Multi-Omics Integration Reveals Three Potential Therapeutic Targets for Spina Bifida: NPC2, PCDH12 and IL10RB

Technical Summary: Multi-Omics Integration Reveals Three Potential Therapeutic Targets for Spina Bifida

Problem Statement
Spina bifida (SB), a severe neural tube defect (NTD) resulting from failed neural tube closure, remains a significant global health burden with a prevalence of approximately 1 per 1,000 live births. Despite advances in prenatal diagnosis (ultrasound, maternal serum AFP) and surgical intervention, current therapies cannot reverse neurological damage or fully address the underlying pathophysiology. The etiology is complex, involving folate metabolism, planar cell polarity (PCP) signaling, and environmental factors. A critical gap exists in identifying reliable biomarkers and novel therapeutic targets, as phenotype-to-protein observations alone often fail to establish causal relationships or uncover specific disease mechanisms.

Methodology
This study employed a multi-omics integrative strategy combining proteome-wide Mendelian randomization (MR) with transcriptomic data analysis to identify causal biomarkers and druggable targets for SB.

1. Data Sources:

   • Proteomics: Three large-scale protein quantitative trait locus (pQTL) datasets were utilized: Decode (SomaScan, 35,559 individuals), UKB_PPP (Olink, 54,219 individuals), and Prot_a (SomaScan, 3,301 individuals).
   • Genetics: SB genetic association data were sourced from the Pan-UK Biobank (PAN_UKB), comprising 210 cases and 408,394 controls.
   • Transcriptomics: Two Gene Expression Omnibus (GEO) datasets (GSE252552 and GSE33111) containing placental, fetal, brain, and spinal cord tissue samples from SB cases and controls were analyzed.

2. Analytical Pipeline:

   • Mendelian Randomization (MR): Instrumental variables (SNPs) were selected based on genome-wide significance ($p < 5 \times 10^{-8}$) and linkage disequilibrium pruning ($R^2 < 0.1$). Causal effects were estimated using Inverse Variance Weighted (IVW), Wald ratio, weighted median, and MR-Egger methods. Sensitivity analyses included Cochran's Q test for heterogeneity, MR-PRESSO for pleiotropy, and Steiger directionality tests.
   • Transcriptomic Analysis: Differential expression analysis (DEG) was performed using the limma package. Genes were filtered by $|log2FC| > 0.25$ and $p < 0.05$. Functional enrichment (GO/KEGG) was conducted using clusterProfiler.
   • Gene Prioritization: Candidate genes were identified by intersecting MR-significant proteins (Bonferroni corrected $p < 0.05$) with differentially expressed genes (DEGs) showing concordant expression directions (downregulated genes with OR < 1).
   • Validation and Mechanism: Generalized Summary-data-based Mendelian Randomization (GSMR) was used to adjust for pleiotropy. Single-gene Gene Set Enrichment Analysis (GSEA) explored biological pathways. Phenome-wide Association Studies (PheWAS) assessed potential adverse effects.
   • Drug Discovery: Potential therapeutic compounds were screened using the Enrichr platform (DSigDB). Molecular docking (Surflex-Dock in SYBYL-X) and Molecular Dynamics (MD) simulations (GROMACS 2022.2) were performed to evaluate binding affinity and stability of candidate compounds with target proteins (NPC2 and IL10RB).

Key Results

• Causal Protein Identification: Proteome-wide MR identified 34 proteins with statistically significant causal associations with SB.
• Candidate Gene Prioritization: By intersecting MR results with transcriptomic data, three downregulated candidate genes were prioritized:
  • NPC2: OR = 0.753 (95% CI: 0.646–0.879); involved in cholesterol metabolism.
  • PCDH12: OR = 0.636 (95% CI: 0.507–0.798); involved in cell adhesion.
  • IL10RB: OR = 0.665 (95% CI: 0.549–0.805); involved in immune regulation.
• Pathway Enrichment: Single-gene GSEA revealed significant enrichment in the JAK–STAT, Cytokine–Cytokine Receptor Interaction (CCRI), and Cell Adhesion Molecules (CAMs) pathways.
• GSMR Validation: GSMR analyses confirmed significant causal associations for all three genes ($p < 0.05$), reinforcing their role in SB pathogenesis.
• Drug Target Prediction:
  • Tamibarotene was identified as a top candidate compound with significant predicted interactions for both NPC2 and IL10RB.
  • Decitabine also showed overlap for both genes.
  • Cholesterol served as an endogenous positive control for NPC2.
• Structural Validation:
  • Molecular Docking: NPC2 showed strong binding to cholesterol (Total Score: 7.262), validating the computational approach. Tamibarotene showed favorable binding to IL10RB (Total Score: 6.019).
  • MD Simulations: The NPC2-cholesterol complex reached rapid stability. The IL10RB-tamibarotene complex remained stable for the majority of the 50 ns simulation but exhibited structural fluctuations in the later stages (34–50 ns), indicating lower stability compared to the endogenous ligand.

Significance and Claims
The authors claim that this study successfully identifies NPC2, PCDH12, and IL10RB as novel, downregulated protective factors and potential therapeutic targets for Spina Bifida through a rigorous multi-omics integration strategy.

• Mechanistic Insight: The findings implicate specific biological mechanisms in SB, including cholesterol transport (NPC2), cell adhesion during neural tube closure (PCDH12), and immune regulation via the JAK-STAT and CCRI pathways (IL10RB).
• Methodological Validation: The study validates its computational workflow by confirming the known NPC2-cholesterol interaction as a positive control.
• Therapeutic Potential: The paper highlights tamibarotene as a promising candidate compound for targeting IL10RB (and potentially NPC2), suggesting a new avenue for drug development.
• Cautious Outlook: The authors explicitly state that their findings are primarily bioinformatic. They emphasize that the clinical applicability of tamibarotene remains speculative and requires further in vitro and in vivo validation to confirm pharmacological activity and mechanistic roles. They also acknowledge limitations regarding statistical power (small case numbers in MR and GEO datasets) and the lack of non-European population data, which may affect the generalizability of the results.