Systemic inflammation as a predictor of brain aging: Contributions of physical activity, metabolic risk, and genetic risk

Abstract

Inflammatory processes may contribute to risk for Alzheimer's disease (AD) and age-related brain degeneration. Metabolic and genetic risk factors, and physical activity may, in turn, influence these inflammatory processes. Some of these risk factors are modifiable, and interact with each other. Understanding how these processes together relate to brain aging will help to inform future interventions to treat or prevent cognitive decline. We used brain magnetic resonance imaging (MRI) to scan 335 older adult humans (mean age 77.3 ± 3.4 years) who remained non-demented for the duration of the 9-year longitudinal study. We used structural equation modeling (SEM) in a subset of 226 adults to evaluate whether measures of baseline peripheral inflammation (serum C-reactive protein levels; CRP), mediated the baseline contributions of genetic and metabolic risk, and physical activity, to regional cortical thickness in AD-relevant brain regions at study year 9. We found that both baseline metabolic risk and AD risk variant apolipoprotein E ε4 (APOE4), modulated baseline serum CRP. Higher baseline CRP levels, in turn, predicted thinner regional cortex at year 9, and mediated an effect between higher metabolic risk and thinner cortex in those regions. A higher polygenic risk score composed of variants in immune-associated AD risk genes (other than APOE) was associated with thinner regional cortex. However, CRP levels did not mediate this effect, suggesting that other mechanisms may be responsible for the elevated AD risk. We found interactions between genetic and environmental factors and structural brain health. Our findings support the role of metabolic risk and peripheral inflammation in age-related brain decline.

Keywords: AD; Alzheimer's disease; Brain structure; Brain volume; C-reactive protein; CRP; Cholesterol; Cognitive aging; Cortical thickness; Gene; Genetic risk; Inflammation; Insulin; Metabolic risk; Neuroinflammation; Obesity; Physical activity; Polygenic risk score.

Figure 1

FreeSurfer regions selected in our analysis. Slice A colors illustrate anterior (ACC), posterior (PCC), and isthmus (ICC) of the cingulate cortex, and precuneus (PRE). Slice B shows the hippocampus (HIPP), the parahippocampal gyrus (PARA), and entorhinal gyrus (ENT), the supramarginal gyrus (SM) and the middle frontal gyrus (*), which is divided into its rostral (RMF) and caudal (CMF) sections.

Figure 2

Path diagram representing the structural model. A: theoretical model initially tested before the iterative pruning procedure that led to Fig. 2b. Ellipse-shaped variables represent latent variables; square-shaped variables represent manifest variables. Abbreviations: BS = Brain Structure; MR = Metabolic Risk; PA = Physical Activity B: final most parsimonious model. Quality of fit: CFI = 0.999; TLI = 0.999; RMSEA = 0.004; SRMR = 0.05. Solid lines indicate significant loadings. Dashed lines indicate relationships that are not statistically significant (Z score < 1.96) but were taken into account in the overall fit indices. Brain_structure represents the latent process inferred by the additive inverse of cortical thickness in the regions identified through multiple regression (i.e., positive loadings on the Brain_structure latent variable indicate that a higher value for the independent variable is associated with thinner cortex). Metabolic_risk was derived from heart pulse pressure (PulseP), blood fasting insulin (ins) and triglycerides (tri), and body mass index (bmi). In the final model, PulseP was not included (see Model construction procedure in the Methods section); Physical_activity was derived from the number of blocks walked (blocks) and exercise intensity (exint). Edu: education (in years), APOE, Apolipoprotein E associated risk score for late onset Alzheimer's disease (see Table 2), GRS, genetic risk score obtained by cumulating the specific late onset AD – odds ratios related to single nucleotide polymorphisms in CR1 (rs6701713), MS4A4E and 6A (rs670139 and rs610932 respectively), CD33 (rs3865444), ABCA7 (rs3764650), EPHA1 (rs11767557), IL1a (rs1800587), IL1b (rs1143634), CLU (rs11136000), for individuals of European descent (Alzgene meta-analysis).