UC Irvine

Alzheimer’s disease (AD) is a devastating, progressive neurodegenerative disorder that results in dementia, with care for those with dementia estimated to cost the U.S. $321 billion in 2022. Although many years of research have uncovered notable findings about AD biology, it is clear that we still have a limited understanding of the disease as evidenced by the lack of effective therapeutics against AD. For example, genome-wide association studies (GWAS) have uncovered multiple genetic risk variants, revealing novel genes and pathways for study in AD. However, it remains a challenge to ascertain the functional significance of these risk variants. Multiple studies have attempted to clarify the role of AD risk variants with “bulk”-tissue RNA- sequencing (RNA-seq), but they have been hindered by the vast cellular heterogeneity of the brain. Single-cell (scRNA-seq) and single-nucleus RNA-seq (snRNA-seq) performed on the 5XFAD mouse model of AD, as well as human AD samples, identified disease-associated glial subpopulations, but it is unclear what regulates these subtypes. Additionally, studies have revealed region-specific glial subpopulations existing in the healthy brain, suggesting potential regional differences in the disease phenotype. This dissertation aimed to clarify the molecular landscape of AD with spatial and cellular resolution with an emphasis on the analysis and integration of different data modalities and model systems.We first sought to define the cell-type specific gene regulatory programs dysregulated in AD to identify potential regulators of disease-associated cell subpopulations and to further unravel AD genetic risk (Chapter Two). Recent advances in sequencing methods now allow interrogation of the transcriptome and epigenome at the single cell resolution, and AD epigenetic data has been limited. We generated paired single-nucleus transcriptomic and epigenomic data from postmortem human brain tissue of late-stage AD and cognitively healthy controls. In addition to being the first epigenetic dataset of human AD with single-cell resolution, we directly integrated the two different data modalities, allowing us to define disease-associated glial subpopulations at the transcriptome and epigenome. We identified cell-type specific, disease-associated candidate cis-regulatory elements (cCREs) and their candidate target genes. Although this is possible with single-cell epigenetic data alone, paired gene expression data provides additional functional evidence of cCREs. We also revealed cell-type specific transcription factors dysregulated with disease, like SREBF1 in oligodendrocytes, altogether identifying both cis- and trans-gene regulatory mechanisms that may regulate AD cell states. Furthermore, we characterized the cis- regulatory landscape at AD GWAS loci in specific cell-types, providing insight into the cell-types relevant to specific AD risk variants. On the other hand, while recent transcriptomic studies have revealed both brain region- and cell- type-specific gene expression changes in AD, “bulk”-tissue and scRNA-seq do not retain spatial information for gene expression changes without careful microdissection and sequencing separate samples. The human brain’s spatially complexity at both macro- and microscopic levels underlies brain function and thus is a critical feature to consider in disease pathophysiology. We generated spatial transcriptomic data from postmortem human brain tissue from cognitively healthy controls, early-, and late-stage AD to investigate the spatial relationship of disease-associated transcriptomic changes (Chapter Three). Additionally, we performed a comparative analysis of AD in Down Syndrome (DS) and the general population by generating both spatial and single-nucleus transcriptomic data from AD in DS. To date, no published spatial or single- nucleus studies have explored the concordance between these two populations, although AD in DS may serve as an advantageous group for preclinical and clinical studies of AD. We identified regional and cell-type specific transcriptomic changes shared between both AD populations. Further, we present a time-course analysis of the spatial transcriptome of the amyloid mouse model 5XFAD to assess the AD transcriptome across multiple brain regions and identify evolutionary-conserved gene expression changes. In addition to surveying different model systems, we integrated imaging data with spatial transcriptomic data. Amyloid beta (Aβ) pathology is one of the classical hallmarks of AD and proposed as a critical driver of AD pathogenesis (amyloid cascade hypothesis). We identified transcripts spatially localized to Aβ pathology conserved between both human and mouse. We also integrated spatial and single-nucleus transcriptomic data to discover spatially defined cell signaling pathways dysregulated with disease and highlight regional heterogeneity of astrocytes. Altogether, this dissertation reveals regional and cellular molecular changes occurring in AD and contextualizes them in a systems-level framework to uncover pathways for further study in AD.