathologic Aβ staging classification. We also observed a possible ceiling effect of GFAP in the mid‐to‐late stages of the AD disease course. Furthermore, our results confirm the role of AD plasma markers in promoting Aβ deposition at an early stage, particularly in females with subjective cognitive decline. The overlapping brain regions of plasma p‐tau181, GFAP, and neurofilament light for Aβ deposition in the brain in early AD were distributed across various regions, including the precuneus, rectus gyrus, and inferior temporal gyrus. It is further emphasized that the staging of amyloid pathology may become more important as clinical trials focus on interventions targeting more specific stages of AD.

3. Future directions: Our study delved into AD clinical, pathological, and biomarker profiles to further elucidate the underlying mechanisms underlying the early predictive performance of plasma in AD.

2.2. Image data acquisition and processing

Aβ PET imaging at Xuanwu Hospital was performed using ¹⁸F‐florbetapir (¹⁸F‐AV‐45). In addition, both PET and MRI image scans were performed on a 3.0T TOF PET/MR integrated simultaneous scanner located at Xuanwu Hospital of Capital Medical University, Beijing, China. During PET scanning, participants received 7 to 10 mCi ¹⁸F‐florbetapir radiotracer intravenously and then rested for approximately 40 minutes before undergoing a 20‐minute static PET scan. A time‐of‐flight ordered subset expectation maximization (TOF‐OSEM) algorithm was used to obtain PET data with the following parameters: iterations = 8, field of view (FOV) = 350 × 350 mm², 32 subset matrices = 192 × 192, and half‐width height = 3. In the MRI scan, the parameters were as follows: spoiled gradient‐recalled sequence, FOV = 256 × 256 mm², slice thickness = 1 mm, matrix size = 256 × 256, number of slices = 192, gap = 0, repetition time (TR) = 6.9 ms, inversion time (TI) = 450 ms, echo time (TE) = 2.98 ms, flip angle = 12°, voxel size = 1 × 1 × 1 mm³.

The PET and T1 MRI images were processed using the SPM12 toolbox (http://www.fil.ion.ucl.ac.uk/spm/software/spm12), as follows: (1) convert DICOM (Digital Imaging and Communications in Medicine) files to NIfTI (Neuroimaging Informatics for Technology Initiation) files using the MRIcron toolbox (dcm2niigui.exe); (2) correct the origin node to anterior commissure–posterior commissure (AC‐PC); (3) co‐registration between PET image and the corresponding T1 MRI image; (4) T1 MRI segmentation and spatial normalization, with PET individual space converted into Montreal Neurological Institute (MNI) space using the transformation parameters of T1 MRI image standardization to MNI space; (5) smooth with 8 mm full‐ width at half maximum (FWHM).

For the subsequent Aβ staging, we selected the whole cerebellum as the reference brain region to calculate the standardized uptake value ratio (SUVR) of PET images and selected 32 brain regions and five meta‐regions as regions of interest (ROI) from the Automated Anatomical Labeling atlas 3 (AAL3) according to the previous studies on Aβ stage. ¹⁵ , ¹⁶

2.3. Measurement of plasma biomarkers

Blood samples (2 mL venous blood) were taken between 7:00 and 8:00 in the morning after an overnight fast using ethylenediaminetetraacetic acid tubes, plasma samples were thawed at room temperature and centrifuged at 4°C and 10,000 × g for 5 minutes, and the resultant supernatant was transferred to 96‐well plates and diluted 4‐fold with sample diluent, followed by two‐step digital immunoassay. In this study, the interval between blood sample collection and Aβ‐PET examination for all participants did not exceed 30 days.

