Researchers in South Korea have carried out the first genome-wide search for genetic variants that modify the relationship between long-term exposure to particulate matter and thyroid-stimulating hormone, and they report several suggestive loci that could explain why people respond differently to the same polluted air. The study, published in the journal Genes and Environment, analyzed nearly 3,300 Korean adults across a discovery and a replication cohort, scanning more than 4.6 million genetic variants for interactions with annual average exposure to particulate matter with a diameter of 10 micrometers or smaller, known as PM10.
The central result is nuanced rather than dramatic. No single variant crossed the stringent genome-wide significance threshold for interaction, set at a P value below 5 × 10⁻⁸. However, one variant near the gene MSRA reached a suggestive level of significance, and two genetic susceptibility loci, one near FAM84B and PCAT1 and another within STARD13, were replicated in an independent cohort at a nominal significance level. A third variant, rs7169081, located between the genes CGNL1 and GCOM1 on chromosome 15, emerged as functionally interesting because it sits in a predicted enhancer region active in atrial cardiomyocytes, cells that are strongly influenced by thyroid hormones.
Previous research has linked air pollution exposure to changes in thyroid hormone levels, including in pregnant women and their fetuses, and some studies suggest the association is stronger in obese individuals. What remained unknown until now was whether genetic variation could explain why people exposed to similar levels of particulate matter show different thyroid responses. The Korean team set out to answer that question not by testing a handful of candidate genes, as earlier gene-environment studies of air pollution had done, but by interrogating the entire genome.
The participants came from health checkup centers at Seoul National University Hospital. The discovery cohort was recruited between December 2009 and December 2013 and, after exclusions for missing data, invalid DNA samples, lack of residential zip code information, or use of thyroid medications, comprised 1,863 adult men with a mean age of 48.9 years. The replication cohort, recruited between 2014 and 2015 at the same hospital, included 1,458 individuals, 58 percent of them men, with a mean age of 54.4 years. Genotyping was performed on Illumina and Axiom Korean microarray platforms, and imputation was carried out using the Northeast Asian Reference Database 2 panel, an Asian-based reference that the authors say minimized the loss of genetic information compared with predominantly European panels.
Exposure assessment relied on roughly 300 air quality monitoring stations across South Korea, with data collected from January 2009 through December 2015. For each participant, the researchers estimated the annual mean PM10 concentration for the year of the examination using the monitoring station nearest to the residential postal code. In the discovery cohort, mean PM10 exposure was 49.3 micrograms per cubic meter and mean TSH was 1.9 milli-international units per liter. For the interaction analyses, exposure was treated both as a continuous variable and as a binary variable, with a cutoff of 53.5 micrograms per cubic meter corresponding to the 75th percentile of PM10 levels in the discovery cohort. That split placed 1,398 participants in a low-to-moderate exposure group and 465 in a high-exposure group ranging up to 82.6 micrograms per cubic meter.
Statistically, the genome-wide interaction study used linear regression models that included age, examination center, and body mass index as covariates in the discovery cohort, and age, sex, and body mass index in the replication cohort. The interaction term between genotype and PM10 exposure was the quantity of interest, and beta coefficients were scaled to reflect a 1 microgram per cubic meter increase in PM10. As an internal check, the team also ran a conventional genome-wide association study of TSH alone in the discovery cohort. That analysis recovered a genome-wide significant signal at VAV3 on chromosome 1, with a lead P value of 1.31 × 10⁻⁸, plus suggestive signals at IGFBP5 and B4GALNT3, all three of which had been reported in a previous TSH genome-wide association study of 343,604 individuals, primarily of European ancestry. This consistency supported the validity of the dataset.
In the interaction analysis itself, the continuous PM10 model produced the most robust findings. Three variants, rs6998646 between FAM84B and PCAT1, and two highly linked intronic variants in STARD13, rs79013868 and rs9597585, showed significant interactions in both the discovery and replication cohorts. Stratified analyses revealed that the direction of the association between PM10 and TSH reversed depending on genotype. For rs6998646, the beta coefficient shifted from −0.010 in one genotype group to 0.032 in another; for the two STARD13 variants, the coefficients flipped from positive to negative. All of these sign reversals were statistically significant, a pattern consistent with a genuine interaction rather than a uniform effect.
