Molecular estimation of neurodegeneration pseudotime in older brains

Sumit Mukherjee,Suman Jayadev,Gregory W Carter,Lara M Mangravite,Philip L De Jager,Gwenn A Garden,Anna K Greenwood,Solveig K Sieberts,Benjamin A Logsdon,Christoph Preuss,Nilüfer Ertekin-Taner,Laura Heath

doi:10.1038/s41467-020-19622-y

Abstract

The temporal molecular changes that lead to disease onset and progression in Alzheimer’s disease (AD) are still unknown. Here we develop a temporal model for these unobserved molecular changes with a manifold learning method applied to RNA-Seq data collected from human postmortem brain samples collected within the ROS/MAP and Mayo Clinic RNA-Seq studies. We define an ordering across samples based on their similarity in gene expression and use this ordering to estimate the molecular disease stage–or disease pseudotime-for each sample. Disease pseudotime is strongly concordant with the burden of tau (Braak score, P = 1.0 × 10−5), Aβ (CERAD score, P = 1.8 × 10−5), and cognitive diagnosis (P = 3.5 × 10−7) of late-onset (LO) AD. Early stage disease pseudotime samples are enriched for controls and show changes in basic cellular functions. Late stage disease pseudotime samples are enriched for late stage AD cases and show changes in neuroinflammation and amyloid pathologic processes. We also identify a set of late stage pseudotime samples that are controls and show changes in genes enriched for protein trafficking, splicing, regulation of apoptosis, and prevention of amyloid cleavage pathways. In summary, we present a method for ordering patients along a trajectory of LOAD disease progression from brain transcriptomic data.

Highlights

The temporal molecular changes that lead to disease onset and progression in Alzheimer’s disease (AD) are still unknown
The data from the ROS/MAP cohort are sampled from the dorsolateral prefrontal cortex (DLPFC), and the data from the Mayo Clinic cohort are sampled from the temporal cortex (TCX)
Adjusting for postmortem interval (PMI) (Supplementary Figs. 3A and 4A), ten principal components from a principal component analysis (PCA) of genotype data to account for ancestry effects (Supplementary Figs. 3B and 4B), RNA integrity number (RIN) (Supplementary Figs. 3C and 4C), or all of these variables (Supplementary Figs. 3D and 4D) did not materially change the overarching ordering of patients for either the TCX or DLPFC regions

Summary

Introduction

The temporal molecular changes that lead to disease onset and progression in Alzheimer’s disease (AD) are still unknown. We propose an approach to analyze population level RNA-seq data from postmortem brain tissue to learn a tree structured progression (Fig. 1) that represents distinct subtypes of disease and the relative progression of disease across patients With this approach, we identify potentially generalizable trajectories of LOAD across heterogeneous patient populations at all stages of disease. We refer to manifold learning and lineage inference interchangeably in reference to the construction of a disease progression tree We demonstrate that these tools can estimate the disease staging and progression tree (Fig. 2) from bulk RNA-Seq data collected from postmortem brain tissues in a case/control cohort. These trees show clear LOAD staging, enable the study of cell-type-specific effects of LOAD, and allow the identification of genetic factors driving disease progression

Methods

Results

Conclusion