Quantification of biomarkers plasma levels of Aβ42, Aβ40, NfL, GFAP, t‐tau, and p‐tau181 were measured via immunoassay according to manufacturer protocols using an automatic Simoa HD‐X analyzer (Quanterix). Specifically, plasma samples were thawed at room temperature and centrifuged at 4°C and 10,000 × g for 5 minutes, and the resultant supernatant was transferred to 96‐well plates and diluted 4‐fold with sample diluent, followed by two‐step digital immunoassay. The Simoa Neurology 4‐Plex E Advantage Kit (Quanterix; Cat# 103670) was used to assay the concentrations of Aβ42, Aβ40, NfL, and GFAP. The Simoa Tau Advantage Kit (Quanterix; Cat# 101552) was used to assay the concentrations of t‐tau. The Simoa p‐Tau‐181 Advantage V2 Kit (Quanterix; Cat# 103714) was used to assay the concentrations of p‐tau181. Assays were completed using kits from the same lot, and all samples were run in duplicate. Seven or eight calibrators (2×) and two quality controls (2×) were run on each plate for each analyte. The lower limits of detection of the Aβ42, Aβ40, NfL, GFAP, t‐tau, and p‐tau181 assays were 0.136, 0.384, 0.090, 0.441, 0.019, and 0.028 pg/mL, respectively, and the lower levels of quantification were 0.378, 1.02, 0.400, 2.89, 0.061, and 0.338 pg/mL, respectively. Intra‐assay coefficients of variation for controls ranged from 2% to 13% for Aβ42, 1% to 5% for Aβ40, 2% to 12% for NfL, 1% to 8% for GFAP, 2% to 11% for t‐tau, and 1% to 10% for p‐tau181.

2.4. Aβ staging

To better examine the stage characteristics of Aβ PET images, subjects were divided into five stages (stage A0, A1, A2, A3, A4) based on Aβ PET SUVR values. Briefly, according to a previous study, ¹⁵ the brain region that tends to accumulate Aβ tracers was divided into five meta‐ROIs, with a threshold of 1.57 used to classify the negative and positive, ¹⁵ , ¹⁶ and then the number of positive ROIs was used to classify the Aβ stage. Finally, to better accommodate the data distribution in our study, we merged stages A2, A3, and A4 into one subgroup (stages A2–4) for correlation analysis.

2.5. Statistics

Demographic information and plasma biomarkers between the different stage subgroups were analyzed using one‐way analysis of variance (ANOVA) with post hoc tests. A two‐sample rank‐sum test was used to check for significant differences between the groups when analyzing plasma biomarkers in the male–female subgroups of SCD alone.

A sigmoidal 4PL model was used to fit the changes in each of the four plasma markers with meta‐ROI SUVR (32 ROIs combined). In addition, to analyze the classification ability of plasma markers for different Aβ stages, the study used the meta‐ROI SUVR values to fuse the features with each of the four plasma markers (two features were weighted after z score normalization), and receiver operating characteristic (ROC) analyses were performed using the individual plasma markers and the fused features.

Pearson correlations were used in all correlation analyses in this study, including voxel‐level SUVR maps with plasma biomarkers, ROI‐level SUVRs with plasma biomarkers, and plasma biomarkers with scores on the clinical neurological scale. Correlations between ROI‐level SUVRs and plasma biomarkers, plasma biomarkers, and clinical neurological scale scores were tested for multiplicity using false discovery rate (FDR) correction. To further explore the relationships among Aβ PET images, plasma biomarkers, and clinical neurological scales, mediation analysis was conducted using the simple mediation model in SPSS PROCESS (X‐M‐Y model), with the bootstrap method for testing. The age differences were adjusted in ANOVA, correlation analysis, and mediation analysis. The above analyses were done using MATLAB 2017a, SPSS v25 software; two‐tailed P < 0.05 indicates significance.

3. RESULTS

3.1. Demographic and Aβ stage results

The distribution of all subjects in each A stage along with their demographic and clinical characteristics are shown in Table 1. Stage A0 to 1 exhibited the highest percentage of CN and SCD, while stages A2 to 4 showed an elevated percentage of MCI and AD. ANOVA results showed between‐group differences in all three plasma biomarkers, except Aβ42/Aβ40, with GFAP significantly different across all stages. Between‐group differences were also found in the neurological scales scores (MMSE, N5, N7, STT‐A, VFT, MoCA‐B). Table S1 in supporting information provides further detailed information on clinical subgroups.

TABLE 1.

Characteristics by Aβ stage.

