Association of lipidomics signatures in blood with clinical progression in preclinical and prodromal Alzheimer’s disease.

Sakr F, Dyrba M, Bräuer AU, and Teipel S. (2021) Association of lipidomics signatures in blood with clinical progression in preclinical and prodromal Alzheimer’s disease. Journal of Alzheimer’s Disease

Alzheimer’s disease (AD) is a neurodegenerative disorder known for causing cognitive decline and memory loss. Research to determine the cause of AD has been ongoing for many decades, but no singular cause has been found. Instead, AD pathology has been associated with many risk factors including accumulation of toxic proteins such as beta-amyloid (Aβ) and hyperphosphorylated tau (p-tau), impaired cholesterol transport, poor diet, heart disease, high alcohol consumption, mental illness, lack of physical activity, and reduced plasmalogen levels. Plasmalogens are a class of lipids that contain a vinyl-ether bond at sn-1 which causes the fatty acid chains at sn-1 and sn-2 to have a more compact structure. This vinyl-bond is responsible for their unique roles in membrane structure and fluidity, antioxidative properties, lipid rafts, cholesterol transport, and vesicular fusion. As the roles of plasmalogens are involved in the upstream and downstream pathology of AD, determining changes in the lipidomic signatures of people with stages of AD could be used as a biomarker of the disease. Sakr et al utilized targeted lipidomic data from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) to analyze lipidomic signatures in people with preclinical or prodromal AD and whether this could be used as a predictor for future diagnosis.

When looking at the demographics (more information about the demographics below*) of the cohort studied, the preclinical group (those who were cognitively normal, but with CSF p-tau/Aβ above 0.025) had the greatest proportion of the low category for BMI (average weight, 18.5 – 24.9). APOE ε4 was found at the highest levels in the preclinical and prodromal groups (MCI with CSF p-tau/Aβ above 0.025) at >60% while it was only 18% of the normal control group. As expected, CSF p-tau/Aβ levels were higher in the prodromal group than the preclinical group.

After lipidomic analysis was performed on plasma from the participant cohort, Bayesian elastic net regularized logistic regression was used to determine which lipids were associated with CSF p-tau/Aβ ratios. An initial filtering step was used to only include the top 60% of lipid composite scores that correlated with CSF p-tau/Aβ in the regression models. To test the model, 80% of the cohort was used to estimate the parameters, and these estimates were used to predict the outcome of the 20% of the cohort remaining. Three different models were created: a Reference model that used age and sex, Lipid model using lipid composite W-scores (calculated using the regression models), and Lipid model + APOE ε4 using the lipid model and their APOE status.

When tested, the Reference model was not any better than random prediction. While the Lipid model improved the prediction accuracy [cross validated (CV) area under the receiver operating curves (CV-AUC), CV-Accuracy, CV-Sensitivity, and CV-Specificity were 0.65, 0.66, 0.68, and 0.61, respectively] the Lipid + APOE ε4 was the most accurate (0.76, 0.71, 0.69, and 0.77, respectively). When determining associations between lipid classes and CSF p-tau/Aβ ratio, ethanolamine plasmalogens, choline plasmalogens, and sphingomyelins demonstrated negative associations with the biomarker ratio, while phosphatidylcholine and ceramides were positively associated with CSF p-tau/Aβ. Also, free fatty acids, diacylglycerol, and alkyl diacylglycerol were negatively associated with the AD biomarkers, but cholesterol esters and long-chain acyl-carnitines were positively associated.

Sakr et al analyzed the lipidome of 529 ADNI participants to determine if there is a lipidomic signature between preclinical and prodromal AD and whether this could be used as a biomarker for the disease or its progression. Similar with previous data, plasmalogens were found in lower levels in the patients in the preclinical and prodromal AD groups compared to the normal control group. Contrasting this, arachidonic acid-containing lysophosphatidylcholine and lyso-alkylphosphatidylcholine were both increased in the preclinical and prodromal groups. As lysophosphatidylcholine and lyso-alkylphosphatidylcholine are degradative products resulting from oxidative damage, this increase would be expected. Sakr et al suggest that the arachidonated phosphatidylcholines may play an early role in AD and can be detectable in cognitively normal controls who have CSF p-tau/Aβ biomarkers. Similarly, sphingomyelins, sphingolipid found in cell membranes and especially in myelin sheath, were decreased while ceramides, a bioactive molecule in sphingolipid metabolism were increased, which suggests apoptotic cell death of neurons and oligodendrocytes causing metabolism of sphingomyelin and producing ceramide. This work suggests that lipid levels could be used as a biomarker for the early detection of AD due to the associations found between other AD biomarkers and specific lipids. More work refining these lipidomic changes will allow for better determination of MCI or AD staging and could provide new routes for preventative treatments.

*The final cohort used in this study was 529 ADNI participants that had a baseline diagnosis of cognitively normal or mild cognitive impairment (MCI). As well, their cerebral spinal fluid (CSF) biomarkers, lipidomics, and body mass index (BMI) data was collected. The biomarker data was classified into three diagnostic groups based on the level of CSF p-tau/Aβ: cognitively normal were cognitively normal participants with CSF p-tau/Aβ below 0.025, preclinical group was cognitively normal but with CSF p-tau/Aβ above 0.025, and the prodromal group had MCI and CSF p-tau/Aβ above 0.025. BMI was also designated at low (average weight, 18.5-24.9), medium (overweight, 25-29.9), and high (at least moderately obese, >30). As the APOE (apolipoprotein) gene, specifically the ε4 allele, is the risk factor with the greatest association with AD diagnosis, the participants were genotyped for the presence of the ε4 allele and was designated into two groups: either no ε4 allele, or 1 or 2 ε4 alleles.