Polygenic risk score-based phenome-wide association study identifies novel associations for Tourette syndrome

Pritesh Jain, Tyne W. Miller‐Fleming, Apostolia Topaloudi, Dongmei Yu, Petros Drineas, Marianthi Georgitsi, Zhiyu Yang, Renata Rizzo, Kirsten Müller‐Vahl, Zeynep Tümer, Nanette Mol Debes, Andreas Hartmann, Christel Depienne, Yulia Worbe, Pablo Mir, Daniëlle C. Cath, Dorret I. Boomsma, Veit Roessner, Tomasz Wolańczyk, Piotr Janik, Natalia Szejko, Cezary Żekanowski, Csaba Barta, Zsófia Nemoda, Zsanett Tárnok, Joseph D. Buxbaum, Dorothy E. Grice, Jeffrey Glennon, Hreinn Stefánsson, Bastian Hengerer, Noa Benaroya-Milshtein, Francesco Cardona, Tammy Hedderly, Isobel Heyman, Chaim Huyser, Ástrid Morer, Norbert Mueller, Alexander Münchau, Kerstin Jessica Plessen, Cesare Porcelli, Susanne Walitza, Anette Schrag, Davide Martino, Andrea Dietrich, Carol A. Mathews, Jeremiah M. Scharf, Pieter J. Hoekstra, Lea K. Davis, Peristera PaschoumedRxiv 2022 · DOI

TL;DR

Researchers used polygenic risk scores (PRS) to conduct a massive "phenome-wide" scan of over 330,000 individuals, identifying 57 traits associated with the genetic liability for Tourette Syndrome (TS). The study reveals that TS genetic risk extends far beyond tics, linking to non-psychiatric conditions like Type 2 diabetes and respiratory issues, while highlighting a fascinating "tug-of-war" in genetic directionality between TS and Obsessive-Compulsive Disorder (OCD).

What Problem Does This Paper Address?

Tourette Syndrome (TS) is a highly heritable neurodevelopmental disorder, yet it remains underrepresented in large-scale biobanks. Because only about 0.6% of the population has a clinical diagnosis, traditional case-control studies in biobanks often lack the statistical power to find new associations.

The authors address this "data scarcity" by using Polygenic Risk Scores (PRS)—a single value that summarizes an individual's genetic predisposition for a disease based on thousands of tiny genetic variants. By using PRS as a proxy for TS risk, the researchers could look across the entire "phenome" (thousands of measured traits) in the UK Biobank to see what else travels with the genetic risk of TS, even in people without a formal diagnosis.

Methods

The team employed a PRS-based Phenome-Wide Association Study (PheWAS).

  1. Score Calculation: They used the latest TS meta-analysis summary statistics to calculate risk scores for 330,841 participants in the UK Biobank using PRScs, a Bayesian regression framework that accounts for linkage disequilibrium (how genes are inherited together).
  2. Phenotype Mapping: They analyzed 2,242 distinct phenotypes, ranging from blood markers and disease diagnoses to lifestyle and socioeconomic factors.
  3. Statistical Rigor: To avoid "fishing" for results, they applied a strict Bonferroni correction ($p < 2.3 \times 10^{-5}$) and used the PHESANT package to automate the association testing across different data types (continuous, binary, etc.).
  4. Cross-Disorder Comparison: They repeated the process for ADHD, Autism Spectrum Disorder (ASD), and OCD to see where the genetic risks overlapped or diverged.

Datasets

  • UK Biobank (UKB): The primary dataset, filtered to ~330k individuals of European ancestry to ensure genetic homogeneity.
  • Psychiatric Genomics Consortium (PGC): Provided the GWAS summary statistics used to "train" the risk scores for TS, ADHD, ASD, and OCD.
  • 1000 Genomes Project: Used as an external reference panel for genetic mapping and quality control.

Key Insights and Ideas

  • The Systemic Reach of TS: Genetic risk for TS isn't just "in the head." The study found significant associations with physical ailments like Type 2 diabetes, heart palpitations, and respiratory diseases. This suggests that the biological pathways underlying TS might have systemic physiological effects.
  • The Impulsivity-Compulsivity Spectrum: A major "aha moment" came from comparing TS with other disorders. While TS, ADHD, and ASD generally showed the same direction of effect for most traits, OCD often showed the opposite. For example, higher TS risk was associated with lower educational attainment, whereas higher OCD risk was associated with higher educational attainment. This highlights a genetic divergence between the impulsive nature of TS/ADHD and the compulsive nature of OCD.
  • Sex-Specific Manifestations: TS is famously more common in males. This study found that the manifestation of genetic risk differs by sex: Type 2 diabetes and heart palpitations were significant markers of TS risk in males, while respiratory issues were the primary physical association in females.
  • Socioeconomic Signatures: Higher genetic risk for TS was associated with higher "deprivation indices" and lower age at first birth, suggesting that the genetic architecture of the disorder may influence life trajectories through behavioral or cognitive intermediaries.

