Abstract
Alzheimer’s Disease (AD) is characterized by its complex and heterogeneous etiology and gradual progression, leading to high drug failure rates in late-stage clinical trials. In order to better stratify individuals at risk for AD and discern potential therapeutic targets, we employed a novel procedure utilizing cell-based co-regulated gene networks and polygenic risk scores (cbPRSs). After defining genetic subtypes using extremes of cbPRS distributions, we evaluated correlations of the genetic subtypes with previously defined AD subtypes defined on the basis of domain-specific cognitive functioning and neuroimaging biomarkers. Employing a PageRank algorithm, we identified priority gene targets for the genetic subtypes. Pathway analysis of priority genes demonstrated associations with neurodegeneration and suggested candidate drugs currently utilized in diabetes, hypertension, and epilepsy for repositioning in AD. Experimental validation utilizing human induced pluripotent stem cell (hiPSC)-derived astrocytes demonstrated the modifying effects of estradiol, levetiracetam, and pioglitazone on expression of APOE and complement C4 genes, suggesting potential repositioning for AD.
Introduction
Alzheimer’s Disease (AD) remains one of the most significant challenges in modern medicine, largely due to its profound biological heterogeneity. While Genome-Wide Association Studies (GWAS) have identified over 75 genetic risk loci, translating these findings into effective therapies has been hindered by a "one-size-fits-all" approach to drug development. This lack of precision contributes to a staggering 99% failure rate in clinical trials. To address this, researchers are moving beyond whole-genome analysis toward cell-specific genetic subtyping, aiming to understand how genetic risk manifests within the distinct environments of neurons, astrocytes, and oligodendrocytes.
Methodology: From Whole-Genome to Cell-Specific Risk
The study introduced a multi-stage computational and experimental pipeline designed to stratify patients based on cell-specific genetic risk.
1. Network Construction and cbPRS
Using single-nuclei RNA sequencing (snRNA-seq) from the ROSMAP and SWDBB datasets, the researchers identified co-expressed gene modules specific to seven brain cell types. They then developed Cell-Based Polygenic Risk Scores (cbPRS). Unlike standard PRS, which aggregates risk across the entire genome, cbPRS focuses exclusively on genetic variants located within these cell-specific networks. This allows for a more granular view of how an individual's genetic makeup affects specific biological "neighborhoods" in the brain.
2. PageRank Prioritization
To identify the most critical genes within these networks, the team applied the PageRank algorithm—the same logic used by search engines to rank web pages. By treating genes as nodes in a network, PageRank identified "hub" genes (such as APOE and GFAP) that are central to the structure and flow of information within disease-associated modules.
3. Drug Repositioning and Validation
Using the enrichr platform, the researchers screened for FDA-approved drugs capable of modulating these hub genes. The most promising candidates were then validated in a laboratory setting using human induced pluripotent stem cell (hiPSC)-derived astrocytes to observe their effects on key AD risk markers.
Key Insights and Results
The Glial Connection
The study found that the most significant genetic risk was concentrated in astrocytes (support cells) and oligodendrocytes (insulation for axons). Specifically, patients in the highest quartile of risk for the "Ast-M2" (astrocyte) and "Oli-M45" (oligodendrocyte) modules showed a Hazard Ratio (HR) greater than 4.0 for progressing from mild cognitive impairment to AD. This reinforces the theory that AD is not merely a "neuron problem" but a systemic failure of glial support networks.
Clinical and Neuroimaging Correlations
High cbPRS scores were not just statistical abstractions; they correlated strongly with physical biomarkers. Individuals with high cell-specific risk scores exhibited visible brain atrophy in the temporal lobe and hippocampus, alongside higher levels of amyloid deposition, as confirmed by ADNI and Framingham Heart Study data.
Therapeutic Candidates
The pipeline identified 54 candidate drugs for repositioning. Laboratory validation highlighted three notable compounds:
- $\\beta$-Estradiol (Hormone Therapy): Reduced the expression of APOE and complement genes C4A and C4B by more than two-fold.
- Levetiracetam (Anti-convulsant): Similarly reduced expression of these inflammatory and plaque-related markers.
- Pioglitazone (Diabetes Medication): Demonstrated a genotype-dependent effect, working effectively in cells with the APOE3 genotype but showing significantly reduced efficacy in APOE4 cells.
Discussion and Limitations
This research provides a roadmap for precision medicine in AD, suggesting that future clinical trials should stratify participants based on their cell-specific genetic profiles. However, several limitations remain. The initial gene networks were derived from a relatively small sample of 48 brains, and the genetic data is primarily of European ancestry, which may limit generalizability. Furthermore, the study focused on the prefrontal cortex; future research must account for regional variations in the brain, such as the entorhinal cortex, where AD pathology often begins.
Conclusion
By combining cell-specific transcriptomics with network science and experimental validation, this study identifies new pathways for AD intervention. The discovery that existing medications for epilepsy, diabetes, and hormone replacement can modulate AD-related gene expression offers a promising and accelerated path toward effective treatment for specific genetic subtypes of the disease.
Summary updated