Pathway enrichment in genome‐wide analysis of longitudinal Alzheimer's disease biomarker endophenotypes

Thea Rosewood, Kwangsik Nho, Shannon L. Risacher, Shiwei Liu, Sujuan Gao, Li Shen, Tatiana M. Foroud, Andrew J. Saykin, for the Alzheimer's Disease Neuroimaging InitiativeAlzheimer 2024 · DOI

TL;DR

This study leverages genome-wide association studies (GWAS) on longitudinal Alzheimer's disease (AD) biomarker data to identify genetic pathways influencing disease progression. By analyzing how genetic variations correlate with changes in key AD indicators over time, the research uncovers novel biological pathways that could be targeted for future therapeutic interventions.

What Problem Does This Paper Address?

Alzheimer's disease (AD) is characterized by significant heterogeneity in its progression, meaning individuals experience the disease differently. This variability makes it challenging to understand the underlying biological mechanisms and develop effective treatments. While previous studies have identified genetic associations with AD, many have focused on cross-sectional data, capturing a single snapshot in time. This paper aims to address this gap by investigating how genetic variations influence the longitudinal trajectory of AD biomarkers, providing a more dynamic view of disease progression and its genetic underpinnings. The researchers are looking for genetic pathways that not only correlate with AD but also explain how the disease changes over time in individuals.

Methods

The study employed a comprehensive genome-wide association analysis (GWAS) using a linear mixed model (LMM) approach. This statistical method is well-suited for analyzing longitudinal data, where individuals are measured multiple times, accounting for the correlation between repeated observations.

Here's a breakdown of their approach:

  • Endophenotype Selection: They focused on 11 key AD biomarkers, categorized by the NIA-AA research framework: amyloid-beta (Aβ) deposition (A), tau pathology (T), neurodegeneration (N), and cognition (C). These included measures like PET scans for amyloid and tau, cerebrospinal fluid (CSF) levels of Aβ and tau, MRI-based atrophy measures, and cognitive composite scores.
  • Genome-Wide Association Study (GWAS): For each endophenotype, they performed a GWAS using LMMs. This involved analyzing millions of single nucleotide polymorphisms (SNPs) to identify genetic variants associated with the intercept (baseline level) and the longitudinal trajectory (rate of change over time) of these biomarkers. They specifically looked for "SNP by Time" (SNPxTime) interactions to pinpoint genetic effects that influence how biomarkers change over time.
  • Gene and Pathway Enrichment: To interpret the vast number of significant SNPs, they used a two-step SNP-to-gene assignment method. This involved identifying SNPs located within gene boundaries and those with significant expression quantitative trait locus (eQTL) effects. These genes were then analyzed for enrichment in biological pathways using the MAGMA software and curated gene set collections (like those from GSEA MSigDB). This allowed them to move beyond individual SNPs to understand the broader biological processes implicated.
  • Statistical Thresholds: They used a stringent study-wide significance threshold (p ≤ 8.33 x 10⁻⁹) for identifying key associations, along with more conventional genome-wide (p ≤ 5 x 10⁻⁸) and suggestive (p ≤ 1 x 10⁻⁵) thresholds for broader interpretation.

Datasets

The primary dataset used in this study was the Alzheimer's Disease Neuroimaging Initiative (ADNI). Specifically, they utilized data from ADNI Phase 1 and its subsequent extensions (ADNI-GO/2/3).

  • ADNI: This is a well-established, large-scale, multi-site longitudinal study that collects clinical, imaging, and biomarker data from individuals with normal cognition, mild cognitive impairment (MCI), and AD. The ADNI database is a crucial resource for AD research, providing rich longitudinal data that enables studies like this one.
  • Genotyping Data: Participants were genotyped using various Illumina platforms.
  • Imputation: Genotype imputation was performed using the Haplotype Reference Consortium Panel r1.1 to increase the density of genetic markers.
  • Population Stratification Control: To mitigate confounding from population differences, only non-Hispanic European ancestry participants were selected, and genetic clustering was performed using HapMap 3 data and multidimensional scaling (MDS).

The paper does not introduce new benchmark datasets.

