CSF total tau as a proxy of synaptic degeneration

Abstract

Cerebrospinal fluid (CSF) total tau (t-tau) is considered a biomarker of neuronal degeneration alongside brain atrophy and fluid neurofilament light chain protein (NfL) in biomarker models of Alzheimer's disease (AD). However, previous studies show that CSF t-tau correlates strongly with synaptic dysfunction/degeneration biomarkers like neurogranin (Ng) and synaptosomal-associated protein 25 (SNAP25). Here, we compare the association between CSF t-tau and synaptic degeneration and axonal/neuronal degeneration biomarkers in cognitively unimpaired and impaired groups from two independent cohorts. We observe a stronger correlation between CSF t-tau and synaptic biomarkers than neurodegeneration biomarkers in both groups. Synaptic biomarkers explain a greater proportion of variance in CSF t-tau levels compared to neurodegeneration biomarkers. Notably, CSF t-tau levels are elevated in individuals with abnormalities only in synaptic biomarkers, but not in individuals with abnormalities only in neurodegeneration biomarkers. Our findings suggest that CSF t-tau is a closer proxy for synaptic degeneration than for axonal/neuronal degeneration.

Fig. 1. Correlation between Alzheimer’s disease biomarkers underscores the link between CSF t-tau levels and synaptic degeneration.

Correlation matrix shows significant two-sided Pearson’s coefficient (r) correlations after correcting for multiple comparisons (*Padj-value < 0.05) among established biomarkers of neurodegeneration (HCV, CSF NfL and t-tau), synaptic (CSF Ng and SNAP25), and tau tangles (tau PET) and Aβ (CSF Aβ_1-42) pathologies in (left) cognitively unimpaired (n = 760) and (right) cognitively impaired (n = 932) individuals. All variables were log-transformed and z-scored, and HCV and CSF Aβ_1-42 were inverted to indicate higher values indicate more pathology. Data points with sample sizes of 10 or fewer are omitted and displayed in white. Amyloid-β (Aβ). CSF (cerebrospinal fluid). Total-tau (T-tau). Classical neurodegeneration biomarkers: Hippocampal volume (HCV) and Neurofilament light chain protein (NfL). Classical synaptic biomarkers: Neurogranin (Ng) and Synaptosomal-associated protein 25 (SNAP25). Positron emission tomography (PET). Source data are provided as a Source Data file.

Fig. 2. CSF t-tau is more strongly associated with synaptic than neurodegeneration biomarkers.

Mean β estimates with 95% confidence intervals from linear regressions, adjusted for age, sex, and cohort, showing the association of CSF t-tau with biomarkers of neurodegeneration and synaptic degeneration in (left) cognitively unimpaired (n = 760) and (right) cognitively impaired (n = 932) individuals. Non-overlapping confidence intervals indicate statistically significant differences. All variables were log-transformed and z-scored. HCV was inverted, so higher values indicate more pathology. Cerebrospinal fluid (CSF). Total-tau (t-tau). Classical neurodegeneration biomarkers: Hippocampal volume (HCV) and CSF Neurofilament light chain protein (NfL). Classical synaptic biomarkers: CSF Neurogranin (Ng) and Synaptosomal-associated protein 25 (SNAP25). Source data are provided as a Source Data file.

Fig. 3. Synaptic biomarkers explain a greater proportion of variance in CSF t-tau than neurodegeneration biomarkers.

Bar charts show mean (±s.d.) the partial R² in (left) cognitively unimpaired (n = 188) and (right) cognitively unimpaired groups (n = 84). Partial R² values were estimated only for individuals who had complete data for all the following biomarkers: CSF t-tau as the outcome and biomarkers reflecting neurodegeneration (HCV and CSF NfL) and synaptic degeneration (CSF Ng and SNAP25) as predictors in linear regression models. Models were adjusted for age, sex, and cohort, and compared using AIC. Bootstrapping (n = 1000 iterations) was applied to assess the stability of the partial R² values. *indicate Akaike Information Criterion difference (ΔAIC) between neurodegeneration and synaptic degeneration >15. Cognitively unimpaired ΔAIC = 339; cognitively impaired ΔAIC = 95. Cerebrospinal fluid (CSF). Total-tau (t-tau). Hippocampal volume (HCV). Neurofilament light chain protein (NfL). Neurogranin (Ng). Synaptosomal-associated protein 25 (SNAP25). Standard deviation (s.d.). Source data are provided as a Source Data file.

Fig. 4. CSF t-tau is increased in individuals with abnormal synaptic degeneration regardless of the concomitant presence of neurodegeneration.

a Bar graphs show the distribution of CU, MCI, and dementia across synaptic degeneration (S) and neurodegeneration (N) groups. N positivity was based on HCV, and S positivity was based on CSF Ng (N⁻S⁻: n = 85; N⁻S⁺: n = 137; N⁺S⁻: n = 110; S⁺N⁺: n = 252). Cutoffs for biomarker positivity were calculated anchored in the CU Aβ^- individuals. b Violin plots show CSF t-tau levels in individuals with abnormal S and/or N in the whole population (CU and CI). The median is shown by the middle dashed line, and the quartiles by the top and bottom dashed lines. CSF t-tau levels were compared across groups using a linear regression model with dummy variables, adjusted for age, sex, cognitive status, amyloid burden, and cohort. Pairwise comparisons were corrected for multiple testing using Tukey’s method. * Padj-value < 0.001. Amyloid-β (Aβ). Cognitively unimpaired (CU). Mild cognitive impairment (MCI). Neurodegeneration (N). Neurogranin (Ng). Synaptic degeneration (S). Cerebrospinal fluid (CSF). Total-tau (t-tau). Hippocampal volume (HCV). Neurofilament light chain protein (NfL). Neurogranin (Ng). Synaptosomal-associated protein 25 (SNAP25). Source data are provided as a Source Data file.