Amanda J. Myers, Ph.D. - US grants

This subproject is one of many research subprojects utilizing the resources provided by a Center grant funded by NIH/NCRR. The subproject and investigator (PI) may have received primary funding from another NIH source, and thus could be represented in other CRISP entries. The institution listed is for the Center, which is not necessarily the institution for the investigator. Although late onset Alzheimer's disease (LOAD) is one of the most common neurodegenerative disorders its molecular etiology is far from being completely understood. In addition LOAD does not show any obvious inheritance pattern, complicating the diagnostics based on genetic background. To study the molecular mechanisms of the LOAD and potential links to the genetic background we have performed whole genome SNP genotyping using the Affymetrix 500K chips and transcriptome expression analysis using the Illumina ref-seq 8 chips on a series of 193 neuropathologically normal human brains and 177 samples from LOAD brains. So far we have analyzed and described the correlation of SNP polymorphism with gene expression profiles for brains from normal individuals and currently in the process of examining the LOAD series to look for DNA variants controlling RNA expression that might be involved in disease processes. However, it is very important to analyze this data in conjunction with protein abundance measurements, as the main molecular hallmark of Alzheimer's disease is the protein aggregation, which is pointing to dis-regulation of protein biosynthesis and degradation. The protein aggregates: senile plaques and neurofibrillary tangles predominantly consist of amyloid beta protein. However the recent LC-MS/MS proteomic profiling studies of senile plaques and neurofibrillary tangles obtained by laser capture microdissection have shown that those protein aggregates are more complex with estimated number of proteins ~25 and ~60, respectively. We propose for the current study to quantitatively analyze the human proteome with high throughput nano-LC FTICR MS in a limited subset normal and LOAD brains to investigate the potential links between genetic polymorphism, gene expression profiles, and protein abundance profiles with emphasis on protein aggregation. The quantitative proteome profiling is going to be a highly valuable addition to already collected data on genome and transcriptome profiling of normal and LOAD human brains.

Description (provided by applicant): Our Human Brainome project seeks to define the genome-transriptome-proteome- phenome interactions in the cortexes of normally aged human brains and brains affected by neurodegenerative disease. We hypothesize that the current accepted approach for discovering novel genetic risk loci by looking at a single layer of information (genotypes) is lacking power and taking a more systems-wide approach might increase the success of finding novel targets. We intend upon comparing our genotype information (~ 1.8 million single nucleotide polymorphisms) and our expression information (~ 46,000 transcripts) with a novel proteomics dataset generated by running Liquid Chromatography coupled online with high mass accuracy Mass Spectrometry (LC- MS, providing quantification of ~2000-3000 proteins). We will look at both single correlative cis and trans relationships (i.e. DNA change affects downstream regulation of one transcript or protein), as well as perform analyses to understand the networks of relationships occurring both at the transcriptome and proteome level. We are well situated to perform this work. First, we have an extensive collection of frozen human brain samples (n~1500, ~60% late onset Alzheimer's disease samples) for which there is genotype and expression data available. Large sample sizes are needed to obtain sufficient power to accurately assess the human genome as well as overcome some of the noise and other issues inherent to transcriptomics and proteomics sample analysis. These existing genotype and gene expression datasets are essential to success in this grant. Second, to accomplish the proteome analyses we will utilize the accurate mass and time (AMT) tag approach developed at the Pacific Northwest National Laboratories to avoid the sensitivity constraints of conventional approaches and improve the throughput of measurements providing broad proteome coverage. While having the same coverage of the proteome, the AMT tag approach typically reduces by 1-2 orders (e.g. 1 hour vs normally 24 hours) of magnitude the instrument time per sample analysis. Thus, the AMT tag approach is the only reasonable option to provide the sensitivity and measurement throughput essential to this project. Finally, we expect to achieve an additional 10-fold increase in sensitivity using our novel de-noising algorithm that will allow for a more accurate assessment of the complete proteome of the human brain cortex. By developing a more global view of the processes involved in human brain expression we will be able to relate new genetic findings to their downstream neuro-pathobiological relevance. This should aid in the development of novel genetic and molecular biomarkers of neurodegenerative disease. Identifying biomarkers that could further classify pre-clinical subgroups and identify sub-classes of rapid converters would help to significantly reduce the cost of drug trials. These biomarkers will have the added benefit that they are not only molecular, but in addition have mapped genotype profiles, which should be easier to assay than a molecular marker.

