Graves’ disease is an organ-specific autoimmune disorder in which the immune system generates stimulatory autoantibodies against the thyroid-stimulating hormone receptor. When these antibodies bind the receptor, they activate the adenylate cyclase signaling pathway inside thyroid cells, driving the gland to produce hormone relentlessly. The result is diffuse goiter, hyperthyroidism, and in many patients infiltrative ophthalmopathy, the eye disease that can accompany the condition. The disorder most commonly strikes women between the ages of thirty and sixty, and its incidence has been rising in recent years while the average age of patients has been falling. Current treatments, which include anti-thyroid drugs, radioactive iodine, and thyroidectomy, suffer from low remission rates, high relapse rates, and meaningful complications, which is precisely why researchers are keen to understand the upstream events that set the disease in motion.

The suspicion that infections might be involved is not fringe science. Many infectious agents have been implicated in the onset and progression of noncommunicable diseases, particularly autoimmune conditions, and antibody-mediated immune responses represent the body’s principal defense against pathogenic microorganisms. Patients with Graves’ disease, whose immune systems run persistently hot, might respond differently to certain pathogens than healthy individuals. Some patients, for example, have tested positive for hepatitis C virus antibodies. Observational and mechanistic studies have repeatedly suggested that infectious agents may contribute to autoimmune thyroid disease, but the direction of the relationship, and the immune pathways underneath it, have remained stubbornly uncertain.

This is where Mendelian randomization enters the picture. The method exploits the random assortment of genetic variants at conception as a kind of natural randomized controlled trial. Because variants inherited at conception are fixed for life and largely independent of environmental confounders, they can serve as instrumental variables to test whether an exposure plausibly causes an outcome, or vice versa, without the confounding and reverse causation that plague traditional observational studies. The design rests on three assumptions: the genetic variants must be strongly associated with the exposure, they must be independent of confounders, and they must influence the outcome only through the exposure. When these assumptions hold, the technique offers causal inference at a fraction of the cost and duration of a clinical trial.

The research team, led by investigators affiliated with Shanghai University of Traditional Chinese Medicine and the Shanghai University of Medicine and Health Sciences Affiliated Zhoupu Hospital, assembled genetic instruments for forty-six antibody-mediated immune-response phenotypes spanning thirteen pathogens, drawing on a large genome-wide association study within the UK Biobank. These phenotypes included fifteen case-control measures of seropositivity and thirty-one quantitative assessments of antibody levels. Graves’ disease data came from the Finnish FinnGen database, specifically the strict definition encompassing 3,176 cases and 409,005 controls, all of European ancestry. The analysts applied stringent filtering criteria, setting a genome-wide significance threshold of five times ten to the minus eight, pruning variants for linkage disequilibrium, and excluding weak instruments with F-statistics below ten to guard against bias.

In the forward analysis, which asked whether antibody responses cause Graves’ disease, the answer after rigorous multiple-testing correction was a resounding no. None of the forty-six phenotypes was significantly associated with Graves’ disease risk after Bonferroni or false discovery rate correction. Three Epstein-Barr virus antibody phenotypes showed nominal inverse associations under the inverse-variance weighted method, the workhorse estimator of the field, but these signals dissolved under correction. Worse, sensitivity analyses revealed heterogeneity among the genetic variants for all three, and the ZEBRA antibody phenotype additionally showed evidence of horizontal pleiotropy, meaning the variants may influence the disease through pathways other than antibody levels. In plain terms, the tantalizing hints linking viral antibody responses to Graves’ risk do not survive scrutiny.

