AI-Designed Nanoparticles Reverse Liver Scarring by Restoring Tiny Pores in Liver Cells

Liver fibrosis, the excessive buildup of scar tissue in the liver, is driven in part by the transformation of specialized liver sinusoidal endothelial cells (LSECs). Healthy LSECs contain characteristic nanoscale pores called fenestrations organized into sieve plates, which facilitate metabolic exchange. During fibrosis, LSECs lose these structures in a process called capillarization, shifting to a pro-fibrogenic state that accelerates scarring. Reversing capillarization is therefore a compelling but underexplored therapeutic strategy.

Mesenchymal stem cell (MSC) transplantation is one of the few approaches known to reverse LSEC capillarization, but safety concerns (tumorigenicity, immune rejection), ethical issues, and high costs limit its clinical translation. This study set out to identify the active molecular agent MSCs use to restore LSECs, then replicate that effect with a safer, scalable nanoparticle system.

Using a co-culture system that separated physical contact between cells, researchers confirmed that MSC-conditioned medium restores LSEC fenestrations, and that this effect is abolished when extracellular vesicles are removed or RNase is added—pointing to a miRNA payload as the active factor. An AI model built on ChatGPT 4.0, trained on human and mouse miRNA databases (GSEA and TargetScan), was used to analyze RNA-seq data from fibrotic mouse LSECs treated with MSCs and proteomics data from cirrhotic human liver tissue. Both analyses independently converged on miR-325-3p as the key candidate. Validation in human fibrotic LSECs and mouse models confirmed that miR-325-3p mimics restore fenestrations within 48 hours, while blocking miR-325-3p in MSCs abolished their anti-fibrotic efficacy.

Mechanistically, miR-325-3p was found to suppress its target gene Ptprm (protein tyrosine phosphatase receptor type M), which regulates actin cytoskeleton dynamics. By downregulating Ptprm, miR-325-3p allows LSEC cytoskeletal remodeling that re-establishes fenestrations and sieve plates. The researchers then engineered spherical nucleic acid (SNA) nanoparticles by assembling sulfhydryl- and gold(I)-modified miR-325-3p into ~39 nm polyhedra via aurophilic interactions. These SNAs exploit the overexpression of scavenger receptor A (Scara) on fibrotic LSECs for selective cellular uptake. In three mouse models—CCl4-induced oxidative fibrosis, thioacetamide (TAA) drug-toxicity fibrosis, and bile duct ligation (BDL) cholestatic fibrosis—SNA treatment significantly restored LSEC fenestrations, reduced fibrotic markers (α-SMA, Collagen I, F4/80), increased differentiated LSEC markers (Lyve-1), and showed no adverse biosafety signals.

This work establishes a mechanistic basis for MSC-mediated reversal of liver capillarization and introduces a targeted nanoparticle platform as a clinically actionable alternative. By directly addressing the LSEC phenotype as a root driver of fibrosis, rather than downstream scarring, this approach represents a conceptual advance in liver disease therapy.