DESCRIPTION (provided by applicant): Recently, several genome-wide association studies (GWAS) have been published for late onset Alzheimer's disease (LOAD). Each GWAS that has been published to date has mapped the Apolipoprotein E (APOE) locus as the strongest LOAD risk signal within the human genome. This has led us and others to hypothesize that this large signal can 1. Overwhelm other smaller effects that are in epitasis with the APOE E4 locus and 2. Be de-convoluted into multiple risk loci mapping to the same area of the genome. The first hypothesis is plausible, considering that in our previous GWAS we mapped an effect that was only present in APOE E4 positive individuals. The second hypothesis is also valid in that this type of effect (i.e. multiple risk loci mapping to the same region of the genome) has been seen in other neurological diseases. For example, for Fronto-temporal Dementia Linked to Chromosome 17 (FTDP-17), mutations in both the microtubule associated protein Tau gene (MAPT) as well as Progranulin (PRGN) have been found. Both MAPT and PRGN map within the same linkage peak on chromosome 17. Thus, we propose to leverage our genome-wide and Next Generation Sequencing (NGS) genetic data, as well as transcriptome and proteome datasets to map novel risk loci for LOAD that are acting either in epistasis with or independently of the APOE E4 allele. We propose the following: To test our APOE E4 independent effects, we will perform additional NGS sequencing within the same region of chromosome 19 to capture additional effects (Aim 1a). We will also genotype the variants we found from our NGS in additional case control samples to determine whether they act independently of APOE E4 to increase risk for disease (Aim 1b). To follow our APOE E4 epistatic effects, SNPs which we found to act in conjunction with APOE E4 will be followed by examining an additional cohort (Aim 2a) as well as sequencing within the region to find additional variants (Aim 2b). Finally, we will map the downstream effects of any variants we map in Aims 1 or 2 by examining transcript expression and protein profiles (Aim 3), Our collaboration possesses the unique skills and datasets to perform this work. Drs. Huentelman and Myers have worked together for the greater part of their careers and have co-authored many publications using similar techniques as those proposed in this application. They have access to a unique cohort of ~ 1600 neuropathologically verified individuals, which will allow for both the analysis of risk variation as well as the downstream changes of those variants. They have recruited an additional cohort of ~ 18,000 clinically characterized samples from the University of Cardiff to replicate any effects. They also have access to both the computational power (48-core / 576GB memory computer and a separate 2,700-core cluster through Tgen and one of 5000 CPU at the University of Miami) as well as the expertise to execute the bioinformatics analysis involved in all Aims.

