Different inflammatory signatures based on CSF biomarkers relate to preserved or diminished brain structure and cognition

Abstract

Neuroinflammation is a hallmark of Alzheimer's disease (AD) and both positive and negative associations of individual inflammation-related markers with brain structure and cognitive function have been described. We aimed to identify inflammatory signatures of CSF immune-related markers that relate to changes of brain structure and cognition across the clinical spectrum ranging from normal aging to AD. A panel of 16 inflammatory markers, Aβ42/40 and p-tau181 were measured in CSF at baseline in the DZNE DELCODE cohort (n = 295); a longitudinal observational study focusing on at-risk stages of AD. Volumetric maps of gray and white matter (GM/WM; n = 261) and white matter hyperintensities (WMHs, n = 249) were derived from baseline MRIs. Cognitive decline (n = 204) and the rate of change in GM volume was measured in subjects with at least 3 visits (n = 175). A principal component analysis on the CSF markers revealed four inflammatory components (PCs). Of these, the first component PC1 (highly loading on sTyro3, sAXL, sTREM2, YKL-40, and C1q) was associated with older age and higher p-tau levels, but with less pathological Aβ when controlling for p-tau. PC2 (highly loading on CRP, IL-18, complement factor F/H and C4) was related to male gender, higher body mass index and greater vascular risk. PC1 levels, adjusted for AD markers, were related to higher GM and WM volumes, less WMHs, better baseline memory, and to slower atrophy rates in AD-related areas and less cognitive decline. In contrast, PC2 related to less GM and WM volumes and worse memory at baseline. Similar inflammatory signatures and associations were identified in the independent F.ACE cohort. Our data suggest that there are beneficial and detrimental signatures of inflammatory CSF biomarkers. While higher levels of TAM receptors (sTyro/sAXL) or sTREM2 might reflect a protective glia response to degeneration related to phagocytic clearance, other markers might rather reflect proinflammatory states that have detrimental impact on brain integrity.

Fig. 1. Overview of methods and derived summary inflammation and brain measures.

A A panel of 16 immune-related markers was measured in CSF at baseline in the DELCODE sample. FLAIR and T1-weighted MR images at baseline were acquired in a subsample of participants. White matter hyperintensity (WMH) lesion maps were derived by segmentation of FLAIR images (LST toolbox). Gray matter (GM) and white matter (WM) volume probability maps were derived by segmentation of T1 images (CAT/SPM toolbox). As global summary measures we derived global WMH volume as well as GM and WM volume. GM rate of change maps were derived for subjects with 3 longitudinal T1-images using Jacobian integration (SPM). A PCA on the GM rate of volume change maps was performed within an AD vulnerable atrophy mask (see methods) to create subject-level summary measure (first eigenvariate) of atrophy over follow-ups. B A PCA on the immune markers resulted in 4 principal components (PCs). For each PC, the individual markers with highest loadings are depicted, which were all positive. See Supplementary Fig. 2 for the loadings.

Fig. 2. Associations of PC1 and PC2 with brain structural and cognitive measures.

Scatter plots for the associations of PC1 (A) and PC2 (B) with global gray matter (GM) volume, global white matter (WM) volume, global white matter hyperintensity (WMH) volume, GM rate of change over follow-ups in AD vulnerable areas (first eigenvariate form PCA, high values denote faster shrinkage), memory performance at baseline and PACC5 change over follow-ups (slope extracted from LME models). Standardized residuals are shown after regressing out covariates (age, gender, p-tau181, Aß42/40 and ICV for brain measures). * denotes significant associations in the full regression models (Tables 2 and 3). CU cognitively unimpaired, CI cognitively impaired.

Fig. 3. Whole-brain voxel-wise regression of brain volumes and rate of volume change on inflammatory components.

A Higher levels of inflammatory component PC1 were related to higher GM volume (orange) at baseline in bilateral hippocampus, left amygdala and putamen and higher WM volume (green) in temporal regions and brain stem. N = 258. B Higher levels of the inflammatory component PC2 were related to lower GM volume (light blue) at baseline in bilateral hippocampus, right thalamus and right putamen and lower WM volume (dark blue) in frontal and other regions. N = 258. C Levels of the PC1 were positively related to rate of change in GM volume (i.e., less/slower atrophy over follow-ups) in anterior medial temporal lobe, posterior inferior temporal cortex and posterior cingulate cortex (orange). N = 174. Results are depicted at p < 0.05 (FWE, cluster-level, cluster forming voxel-level threshold p = 0.001). See Supplementary Fig. 3 for results at voxel-level threshold p = 0.005. All analyses were accounting for age, gender, ICV, p-tau181, Aβ42/40. L Left, R Right, A Anterior, P Posterior.