| [Background] Morphological fiber phenotypes can reflect diverse molecular dysregulation of metabolic dysfunction-associated steatotic liver disease (MASLD); however, current histological assessment evaluates only semi-quantitative fibrosis stage. [Methods] Liver biopsy specimens from 94 MASLD patients were subjected to AI-based FibroNest algorithm, using H&E and Masson's Trichrome staining, for morphological fiber phenotyping and hepatic bulk transcriptome analyses. To reveal the microenvironment of pathological fiber phenotypes, single-cell spatial transcriptomics using CosMx (NanoString) was applied to four MASLD-affected liver samples. [Results] FibroNest identified 331 morphological fiber phenotypes, and their principal components 1 to 8 explained >75 percent of the variance in the space of fiber phenotypes; therefore, we termed them FibroPCs 1 to 8, respectively. FibroPCs captured distinct morphological fiber phenotypes. For instance, FibroPCs 1 and 4 represented the abundance of long-assembled and branched fine fibers, respectively. Expectedly, FibroPC1 was highly correlated with histological fibrosis stage (rho=0.59; p <0.001). Comparisons between FibroPCs and clinicohistological characteristics revealed high FibroPC4 in MASLD patients with concurrent hepatocellular carcinoma (HCC). Consistently, only FibroPC4 was associated with high-risk prediction of the hepatic transcriptome signature, PLS-MASLD, associated with HCC development, suggesting that a niche with high FibroPC4 represented an HCC-prone microenvironment. Among >200,000 cells detected by CosMx, IGFBP7-positive hepatic stellate cells accumulated in FibroPC4-high areas and interacted with neighboring cells, likely promoting HCC development. [Conclusion] Morphological fiber phenotypes inform molecular dysregulation of MASLD. |
