Multi-omic ADNI CSF and plasma data integration identifies distinct metabolic transitions in disease progression in Alzheimer’s Disease

Age and APOE allele status are the two greatest risk factors for Alzheimer’s disease, with microglial processing, recycling of debris and cholesterol transport emerging as other key genetic determinants from AD GWAS. How risk factors contribute to neuronal loss is an unanswered question. Using the ADNI dataset, we provide evidence for distinct processes driving neurodegeneration in AD across a range of ADAS13 scores binned into quartiles (Q1-4) found in ADNI. In Q1 individuals, we see evidence of white matter hyperintensities (WMH) associated with complement and HDL proteins, independent of tau and hippocampal atrophy rates. In Q2 individuals, we see specific protein/lipid correlation networks that are associated with tau and hippocampal atrophy. Also in Q2, we see variations in enzyme levels (LPA2G7, LPCAT2) that impact phosphatidylcholine availability for cholesterol efflux onto CNS lipoproteins. Low LPA2G7 and high LPCAT are associated with lower disruption of ANLS, while high LPA2G7 and low LPCAT are associated with higher disruption. In Q3 individuals, DHA and plasmalogens appear to be protective of astrogliosis, presumably by decreasing lipid droplet size, and increasing their lipid debris flux capacity. The networks we identify in this analysis provide evidence that progressively elevated cellular cholesterol levels, likely the result of accelerating neuronal and myelin debris, are perturbing homeostatic processes, resulting in disruption of astrocyte-neuron lactate shuttle (ANLS), contributing to lipid droplet formation, and astrogliosis. Existing studies have already identified elevated cellular cholesterol as the driver of increased access of APP to β- and γ-secretase for β-amyloid production in late-onset AD. Importantly, we provide evidence that points to actionable variations in lipids associated with slower disease progression and decreased metabolic disruption and inflammation. We believe this positive feedback loop (neurodegeneration → lipid debris processing → metabolic disruption → neurodegeneration) is a key destabilizing cycle responsible for disease progression in AD, reinforcing the perspective that Alzheimer’s is fundamentally a metabolic disease. Our findings are consistent with established lipid risk factor associations. We anticipate the mechanistic insights from our analysis will advance nutritional and pharmacological interventions and slow cognitive decline.