PheWAS and cross-disorder analyses reveal genetic architecture, pleiotropic loci and phenotypic correlations across 11 autoimmune disorders

Apostolia Topaloudi, Pritesh Jain, Melanie B. Martinez, Josephine K. Bryant, Grace L. Reynolds, Petros Drineas, Peristera PaschoumedRxiv 2022 · DOI

This study, recently published as a preprint, provides a comprehensive map of the shared genetic architecture and phenotypic consequences of 11 major autoimmune disorders (ADs). By leveraging Polygenic Risk Scores (PRS) and the massive UK Biobank dataset, the researchers identified hundreds of traits linked to autoimmune risk and discovered a novel genetic locus shared across three distinct diseases.

TL;DR

Researchers mapped the genetic liability of 11 autoimmune disorders against over 3,000 traits, identifying 554 significant associations and five shared genetic "factors." The study highlights the massive influence of the HLA region while uncovering a novel shared genetic locus (near the CLNK gene) for Type 1 Diabetes, Myasthenia Gravis, and Primary Sclerosing Cholangitis.

What Problem Does This Paper Address?

Autoimmune diseases (ADs)—such as Multiple Sclerosis, Rheumatoid Arthritis, and Type 1 Diabetes—affect millions, yet they are notoriously heterogeneous. While we know they often cluster in families and share a major genetic driver in the Human Leukocyte Antigen (HLA) region, the "non-HLA" genetic landscape remains murky.

Furthermore, we don't fully understand how the genetic predisposition to these diseases influences a person's broader "phenome"—their lifestyle, biomarkers, and other health conditions. This paper seeks to move beyond looking at one disease at a time to understand the systemic, cross-disorder genetic relationships that drive autoimmunity.

Methods

The authors employed a multi-stage genomic pipeline:

  1. Polygenic Risk Scores (PRS): Using summary statistics from previous large-scale studies, they calculated an individual's genetic liability for 11 ADs. Crucially, they performed these calculations both including and excluding the dominant HLA region to isolate "hidden" genetic signals.
  2. PheWAS (Phenome-Wide Association Study): They tested these PRSs against 3,281 outcomes in the UK Biobank to find associations with everything from diet to mental health.
  3. Genetic Correlation & Factor Analysis: Using Linkage Disequilibrium Score Regression (LDSC) and Genomic Structural Equation Modeling (GenomicSEM), they identified how these 11 diseases correlate genetically and grouped them into five underlying "factors" or clusters.
  4. Cross-Disorder Meta-Analysis: For each cluster, they performed a meta-analysis to find specific genetic variants (SNPs) that contribute to multiple diseases simultaneously (pleiotropy).

Datasets

  • Discovery GWAS: Publicly available and in-house summary statistics for 11 ADs (e.g., Rheumatoid Arthritis, SLE, Celiac Disease).
  • UK Biobank (UKB): A dataset of 330,841 individuals of European ancestry. This provided the "testing ground" for the PheWAS, covering 3,281 phenotypes across categories like biomarkers, lifestyle, and mental health.

Key Insights and Ideas

  • The HLA "Mask": The HLA region is so genetically "loud" that it can drown out other signals. For Celiac Disease and Rheumatoid Arthritis, the researchers actually found more phenotypic associations after removing the HLA region, suggesting that non-HLA risk variants have broad, distinct effects on health.
  • The Celiac Hub: Genetic risk for nearly all 11 ADs was significantly associated with Celiac disease or gluten sensitivity phenotypes. This reinforces the idea that gluten intolerance may be a common phenotypic "canary in the coal mine" for general autoimmune liability.
  • Mental Health Divergence: The study found a surprising split in mental health associations. Genetic risk for Systemic Lupus Erythematosus (SLE) and Primary Sclerosing Cholangitis (PSC) was associated with better reported mental health outcomes (lower depression/anxiety), while risk for Vitiligo and Psoriasis was linked to poorer mental health. This suggests that the biological pathways for different ADs interact with the brain in vastly different ways.
  • Sun Exposure & Vitiligo: While clinical observation often links Vitiligo to sun sensitivity, the genetic risk for Vitiligo was associated with darker skin and a lower likelihood of tanning or burning, suggesting a complex interplay between genetics and behavior.

