U-M researchers used an implant beneath the skin to uncover early immune signals that could help predict autoimmune type 1 diabetes progression and response to immunotherapy.
A biomaterial implant developed by University of Michigan researchers could help answer two challenging questions in type 1 diabetes: What are the early warning signs of the disease, and what can those early signals tell us about who might benefit from treatments targeting the immune system?
The study, recently published in The Journal of Clinical Investigation, builds on earlier work from the laboratory of Lonnie Shea, the Steven A. Goldstein Collegiate Professor of Biomedical Engineering. His team developed a microporous scaffold that, when implanted beneath the skin, recruits immune cells to form a synthetic immunological niche. By sampling the implanted niche and examining the cells and their gene expression, researchers can learn about disease-related immune activity without biopsying vital organs like the pancreas. The latest work uses the implant to study immune changes as type 1 diabetes progresses in mice and assess whether these changes could help guide treatment decisions.
“We wanted to use the implant to better understand how type 1 diabetes develops,” said Jyotirmoy Roy, a Ph.D. student in Dr. Shea’s lab and lead author of the study. “Could we identify the immune changes contributing to the disease at its earliest stages, and could we use that biological information to guide immunotherapy decisions?”
Type 1 diabetes is an autoimmune disease that develops when the immune system attacks beta cells in the pancreas that make insulin, the hormone that helps control blood sugar. This attack can begin well before blood sugar rises enough for clinical diagnosis.
The researchers used machine learning to look for patterns of gene activity in the implants before the mice showed symptoms. These patterns helped them identify which mice would later develop diabetes, well before their blood sugar levels began to rise.
The researchers found early signs of disease-related inflammation linked to macrophages, innate immune cells that help defend the body. Changes involving other immune cells, including T cells, became more prominent as the disease progressed. These findings helped reveal which parts of the immune system were involved at different stages of the disease.
“We were able to use the implant to identify signs of type 1 diabetes progression weeks before the mice became diabetic,” Roy said. “This gave us a chance to understand what was happening early in the disease, when there may still be an opportunity to intervene.”
To explore whether the findings might be relevant to people, the team worked with Jie Liu, U-M associate professor of computational medicine and bioinformatics, to analyze existing human datasets, including data from the Human Pancreas Analysis Program and a European clinical trial.
The patterns of gene activity identified in the implants were less apparent in blood samples from mice and humans. However, comparable patterns were detected in human tissues, including the spleen and lymph nodes, which helped distinguish individuals with type 1 diabetes from those without the disease.
“The implant captured immune signals associated with tissues that were less apparent in blood,” Roy said. “This may help explain why it could identify which mice would later develop diabetes at such an early stage.”
Because routine biopsies of the human pancreas are not practical, the implant could potentially serve as a more accessible site for capturing immune activity associated with disease-affected tissues.
“In patients, we cannot biopsy the pancreas to assess whether the disease is developing,” Dr. Shea said. “A biopsy of the implant could provide an accessible site to investigate tissue-level biology that is not captured through a blood draw alone.”
For Dr. Liu, the findings also support a broader effort to bring artificial intelligence, or AI, into type 1 diabetes research.
“One of the biggest challenges in type 1 diabetes is understanding why the disease progresses differently and why the same treatment does not work for everyone,” Dr. Liu said. “By combining the engineered implants with computational analysis, this study reveals early immune changes associated with disease progression and treatment responses in mice, with related signals found in human tissues. This also aligns with our MAI-T1D project’s vision: connecting multimodal AI and immunology to move toward earlier, more personalized intervention for T1D.”
After using the implant to identify mice likely to develop diabetes, the researchers took a closer look at immune cells in the pancreas. They found increased activity in the TNF-alpha signaling pathway, which promotes inflammation, in macrophages. Similar inflammatory signals were also found in pancreatic macrophages from people with type 1 diabetes.
The team then tested whether blocking TNF-alpha could help prevent diabetes in mice. Some mice benefited from the treatment, while others developed diabetes despite receiving it.
To investigate this difference, the researchers used computational tools to analyze immune signals in implants collected before treatment and develop a response score. The score reflected differences in the animals’ immune activity before they received the therapy.
Most mice with high response scores remained diabetes-free after treatment, while most with low scores developed diabetes despite receiving the same therapy. These findings suggest that the immune signals captured by the implant before treatment could help predict which mice would benefit from the treatment.
“By looking at immune activity before treatment, we could identify which mice were most likely to benefit,” Roy said. “In the future, the implant could help clinicians identify which patients are most likely to benefit from a particular treatment.”
Understanding a complex disease like type 1 diabetes takes more than one field of expertise. This study brought together researchers across the University of Michigan to connect advances in biomaterials, immunology, computational analysis and molecular profiling—turning signals from an implant into insights about disease progression and treatment response.
“One of the things that made this work possible was the collaborative nature of Michigan,” Roy said. “Each team helped us understand a different part of the picture. Working together allowed us to connect what we measured in the implant with what was happening in the disease.”
Costas Lyssiotis, the Maisel Research Professor of Oncology and professor of molecular and integrative physiology, whose group contributed to the implant’s molecular profiling, emphasized the importance of that collaboration.
“This work offers a platform to better understand the biology driving type 1 diabetes and could help identify who is most likely to benefit from treatment,” he said. “These insights grew out of close collaboration among colleagues in Biomedical Engineering, Bioinformatics and Michigan Medicine, each bringing a different perspective and expertise.”
Immunotherapies can slow type 1 diabetes, but their benefits vary from person to person. Predicting who is likely to respond could help doctors choose treatments while sparing others unnecessary side effects.
“We want to help the clinicians choose the treatment most likely to work for each patient,” Roy said. “Understanding a person’s immune activity using the implant could help make that choice more informed.”
With support from Breakthrough T1D, the researchers plan to investigate whether the implant can monitor responses to anti-CD3 immunotherapy, a treatment already used in clinics to delay type 1 diabetes in eligible patients.
The findings remain preclinical, and studies in people will be needed to establish the implant’s safety and reliability. Still, the work points toward a future in which biomaterial technologies could help researchers observe otherwise inaccessible immune activity and support more individualized approaches to type 1 diabetes care.