Plasma GFAP sits at the intersection of amyloid and tau biology in Alzheimer disease
Plasma GFAP is increasingly treated as a marker of astrocytic reactivity, but this preprint pushes the biomarker into a more specific role: a statistical link between the amyloid cascade and downstream tau pathology in Alzheimer disease. The analysis drew on two well-characterized cohorts, the Knight Alzheimer Disease Research Center discovery sample and an independent Alzheimer’s Disease Neuroimaging Initiative replication sample, and integrated plasma or cerebrospinal fluid amyloid measures, amyloid PET, tau PET, structural MRI, and clinical severity. Across both cohorts, plasma GFAP mediated the association between soluble Abeta42/40 and amyloid PET, and it also mediated the association between amyloid PET and tau PET. Both mediation signals survived false discovery rate correction in the overall samples, which places GFAP in a biologically credible position between early amyloid imbalance and later tau accumulation.
The result fits with what astrocytes do when the brain’s proteostatic balance begins to fail. Astrocytic activation can reflect injury, synaptic stress, and plaque-adjacent inflammatory responses, all of which may intensify once soluble amyloid dynamics shift toward plaque deposition. A biomarker such as GFAP may therefore be more than a passive injury readout; it may capture a reactive glial state that travels with the transition from soluble amyloid abnormalities to fibrillar amyloid burden and then to tau spread. In practical terms, the data suggest that GFAP could help map where a patient sits along the amyloid-to-tau continuum rather than simply indicating that neurodegeneration is underway.
Stage-specific patterns were present, but they did not replicate cleanly across cohorts
The stage-stratified findings were more selective. In tau-negative individuals, the indirect association between soluble Abeta42/40 and amyloid PET remained significant after correction only in Knight ADRC. In amyloid-positive individuals, the indirect association between amyloid PET and tau PET remained significant after correction only in ADNI. Interaction findings did not replicate consistently across cohorts, before or after correction. Those differences point to a biomarker effect that is real enough to appear in two independent datasets, yet sensitive to sample composition, disease stage, and cohort-specific measurement context.
A single sentence captures the clinical tension here: GFAP appears biologically connected to the amyloid cascade, but the exact stage at which that connection is strongest shifts between cohorts.
That variability matters because Alzheimer disease is not a single static state. Once tau is absent, the relationship between soluble amyloid dynamics and plaque burden may be easier to detect through astrocytic reactivity. After amyloid positivity is established, the same GFAP signal may be tracking a different process: plaque-associated glial activation that accompanies or facilitates tau propagation. This is the kind of stage dependence that can disappear if biomarker analyses collapse all participants into one broad case group.
Plasma proteomics points to matrix remodeling and immune signaling
The proteomic analysis adds an important layer. Plasma proteins associated with GFAP clustered in pathways enriched for extracellular matrix biology and immune-related signaling. Those signatures are not random background noise. They fit a model in which astrocytic activation is coupled to tissue remodeling, vascular interface changes, and inflammatory signaling around amyloid deposits. Extracellular matrix remodeling can alter diffusion, cellular adhesion, and synaptic environment, while immune pathway enrichment suggests that GFAP may be reporting on a broader inflammatory network rather than a narrow astrocyte-only process.
This network view matters for Alzheimer biology because amyloid deposition and tau spread do not occur in isolation. Plaque formation can alter local microenvironmental mechanics, provoke glial activation, and shift cytokine and complement activity; those processes can then influence how tau pathology emerges and spreads across connected brain regions. The proteomic data do not prove causation, but they strengthen the idea that plasma GFAP is embedded in a larger signaling architecture tied to neuroinflammation and tissue remodeling.
What the biomarker can and cannot say today
The sample sizes were moderate, with 278 participants in the discovery cohort and 236 in the replication cohort, and the work was observational rather than interventional. That design is well suited to biomarker mapping, but it cannot establish whether GFAP is a driver of amyloid or tau biology, a downstream reflection of those processes, or both. The mediation models are compelling because they survived correction in the overall cohorts, yet the stage-specific effects were not uniformly reproduced. Replication remained incomplete when the cohorts were split by biomarker-defined status, which limits immediate clinical translation.
Even so, the findings sharpen a useful question for the field: can plasma GFAP help distinguish early amyloid dysregulation from later tau-linked neurodegeneration? If so, it could become a component of longitudinal risk staging, especially when paired with amyloid and tau measurements rather than used alone. The current data are strongest for association, not diagnosis, and certainly not treatment selection.
Why this preprint matters for biomarker staging in Alzheimer disease
The most concrete clinical implication is not a new cutoff or a ready-made assay algorithm. It is the possibility that plasma GFAP indexes a reactive astrocytic response that emerges at different points along the Alzheimer trajectory and behaves differently before and after biomarker-defined disease stages are established. That interpretation is consistent with the observed mediation from soluble amyloid to plaque burden and from plaque burden to tau burden. It also aligns with the proteomic enrichment for extracellular matrix and immune pathways, which suggests that astrocytic reactivity is embedded in a broader neurobiological response rather than standing alone.
Because this is a preprint, the findings still need peer review and broader validation in additional cohorts. Stage-stratified analyses, especially those based on amyloid and tau status, will need to be tested prospectively and across platforms with harmonized biomarker assays. The question is no longer whether GFAP rises in Alzheimer disease; it clearly does. The next step is determining when that rise is most informative, which competing biological pathways it tracks, and whether those signals can be used reliably across independent populations.