Investigation of Stress Granule-Associated Biomarkers (IGF1 and CDK2) and Their Regulatory Mechanisms in Alzheimer’s Disease: A Combined Transcriptomic, Bioinformatic and Experimental Study

Technical Summary: Investigation of Stress Granule-Associated Biomarkers (IGF1 and CDK2) in Alzheimer's Disease

Problem Statement
Alzheimer's disease (AD) is the primary cause of dementia globally, characterized by amyloid-β (Aβ) plaques and neurofibrillary tangles (NFTs). Despite the centrality of Aβ and tau pathology, early detection remains hindered by a lack of specific biomarkers, and current therapies offer only symptomatic relief. Recent evidence suggests that Stress Granules (SGs)—dynamic, membraneless condensates formed during cellular stress—play a role in AD pathogenesis, potentially facilitating tau oligomerization and synaptic dysfunction. However, the regulatory circuitry linking SGs to AD and the identification of specific SG-associated biomarkers remain insufficiently defined. This study aims to bridge this gap by identifying SG-related transcriptional biomarkers in AD using a combined bioinformatic and experimental approach.

Methodology
The study employed a multi-stage pipeline integrating public transcriptomic data with experimental validation:

1. Data Acquisition and Preprocessing: Two peripheral blood transcriptomic datasets (GSE97760 as the training cohort and GSE140831 as the validation cohort) were retrieved from the GEO database. A list of 844 Stress Granule-Related Genes (SGRGs) was compiled from GeneCards.
2. Candidate Gene Screening: Differentially expressed genes (DEGs) were identified in the training set (|log₂FC| > 0.5, p < 0.05) and intersected with the SGRG list to yield 287 candidate genes (CGs).
3. Network and Functional Analysis: Protein-Protein Interaction (PPI) networks were constructed using STRING and Cytoscape to identify key nodes. Gene Ontology (GO) and KEGG enrichment analyses were performed to characterize biological functions.
4. Machine Learning Screening: Two algorithms, LASSO regression and Support Vector Machine-Recursive Feature Elimination (SVM-RFE), were applied to the training set to select feature genes. The intersection of these algorithms identified seven core genes.
5. Diagnostic Modeling: A Fully Connected Neural Network (FCNN) was constructed to evaluate the diagnostic performance of the selected biomarkers. Receiver Operating Characteristic (ROC) curve analysis was used to assess sensitivity and specificity.
6. Mechanistic Exploration: Gene Set Enrichment Analysis (GSEA), immune infiltration profiling (CIBERSORT), and molecular network regulation (miRNA-lncRNA-mRNA) were conducted. Molecular docking was performed to explore potential ligand-protein interactions.
7. Experimental Validation: Reverse transcription quantitative PCR (RT-qPCR) was performed on 10 clinical samples (5 AD, 5 controls) from the Fifth Affiliated Hospital of Xinjiang Medical University to verify biomarker expression.

Key Results

• Biomarker Identification: Machine learning analysis refined the candidate genes to seven features, from which IGF1 and CDK2 were retained as core biomarkers based on consistent expression trends and diagnostic performance (AUC > 0.7).
• Expression Patterns: Across datasets, IGF1 was consistently downregulated, while CDK2 was upregulated in AD patients compared to controls (p < 0.05).
• Diagnostic Performance: The FCNN model utilizing IGF1 and CDK2 achieved an AUC of 0.989 in the training set and 0.773 in the external validation set, demonstrating strong discriminative ability.
• Regulatory Mechanisms:
  • miRNA Interactions: Regulatory networks identified specific miRNA associations: IGF1 with hsa-miR-19b-3p and CDK2 with hsa-miR-302a-3p.
  • Pathway Enrichment: Both biomarkers were enriched in the KEGG T-cell receptor (TCR) signaling pathway, alongside metabolic pathways (alpha-linolenic acid and linoleic acid metabolism).
  • Immune Infiltration: AD samples showed decreased M0 macrophages, neutrophils, and Tregs, and increased activated CD4+ memory T cells. IGF1 correlated positively with M0 macrophages and Tregs, whereas CDK2 correlated negatively with Tregs.
• Molecular Docking: Computational docking predicted binding affinities of -7.3 kcal/mol for IGF1 with N,N-bis(3-(D-gluconamido) propyl) deoxycholamide and -11.2 kcal/mol for CDK2 with Alvocidib.
• Experimental Validation: RT-qPCR confirmed the significant downregulation of IGF1 in clinical AD samples. However, the upregulation of CDK2 was not statistically replicated in this specific clinical cohort, a discrepancy the authors attribute to potential peripheral blood detection limitations or cohort heterogeneity.

Significance and Claims
The study claims to provide preliminary evidence linking SG-related regulatory networks to AD pathogenesis, moving beyond the traditional Aβ/tau framework. By identifying IGF1 and CDK2 as potential SG-associated biomarkers, the work suggests a "growth factor-cell cycle-immunity axis" that may drive neuroinflammation and disease progression. The authors posit that the combined biomarker pattern (CDK2↑ + IGF1↓) could serve as a basis for non-invasive early diagnosis.

The paper modestly frames its findings as "preliminary clues" for diagnostic markers and therapeutic targets. It explicitly states that the molecular docking results are exploratory and should not be interpreted as pharmacological recommendations without further preclinical validation. The authors acknowledge limitations, including the restriction to peripheral blood, potential selection bias in public datasets, and the lack of replication for CDK2 in the clinical validation cohort. They conclude that further validation in larger, multicenter cohorts and at the protein level is required to confirm the clinical utility of these biomarkers.