Multi-tissue transcriptome-wide association study identifies 29 risk genes associated with attention-deficit/hyperactivity disorder

This multi-tissue transcriptome-wide association study integrates brain expression data with large-scale ADHD GWAS to identify 29 risk genes, including six novel candidates, thereby refining the disorder's genetic architecture and highlighting potential targets for drug discovery.

The Gist

Imagine your brain is a massive, bustling city. In this city, there are billions of workers (genes) who build and maintain the roads, power plants, and communication lines. Sometimes, due to a glitch in the city's blueprint (your DNA), certain workers don't get the right instructions on how much of their product to make. If a worker makes too much or too little, it can cause traffic jams or blackouts, leading to problems like ADHD (Attention-Deficit/Hyperactivity Disorder).

For a long time, scientists have been trying to find these "glitched blueprints" by looking at the DNA itself. They found many broken pieces of the blueprint, but they didn't know which workers were actually affected by those breaks. It was like knowing a pipe burst somewhere in the city, but not knowing which specific factory was flooding.

This paper is like sending out a team of detectives to check the actual output of the factories in three specific, critical districts of the brain city: the Cortex (the city's main command center), the Basal Ganglia (the traffic control hub), and the Cerebellum (the coordination and balance center).

Here is what they found, explained simply:

1. The Detective Tool: "The Omnibus Test"

Instead of just checking one factory at a time, the researchers used a super-smart tool called OTTERS. Think of this tool as a "super-scope" that combines five different ways of looking at the data. It takes the "blueprint glitches" (DNA) and matches them to the "factory output" (gene expression) to see which workers are actually misbehaving.

2. The Big Discovery: 29 Suspects

By scanning these three brain districts, the team identified 29 specific genes (workers) that are likely causing trouble in people with ADHD.

• 11 were found in the Command Center (Cortex).
• 4 were found in the Traffic Hub (Basal Ganglia).
• 14 were found in the Coordination Center (Cerebellum).

3. The "New" Suspects

Out of these 29, 6 were brand new suspects that scientists had never linked to ADHD before.

• MPL: This is a very interesting one. It was found in all three districts. Usually, this gene is known for managing blood cells (like platelets). It's like finding out that a blood cell manager is also secretly running the city's traffic lights. This suggests that ADHD might be connected to how the brain's "plumbing" (blood vessels) and immune system work, not just the neurons themselves.
• NKX2-2: This worker helps build the brain's structure. It was found in the Traffic Hub and Coordination Center. Interestingly, this same worker has been linked to Autism, suggesting that ADHD and Autism might share some of the same broken blueprints.

4. The "Old" Suspects (Confirmed)

The study also confirmed 23 other genes that other scientists had suspected before. Finding them again in this new, more detailed investigation is like seeing the same fingerprints at a crime scene three different times—it proves they are definitely involved.

5. Why This Matters

Before this study, most research only looked at the "Command Center" (Cortex). This study is like finally sending detectives to the other parts of the city, too.

• The "Blood" Connection: The discovery of genes like MPL suggests that ADHD isn't just about "brain chemistry" in the traditional sense; it might also be about how the brain's blood supply and immune system interact with brain development.
• New Targets for Medicine: Now that we know exactly which 29 workers are misbehaving, pharmaceutical companies can design medicines to fix those specific workers rather than guessing.

The Bottom Line

Think of this study as a massive upgrade to the city's maintenance manual. By looking at three different brain districts and using a powerful new detective tool, the researchers found 29 specific genes that are likely the root cause of ADHD symptoms. They found some familiar faces, but also discovered 6 new suspects, including one that links blood health to brain health. This gives doctors and scientists a much clearer map to find better treatments and cures in the future.

Technical Summary

1. Problem Statement

Attention-deficit/hyperactivity disorder (ADHD) is a highly heritable neurodevelopmental disorder affecting approximately 11.4% of children in the U.S. Despite recent large-scale Genome-Wide Association Studies (GWAS) identifying 27 risk loci, the underlying genetic architecture remains largely unexplained.

• The Gap: Most GWAS hits are located in non-coding regions, making functional interpretation difficult. Previous Transcriptome-Wide Association Studies (TWAS) of ADHD have been limited by:
  • Tissue Bias: Primarily focusing on the cerebral cortex while neglecting other functionally relevant brain regions like the basal ganglia and cerebellum.
  • Data Limitations: Relying on individual-level expression datasets (e.g., GTEx) with relatively small sample sizes, which reduces statistical power compared to summary-level meta-analyses.
• Objective: To refine the genetic architecture of ADHD by conducting a multi-tissue TWAS using larger, summary-level eQTL reference data to identify genes whose genetically regulated expression (GReX) is associated with ADHD risk.

2. Methodology

The study employed a two-stage analytical framework using the OTTERS (Omnibus Transcriptome Test using Expression Reference Summary Data) pipeline.

