Distinct cytokine profiles in human brains resilient to Alzheimer's pathology

Abstract

Our group has previously studied the brains of some unique individuals who are able to tolerate robust amounts of Alzheimer's pathological lesions (amyloid plaques and neurofibrillary tangles) without experiencing dementia while alive. These rare resilient cases do not demonstrate the patterns of neuronal/synaptic loss that are normally found in the brains of typical demented Alzheimer's patients. Moreover, they exhibit decreased astrocyte and microglial activation markers GFAP and CD68, suggesting that a suppressed neuroinflammatory response may be implicated in human brain resilience to Alzheimer's pathology. In the present work, we used a multiplexed immunoassay to profile a panel of 27 cytokines in the brains of controls, typical demented Alzheimer's cases, and two groups of resilient cases, which possessed pathology consistent with either high probability (HP, Braak stage V-VI and CERAD 2-3) or intermediate probability (IP, Braak state III-IV and CERAD 1-3) of Alzheimer's disease in the absence of dementia. We used a multivariate partial least squares regression approach to study differences in cytokine expression between resilient cases and both Alzheimer's and control cases. Our analysis identified distinct profiles of cytokines in the entorhinal cortex (one of the earliest and most severely affected brain regions in Alzheimer's disease) that are up-regulated in both HP and IP resilient cases relative to Alzheimer's and control cases. These cytokines, including IL-1β, IL-6, IL-13, and IL-4 in HP resilient cases and IL-6, IL-10, and IP-10 in IP resilient cases, delineate differential inflammatory activity in brains resilient to Alzheimer's pathology compared to Alzheimer's cases. Of note, these cytokines all have been associated with pathogen clearance and/or the resolution of inflammation. Moreover, our analysis in the superior temporal sulcus (a multimodal association cortex that consistently accumulates Alzheimer's pathology at later stages of the disease along with overt symptoms of dementia) revealed increased expression of neurotrophic factors, such as PDGF-bb and basic FGF in resilient compared to AD cases. The same region also had reduced expression of chemokines associated with microglial recruitment, including MCP-1 in HP resilient cases and MIP-1α in IP resilient cases compared to AD. Altogether, our data suggest that different patterns of cytokine expression exist in the brains of resilient and Alzheimer's cases, link these differences to reduced glial activation, increased neuronal survival and preserved cognition in resilient cases, and reveal specific cytokine targets that may prove relevant to the identification of novel mechanisms of brain resiliency to Alzheimer's pathology.

Keywords: Alzheimer's disease; Neuroinflammation; Partial least squares regression; Resilience.

Fig. 1.. Multivariate analysis of resilient cases.

Color map indicates the mean value of each pathological measure, over each column, from the EC and STS in control, HP, IP, and AD groups. The mean values demonstrate that in contrast to AD cases, resilient cases demonstrate no neuronal loss and no astrocyte or microglial activation markers despite Aβ deposits and neurofibrillary tangles at levels that are comparable to demented AD cases in the EC.

Fig. 2.. Analysis of pathological markers in the entorhinal cortex distinguishes control, AD, and Resilient cases.

(A) z-scored data for 12 pathological markers. (B) Dot plots for six pathological markers that distinguish control, resilient, and AD cases (mean±SD).

Fig. 3.. Multiplexed immunoassay cytokine analysis of postmortem EC tissues.

z-scored cytokine panel data plotted together with key pathological measures from postmortem EC samples. (Gray pathological measures indicate no measurement was available). Note that VEGF was not detected in any samples.

Fig. 4.. D-PLSR analysis of EC cytokines separates control from AD cases along LV1. D-PLSR analysis of EC cytokines separates control from AD cases along LV1.

(A) Our D-PLSR analysis identifies a profile of weights for each cytokine maker that transforms the original cytokine data into a new space displayed on two axes called LV1 and LV2. The values of the sample data in the new space are called Scores, which better distinguish between groups than the original data space. (B) D-PLSR analysis separates controls to the left and AD cases to the right along LV1 (C) The new axis LV1 consisted of a profile of cytokines that correlated with either control (negative) or AD (positive) cases.

Fig. 5.. D-PLSR analysis of EC cytokines separates HP or IP resilient groups from AD cases along LV1.

(A) D-PLSR analysis identified a cytokine profile axis, LV1, that separated HP cases to the left and AD cases to the right. The axis consisted of a profile of cytokines that were up-regulated in HP (negative) or AD cases (positive). (B) D-PLSR analysis identified a cytokine profile axis, LV1, that separated IP cases to the left and AD cases to the right. The axis consisted of a profile of cytokines that were up-regulated in IP (negative) or AD cases (positive). Errors bars on each LV1 profile were computed by PLSR model regeneration using iterative leave-one-out cross validation (mean ± SD).

Fig. 6.. D-PLSR analysis of EC cytokines separates HP or IP resilient groups from control cases along LV1.

(A) D-PLSR analysis identified a cytokine profile axis, LV1, that separated control cases to the left and HP cases to the right. The axis consisted of a profile of cytokines that were up-regulated in control (negative) or HP cases (positive). (B) D-PLSR analysis identified a cytokine profile axis, LV1, that separated control cases to the left and IP cases to the right. The axis consisted of a profile of cytokines that were up-regulated in control (negative) or IP cases (positive). Errors bars on each LV1 profile were computed by PLSR model regeneration using LOOCV (mean ± SD).