PROJECT SUMMARY The goal of the NIA Research Career Enhancement Award is to bridge expertise gaps in data science and drug discovery by allowing Alzheimer's disease (AD) researchers to expand their expertise. This application will focus on obtaining both the skillset and preliminary data to transition targets from our innovative approach (The Human Brainome) to mapping risk loci for late onset Alzheimer's disease (LOAD). Rather than looking at a single layer of information as in most genome-wide association (GWAS) studies, our approach involves mapping genomic variation in the context of downstream transcriptomic and proteomic expression. This allows for not only mapping the crucial variation involved in LOAD, but also mapping downstream effects and their directions. Additionally, it allows for building networks of multiple players crucial for disease processes. What is crucial is to obtain models and systems for following the Brainome targets outside of the context of human tissues so that hypothesis that have been generated from the Brainome work can be further tested. These models are crucial for the development of drug screening, as we need manipulatable systems for testing no-go/go outcomes in any de novo drug screen. We proposed to use a novel, innovative approach to iPSC development, by using well-characterized control iPSC lines and both viral overexpression and CRISPR technology to insert pathogenic mutations. The innovation lies in our access to lines from autopsy subjects where pathologically confirmed brain tissue is available for confirmation of central profiles as well as the extensive output data from our screen of the Human Brainome cohorts (AG034504 and AG041232). In Aim 1, we will expand our toolkit of iPSC lines to include 3 additional control lines and 3 late onset Alzheimer's Disease (LOAD) lines. All lines will be stably modified with Cas9, so that CRISPR technologies can be exploited in future work. In Aim 2, the top target from the Brainome screen will be introduced into an additional line from Aim 1 to determine whether our preliminary results replicate. These research aims will be combined with a drug development didactic component with our final project goal of obtaining the necessary skills and experience to apply for further funding to develop these models. This project will be unique from prior iPSC modeling for the following reasons: first, developing lines from postmortem confirmed donors will increase rigor, reproducibility and transparency because background effects and baselines can be determined by chasing back to brain tissue and second, the targets for Aim 2 are innovative since they have never been studied in this context. While to date, the drug development space for LOAD has been rife with failures, given the estimated human health cost of $20 trillion over the next 40 years, it is necessary to continue to pursue treatments. Our targets from the Brainome screen present a unique opportunity to follow outcomes from a hypothesis-free screen in an internationally-based, multiple stage screen of appropriate human brain tissues.

PROJECT SUMMARY Previously, we have taken an innovative approach (The Human Brainome; [1-6]) to mapping risk loci for late onset Alzheimer's disease (LOAD). Rather than looking at a single layer of information as in most genome- wide association (GWAS) studies, we have mapped genomic variation in the context of downstream transcriptomic and proteomic expression. This allows for mapping both the crucial variation involved in LOAD, as well as the downstream effects and their directions. Additionally, it allows for building networks of multiple players crucial for disease processes. One shortcoming of the current work is that we have mapped DNA-expression relationships that are subtly changed in Alzheimer's, but we have yet to fully understand why those DNA-expression relationships are altered. We know that expression is altered by specific alleles, but there must be added regulation given our mapped outputs. One target that can alter pathways are natural antisense transcripts (NATs), which can bind to oligonucleotide products and alter their expression and degradation. In our application, we propose to use long read sequencing technology (SMRT; Single Molecule, Real-Time) and fully profile RNA from our human brain bank samples. We will examine where these outputs are located and perform preliminary work to determine if any of these new hits can act on our existing results. We propose to follow these targets through 3 Aims. Aim 1 will involve following hits from public databases. Aim 2 will involve collecting additional RNA profiling data. Finally, Aim 3 will seek to validate and order all novel findings from Aims 1 and 2. It is important to use technologies appropriate to our hypothesis for the new data collection. The majority of non-coding RNA belongs to the class of transcripts called long non-coding RNA (lncRNA), which can span from 1000-10,000 bp [7]. Typical short-read RNA sequencing (SRS) technologies are based on the capture of short sequences of ~150 bp, and therefore, SRS has difficulty in capture and alignment of longer products. We are working with Robert Sebra at the Icahn Institute for Genomics and Multicale Biology, who is an expert in SMRT sequencing [8]. This technology offers longer read lengths and will be unique-in-field, since most human RNA profiling involves SRS. By the completion of these Aims, we will have 1. A map of novel long read sequencing in human brain tissues, which will be a significant add-to-field, given most technologies used to date are focused on short read sequencing, and there is limited profiling in pathologically defined human brain tissues, 2. An understanding of how these novel hits are affecting both the direct sense transcript of interest as well as the known LOAD pathologies, 3. Multi-level mapping of rigor and reproducibility of targets through the use of multiple capture techniques and 4. Validation of the effect of hits on the known LOAD pathogenic targets by measuring both expression and protein levels in cell culture.