• Data Sources:

  • GWAS Data: Summary statistics from the 2023 Psychiatric Genomics Consortium (PGC) ADHD meta-analysis ($N=225,534$; 38,691 cases, 186,843 controls), restricted to European ancestry.
  • eQTL Reference Data: Summary-level cis-eQTL data from MetaBrain for three specific brain regions:
    • Cortex ($n=2,683$)
    • Basal Ganglia ($n=208$)
    • Cerebellum ($n=492$)
  • LD Reference: GTEx V8 whole-genome sequencing data (European ancestry).

• Analytical Pipeline:

  1. Model Training (Stage 1): OTTERS trained five distinct gene expression imputation models using five different Polygenic Risk Score (PRS) methods:
     • Pruning and Thresholding (P+T) at $p < 0.05$ and $p < 0.001$.
     • Frequentist LASSO (lassosum).
     • Nonparametric Bayesian Dirichlet process (SDPR).
     • Bayesian multiple linear regression with continuous shrinkage prior (PRS-CS).
  2. Association Testing (Stage 2): The trained models were applied to the ADHD GWAS summary statistics to generate gene-based association p-values.
  3. Omnibus Testing: An Aggregated Cauchy Association Test (ACAT-O) was used to combine the p-values from the five methods into a single, robust OTTERS p-value for each gene.
  4. Correction: P-values were adjusted for genomic inflation and corrected for multiple testing using a Bonferroni threshold ($\alpha = 2.7 \times 10^{-6}$).
  5. Downstream Analyses:
     • Fine-mapping: Used GIFT (Gene-based Integrative Fine-mapping) to distinguish independent causal signals from linked genes in genomic regions with multiple hits.
     • Colocalization: Used COLOC (SuSIE-based) to determine if ADHD GWAS signals and eQTL signals share the same causal variant (Posterior Probability $H4 > 0.8$).
     • Functional Enrichment: Gene Ontology (GO) analysis via clusterProfiler and Protein-Protein Interaction (PPI) network analysis via STRING.

3. Key Results

• Significant Risk Genes: The study identified 29 significant TWAS risk genes for ADHD across the three tissues:
  • Cortex: 11 genes.
  • Basal Ganglia: 4 genes.
  • Cerebellum: 14 genes.
• Novelty: Six genes were identified as novel candidates for ADHD risk: MPL, C1orf210, MDFIC, NKX2-2, FAM183A, and HIGD1A.
• Cross-Tissue Overlap:
  • MPL was significant across all three tissues.
  • TIE1, IPO13, PTPRF, and MED8 were found in both cortex and cerebellum.
  • NKX2-2 was found in both basal ganglia and cerebellum.
• Fine-Mapping: Of the 29 genes, 23 were prioritized as putatively causal signals after GIFT fine-mapping and FDR correction.
• Colocalization: Strong evidence of shared causal variants (PP.H4 > 0.8) was found for:
  • Cortex: TIE1, PTPRF, KIZ, MED8, XRN2.
  • Cerebellum: HYAL3 (across 10 credible sets).
  • Basal Ganglia: No genes met the threshold (likely due to smaller sample size).
• Functional Insights:
  • Pathway analysis revealed enrichment in neurodevelopmental pathways and cell development regulation in the cortex and basal ganglia.
  • The PPI network was statistically significant ($p=1.12 \times 10^{-4}$), showing connections to ADHD and Autism Spectrum Disorder (ASD) literature.
  • Pleiotropy: Several genes (NKX2-4, KIZ, XRN2, NKX2-2) were previously implicated in ASD, supporting the known genetic correlation between ADHD and ASD.

4. Significance and Contributions

• Methodological Advancement: Demonstrated the utility of OTTERS combined with MetaBrain summary data to increase power over traditional individual-level TWAS approaches.
• Multi-Tissue Insight: Moved beyond the cortex to implicate the basal ganglia and cerebellum, highlighting that ADHD risk involves dysregulation across multiple brain regions.
• Novel Biological Mechanisms:
  • Identified genes involved in vascular and immune processes (e.g., MPL, TIE1), suggesting that endothelial function and inflammation may play a role in ADHD etiology, expanding the scope beyond traditional neurotransmission hypotheses.
  • Validated the genetic overlap between ADHD and ASD through shared risk genes.
• Translational Potential: The identification of 29 risk genes, including six novel candidates, provides a refined list of targets for future mechanistic studies and drug discovery. Specifically, MPL was highlighted as a potential druggable target.

5. Limitations

• Sample Size Disparity: The basal ganglia eQTL reference ($n=208$) was significantly smaller than the cortex, potentially limiting power to detect signals in that region.
• Ancestry: The study was restricted to European ancestry, limiting generalizability to other populations.
• Bulk Tissue Data: The use of bulk RNA-seq data may obscure cell-type-specific effects, suggesting a need for future single-cell resolution studies.