Astrocyte biomarker signatures of amyloid-β and tau pathologies in Alzheimer's disease

Abstract

Astrocytes can adopt multiple molecular phenotypes in the brain of Alzheimer's disease (AD) patients. Here, we studied the associations of cerebrospinal fluid (CSF) glial fibrillary acidic protein (GFAP) and chitinase-3-like protein 1 (YKL-40) levels with brain amyloid-β (Aβ) and tau pathologies. We assessed 121 individuals across the aging and AD clinical spectrum with positron emission tomography (PET) brain imaging for Aβ ([¹⁸F]AZD4694) and tau ([¹⁸F]MK-6240), as well as CSF GFAP and YKL-40 measures. We observed that higher CSF GFAP levels were associated with elevated Aβ-PET but not tau-PET load. By contrast, higher CSF YKL-40 levels were associated with elevated tau-PET but not Aβ-PET burden. Structural equation modeling revealed that CSF GFAP and YKL-40 mediate the effects of Aβ and tau, respectively, on hippocampal atrophy, which was further associated with cognitive impairment. Our results suggest the existence of distinct astrocyte biomarker signatures in response to brain Aβ and tau accumulation, which may contribute to our understanding of the complex link between reactive astrogliosis heterogeneity and AD progression.

Fig. 1. GFAP associates with Aβ and YKL-40 with tau accumulation.

The panels show box-and-whisker plots of A CSF GFAP and B CSF YKL-40 levels adjusted for age-, sex-, and APOE ε4 status across AT groups. The horizontal line in each box represents the median; box ends represent the 25th and 75th percentiles. Shape of the dots depicts the clinical diagnosis (CU: 73.3% A-T-, and 26.7% A + T−; MCI: 34.5% A − T−, 34.5% A + T−, and 31.0% A + T+; AD: 23.5% A + T−, and 76.5% A + T). Groups were compared using analyses of variance with Tukey’s multiple comparison test (*P < 0.05, **P < 0.01, ***P < 0.001). C Partial residual plots of ROI-based linear regressions testing the associations of neocortical Aβ-PET SUVR with CSF GFAP and YKL-40 levels adjusting for temporal meta-ROI tau-PET SUVR. The shape of the dots depicts the AT group. D T-statistical parametric maps show the result of voxel-wise linear regression testing the regional association of Aβ-PET SUVR with CSF GFAP and YKL-40 levels adjusting for tau-PET SUVR. R and L indicate right and left, respectively; A and P denote anterior and posterior, respectively. E Partial residual plots of ROI-based linear regressions testing the associations of temporal meta-ROI tau-PET SUVR with CSF GFAP and YKL-40 levels adjusting for neocortical Aβ-PET SUVR. The shape of the dots depicts the AT group. F T-statistical parametric maps show the result of voxel-wise linear regression testing the regional association of tau-PET SUVR with CSF GFAP and YKL-40 levels adjusting for neocortical Aβ-PET SUVR. R and L indicate right and left, respectively; A and P denote anterior and posterior, respectively. Voxel-wise linear regressions were RFT-corrected for multiple comparisons at a voxel threshold of P < 0.001. Age, sex, cognitive status, and APOE ε4 status were used as covariates for adjustment in all ROI- and voxel-based linear regressions. NS not significant.

Fig. 2. Sensitivity analyses testing the associations of Aβ-PET and tau-PET with reactive astrocyte biomarkers using plasma GFAP instead of CSF GFAP.

A Box-and-whisker plot of plasma GFAP levels adjusted for age, sex, and APOE ε4 status across AT groups. The horizontal line in each box represents the median; box ends represent the 25th and 75th percentiles. Shape of the dots depicts the clinical diagnosis (CU: 73.6% A-T-, and 26.4% A + T−; MCI: 37.0% A − T−, 29.6% A + T−, and 33.3% A + T+; AD: 26.7% A + T−, and 73.3% A + T). Groups were compared using analyses of variance with Tukey’s multiple comparison test (*P < 0.05, **P < 0.01, ***P < 0.001). B Partial residual plots of linear regressions testing the associations of neocortical Aβ-PET SUVR with plasma GFAP and CSF YKL-40 levels adjusting for temporal meta-ROI tau-PET SUVR, age, sex, cognitive status, and APOE ε4 status. The shape of the dots depicts the AT group. C Partial residual plots of linear regressions testing the associations of temporal meta-ROI tau-PET SUVR with plasma GFAP and CSF YKL-40 levels adjusting for neocortical Aβ-PET SUVR, age, sex, cognitive status, and APOE ε4 status. The shape of the dots depicts the AT group. Of note, analyses involving plasma GFAP were conducted in a subset of 114 individuals; from the total study population of 121 subjects, five participants did not have available plasma GFAP measures, and two were excluded because they were considered outliers (plasma GFAP concentrations three SD above the mean of the population). NS not significant.

Fig. 3. Reactive astrocyte biomarkers mediate the effect of AD pathophysiology on downstream neurodegeneration and cognitive impairment.

The figure shows the standardized structural equation model estimates of the associations between A CSF GFAP and B CSF YKL-40 levels, hippocampal volume, and cognition. Solid black lines represent significant associations, whereas dashed gray lines represent non-significant effects. The fact that the presented estimates are standardized allows direct comparison between effects. Aβ pathology was measured with neocortical Aβ-PET SUVR and tau pathology using temporal meta-ROI tau-PET SUVR. Cognition was indexed using the MMSE score. All associations were further adjusted for age, APOE ε4 status, and years of education. NS not significant.