Key Results

  • 57 Significant Associations: Out of 2,242 traits, 57 passed the strict significance threshold.
  • Mental Health: The strongest association was with "Seen doctor for nerves, anxiety or depression" ($\beta = 0.031, p = 9.2 \times 10^{-17}$). Other links included neuroticism, mood swings, and irritability.
  • Physical Health: Significant positive associations were found for chronic pain (back, neck, hip), sensitive stomach, and mouth ulcers.
  • Cross-Disorder Overlap: 16 phenotypes were significantly associated with the genetic risk of all four disorders (TS, ADHD, ASD, OCD), primarily in the realm of cognition and mental health.

Limitations and Future Work

  • Correlation vs. Causation: As with all PheWAS, these results show associations. The authors note that techniques like Mendelian Randomization are needed to determine if TS genetic risk causes these outcomes or if they share a common underlying cause.
  • Case Validation: The UK Biobank only contained about 30 individuals with actual tic-related diagnoses. This prevented the authors from validating exactly how well the PRS predicts a clinical TS diagnosis within this specific cohort.
  • Ancestry Bias: The study was limited to individuals of European ancestry. Future work must expand to more diverse populations to ensure these genetic insights are globally applicable.
  • SNP Overlap: The number of genetic variants (SNPs) available for the TS risk score was lower than for ADHD or ASD, which may have led to an underestimation of some correlations.

Summary updated

Abstract

Abstract Tourette Syndrome (TS) is a complex neurodevelopmental disorder characterized by vocal and motor tics lasting more than a year. It is highly polygenic in nature with both rare and common previously associated variants. Epidemiological studies have shown TS to be correlated with other phenotypes, but large-scale phenome wide analyses in biobank level data have not been performed to date. In this study, we used the summary statistics from the latest meta-analysis of TS to calculate the polygenic risk score (PRS) of individuals in the UK Biobank data and applied a Phenome Wide Association Study (PheWAS) approach to determine the association of disease risk with a wide range of phenotypes. A total of 57 traits were found to be significantly associated with TS polygenic risk, including multiple psychosocial factors and mental health conditions such as anxiety disorder and depression. Additional associations were observed with complex non-psychiatric disorders such as Type 2 diabetes, heart palpitations, and respiratory conditions. Cross-disorder comparisons of phenotypic associations with genetic risk for other childhood-onset disorders (e.g.: attention deficit hyperactivity disorder [ADHD], autism spectrum disorder [ASD], and obsessive-compulsive disorder [OCD]) indicated an overlap in associations between TS and these disorders. ADHD and ASD had a similar direction of effect with TS while OCD had an opposite direction of effect for all traits except mental health factors. Sex-specific PheWAS analysis identified differences in the associations with TS genetic risk between males and females. Type 2 diabetes and heart palpitations were significantly associated with TS risk in males but not in females, whereas diseases of the respiratory system were associated with TS risk in females but not in males. This analysis provides further evidence of shared genetic and phenotypic architecture of different complex disorders.
BibTeX
@misc{jain2022polygenic,
  author = {Pritesh Jain and Tyne W. Miller‐Fleming and Apostolia Topaloudi and Dongmei Yu and Petros Drineas and Marianthi Georgitsi and Zhiyu Yang and Renata Rizzo and Kirsten Müller‐Vahl and Zeynep Tümer and Nanette Mol Debes and Andreas Hartmann and Christel Depienne and Yulia Worbe and Pablo Mir and Daniëlle C. Cath and Dorret I. Boomsma and Veit Roessner and Tomasz Wolańczyk and Piotr Janik and Natalia Szejko and Cezary Żekanowski and Csaba Barta and Zsófia Nemoda and Zsanett Tárnok and Joseph D. Buxbaum and Dorothy E. Grice and Jeffrey Glennon and Hreinn Stefánsson and Bastian Hengerer and Noa Benaroya-Milshtein and Francesco Cardona and Tammy Hedderly and Isobel Heyman and Chaim Huyser and Ástrid Morer and Norbert Mueller and Alexander Münchau and Kerstin Jessica Plessen and Cesare Porcelli and Susanne Walitza and Anette Schrag and Davide Martino and Andrea Dietrich and Carol A. Mathews and Jeremiah M. Scharf and Pieter J. Hoekstra and Lea K. Davis and Peristera Paschou and Lea K. Davis and Peristera Paschou},
  title = {{Polygenic risk score-based phenome-wide association study identifies novel associations for Tourette syndrome}},
  year = {2022},
  doi = {10.1101/2022.09.01.22279340},
  howpublished = {medRxiv},
}
Figure