Analysis of brain region-specific co-expression networks reveals clustering of established and novel genes associated with Alzheimer disease

Daniel Lancour, Josée Dupuis, Richard Mayeux, Jonathan L. Haines, Margaret A. Pericak‐Vance, Gerard C. Schellenberg, Mark Crovella, Lindsay A. Farrer, Simon KasifAlzheimer 2020 · DOI

TL;DR

This study leverages gene co-expression networks and genome-wide association study (GWAS) data to identify novel genes associated with Alzheimer's disease (AD). By analyzing gene expression across different brain regions, they found that established AD genes are more interconnected in brain areas most affected by the disease, suggesting region-specific genetic influences.

What Problem Does This Paper Address?

Alzheimer's disease (AD) is a complex neurodegenerative disorder where genetic risk factors play a significant role. However, understanding these genetic influences has been challenging due to the variability of gene expression across different brain regions and the confounding effects of age-related neurodegeneration. Previous studies have often treated the brain as a single entity, potentially overlooking crucial region-specific patterns of gene interaction that might be key to AD pathogenesis. This paper aims to address this gap by investigating how gene co-expression networks differ across brain regions and how these differences relate to established AD risk genes, with the ultimate goal of identifying novel AD-associated genes.

Methods

The researchers employed a multi-pronged approach:

  • Gene Co-expression Network Construction: They built gene co-expression networks using microarray data from the Allen Brain Atlas, focusing on three high-level brain structures: cerebrum, cerebellum, and brain stem. To account for the known progression of AD pathology, the cerebrum was further divided into "early-stage" and "late-stage" regions based on the timing of protein aggregation.
  • Network-Based Ranking: They started with a set of 52 established AD (RAD) genes. They then calculated the correlation between these RAD genes and all other genes within each brain region network. A novel metric, "median ranking by correlation" (MRC), was used to normalize these correlations and allow for comparisons across different networks.
  • Integration with GWAS Data: Summary statistics from a large GWAS for AD risk were integrated with the network-derived scores. This was done using a modified Stouffer method to combine genetic association Z-scores and network scores, creating a "combined score" for ranking genes.
  • Phenotypic Validation (Drosophila): To further refine their candidate gene list, they incorporated phenotypic information from gene knockout studies in Drosophila. Genes whose knockouts in flies mimicked AD-related phenotypes (e.g., defective memory, aging, oxidative stress) were prioritized.

Datasets

  • Allen Brain Atlas (ABA): Microarray data from six neuropathologically healthy human brains (ages 24-57) were used for gene expression analysis across multiple brain regions.
  • Genome-Wide Association Study (GWAS) Data: Summarized results from a GWAS for AD risk involving 16,175 AD cases and 17,175 controls of European ancestry were utilized.
  • Drosophila Phenotype Data: Information from gene knockout studies in Drosophila, specifically focusing on phenotypes related to aging, memory, and oxidative stress, was used for functional validation.

Key Insights and Ideas

  • Brain Region Specificity Matters: The study strongly suggests that AD-related genetic risk factors are not uniform across the brain. Gene co-expression patterns are significantly different between brain regions, with higher co-expression of established AD genes observed in regions most severely affected by AD (like the cerebrum).
  • Clustering of Established AD Genes: Established AD genes tend to cluster together more strongly in the cerebrum compared to the cerebellum and brain stem. This supports the idea that AD pathology is driven by coordinated gene networks that are more active or interconnected in vulnerable brain areas.
  • Novel Gene Discovery Potential: By combining network analysis with GWAS data and functional validation, the study successfully identified potential novel genes associated with AD. This integrative approach is more powerful than analyzing genetic association or gene expression in isolation.

Key Results

  • Higher Correlation in Cerebrum: The correlation rank (CR) of RAD genes was significantly higher in the cerebrum (0.748) compared to the brain stem (0.648) and cerebellum (0.574).
  • Novel AD Genes Identified: The study identified two novel AD-related genes: EPS8 (FDR p = 8.77 × 10⁻³) and HSPA2 (FDR p = 0.245). EPS8, in particular, showed a strong association.
  • Consistency Across Brains: Pairwise correlations of RAD co-expression networks among the six individual brains showed moderate to high consistency, suggesting that the observed patterns are robust.
  • Phenotypic Linkages: Several top-ranked genes, including EPS8, were linked to AD-related phenotypes through Drosophila knockout studies, providing functional support for their candidacy.

Limitations and Future Work

  • Small Sample Size: The study used a limited number of individuals (six brains) for gene expression analysis, which could affect the statistical power and generalizability of findings.
  • Absence of AD Brains: The study utilized brains from neurologically healthy individuals. The authors acknowledge that comparing gene expression patterns between healthy and AD brains, and correlating them with disease severity, is crucial for future work.
  • Multiple Testing Correction: While FDR was applied, the relatively small sample size meant that only a few genes remained significant after correction, highlighting the need for larger datasets.

Future work should focus on analyzing expression in more fine-grained brain regions, using larger sample sizes, and importantly, including brains from individuals with pathologically confirmed AD at various stages. This will allow for more definitive conclusions about the role of region-specific gene co-expression in AD progression and the development of targeted therapeutic strategies.

Summary updated

Abstract

Our findings indicate that AD-related risk factors are potentially associated with brain region-specific effects on gene expression that can be detected using a gene network approach.
BibTeX
@inproceedings{lancour2020analysis,
  author = {Daniel Lancour and Josée Dupuis and Richard Mayeux and Jonathan L. Haines and Margaret A. Pericak‐Vance and Gerard C. Schellenberg and Mark Crovella and Lindsay A. Farrer and Simon Kasif},
  title = {{Analysis of brain region-specific co-expression networks reveals clustering of established and novel genes associated with Alzheimer disease}},
  booktitle = {Alzheimer},
  year = {2020},
  doi = {10.1186/s13195-020-00674-7},
}
Figure