	Stage A0	Stage A1	Stage A2–4	Group test (p‐value)	Pairwise difference (p < 0.05)
CN/SCD/MCI/AD (%)	48/49/3/0	56/26/11/7	11/11/31/47	/	/
Sex (M/F)	33/69	14/13	5/14	0.1315	/
Age (years)	65.346 ± 5.542	67.593 ± 8.191	71.059 ± 9.464	0.0032	0 vs. 2‐4
Education (years)	12.692 ± 3.428	13.667 ± 4.323	11.588 ± 3.465	0.1748	/
Plasma Aβ42/Aβ40	0.063 ± 0.018	0.057 ± 0.012	0.053 ± 0.012	0.0980	/
Plasma p‐tau181	1.91 ± 0.938	2.376 ± 1.084	4.033 ± 1.335	<0.0001	0 vs. 2‐4; 1 vs. 2‐4
Plasma NfL	15.865 ± 8.054	22.351 ± 15.315	29.813 ± 15.249	<0.0001	0 vs. 1,2‐4
Plasma GFAP	111.729 ± 49.773	155.304 ± 73.973	240.169 ± 110.037	<0.0001	0 vs. 1,2‐4;1 vs. 2‐4
SCD‐9	4.552 ± 2.243	4.318 ± 2.547	5.8 ± 1.859	0.2091	/
HAMD	3.899 ± 4.127	3.522 ± 5.062	3.182 ± 2.359	0.8230	/
HAMA	4.576 ± 4.333	3.913 ± 3.19	3.727 ± 2.453	0.6629	/
MMSE	28.68 ± 1.941	27.259 ± 4.302	22.765 ± 5.869	<0.0001	0 vs. 2‐4; 1 vs. 2‐4
N5	7.242 ± 2.295	7.13 ± 3.123	3.091 ± 3.506	<0.0001	0 vs. 2‐4; 1 vs. 2‐4
N7	21.96 ± 2.19	21.87 ± 3.252	18.091 ± 3.081	<0.0001	0 vs. 2‐4; 1 vs. 2‐4
STT‐A	60.936 ± 21.948	62.714 ± 34.034	87.1 ± 58.202	0.0234	0 vs. 2‐4
STT‐B	137.747 ± 42.283	140.4 ± 75.794	177.333 ± 52.851	0.0781	/
VFT	18.792 ± 4.764	19.5 ± 5.73	14.091 ± 5.612	0.0092	0 vs. 2‐4; 1 vs. 2‐4
BNT	24.896 ± 3.438	25.909 ± 4.363	23.091 ± 3.477	0.1110	/
MoCA‐B	25.505 ± 3.403	25.261 ± 4.702	19 ± 4.359	<0.0001	0 vs. 2‐4; 1 vs. 2‐4

Note: Continuous variables are expressed as mean ± SD, and sex as number (male/female). Unadjusted post hoc differences are reported for group tests with p < 0.05.

Abbreviations: Aβ, amyloid beta; AD, Alzheimer's disease; BNT, Boston Naming Test; CN, cognitively normal; GFAP, glial fibrillary acidic protein; HAMA, Hamilton Anxiety Rating Scale; HAMD, Hamilton Depression Rating Scale; MCI, mild cognitive impairment; MMSE, Mini‐Mental State Examination; MoCA‐B, basic version of Montreal Cognitive Assessment; N5, long‐term delayed recall; N7, long‐term delayed recognition; NfL, neurofilament light chains; p‐tau, phosphorylated tau; SCD, subjective cognitive decline; SCD‐9, Subjective Cognitive Decline Questionnaire 9; SD, standard deviation; STT‐A, Shape Trail Test A; STT‐B, Shape Trail Test B; VFT, Verbal Fluency Task.

3.2. Biomarkers in plasma and PET images

Figure 1A,B demonstrate the changes in whole‐brain SUVR values and plasma Aβ42/Aβ40 at the Aβ stage. Plasma p‐tau181 levels were significantly lower in subjects at the A0 and A1 stages than at the A2 and A3 stages (Figure 1C). Similarly, GFAP levels were significantly lower in stage A0 than in A1, and in addition, plasma GFAP was significantly higher in stage A3 than in stage A4 (Figure 1E). In each Aβ phase, plasma p‐tau181 and plasma GFAP formed a trend of increasing and then decreasing. This phenomenon was not found in plasma NfL (Figure 1D).

FIGURE 1.

Plasma biomarkers and Aβ PET SUVR profiles across Aβ stage. (A) Whole Brain SUVR; (B) Plasma Aβ 42/40; (C) Plasma p‐tau 181; (D) Plasma NfL; (E) Plasma GFAP. Aβ, amyloid beta; PET, positron emission tomography; SUVR, standardized uptake value ratio.