Key Results

  • 554 significant associations were found between AD risk scores and various traits.
  • 300 associations remained significant even after the HLA region was excluded.
  • Five Genetic Factors: The 11 diseases clustered into five distinct groups based on shared genetic risk.
  • Novel Pleiotropic Locus: The meta-analysis of the Type 1 Diabetes, Myasthenia Gravis, and PSC cluster revealed a novel genome-wide significant locus on chromosome 4, located 23kb downstream of the CLNK gene. CLNK is involved in immunoreceptor signaling, making it a prime candidate for future functional studies.

Limitations and Future Work

  • Ancestry Bias: The study was limited to individuals of European ancestry due to the constraints of the UK Biobank. Results may not generalize to other global populations.
  • Statistical Power: Some disorders, like Myasthenia Gravis, had smaller discovery datasets, which may have limited the ability to detect subtler genetic correlations.
  • Case Numbers: The number of actual diagnosed cases for some rare ADs within the UK Biobank is relatively low, necessitating the use of PRS as a proxy for liability.
  • Future Directions: The authors suggest that the novel CLNK locus warrants deep functional investigation to understand how it triggers three very different diseases. Additionally, clinicians might eventually use these broad phenotypic associations to better screen high-risk individuals for comorbid conditions.

Summary updated

Abstract

Abstract Autoimmune diseases (ADs) are a group of more than 80 heterogeneous disorders that occur when there is a failure in the self-tolerance mechanisms triggering self-attacking autoantibodies. Most autoimmune disorders are polygenic and associated with genes in the human leukocyte antigen (HLA) region. However, additional non-HLA genes are also found to be associated with different ADs, and often these are also implicated in more than one disorder. Previous studies have observed associations between various health-related and lifestyle phenotypes and ADs. Polygenic risk scores (PRS) allow the calculation of an individual’s genetic liability to a phenotype and are estimated as the sum of the risk alleles weighted by their effect sizes in a genome-wide association study (GWAS). Here, for the first time, we conducted a comparative PRS-PheWAS analysis for 11 different ADs (Celiac Disease, Juvenile Idiopathic Arthritis, Multiple Sclerosis, Myasthenia Gravis, Primary Sclerosing Cholangitis, Psoriasis, Rheumatoid Arthritis, Systemic Lupus Erythematosus, Type 1 Diabetes, Vitiligo Early Onset, Vitiligo Late Onset) and 3,281 outcomes available in the UK Biobank that cover a wide range of lifestyle, socio-demographic and health-related phenotypes. We also explored the genetic relationships of the studied ADs, estimating their genetic correlation and performing cross-disorder GWAS meta-analyses for the identified AD clusters. In total, we observed 554 outcomes significantly associated with at least one disorder PRS, and 300 outcomes were significant after variants in the HLA region were excluded from the PRS calculations. Based on the genetic correlation and genetic factor analysis, we observed five genetic factors among studied ADs. Cross-disorder meta-analyses in each factor revealed genome-wide significant loci that are pleiotropic across multiple ADs. Overall, our analyses confirm the association of different factors with genetic risk for ADs and reveal novel observations that warrant further exploration.
BibTeX
@misc{topaloudi2022phewas,
  author = {Apostolia Topaloudi and Pritesh Jain and Melanie B. Martinez and Josephine K. Bryant and Grace L. Reynolds and Petros Drineas and Peristera Paschou},
  title = {{PheWAS and cross-disorder analyses reveal genetic architecture, pleiotropic loci and phenotypic correlations across 11 autoimmune disorders}},
  year = {2022},
  doi = {10.1101/2022.10.16.22281127},
  howpublished = {medRxiv},
}
Figure