Objective
This proposal’s overall goal is to harness control over the functional maturation of human stem cell-derived islets (SC-islets), using artificial intelligence-driven bioelectronics as a radically different approach to building fully functional SC-islets for replacement therapy to cure T1D. Our specific objectives are to infer SC-islet cell maturation stages from continuous electrical activity readouts, and to then train AI algorithms to find optimal electrical stimulation conditions that accelerate progression through these stages.
Background Rationale
Type 1 diabetes (T1D) afflicts 8.75 million people worldwide, including 2 million Americans. Chronic islet autoimmunity prompts selective death of β cells, causing insulin deficiency and hyper-glycemia. T1D patients suffer persistent glucose dysregulation, which ultimately reduces their life expectancy. Daily insulin injections can help control glucose, but are expensive, onerous, and carry risk of ketoacidosis and coma.
Pancreatic islet transplants can cure T1D but are limited by a scarcity of acceptable islets from cadaveric donors. Islets differentiated from human pluripotent stem cells (SC-islets) represent an unlimited cell supply to meet the demand for transplantable islets as T1D cell therapy. Alas, they face bioprocessing (consistency, scalability, durability) and physiologic (composition, heterogeneity, maturity) challenges. These challenges limit the potential of SC-islets as a broadly translatable curative therapy.
Others and we have devised methods to foster SC-islet maturation in vitro that enhance their ability to reverse diabetes in vivo. Through such approaches, SC-islets offer an unprecedented human model to recapitulate and study islet maturation.
Islet maturation studies have been historically hindered by a lack of methods to stably trace single-cell activities across an islet. Recently, in collaboration with Prof. Jia Liu, we devised strategies to implant electrode arrays across SC-islet volumes, effectively building “cyborg” organoids that pioneer chronic 3D electrophysiology in intact islets with cell-level resolution and minimal impact on cell identity and function.
The gene programs underlying islet functional states are poorly understood. We have developed methos that combine tissue clearing and immunostaining with in situ single-cell transcriptomics to map gene and protein expression in cyborg SC-islets. Using electronic barcodes to map sensor positions, we are able to link these data to electrical recordings, bridging functional and molecular states at the single-cell level.
The ability of electrical stimulation to promote maturation of stem cell-derived organoids has been recently demonstrated. Yet the numerous tunable parameters for stimulators make conventional control methods impractical. Recent advances in AI, particularly reinforcement learning, offer unprecedented solutions. Here, we hypothesize that integrating cyborg SC-islets with AI into an AI-cyborg system that uses continuous electrical sensing to guide stimulation can achieve real-time, closed-loop modulation of SC-islet function. To test this hypothesis, our rationale is to combine reinforcement learning with organoid-embedded bioelectronics to achieve closed-loop control of SC-islet maturity, a new frontier in T1D cell therapy.
Description of Project
Transplantable pancreatic islets differentiated from human pluripotent stem cells (SC-islets) are likely to become a broad therapeutic option for severe type 1 diabetes (T1D) but suffer from underdeveloped function. This project seeks to control the functional maturation of SC-islets using tissue-embedded bioelectronics driven by artificial intelligence (AI) as a radically different approach to building fully functional SC-islets for curative T1D replacement therapy. The approach combines soft, flexible nanoelectronics with learning algorithms to pioneer a bidirectional interface that allows long-term electrical sensing and stimulation. When integrated with SC-islets, this AI-driven “cyborg” system will continuously adapt and optimize stimulation parameters based on real-time mapping of cell molecular and functional states, making closed-loop control of SC-islet function possible for the first time.
Anticipated Outcome
We anticipate that combining flexible nanoelectronics with learning algorithms will result in a bidirectional interface for long-term electrical sensing and stimulation that, when integrated with SC-islets, will be able to continuously adapt and optimize stimulation parameters based on real-time cell-state mapping, enabling closed-loop control of SC-islet function for the first time. We further anticipate that such AI-cyborg system will allow us to not just accelerate, but to standardize and expand functional maturation of SC-islets to realize the goal of broadly applicable T1D cell therapeutics.
Relevance to T1D
Islets differentiated from human pluripotent stem cells (SC-islets) represent an unlimited cell supply with the potential to solve the scarcity of transplantable islets for T1D cell therapy. However, SC-islets often exhibit immature function, characterized by poor insulin stimulation capacity, which limits their use as medicine. Thus, we believe that harnessing artificial intelligence to drive the functional maturation of SC-islets through electrical feedback stimulation will usher in next-generation SC-islet products that will transform T1D cell therapy.