3.3. Curve fitting results

Figure 2A–D shows the results of the fitted curves for the four plasma markers with changes in meta‐ROI SUVR. Figure 2E combines the four normalized curves while combining them and divides the different A‐stage regions according to the distribution of scatter points. The summary curves comparison shows that GFAP showed the earliest significant increase in stage A0, followed by NfL and p‐tau181 and then began to show an upward trend in stage A1. Meanwhile, Aβ42/Aβ40 showed a slow downward trend in the whole stage.

FIGURE 2.

Curve fitting. Curve fitting results of Plasma Aβ42/Aβ40 (A), Plasma p‐tau 181 (B), Plasma NfL(C) and Plasma GFAP (D) with meta‐ROI SUVR; (E) Curve Fitting Summary. Aβ, amyloid beta; GFAP, glial fibrillary acidic protein; NfL, neurofilament light; PET, positron emission tomography; p‐tau, phosphorylated tau; SUVR, standardized uptake value ratio.

3.4. ROC results

Table 2 shows the area under the curve (AUC) results for the ROC analysis of single and fused features. Combining the two results, after fusing the meta‐ROI SUVR and plasma markers with the features, GFAP displayed the highest classification efficacy in the early stage (stage A0 vs. stage A1: AUC = 0.878 ± 0.036), as well as the highest AUC in distinguishing stage A0 vs. stages A2 to 4 (0.994 ± 0.005), and in stage A1 versus stages A2 to 4, p‐tau181 showed the highest AUC (0.928 ± 0.043). In addition, in the supplementary ROC analysis to distinguish A0 versus A1 to 4 stages, it was still the fusion feature of meta‐ROI SUVR with GFAP that had the highest AUC (0.915 ± 0.027).

TABLE 2.

Results of ROC analysis of single and fused features.

	Stage A0 vs. stage A1	Stage A0 vs. stage A2–4	Stage A1 vs. stage A2–4	Stage A0 vs. stage A1–4
Single feature
Meta‐ROI SUVR	0.871 ± 0.034	0.966 ± 0.021	0.848 ± 0.067	0.909 ± 0.024
Aβ42/Aβ40	0.582 ± 0.067	0.727 ± 0.079	0.629 ± 0.101	0.628 ± 0.056
p‐tau181	0.619 ± 0.068	0.921 ± 0.039	0.833 ± 0.072	0.7140 ± 0.055
NfL	0.637 ± 0.071	0.819 ± 0.083	0.682 ± 0.099	0.694 ± 0.059
GFAP	0.692 ± 0.062	0.858 ± 0.068	0.742 ± 0.098	0.744 ± 0.051
Fused feature
Meta‐ROI SUVR & Aβ42/Aβ40	0.683 ± 0.055	0.784 ± 0.067	0.648 ± 0.106	0.715 ± 0.047
Meta‐ROI SUVR & p‐tau181	0.814 ± 0.046	0.985 ± 0.011	0.928 ± 0.043	0.868 ± 0.035
Meta‐ROI SUVR & NfL	0.847 ± 0.038	0.966 ± 0.017	0.803 ± 0.073	0.884 ± 0.030
Meta‐ROI SUVR & GFAP	0.878 ± 0.036	0.994 ± 0.005	0.878 ± 0.053	0.915 ± 0.027

Note: The data presented in the table are the area under the curve (AUC) ± standard. Bold denotes the maximum value in this group.

Abbreviations: Aβ, amyloid beta; GFAP, glial fibrillary acidic protein; NfL, neurofilament light; PET, positron emission tomography; p‐tau, phosphorylated tau; ROC, receiver operating characteristic; ROI, region of interest; SUVR, standardized uptake value ratio

3.5. Correlation of images with plasma biomarkers

According to the voxel‐level correlation results (Figure 3), no significant correlation was observed in subjects at stage A0. However, starting from stage A1, PET images showed significant positive correlations with GFAP, NfL, and p‐tau181 biomarkers, alongside significant negative correlations with Aβ42/Aβ40. Similar results were found in clinical subgroups (Figure S1 in supporting information), where GFAP first showed the most significant positive correlation in specific brain regions in the SCD group. Interestingly, the majority of the correlations in the sex subgroups of SCD were present in the female group (Figure S2 in supporting information).

FIGURE 3.

Voxel‐level correlation between PET images and plasma biomarkers across Aβ stage. Aβ, amyloid beta; GFAP, glial fibrillary acidic protein; NfL, neurofilament light; p‐tau, phosphorylated tau.

