Self-Feedback Synaptic Networks Program In Vivo Stem Cells Multi-differentiation and Repair

Stem cell therapy is rapidly advancing as a transformative approach in regenerative medicine. Due to self-renewal capabilities and plenipotentiary differentiation potential—with low risk of immune rejection—stem cells (SCs) offer the possibility of regenerating specific damaged tissues and organs. Traditional strategies to guide directed SC differentiation primarily includeinduction factors, microenvironment simulation, gene editing, and material scaffolding. Current approaches are capable of inducing the directed differentiation of SCs into a single lineage for tissue repair. However, tissue repair constitutes a complex process including tissue healing and vascular regeneration. Thus, these methods exhibit significant limitations regarding in vivo precision, the efficiency of inducing multi-directional differentiation, real-time monitoring capabilities and biosafety caused by animal-derived products. In contrast, functional stimulation (FS) with external interventions (namly electrical signals in this research) in SCs differentiation presents unique advantages: it offers in vivo modulation, possesses the potential to guide multi-lineage differentiation, high-resolution dynamic sensing of the microenvironment, and maintains excellent bio-compatibility. Consequently, FS is regarded as one of the most promising technological pathways for achieving precise in vivo stem cell-based tissue repair. Despite the evident advantages of FS, existing technologies have largely failed to integrate and improve the above advantages to achieve the multicellular synergistic differentiation and in vivo controllability required for complex process of tissue regeneration. The root cause lies in the inability of current methods to satisfy the core requirements of in vivo repair: 1) the regulation of SCs differentiation is inherently multidimensional, involving the synergistic action of parameters such as voltage, frequency, and waveform, while existing low-dimensional parameter screening strategies fail to cover this complex differentiation lineage; 2) during the repair process, the bioimpedance of the injury micro-environment evolves dynamically due to changes in repair degree, micro-environmental composition, and ion migration, leading to stimulation deviations, while current devices only senses changes in bioimpedance without further regulation; 3) complex injury situations require the induction of distinct differentiation endpoints (e.g. cells that regenerate defective tissues and endothelial cells that form new blood vessels) at specific locations, while existing technologies remain unable to achieve simultaneous repair of multiple tissues. Therefore, there is an urgent need for a spatiotemporally customized, multi-dimensional FS technology capable of dynamically sensing micro-environmental changes and performing self-feedback regulation to replace current static, single-mode stimulation paradigms. To address these challenges, we developed a neuromorphic closed-loop system integrated with deep learning algorithms—a Self-Feedback Synaptic Network(SFSN) capable of real-time bio-computation and multi-dimensional functional stimulation. Utilizing an implantable 1-kb synaptic array, the system senses biological signals and delivers customized stimulation based on the dynamic state of the tissue micro-environment. achieved high-dimensional, precise control over rat bone marrow mesenchymal stem cells (rBMSCs) differentiation into multiple lineages. By mapping the closed-train synaptic network weights onto the implantable 1-kb synapse array, the system performs in-situ hardware inference. This approach successfully guides hydrogel-delivered SCs toward target lineages at precise injury sites—such as fibroblasts, osteoblasts, and tenocytes—while simultaneously promoting endothelial differentiation in regions requiring vascular ingrowth to support site-specific vascularization, ultimately achieving integrated regeneration of structural and functional tissue with in vivo controllability. Overall design: There are 7 groups: Control; SFSN-FB (Fibroblasts), SFSN-Ot (osteoblasts), SFSN-EC (endothelial cells), SFSN-TC (tendon cells), SFSN-Neu (neuron-like cells) and SFSN-Adipo( adipocytes).