Researchers analyzed two Gene Expression Omnibus datasets separately before intersecting their differentially expressed genes. The hypothyroidism dataset, GSE176153, contained four euthyroid controls and four patients, but only one man in each group. The researchers therefore restricted their main comparison to women, leaving three hypothyroid cases and three controls. The subclinical atherosclerosis dataset, GSE221615, included 391 people, of whom only 27 were women: 13 with subclinical atherosclerosis and 14 controls. In this female-restricted intersection, 15 shared genes were upregulated in both disorders, two were downregulated and three changed in opposite directions. A stricter analysis using Benjamini–Hochberg-adjusted significance found no shared genes, making the 20-gene signature exploratory rather than statistically robust.

Functional enrichment connected the shared signature to B-cell proliferation and activation, a result that differed from the interferon, antiviral and antimicrobial pathways found when the authors compared hypothyroidism with rheumatoid arthritis and systemic lupus erythematosus. A high-confidence protein-interaction network linked CD19, BLK, FCRLA and TNFRSF13C. CD19 had the highest degree of connectivity, while BLK and FCRLA were also central nodes. CD180 and SPIB were added to the six-gene pivotal group through the combined evidence from enrichment, functional-network and protein-interaction analyses. These genes encode components associated with B-cell signaling, maturation or regulation, providing a plausible molecular basis for the shared signature. However, whole-blood measurements cannot determine whether the pattern arises from altered activity within cells or from subtle changes in the proportions of circulating immune-cell subsets.

The study used CIBERSORTx to estimate the relative abundance of 22 immune-cell types from bulk RNA-sequencing data. It found no significant differences in estimated immune-cell proportions between diseased and control groups. Expression levels nevertheless correlated with inferred cell signatures: BLK and SPIB were strongly positively associated with the CD8-positive T-cell signature in the female hypothyroidism cohort, while FCRLA was negatively associated with the naïve CD4-positive T-cell signature. In the female atherosclerosis cohort, IFRD1 correlated negatively with the memory B-cell signature, and several pivotal genes correlated positively with activated natural killer cell, M2 macrophage and resting mast-cell signatures. Because these were computational estimates and the M2 macrophage fraction was relatively small, the correlations do not establish functional relationships between the genes and those cell populations.

ROC analyses suggested that some pivotal genes could discriminate groups in the discovery cohorts. Among women with subclinical atherosclerosis, TNFRSF13C had an area under the ROC curve of 0.835, with a 95% confidence interval of 0.664–1.000, while SPIB reached 0.808, with a 95% interval of 0.626–0.990. All six genes produced an AUC of 1.00 in the hypothyroidism cohort, but that calculation separated only three cases from three controls and therefore cannot be interpreted as evidence of diagnostic utility. The candidate transcription factors were less consistent: all had AUC values below 0.70 in the female subclinical atherosclerosis cohort. These results show why the reported classification performance should be treated as hypothesis-generating rather than a clinically usable biomarker panel.

The researchers sought an external check using GSE202625, a cohort of consecutive patients undergoing coronary CT angiography that represented established atherosclerotic disease rather than subclinical disease. Its female subset contained only 11 participants—six with coronary artery disease and five controls—so it served as a limited proxy rather than a matching validation cohort. CD180 had the highest AUC among the six pivotal genes in this group, at 0.87, followed by SPIB and BLK, both at 0.70. Among predicted regulators, MEF2A reached 0.90, SPI1 0.80 and PPARG 0.73. The wide confidence intervals and the mismatch between advanced and subclinical disease limit interpretation, but the results suggest that parts of the candidate axis may remain detectable beyond the early stage of atherosclerosis.

To explore thyroid-stimulating hormone signaling outside circulating blood, the team examined GSE275079, a dataset of primary bone-marrow-derived macrophages from male mice with myeloid-specific Tshr knockout and their wild-type littermates after TSH stimulation. Downregulated genes in knockout macrophages were enriched for leukocyte and mononuclear-cell proliferation, regulation of inflammatory responses, immune signaling, chemotaxis and migration. Ifrd1 expression was lower in knockout cells, whereas Cd19, Arhgap24, Pparg and Mef2c were higher. These findings connect disruption of TSH-receptor signaling with immune pathways, but they do not validate the B-cell-dominated human signature or demonstrate that the same responses occur in women. The mouse macrophage experiments instead provide a complementary myeloid context for the systemic human analysis.

Human THP-1 monocytes were differentiated into macrophages and exposed to TSH, bacterial lipopolysaccharide, both together, or vehicle for 24 hours. IFRD1 messenger RNA was significantly reduced by 1 mU/mL TSH alone, with an adjusted p-value of 0.001, and by TSH combined with lipopolysaccharide, with an adjusted p-value below 0.001. The same reduction was not observed at the higher TSH concentration of 10 mU/mL, leaving the basis for this concentration-dependent response unresolved. MEF2A expression did not differ significantly from vehicle in any treatment group. The experiment provides initial evidence that TSH exposure can alter IFRD1 expression in a human macrophage model, but a cell line experiment cannot establish a mechanism in patients or show that IFRD1 mediates cardiovascular risk.

Several limitations constrain the findings. Cross-dataset technical differences may have influenced the intersection, and important clinical variables—including thyroid hormone concentrations in the atherosclerosis cohort, vascular measurements in the hypothyroidism cohort, body mass index, glucose, lipids, menopausal status, hormone replacement therapy and thyroid-antibody profiles—were unavailable. The female-only design reduced sex-related heterogeneity but left effects of reproductive stage unresolved. Bulk RNA also lacks the resolution to distinguish cell-intrinsic changes from shifts in rare or subtle immune-cell populations. In addition, the functional experiments concentrated on macrophages rather than the B cells that dominate the human signature, and the male-derived models cannot establish female specificity. Overall, the study identifies a testable candidate network linking thyroid dysfunction with early atherosclerosis in women, but larger independent cohorts, single-cell profiling and direct functional studies in relevant lymphocyte populations are required before its clinical or prognostic significance can be determined.

Subject of Research: Whole-blood transcriptomics of shared molecular mechanisms between subclinical atherosclerosis and hypothyroidism in women

Article Title: A whole-blood transcriptomic analysis of shared molecular mechanisms between subclinical atherosclerosis and hypothyroidism in women

Article References: Torres, E. M., Raffo, D. A., & Tellechea, M. L. (2026). A whole-blood transcriptomic analysis of shared molecular mechanisms between subclinical atherosclerosis and hypothyroidism in women. Frontiers in Bioinformatics, 6, Article 1932220. https://doi.org/10.3389/fbinf.2026.1932220

Image Credits: AI Generated

DOI: 10.3389/fbinf.2026.1932220

Keywords: hypothyroidism, subclinical atherosclerosis, whole-blood transcriptomics, women's health, B-cell activation, immune regulation, TSH signaling, TSHR, IFRD1, macrophages, whole-blood, transcriptomic