Recent studies reveal significant genetic heterogeneity within type 1 diabetes, indicating it could be a family of related disorders with different biological pathways, paving the way for more personalised diagnosis and treatments.
Type 1 diabetes is beginning to look less like a single autoimmune illness and more like a family of related disorders that can end in the same diagnosis by different biological routes. That argument strengthened on 31 August, when a Diabetologia study split patients by two well-known HLA risk backgrounds, DR3 and DR4, and found only moderate genetic overlap between them. It follows an April Nature Genetics paper from UC San Diego that used machine learning to sort patients into four genetic sub-types with different ages of onset and different patterns of complications.
In the newer Diabetologia work, researchers in the US and UK analysed 9,091 people with type 1 diabetes and 14,157 controls. They reported a genetic correlation of 0.6 between DR3-linked and DR4-linked disease, low enough, they argued, to point to materially different disease processes. Signals tied to DR3 cases were enriched in mast cells, immune cells better known for allergy and inflammation, while DR4 cases showed stronger links to T-cell pathways, and the IL2 locus had a bigger effect in DR4. The paper said that divide was on a par with the genetic distance seen between paired disorders such as schizophrenia and bipolar disorder. That fits with the broader message from the April study, which argued that type 1 diabetes contains more heterogeneity than older scoring systems captured.
The April Nature Genetics study was much larger. According to the UC San Diego account and independent reports by News Medical and Newswise, the researchers examined DNA from 20,355 people of European ancestry with type 1 diabetes and 797,363 without it, then analysed the main immune-gene region in more than 29,000 people. From that they built T1GRS, a machine-learning score using 199 variants. Emily Griffin, one of the co-first authors quoted by UC San Diego, said the MHC contains genetic “blocks” that are heavily concentrated in people with type 1 diabetes: “If you have them, it doesn’t mean that you’re going to get diabetes, but if you don’t have them, it means you have a very low chance of getting diabetes.”
TJ Sears, another co-first author, said the tool picked up people “who get diabetes but don’t have known high-risk genetic regions” far better than earlier tests. Medical Xpress reported that when the model was tried on the US National Institutes of Health’s All of Us programme and on nPOD, a pancreatic tissue research resource, it still predicted risk with about 87% accuracy and recreated the same four groups seen in the training data. In the paper itself, those validation results were given as area-under-curve scores of 0.872 in All of Us and 0.887 in nPOD, suggesting the score remained strong outside the original dataset.
Those four groups go well beyond the older idea of a simple DR3/DR4 split. Clear Sky Science’s summary of the Nature Genetics paper described two clusters dominated by classic MHC risk, one driven more by T-cell genes and one enriched for pancreatic-cell genes. The pancreas-enriched group tended to develop disease later, but it also carried the heaviest burden of complications, with higher rates of kidney disease, nerve damage and cardiovascular problems, a pattern the paper said was reproduced in All of Us. The same summary said the model uncovered 154 interacting pairs of variants, many linking the MHC to the insulin gene, and could also help separate type 1 from type 2 diabetes in independent cohorts.
That has obvious implications for screening and prevention. Carolyn McGrail said in the UC San Diego material that a broader score could catch a wider pool of high-risk children and adults, support closer monitoring and help identify candidates for preventative therapies such as teplizumab. Bioengineer, another report on the same release, added that earlier recognition could reduce the risk of diabetic ketoacidosis at diagnosis and eventually allow care plans to be shaped around the dominant biology in each genetic group.
The April reports did not present the discovery figures in exactly the same way. News Medical, Medical Xpress and Newswise said the team confirmed 79 known loci and found 13 new ones. Clear Sky Science, summarising the journal paper, described 89 previously known regions and eight new ones. The paper itself refers to 160 risk signals and a model built from 102 non-MHC loci plus HLA variants. Those are not necessarily contradictory counts, but they do show how quickly the language shifts when researchers move between broad genomic regions, loci and individual signals.
There are still important limits. The main training set was drawn from people of European ancestry, and Bioengineer noted that the authors called for larger, more ethnically diverse cohorts and for environmental and lifestyle data to be folded into future models. Even so, Clear Sky Science reported that the new score performed comparably to a dedicated tool in African American participants. When Biotech Networks syndicated the UC San Diego study on 3 May, it presented T1GRS as a route to earlier diagnosis and more personalised treatment. The newer Diabetologia paper suggests that ambition may depend on accepting a more awkward truth: type 1 diabetes may share a name while arriving by more than one path.
Disclaimer: This content is for informational purposes only and is not intended to be a substitute for professional medical judgment, advice, diagnosis, or treatment.