According to the ROI‐level correlation results (Figure 4 and Table S2 in supporting information), plasma p‐tau181 initially exhibited the highest number (six ROIs after FDR correction) of SUVR values with significant positive correlation at stage A1, followed by plasma GFAP (three ROIs after FDR correction). Notably, among the three plasma markers, the overlapping brain regions with significant correlation were distributed in the precuneus (PCUN), the rectus gyrus (REC), and inferior temporal gyrus (ITG). The remaining correlations were not significant after FDR correction. In addition, in the SCD sex subgroup, ROI‐level correlation analysis showed that most of the ROI SUVR values exhibited significant correlations with plasma GFAP, NfL, and p‐tau181 levels in females, which were significantly higher than those in males (Figure S3 and Table S3 in supporting information).

FIGURE 4.

ROI‐level correlation (r value) between PET SUVRs and plasma biomarkers across Aβ stage. Aβ, amyloid beta; ANG, angular gyrus; CUN, cuneus; GFAP, glial fibrillary acidic protein; IFG, inferior frontal gyrus; INS, insula; IOG, inferior occipital gyrus; ITG, inferior temporal gyrus; LING, lingual gyrus; MCC, midcingulate cortex; MFG, middle frontal gyrus; MOG, middle occipital gyrus; MTG, medial temporal gyrus; NfL, neurofilament light; OLF, olfactory cortex; PCL, paracentral gyrus; PCUN, precuneus; PET, positron emission tomography; PFC, prefrontal cortex; p‐tau, phosphorylated tau; REC, rectus gyrus; ROL, rolandic operculum; SFG, superior frontal gyrus; SMG, supramarginal gyrus; SUVR, standardized uptake value ratio; TPO, temporo‐parietal‐occipital.

3.6. Correlation of clinical scales with plasma biomarkers and the image metrics

In all subject groups, plasma GFAP exhibited the highest number of clinical scales with significant correlation (Figure 5A). Notably, at stage A0, three clinical scales were the first to show a significant correlation with plasma GFAP (Figure 5B). Specifically, these scales were MMSE (r = −0.370, P < 0.001), STT‐A (r = 0.376, P < 0.001), and MoCA‐B (r = −0.211, P < 0.05). Additionally, N5 (r = −0.575, P < 0.01) and N7 (r = −0.554, P < 0.01) scales showed significant positive correlation with plasma GFAP in stage A1.

FIGURE 5.

Correlation of clinical scale scores with plasma biomarkers at different stages across Aβ stage. (A) All subjects; (B) A0 subjects; (C) A1 subjects; (D) A2‐4 subjects. Circle size and color represent R‐value magnitude. *P < 0.05; **P < 0.01; ***P < 0.005; ****P < 0.001; the red box represents that the result is still significant after passing the false discovery rate correction. Aβ, amyloid beta; BNT, Boston Naming Test; GFAP, glial fibrillary acidic protein; HAMA, Hamilton Anxiety Rating Scale; HAMD, Hamilton Depression Rating Scale; MCI, mild cognitive impairment; MMSE, Mini‐Mental State Examination; MoCA‐B, basic version of Montreal Cognitive Assessment; N5, long‐term delayed recall; N7, long‐term delayed recognition; NfL, neurofilament light; PET, positron emission tomography; p‐tau, phosphorylated tau; SCD‐9, Subjective Cognitive Decline Questionnaire 9; STT‐A, Shape Trail Test A; STT‐B, Shape Trail Test B; SUVR, standardized uptake value ratio; VFT, Verbal Fluency Task.

In addition, Aβ whole‐brain SUVR values were significantly correlated with all four scales (MMSE, AVLT, STT, and MoCA‐B). HAMA levels were higher at the A2 to 4 stages when plasma p‐tau181 levels were also higher.

3.7. Mediating effect analysis

Table 3 showed significant fully mediated effect results in two pathways: the PET SUVR–plasma p‐tau181–MMSE score and the PET SUVR–plasma GFAP–MMSE score, respectively. Neither pathway exhibited a significant direct effect between the independent and dependent variables, but after mediation by plasma biomarkers, a significant fully mediated effect emerged (direct effect bootstrap 95% confidence interval [CI] includes 0, indirect bootstrap 95% CI does not include 0). In the PET SUVR–plasma p‐tau181–MMSE score pathway, all four meta‐ROI SUVRs demonstrated fully mediated effects, whereas in the PET SUVR–plasma GFAP–MMSE score pathway, only meta‐ROI SUVR values in stage A2 demonstrated mediated effects.