Key Insights and Ideas

  • Longitudinal Trajectory Matters: The study strongly emphasizes that analyzing genetic associations with longitudinal changes in AD biomarkers provides unique insights beyond cross-sectional analyses. This dynamic perspective is crucial for understanding disease progression.
  • Pathway-Level Understanding: Moving from individual SNPs to enriched pathways reveals broader biological themes. This approach is more powerful for identifying actionable targets than focusing on single genetic variants.
  • Immune System's Central Role: Immune-related pathways were consistently and strongly associated with both baseline biomarker levels (intercept) and their changes over time (trajectories), highlighting inflammation as a key player throughout AD progression.
  • Metabolic Pathways Link to Multiple Biomarkers: Metabolic pathways showed broad associations across various endophenotypes, including MRI, cognitive, and amyloid measures, suggesting a significant interplay between metabolism and AD pathology.
  • APOE's Broad Influence: The APOE ε4 allele, a well-known AD risk factor, showed significant associations with baseline levels of multiple biomarkers, underscoring its foundational role in AD pathogenesis.
  • Specific Pathways for Trajectory: The SNPxTime analysis identified pathways that are specifically linked to the rate of change in biomarkers, offering a more nuanced view of how genetics influences disease dynamics. For example, glycosylation pathways were linked to changes in brain atrophy.

Key Results

  • 33 pathways were identified as significantly enriched for the intercept (baseline levels) of AD endophenotypes. These pathways fall into eight categories: immune, metabolic, cell growth and survival, DNA maintenance, neuronal signaling, RAS/MAPK/ERK signaling, vesicle and lysosomal transport, and transcription modification.
  • 19 pathways were identified as significantly enriched for the longitudinal trajectory (changes over time) of AD endophenotypes. These pathways fall into five categories: immune, metabolic, cell signaling, cytoskeleton, and glycosylation.
  • APOE ε4 allele was a significant SNP for the intercept across multiple endophenotypes.
  • The study identified several novel SNPs associated with longitudinal changes, including one intronic to RNF152 (linked to tau progression) and others in the HLA Class II region (heavily implicated in AD).
  • Specific pathways like "white adipocyte differentiation" showed cross-endophenotype association for intercept, while "conjugation of benzoate with glycine" was associated with amyloid and MRI measures. For SNPxTime, pathways like "PI3K events in ERBB4 signaling" were linked to neurodegeneration, and glycosylation pathways were broadly associated with changes in atrophy, cognition, and tau.

Limitations and Future Work

The authors acknowledge several limitations:

  • Sample Size: While using ADNI is a strength, GWAS studies are inherently limited by sample size, especially when analyzing complex longitudinal data and multiple phenotypes.
  • Missing Data and Visit-to-Visit Variability: Longitudinal studies often have missing data points and variability in measurements between visits. While LMMs help, these factors can still introduce limitations.
  • Participant Population: The study focused on non-Hispanic European ancestry participants to control for population stratification. This limits the generalizability of findings to other ethnic groups, though efforts are underway to increase diversity in AD research.
  • Tau-PET Data: Tau-PET data were available in ADNI-3 but were not sufficiently powered for genetic analysis at the time of the study.
  • Age as Time Variable: Using age at observation as the time variable is a simplification; alternative models for temporal alignment could offer more power.
  • Age of Onset: Precise age of onset data is often lacking, which can affect the modeling of trajectories.

Future work suggested includes:

  • Larger and Independent Samples: Replicating findings in larger, independent cohorts is crucial.
  • Improved Temporal Modeling: Developing more sophisticated models for temporal alignment and subgroup identification could enhance understanding of longitudinal trajectories.
  • Functional Validation: Further studies, including cell models and animal models, are needed to precisely determine the functional effects of the identified pathways and genetic variants.
  • Multi-ethnic Studies: Expanding AD genetic research to include diverse populations is essential for broader applicability.
  • Incorporating New Data: As more data (e.g., tau-PET) become available in ADNI and other datasets, longitudinal analyses incorporating these measures will be valuable.

Summary updated

Abstract

A systematic genome-wide analysis with longitudinal AD biomarker endophenotypes was performed. Enriched pathways were identified with functionally derived SNP to gene analysis. Fifty-two pathways were associated with longitudinal trajectory and intercept. Many of the identified pathways are specific steps in larger pathways implicated in AD. The identified pathways may provide therapeutic targets and areas for further study.
BibTeX
@inproceedings{rosewood2024pathway,
  author = {Thea Rosewood and Kwangsik Nho and Shannon L. Risacher and Shiwei Liu and Sujuan Gao and Li Shen and Tatiana M. Foroud and Andrew J. Saykin and for the Alzheimer's Disease Neuroimaging Initiative},
  title = {{Pathway enrichment in genome‐wide analysis of longitudinal Alzheimer's disease biomarker endophenotypes}},
  booktitle = {Alzheimer},
  year = {2024},
  doi = {10.1002/alz.14308},
}
Figure