Th17-associated cytokine gene hypomethylation reflects epigenetic dysregulation in graves’ disease

Yanfei Jiang, Kaida Mu, Zhaowei Huang, Xinwei Zhang, Yalin Wang, Wenyu Xu, Ronghua Song, Jin‐an ZhangFrontiers in Immunology 2025 · DOI

TL;DR

Researchers have identified a distinct "epigenetic signature" in Graves’ disease (GD) characterized by the hypomethylation of Th17-associated cytokine genes (IL17, IL21, and IL22). By applying machine learning to DNA methylation data, the team identified specific CpG sites that serve as promising biomarkers for diagnosing GD and monitoring how patients respond to treatment.

What Problem Does This Paper Address?

Graves’ disease is a common autoimmune disorder where the body produces antibodies that overstimulate the thyroid, leading to hyperthyroidism. While we know that Th17 cells—a subset of pro-inflammatory T cells—play a major role in this process by secreting cytokines like IL-17, IL-21, and IL-22, the underlying "volume control" for these genes is poorly understood.

Genetic predisposition only tells part of the story. This paper investigates epigenetics, specifically DNA methylation, which acts as a chemical switch to turn genes on or off. The authors sought to determine if the "brakes" (methylation) on pro-inflammatory genes are missing in GD patients and whether these epigenetic changes could be used as a diagnostic tool.

Methods

The study utilized a cohort of 120 Han Chinese individuals: 60 GD patients (further divided into newly diagnosed and refractory cases) and 60 healthy controls.

Key technical steps included:

  • Targeted Bisulfite Sequencing: The researchers focused specifically on the promoter regions of the IL17, IL21, and IL22 genes in peripheral blood mononuclear cells (PBMCs) to quantify methylation at individual CpG sites.
  • Machine Learning Integration: Rather than relying on simple statistical averages, the team employed four supervised learning algorithms—Random Forest (RF), Support Vector Machine (SVM), Generalized Linear Model (GLM), and eXtreme Gradient Boosting (XGBoost)—to identify which specific methylation sites were most predictive of the disease.
  • Clinical Correlation: They used multivariate logistic regression to see how these methylation levels correlated with clinical markers like TRAb (thyroid-stimulating hormone receptor autoantibodies) and FT4 (free thyroxine).

Datasets

The study relied on a primary dataset of 120 clinical samples collected at Zhoupu Hospital in Shanghai between 2023 and 2025.

  • GD Group (n=60): 35 newly diagnosed (NGD) and 25 refractory (RGD) patients.
  • Control Group (n=60): Age- and sex-matched healthy individuals (NC).
  • Genomic Targets: 16 CpG sites for IL17, 10 for IL21, and 17 for IL22.

Key Insights and Ideas

  • The "Brakes" are Off: GD patients showed significantly lower methylation levels (hypomethylation) across all three cytokine promoters compared to healthy controls ($P = 2.5 \times 10^{-7}$). In epigenetic terms, lower methylation usually leads to higher gene expression, explaining why these patients have such high levels of pro-inflammatory cytokines.
  • Epigenetic Memory: Interestingly, while refractory patients (those undergoing treatment) showed some restoration of methylation, their levels remained lower than healthy controls. This suggests an "epigenetic memory" that might predispose patients to relapse even when their thyroid levels seem stable.
  • Specific Diagnostic "Hotspots": The machine learning models prioritized four specific CpG sites as high-value biomarkers. One site in particular, chr4_123542549_R, was strongly associated with elevated TRAb levels (Odds Ratio = 4.00), suggesting a direct link between epigenetic status and the severity of the autoimmune response.

Key Results

  • Diagnostic Accuracy: The top-performing CpG sites demonstrated solid diagnostic potential, with Area Under the Curve (AUC) values exceeding 0.7. Specifically, chr4_123542549_R reached an AUC of 0.754 and chr12_68647247_R reached 0.752.
  • Clinical Links: Hypomethylation at chr12_68647247_R was significantly associated with increased FT4 levels (OR = 0.29, $P = 0.02$), indicating that these epigenetic markers reflect the actual functional status of the thyroid.
  • Model Performance: The SVM and XGBoost models showed superior performance in classifying patients based on their "methylome" compared to traditional linear models.

Limitations and Future Work

The authors acknowledge several areas for further exploration:

  • Sample Size and Diversity: The study was limited to 120 Han Chinese participants. Larger, multi-ethnic cohorts are needed to validate these biomarkers globally.
  • Correlation vs. Causation: While the study shows a strong link between hypomethylation and GD, it doesn't definitively prove that the methylation changes cause the disease. Functional studies (like CRISPR-based epigenetic editing) would be required to confirm this.
  • Longitudinal Tracking: The study provides a snapshot in time. Future research should track the same patients over several years to see if methylation levels can accurately predict a relapse before clinical symptoms appear.

Summary updated

Abstract

Our findings highlight Th17-related epigenetic dysregulation as a key feature of GD and support the potential of methylation markers for diagnostic and therapeutic monitoring applications.
BibTeX
@inproceedings{jiang2025th,
  author = {Yanfei Jiang and Kaida Mu and Zhaowei Huang and Xinwei Zhang and Yalin Wang and Wenyu Xu and Ronghua Song and Jin‐an Zhang},
  title = {{Th17-associated cytokine gene hypomethylation reflects epigenetic dysregulation in graves’ disease}},
  booktitle = {Frontiers in Immunology},
  year = {2025},
  doi = {10.3389/fimmu.2025.1635883},
}
Figure