Developing Topics.

Background: Amyloid deposition occurs during the preclinical stages of Alzheimer's disease (AD) a decade or more before clinical symptoms emerge. We leveraged blood transcriptomics and positron emission tomography (PET) measures of amyloidosis to identify cell types and gene networks in the blood that relate to amyloid burden in the brain.

Method: Whole blood RNA sequencing and amyloid PET data were leveraged from 1771 participants (62% females, mean age 71, 32% amyloid+) in the Anti-Amyloid Treatment in Asymptomatic Alzheimer's Disease (A4) Study. RNAseq data were aligned with STAR, sorted, counted using Subread, and batch normalized with voom. 19,945 genes were available for analysis following quality control. Cell fractions were quantified using CIBERSORTx (LM22 used as reference) and WGCNA was used to calculate gene co-expression modules (soft threshold = 12). Amyloid PET data were acquired using 18F-florbetapir 50-70 minutes post injection and a global cortical standardized uptake value ratio (SUVR) was quantified (whole cerebellum as reference). Linear regression related gene module expression or cell fraction to mean SUVR, covarying for age, sex, education, APOE ε2 and ε4 status. Correction for multiple comparisons used false discovery rate. Gene enrichment was performed using Gene Ontology.

Result: Two gene modules were associated with amyloid deposition. One included 48 genes whereby higher expression was associated with lower amyloid burden (β = -12.37, p = 0.0001). This module was enriched for DNA damage pathway genes and the top gene associations in the module included histone genes such as H3C3 and H2AC14. The second module included 241 genes whereby higher expression related to higher amyloid burden (β = 7.15, p = 0.001). This module was enriched for leukocyte migration and regulation of RNA pathways. The top gene association in the module was the calcium homeostasis gene CHERP. Finally, a higher proportion of CD4 positive activated memory T cells was associated with higher amyloid burden (β = -1.940, p = 0.04), but did not survive FDR correction.

Conclusion: Our results implicate DNA damage and activated memory T cells as contributing to amyloid deposition in preclinical AD. Future work will seek to identify modifiers of the APOE effect on amyloid and evaluate blood transcription networks as complementary biomarkers of amyloid deposition.