Integrated whole-transcriptome and DNA methylation analysis identifies new gene network in Alzheimer disease. C. E. Humphries1, M. A. Kohli1, P. W. Whitehead1,2, G. Beecham1,2, D. C. Mash3, M. A. Pericak-Vance1,2, J. Gilbert1,2 1) Hussman Inst Human Genomics, Univ Miami, Miami, FL; 2) Dr. John T. Macdonald Department of Human Genetics, Univ Miami, Miami, FL; 3) Department of Neurology, Univ Miami, Miami, FL.
To understand the processes contributing to Late-onset Alzheimer Disease (LOAD), we investigated differences in transcription and DNA methylation in human post-mortem brain tissue. Transcription was examined using total RNA-seq which permits the identification and quantitation of both known genes and non-coding RNAs. DNA methylation was examined using Illuminas Infinium HumanMethylation450 BeadChip Kit. Neuropathological specimens were sampled from age, sex and race-matched temporal poles (BA 38) from ten cases each of LOAD, Diffuse Lewy-Body (DLB) disease and pathologically and clinically normal controls. RNA was extracted using Qiagens miRNeasy kit and the library was prepared with Epicentres Script-Seq protocol. Samples were run at a density of two per lane on Illuminas HiSeq2000, generating 40-65 million reads per library. The program DESeq was used to examine transcriptional differences between Gencode genes, a union of genomic sequences encoding a coherent set of potentially overlapping functional products. Transcriptional analysis between LOAD and Controls revealed 2,700 genes out of a total of 37,000 genes (Gencode) to be differentially expressed (p<0.05). Of these 2700 genes, 60% were protein coding and 40% were non-coding RNAs. Seventeen of these genes survived correction for multiple testing (FDR) of which 6 were non-coding RNAs. Next, the 2,700 genes that were differentially expressed in LOAD were compared between LOAD and DLB. Three hundred of these genes were differentially expressed (p<0.05) with 140 being protein coding. StringDB, a network program, found 49 of the protein coding genes to be functionally connected in a network. The hub of the network was VCAM-1, a gene known to be upregulated in LOAD. This hub had 3 branches consisting of genes involved in: 1) angiogenesis, 2) immune responses, and 3) axonal growth and cell adhesion. DNA methylation within these 49-networked genes was examined for differences between LOAD and CON+DLB brains. 35 of the 49 genes had altered methylation in LOAD. Of the 35 genes, changes of DNA methylation in the promoter region were negatively correlated(r-0.76) with transcription in 24 genes. The study of this network of genes with altered expression and methylation specific to LOAD may offer a fruitful approach to advancing our understanding of the etiology of late-onset Alzheimers disease and the role that methylation changes may play in LOAD gene transcription.