TABLE 3.

Results of mediation effects analyses.

Variable			Effect			Bootstrap 95% CI
X	Y	M	a	b	c	Direct effect of X on Y	Indirect effect of X on Y
Aβ stage 1 SUVR	MMSE	plasma p‐tau181	1.7119**** P < 0.0001	−0.6540** P = 0.0022	−0.0779 P = 0.9355	−1.9797 ∼ 1.8238	−2.8478 ∼ −0.0879
Aβ stage 2 SUVR	MMSE	plasma p‐tau181	2.4070**** P < 0.0001	−0.6432** P = 0.0028	−0.2593 P = 0.8392	−2.7838 ∼ 2.2651	−4.0064 ∼ −0.1258
Aβ stage 3 SUVR	MMSE	plasma p‐tau181	2.0291**** P < 0.0001	−0.6152** P = 0.0032	−0.8566 P = 0.5015	−3.3714 ∼ 1.6581	−3.3454 ∼ −0.0656
Aβ stage 4 SUVR	MMSE	plasma p‐tau181	2.1573**** P < 0.0001	−0.6290** P = 0.0024	−0.7302 P = 0.6133	−3.5825 ∼ 2.1220	−3.6813 ∼ −0.0792
Aβ stage 2 SUVR	MMSE	plasma GFAP	105.1932*** P = 0.0008	−0.0176**** P < 0.0001	0.0439 P = 0.9692	−2.1999 ∼ 2.2877	−3.6042 ∼ −0.0709

Note: SPSS PROCESS toolkit simple Model 4 (X‐M‐Y model) was used; X, independent variable, Y, dependent variable; M, mediating variable. Effect a , the effect size of X → M; Effect b , the effect size of M → Y; Effect c , the effect size of X → Y. Bootstrap 95% confidence intervals (CIs) not including 0 represent significant mediation effects.

Abbreviations: Aβ, amyloid beta; GFAP, glial fibrillary acidic protein; MMSE, Mini‐Mental State Examination; NfL, neurofilament light; p‐tau, phosphorylated tau; SUVR, standardized uptake value ratio.

4. DISCUSSION

The results of the present study support (1) the potential role of the plasma markers p‐tau181 and GFAP in patients with early AD, as evidenced by both the clinical spectrum of AD and the pathologic Aβ stage classification; (2) the discovery of a possible ceiling effect of plasma biomarkers in the middle and late stages of the AD disease course; (3) the confirmation of the role of plasma to promote Aβ deposition in the early stages of the disease, particularly in female SCD patients; and (4) the role of plasma markers on cognitive function in different stages of the Aβ continuum.

Among blood‐based biomarkers, plasma p‐tau 181, 217, and 231 have demonstrated high performance in distinguishing AD from non‐AD individuals. ¹⁷ In previous studies, the ability to discriminate between normal and abnormal amyloid PET has been compared between individual plasma markers and combinations of markers, with combinations resulting in marginally better results. ¹⁸ , ¹⁹ , ²⁰ , ²¹ , ²² For example, Janelidze et al. found that combining plasma p‐tau181 with Aβ42/40 improved the AUC for distinguishing normal from abnormal amyloid PET compared to using each marker alone. ¹⁹ Another study ²² emphasized the diagnostic and longitudinal monitoring potential of GFAP and p‐tau for preclinical AD. These results also flank the p‐tau181–Aβ–PET relationship but rarely distinguish amyloid staging. An obvious explanation is that distinguishing amyloid PET staging is more difficult than the binary distinction between normal and abnormal. Nevertheless, the findings of Jack et al. ¹⁵ suggested that combinations of plasma biomarkers were more accurate than individual plasma analytes in distinguishing amyloid PET stages, providing a scientific basis for plasma prediction of AD pathologic staging. In a longitudinal follow‐up cohort, researchers found that plasma p‐tau181 was significantly correlated with Aβ‐normalized measurements (Centiloid), ²³ and that elevated plasma p‐tau181 was independently correlated with cognitive stage shifts over 3 years and that plasma p‐tau181 and Centiloid alone could accurately predict cognitive stage. Centiloid alone accurately predicted cognitive stage shifts, especially in patients with MCI. ²⁴ In our study, voxel‐level correlation results revealed that no significant correlation was seen in subjects at stage A0, but from stage A1 onward, PET images exhibited significant positive correlations with p‐tau181, GFAP, and NfL biomarkers. This finding explains to some extent why the use of plasma p‐tau181 predicts cognitive stage transition better in MCI patients. In addition, our results also found that HAMA was higher in stages A2 to 4, which coincided with the clinical appearance of anxiety in patients with AD in the middle and late stages, and the plasma p‐tau181 level was also higher at this time.

