Identifying genetic and molecular signatures of multiple sclerosis onset and progression

Multiple sclerosis (MS) is an inflammatory and neurodegenerative disease of the central nervous system. A significant and persisting challenge to diagnostic and therapeutic innovations for MS remains in identifying MS-specific biomarkers that can accurately capture the complex and heterogeneous pathophysiology of MS. Consequently, despite the increase in the number of MS disease-modifying therapies becoming available in recent decades, the search for a preventative agent or cure for MS is still ongoing. Multi-omics is a powerful approach for biomarker discovery through integrating in-depth biological (“multi-omics”) data generated from high?throughput technologies. However, there are few multi-omics studies in MS. In parallel with this, while twin studies have established a strong genetic component for MS, known MS risk variants can only explain a part of the heritability of MS supporting contributions from unknown forms of genetic variation. To address the knowledge gaps in the genetic and molecular basis of MS, this thesis employs a range of analysis methods incorporating large multi-omics datasets and clinical data to identify and characterise novel candidate biomarkers for MS. Firstly, Chapter 1 provides an overview of multi-omics research and its applications, as well as summarising relevant sections from a literature review that I conducted on the role of genetic structural variants (SVs) in clinical neurology. This chapter also outlines the rationale for each of the succeeding chapters and outlines the overall theme for this thesis. In total, this thesis aims to develop a better understanding of the genetic and molecular architecture of MS, and to identify important candidate MS biomarkers that are critical for improving diagnostics, defining biomarkers of disease progression. From this, the goal is to facilitate identification of potential points of therapeutic intervention, leading to precision medicine. Secondly, Chapter 2 examines novel plasma and brain proteins in MS genetic susceptibility. Since proteins are key players in most biological pathways, these are excellent targets for pharmacological interventions. In this study, I integrated plasma proteomics (n=3,301) and brain proteomics (n=376 discovery; n=152 replication) into MS genome-wide association studies (n=14,802 cases and 26,703 controls) by employing established summary-based methods to identify candidate proteins for MS risk. This led to the identification of 39 novel proteins associated with MS risk. These identified protein candidates were reinforced by high concordance in downstream transcriptomic analyses at tissue-level using large gene expression quantitative trait data from whole-blood (n=31,684) and brain (n=1,194) tissue. Further, I assessed and identified cell-specific transcriptional profiles for candidate proteins at cell-level, in immune cell types (dataset 1: n=73 cases and 97 controls; dataset 2: n=31 cases and 31 controls) for identified plasma proteins, and in brain cell types (dataset 1: n=4 cases and 5 controls; dataset 2: n=5 cases and 3 controls) for identified brain proteins. In an independent longitudinal cohort (AusLong study) of 203 MS cases with up to 15 years follow-up, we found gene-level associations for 23 of the candidates with disability worsening outcomes. Thirdly, Chapter 3 assesses and identifies causal metabolites as candidate biomarkers for MS. Metabolites are small molecules underlying the breadth of biological functions and which thus can be used to assess a wide spectrum of biological processes, but the associations of metabolites with MS risk and progression remains unclear. This study first used six Mendelian randomisation methods to assess causal relationship of 174 metabolites with MS, incorporating data from European-ancestry metabolomics (n=8,569-86,507) and MS (n=14,802 MS cases, 26,703 controls) genome-wide association studies. Genetic scores for identified causal metabolite(s) were then computed to predict MS disability progression in an independent longitudinal MS cohort (AusLong). From this study, we found a novel genetic causal effect of serine on MS onset (OR=1.67, P=1.73x10-20), such that individuals whose serine level was one standard deviation above the population mean had 1.67 times higher risk of developing MS. This finding was robust across all sensitivity methods applied (OR ranges from 1.49 to 1.67). In the AusLong study cohort, we then constructed time-dynamic and time-fixed genetic risk scores based on serine genetic instrument SNPs, where higher scores for raised serum serine level were associated with increased risk of disability worsening, especially in the time-dynamic model, with an estimated 25% increase in relative risk of disability worsening per unit increase in the genetic scores (RR=1.25, P=7.52x10-4). Fourthly, Chapter 4 investigates copy-number variants (CNVs), an understudied form of genetic variation in MS. CNVs are duplications and deletions of DNA segments that span greater than 50 base-pairs in the genome. In this study, I used the PennCNV software to call CNVs in three independent cohorts: ANZgene cohort (n=3,307 MS cases and 2,738 controls), UK Biobank (n=1,072 MS cases and 296,583 controls), and WTCCC2 cohort (n=9,772 MS cases and 527 controls). Based on the CNV calls, I first examined the global impact of CNVs in a meta-analysis across the three cohorts and found an estimated 18% increase in the odds of developing MS per 1Mb increase in CNV length burden (OR=1.18, P=6.3x10-5). Next, I encoded CNV-proxy probes and performed a genome-wide analysis of CNVs that identified a total of 30 independent novel genomic regions where CNV events were associated with MS risk (P