The binary exposure analysis told a more tentative story. Seven lead SNPs reached significance in the discovery cohort, including rs11781213 and rs7837316 near MSRA, and variants near GOT2, PRAG1, GCOM1, MKI67, and SGCG. The strongest of these, rs11781213 near MSRA, reached a suggestive interaction P value below 5 × 10⁻⁷, and stratified analysis showed that TSH levels differed significantly between genotype groups when comparing low-to-moderate with high PM10 exposure, with the slope of the association reversing between exposure groups. However, this variant did not replicate, returning a P value of 0.661 in the second cohort. The authors note that no variant surpassed the stringent genome-wide threshold and suggest that such thresholds may be overly conservative for gene-environment interaction studies, which is why they also applied relaxed criteria of P below 1 × 10⁻⁵ for discovery and P below 0.05 for replication.
The functional candidate rs7169081, an intergenic G-to-A variant between CGNL1 and GCOM1, was flagged using activity-by-contact enhancer models and single-cell chromatin accessibility data across 220 cell types. Enhancer activity for this region was observed mainly in atrial cardiomyocytes, and GTEx and Human Protein Atlas data showed elevated GCOM1 expression primarily in cardiomyocytes as well. The variant was also associated with thyroid expression quantitative trait loci signals related to CGNL1, with a P value of 7.50 × 10⁻⁶, and acted as a whole-blood expression quantitative trait locus enhancing GCOM1 expression with a P value of 1.10 × 10⁻¹³. Because thyroid hormones modulate cardiomyocyte genes such as MYH6 and MYH7 and influence ion channel activity, the authors argue that this locus plausibly connects particulate matter exposure, thyroid regulation, and cardiac biology.
The study has clear limitations that temper its conclusions. Exposure was estimated only from residential zip codes, with no information on indoor sources or workplace locations, and PM2.5 could not be analyzed because national monitoring data were unavailable before 2015. The cross-sectional design precludes any causal inference about particulate matter exposure and thyroid hormone levels, and the discovery cohort consisted entirely of men, limiting generalizability. The authors also emphasize that the identified loci, including MSRA, PCAT1, STARD13, and GCOM1, are suggestive rather than definitive, and that studies in other ethnic groups are needed to validate the interactions and clarify the underlying mechanisms. MSRA, which encodes an enzyme that repairs oxidized methionine residues and has been linked to abdominal obesity, offers one plausible biological thread, since its loss increases sensitivity to oxidative stress in mice and promotes inflammatory signaling. As a first genome-wide gene-by-air-pollution interaction analysis of thyroid function, the study maps candidate territory for future confirmation rather than settling the question of how polluted air alters thyroid hormones.
Subject of Research: Genome-wide interaction between particulate matter exposure and genetic variants affecting thyroid-stimulating hormone levels in Korean adults
Article Title: A genome-wide interaction study of thyroid-stimulating hormone levels and particulate matter exposure among Koreans
Article References: Park, Y. J., Kim, H.-J., Son, H.-Y., Kim, J., Yun, J. M., Kwon, H., Cho, B., Park, J.-H., & Kim, J.-I. (2026). A genome-wide interaction study of thyroid-stimulating hormone levels and particulate matter exposure among Koreans. Genes and Environment, 48(1), Article 10. https://doi.org/10.1186/s41021-026-00357-z
Image Credits: AI Generated
DOI: 10.1186/s41021-026-00357-z
Keywords: PM10, thyroid-stimulating hormone, genome-wide interaction study, gene-environment interaction, air pollution, Korean adults, MSRA, PCAT1, STARD13, GCOM1, cross-sectional study, thyroid function
News Source: Juliet Wilcox. (October 29, 2026). Genes May Shape How Air Pollution Touches Thyroid Hormone Levels. Scienmag.