Two powerful liquid biomarkers for measuring astrocyte reactivity in vivo are GFAP and chitinase‐3‐like protein 1 (YKL‐40). ²⁵ Both markers have consistently shown elevation in the dementia stage of AD. ²⁶ , ²⁷ , ²⁸ In particular, plasma GFAP, but not GFAP in CSF, has shown excellent performance in detecting Aβ‐positive cognitively unimpaired (CU) individuals. ²⁹ , ³⁰ , ³¹ Recent studies have shown that astrocyte changes occur early in AD progression, before significant neurodegeneration and cognitive deficits, suggesting upregulation of GFAP levels in Aβ‐positive CU individuals. ⁵ , ²⁹ , ³² , ³³ Previous studies have shown that increased Aβ oligomers are highly correlated with astrocyte reactivity. ³⁴ , ³⁵ , ³⁶ Astrocytes are involved in the clearance and degradation of Aβ, ³⁷ , ³⁸ and there is evidence that astrocytes can internalize Aβ oligomers and protofibrils but may ultimately become overwhelmed and unable to efficiently clear Aβ. This reason may explain the ceiling effect that was found in our study, particularly in the mid‐to‐late stages of the Aβ‐PET grading of GFAP, with plasma GFAP at the Aβ stages forming a phenomenon of increasing and then decreasing, coinciding with a recent study in which GFAP will not increase or even show a decreasing trend in the late Aβ‐PET stage. ¹⁵ In addition, when astrocytes disintegrate, they release their accumulated Aβ, thereby actively contributing to the overall accumulation of Aβ plaques. ³⁹ , ⁴⁰ , ⁴¹

According to the voxel‐level correlation results, from stage A1 onward, PET images showed significant positive correlations with GFAP, NfL, and p‐tau181 biomarkers, and significant negative correlations with Aβ42/Aβ40. Similar results were observed in clinical subgroups, in which GFAP first showed the most significant positive correlation in specific brain regions in the SCD group. Interestingly, in the sex subgroups of SCD, most of the correlations appeared in the female group. Numerous studies have reported ⁴² , ⁴³ , ⁴⁴ that the prevalence and incidence of AD is higher in females than in males, and have shown that AD disproportionately affects females in terms of development, progression, and clinical manifestations. ⁴⁵ , ⁴⁶ , ⁴⁷ Our study identified, for the first time, that at the SCD stage, GFAP‐related Aβ deposition predominantly occurs in the posterior cingulate gyrus (PCC), PCUN, olfactory cortex (OLF), angular gyrus (ANG), ITG, and others. Surprisingly, the hippocampus (HIP) and parahippocampal gyrus (PHG) did not show significant involvement, suggesting that the early symptoms of SCD may be caused by the disturbance of the integrated information processing ability of the whole brain and the blockage of signaling caused by the formation of Aβ in specific brain areas. In the correlation results of ROI level, at stage A1, plasma p‐tau181 exhibited the highest number of significant positive correlation with the most SUVR values first (12 ROI), followed by plasma GFAP (8 ROI), and then plasma NfL (6 ROI). Overlapping brain regions with significant positive correlation among the three plasma markers were distributed in the PCUN, REC, and ITG, all of which have been directly or indirectly associated with cognitive function. ⁴⁸ , ⁴⁹ , ⁵⁰ , ⁵¹ The PCUN is primarily involved in the default mode network (DMN), related to environmental monitoring and introspection. Compared to the MCI group, the SCD group has stronger functional connectivity (FC) in the DMN, with several FC changes similar to the MCI group. SCD also has specific FC patterns related to the DMN. ⁵² Additionally, SCD Aβ+ patients had significantly higher regional homogeneity in the bilateral PCUN, linking functional changes in this region to Aβ burden and indicating early AD pathology. The REC region, part of the frontal lobe, is involved in executive functions, explaining the correlation found with the STT scale in this study. Furthermore, studies show that individuals with SCD have decreased hippocampal tail FC with the right medial prefrontal cortex (mPFC) and left temporoparietal junction (TPJ), and decreased whole hippocampus FC with bilateral mPFC and TPJ. ⁵³