The reverse analysis told a more interesting story. When the researchers treated genetic liability to Graves’ disease as the exposure and antibody responses as the outcome, several associations survived multiple-testing correction. After jointly weighing correction thresholds, heterogeneity, and pleiotropy, five associations emerged without substantial violations of the instrumental-variable assumptions. Genetic liability to Graves’ disease was associated with lower cytomegalovirus pp52 and pp150 antibody levels, higher Helicobacter pylori UreA antibody levels, higher anti-polyomavirus 2 IgG seropositivity, and higher anti-Merkel cell polyomavirus IgG seropositivity. Of these, only the polyomavirus 2 association remained significant after false discovery rate correction, with an odds ratio of 1.156 and a confidence interval of 1.050 to 1.273, though it did not reach the stricter Bonferroni threshold. The remaining four were nominally significant and must be treated as hypothesis-generating rather than definitive.

Steiger directionality testing added a crucial layer of confidence. Applicable to the thirty-one continuous antibody-level phenotypes, the test evaluates which trait explains more genetic variance in the other. For twenty-nine of the thirty-one phenotypes, the variance explained in Graves’ disease exceeded that explained in the antibody phenotype, and the proposed direction, Graves’ liability influencing antibody levels rather than the reverse, was statistically supported. Among the five prioritized associations, directionality was confirmed for the three continuous antibody-level traits, while the two binary seropositivity phenotypes could not be formally assessed because the required counts of seropositive and seronegative participants were unavailable.

The biological context for these findings is rich. Cytomegalovirus, a ubiquitous DNA virus that lurks in tissues and can reactivate during immune deficiency, has been implicated in other autoimmune conditions, with elevated pp52 antibodies observed in systemic lupus erythematosus and reduced levels seen in multiple sclerosis, mirroring the lower pp52 levels the genetic analysis associated with Graves’ liability. Antibodies against the pp150 protein, which is central to viral maturation and release, have been shown to cross-react with the human CIP2A protein, contributing to the death of CD56-bright natural killer cells, a cell population whose depletion is a shared feature of many autoimmune diseases. Helicobacter pylori, best known for gastritis and gastric cancer, has documented links to autoimmune thyroid disease, and antibodies against the bacterium can cross-react with thyroid tissue, while proteins such as urease may share epitopes with thyroid follicular cells, a textbook setup for molecular mimicry. Polyomaviruses add further intrigue, as Simian virus 40-like sequences have been detected in thyroid tissue from a subset of Graves’ disease patients.

The authors are careful to frame their conclusions at the level of genetically predicted traits rather than established immunological mechanisms, and they acknowledge clear limitations. The analyses relied on European-ancestry data, so generalizability across populations remains uncertain, and the Graves’ disease genetic data came from a single Finnish repository. The antibody phenotypes themselves are proxies for humoral immunity and cannot capture acute pathogen exposure, the timing of infection, viral reactivation, cellular immunity, or local immune events within thyroid tissue. Still, the big picture is striking: the study finds no credible evidence that antibody responses to infections cause Graves’ disease, while lending genetic weight to the idea that Graves’ disease, or the biology underlying it, reshapes the antibody landscape against cytomegalovirus, Helicobacter pylori, and polyomaviruses. If those findings hold up in longitudinal and mechanistic work, the field may need to stop asking which infection triggers Graves’ disease and start asking what the disease does to the immune system’s memory of infection.

Subject of Research: Bidirectional Mendelian randomization analysis of antibody-mediated immune responses to infections and Graves' disease

Article Title: The causal relationship between antibody-mediated immune responses to infections and Graves’ disease: a bi-directional Mendelian randomization analysis

Article References: Zhang, X., Xu, W., Liu, Z., Zhang, J., & Yang, X. (2026). The causal relationship between antibody-mediated immune responses to infections and Graves’ disease: a bi-directional Mendelian randomization analysis. Experimental Biology and Medicine, 251, Article 11238. https://doi.org/10.3389/ebm.2026.11238

Image Credits: AI Generated

DOI: 10.3389/ebm.2026.11238

Keywords: Graves' disease, Mendelian randomization, autoimmune thyroid disease, antibody responses, Epstein-Barr virus, cytomegalovirus, Helicobacter pylori, polyomavirus, genome-wide association study, molecular mimicry, humoral immunity, thyroid autoimmunity