Peripheral DNA methylation patterns, methylation age and Alzheimer’s disease risk and related phenotypes

Author Identifier

Lidija Milicic

Date of Award


Document Type



Edith Cowan University

Degree Name

Doctor of Philosophy


School of Medical and Health Sciences

First Supervisor

Simon M Laws

Second Supervisor

Tenielle Porter

Third Supervisor

Michael Vacher



Dementia, including Alzheimer’s disease (AD), is a significant global health challenge, ranked as the seventh leading cause of death with nearly 10 million new cases annually. Despite recent advancements in anti-amyloid therapies such as aducanumab, lecanemab and donenamab, targeting AD in its symptomatic stages has often yielded limited success due to advanced neurodegeneration. Early intervention is crucial for effective treatment, making precise early diagnosis imperative. The current diagnostic process for AD often lacks biomarker support, leading to misdiagnosis rates of around 25-30% in primary care settings. Blood based biomarkers are emerging as a promising solution to enhance AD diagnosis and optimise clinical trial design. These biomarkers can detect core AD pathological indicators like A and phosphorylated tau, as well as neurodegeneration markers such as neurofilament light chain (NfL) and markers of neuroinflammation, such as Glial Fibrillary Acidic Protein (GFAP). Blood-based biomarkers offer the advantages of being minimally invasive, cost-effective, and accessible compared to costly amyloid PET and CSF biomarkers. DNA methylation has gained attention as a potential biomarker due to its association with various diseases, including some neurodegenerative diseases, including AD. While brain tissue studies have identified associations between DNA methylation and AD pathology, translating these findings to peripheral blood remains a challenge, meaning the relationship between brain and blood methylation patterns has not yet been fully elucidated.


The overarching aim of this thesis was to identify markers of DNA methylation associated with AD risk and related phenotypes, such as cognition and neuroimaging modalities. Specifically, this thesis can be divided into two broad aims. Chapters 2 and 3 aim to investigate the association of DNA methylation clocks with AD-related traits and Chapters 4 and 5 aim to investigate the relationship between DNA methylation sites and AD-related traits, using both a targeted (Section 4.2) and an exploratory approach (Section 5.2). Initially, our focus was on evaluating the relationship between methylation age and AD-related phenotypes, with a specific emphasis on the use of DNA methylation clocks, as covered in Chapters 2 and 3. Chapter 2 focussed on using first- (Section 2.2) and second-generation (Section 2.3) clocks to assess whether accelerated ageing is associated with cross-sectional measures of cognition and pathological changes in the brain and determine if an individual’s current methylation age is an indicator of future changes in cognition and brain pathological features. It was then aimed to assess if methylation profiles within a priori candidate AD associated risk genes are associated with cross-sectional measures of cognition and pathological changes in the brain (Chapter 4). Finally, we aimed to undertake epigenome wide association studies with respect to AD-related phenotypes and to develop trait-specific methylation risk scores and test their association with AD-risk and related phenotypes (Section 5.2).


All studies presented in this thesis utilised data from two well-characterised AD cohorts; the discovery cohort; the Australian Imaging, Biomarkers and Lifestyle Study of Ageing (AIBL) and the validation cohort; the Alzheimer’s Disease Neuroimaging Initiative (ADNI). Both the AIBL and ADNI are longitudinal cohort studies that have collected data at intervals over an extended period of greater than 15 years of follow up. All individuals in this thesis, across both cohorts, have available DNA methylation data as generated on the Illumina Infinium HumanMethylation EPIC 850k Chip. Chapters 2 and 3 consisted of assessing the relationship between first- and second-generation clocks with Positron emission topography (PET) derived measures of amyloid- (A) burden, MRI volumetric measures (white matter volume, grey matter volume, hippocampal volume and ventricle volume), cognition (the Pre-Alzheimer’s Cognitive Composite (PACC)) and age at onset of pathological levels of A in the brain (Α ΑAO ). Linear regressions were utilised in Chapter 2 to assess the association between accelerated age and trait outcomes. Cox proportional hazards regressions were utilised in Chapter 3 to assess the time taken to reach an event, where the event of interest was the occurrence of A AAO . For both Chapter 4 and 5, linear regressions were used to assess the relationship between CpG sites and AD-related traits. Chapter 4 relied on a targeted approach for analyses, whereas Chapter 5 utilised an unbiased discovery approach. Further, in Chapter 5, trait-specific methylation risk scores (MRSs) were developed, which relied on utilising the summary statistics from the EWAS to generate MRSs at 6 p-value thresholds and one machine learning based risk score, trained using Elastic Net.


Accelerated biological ageing shows no associations with cross-sectional and longitudinal measures of cognition, brain A burden, or longitudinal measures of brain volume; however, it is associated with cross-sectional measures of brain volume. In cognitively unimpaired individuals with a high amyloid- burden (CU A high ), an inverse association was observed with lower hippocampal volumes seen in those who were experiencing accelerated ageing, measured by the Hannum clock (Chapter 2; Section 2.2).

Chapter 2, Section 2.3, focussed on the same analyses with the addition of second-generation clocks and one extra outcome variable, A AAO . Here, it was observed that accelerated ageing, measured by the PCPhenoAge clock, was associated with hippocampal volume in the whole cohort, reflecting the association seen in Section 2.2. The results were replicated in ADNI, however the effect observed was in the opposite direction. Further, when investigating the utility of second-generation clocks in the prediction of A AAO , associations in AIBL after FDR correction were observed, however these were not validated in ADNI (Chapter 3).

When investigating individual CpG sites from genes previously associated with AD (Chapter 4), we identified numerous associations in the AIBL cohort that were significant after correction for the FDR, although only one novel site was validated in the ADNI cohort. This was located within the gene body region of CR1 (cg00416522) and was associated with ventricle volume in CU A high . Finally, when considering the genome in a set of exploratory epigenome wide association analyses, no individual CpG sites remained significant after correction for the FDR, however, as expected, several MRSs were associated with disease risk.


The work presented in this thesis provides an in-depth investigation of peripheral DNA methylation patterns, methylation age and AD risk and related phenotypes. The results here suggest that while DNA methylation potentially has a role in AD development, further investigation is required to fully elucidate its mechanistic role. Whilst this thesis provides some evidence to support the involvement of DNA methylation in AD risk and related traits and adds to a growing body of literature aiming to describe the utility of DNA methylation as a peripheral biomarker, it also highlights several limitations and proposes future areas for research focus.



Access Note

Access to this thesis is not available

Access to this thesis is restricted. Please see the Access Note below for access details.