Plasma Aβ42/Aβ40 showed a significant negative correlation only in two ROI at stage A1. In addition, in the SCD sex subgroup ROI‐level correlation results, most female ROI SUVR values were significantly correlated with plasma GFAP, NfL, and p‐tau181, and with significantly higher correlations than those of males. As mentioned earlier, female patients may be more likely to have pathologic accumulation during the course of AD.

In all subject groups, plasma GFAP exhibited the highest number of clinical scales significantly correlated, with three clinical scales, MMSE, STT‐A, and MoCA‐B, first showing a significant correlation with plasma GFAP at the A0 stage. Specifically, MMSE, STT‐A, and MoCA‐B, also laterally reflect the scientific validity and feasibility of MMSE and MoCA‐B as early screening scales for AD. At stage A1, the N5 and N7 scores were significantly and positively correlated with plasma GFAP levels. In addition, Aβ whole‐brain SUVR values were significantly correlated with all four scales, namely N5, STT‐A, STT‐B, and MoCA‐B. This may explain, in part, the decline in subjective cognition associated with early AD.

In addition, there is a recurring debate about the Aβ hypothesis. ⁵⁴ , ⁵⁵ In the present study, two pathways showed significant fully mediated effect results: the PET SUVR–plasma p‐tau181–MMSE score and the PET SUVR–plasma GFAP–MMSE score. These pathways showed a significantly fully mediated effect after mediation by plasma biomarkers, highlighting the important role of plasma p‐tau181 and GFAP in AD continuum processes suggestive of cognitive impairment. In the PET SUVR–plasma p‐tau181–MMSE scoring pathway, all four meta‐ROI SUVRs showed fully mediated effects, while in the PET SUVR–plasma GFAP–MMSE scoring pathway, only two meta‐ROI SUVR values, namely stage A2 and stage A4, showed mediated effects. These results support the Aβ hypothesis, suggesting that Aβ influences downstream processes, including tau phosphorylation, leading to neurodegeneration and GFAP augmentation. ⁵⁴ , ⁵⁶ The finding of a close relationship between Aβ‐PET status and plasma p‐tau levels was also consistent with a previous study. ²²

The strengths of the present study include the longitudinal design, comprehensive availability of biomarkers, diagnostics, and cognition assessment, as well as the fairly long follow‐up period. However, our study has several limitations: relatively small sample size and the unequal distribution of male and female proportions. In addition, part of the results of this study were carried out under the clinical definition of AD. However, the data came from the SILCODE cohort, in which < 5% of dementia patients had a negative Aβ PET. Therefore, we believe our results are somewhat generalizable. Finally, only a small number of MCI and AD were included in this study, and we hope to increase the enrollment of such participants in the future. Nevertheless, our results were found in the early stage of AD, so the current results are still reliable.

In conclusion, our findings suggest plasma p‐tau181 and GFAP play a role in early AD, as indicated by both the clinical spectrum and the pathological Aβ staging. We also observed a possible ceiling effect of GFAP in the mid‐to‐late stages of AD. Furthermore, our results confirm the role of AD plasma markers in promoting Aβ deposition at an early stage, particularly in females with SCD. The overlapping brain regions of plasma p‐tau181, GFAP, and NfL for Aβ deposition in the brain in early AD were distributed across various regions, including the PCUN, REC, and ITG. This highlights the increasing importance of amyloid pathology staging in clinical trials targeting specific stages of AD.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest. Author disclosures are available in the supporting information.

CONSENT STATEMENT

This study was approved by the Medical Ethics Committee of Xuanwu Hospital, Capital Medical University, and was conducted in accordance with the Helsinki Declaration. All participants provided written informed consent and authorized the publication of their clinical details. SILCODE is listed on the ClinicalTrials.gov registry (SILCODE: NCT03370744).

Supporting information

Supporting information

Supporting information

Yu X, Shi R, Zhou X, et al. Correlations between plasma markers and brain Aβ deposition across the AD continuum: Evidence from SILCODE. Alzheimer's Dement. 2024;20:6170–6182. 10.1002/alz.14084