Artificial Blood Vessels and AI Will Help Better Model the Human Pancreas

Type 1 diabetes is an autoimmune condition characterized by the immune system attacking and destroying the insulin-producing beta cells in the pancreas, leading to the body’s inability to regulate glucose levels. While researchers know that this immune attack occurs, the exact mechanisms underlying the development of type 1 diabetes remain unclear.

Some of the limitations in understanding this disease include the difficulty of replicating the beta cell environment in vitro and the static nature of most experimental measurements. Most diabetes cell culture models lack the full architecture of the human pancreas, namely the vascularization that sustains these cells, leaving researchers only able to look at isolated snapshots rather than continuous observations of the disease.

These shortcomings frustrated Amin Ardestani, a diabetes researcher at the University of Hull. Ardestani joined the institution in 2023 and met another new faculty member, Amirpasha Moetazedian, an engineer interested in biomaterials for regenerative medicine. Moetazedian told Ardestani about how his group developed in vitro modeling tools and was working on artificial blood vessels. Ardestani recalled thinking, “That could be cool because one of the major issues we have [when modeling] the pancreas in the lab is that lack of vascularization.”

The two struck up a collaboration and invited colleague Mo Lotfollahi, a researcher at the Wellcome Sanger Institute who uses AI in cell engineering, to provide insights into how well the vascularized pancreas model replicated a real human organ.

To advance this work, the team secured funding from Biohub, a nonprofit research group that supports high-risk, high-reward interdisciplinary research projects. The team’s goal is to first optimize the in vitro pancreas model to replicate the function and physiology of the native organ and then incorporate immune cells to study the progression of type 1 diabetes in real time.

“Instead of just taking one static image or endpoint measurement that we have with most of the models we have, we want to follow the biology dynamically, almost like watching this process unfold. That is our ultimate goal,” Ardestani said.

The model starts with the artificial blood vessels developed by Moetazedian’s group. His team builds vasculatures with hydrogels using a microfluidic system, creating a more comparable scaffold for endothelial and smooth muscle cell adhesion than the traditional plastic of many organ-on-a-chip devices.¹ “From the microfluidic point of view, it's difficult to get that vascularization and multicellular aspect all in one platform,” Moetazedian said. He added that most approaches only incorporate a single cell type in vascular systems. “No one has looked at developing multicellular [models] which represent the architecture of pancreatic tissue in a very accurate way while integrating that vascularization to it.”

But this is exactly what the team intends to do. The researchers plan to print Ardestani’s beta cells and additional supporting stromal cells within this system to create a vascularized model of the human pancreas. They intend to use spatial transcriptomics to compare the function of this model to a real organ with the help of AI. Then, based on the gene expression patterns, the researchers will refine their system so that it more closely resembles the actual tissue.

Once they have an accurate pancreas model, the researchers will introduce immune cells, specifically T cells, to study how diabetes unfolds. Using microneedle sampling, the research team plans to study the proteomic and metabolomic changes in this system over time.

Ardestani admitted that the project could take a few years to optimize. However, he and his collaborators are excited about the potential to better understand diabetes and develop drugs for patients with this disease using a more representative cell culture model. “It could give us a clear picture [of] beta cell loss and type 1 diabetes,” Ardestani said.