Maintenance of internal homeostasis is a sophisticated process, during which almost all organs get involved. Liver plays a central role in metabolism and involves in endocrine, immunity, detoxification and storage, and therefore it communicates with distant organs through such mechanisms to regulate pathophysiological processes. Dysfunctional liver is often accompanied by pathological phenotypes of distant organs, including the eyes. Many reviews have focused on crosstalk between the liver and gut, the liver and brain, the liver and heart, the liver and kidney, but with no attention paid to the liver and eyes. In this review, we summarized intimate connections between the liver and the eyes from three aspects. Epidemiologically, we suggest liver-related, potential, protective and risk factors for typical eye disease as well as eye indicators connected with liver status. For molecular mechanism aspect, we elaborate their inter-organ crosstalk from metabolism (glucose, lipid, proteins, vitamin, and mineral), detoxification (ammonia and bilirubin), and immunity (complement and inflammation regulation) aspect. In clinical application part, we emphasize the latest advances in utilizing the liver-eye axis in disease diagnosis and therapy, involving artificial intelligence-deep learning-based novel diagnostic tools for detecting liver disease and adeno-associated viral vector-based gene therapy method for curing blinding eye disease. We aim to focus on and provide novel insights into liver and eyes communications and help resolve existed clinically significant issues.

Concrete epidemiological evidence has suggested the mutually-contributing effect respectively between nonalcoholic fatty liver disease (NAFLD), type 2 diabetes mellitus (T2DM), and periodontitis (PD); however, their shared crosstalk mechanism remains an open issue. The NAFLD, PD, and T2DM-related datasets were obtained from the NCBI GEO repository. Their common differentially expressed genes (DEGs) were identified and the functional enrichment analysis performed by the DAVID platform determined relevant biological processes and pathways. Then, the STRING database established a PPI network of such DEGs and topological analysis through Cytoscape 3.7.1 software along with the machine-learning analysis by the least absolute shrinkage and selection operator (LASSO) algorithm screened out hub characteristic genes. Their efficacy was validated by external datasets using the receiver operating characteristic (ROC) curve, and gene expression and location of the most robust one was determined using single-cell sequencing and immunohistochemical staining. Finally, the promising drugs were predicted through the CTD database, and the CB-DOCK 2 and Pymol platform mimicked molecular docking. Intersection of differentially expressed genes from three datasets identified 25 shared DEGs of the three diseases, which were enriched in MHC II-mediated antigen presenting process. PPI network and LASSO machine-learning analysis determined 4 feature genes, of which the MS4A6A gene mainly expressed by macrophages was the hub gene and key immune cell type. Molecular docking simulation chosen fenretinide as the most promising medicant for MS4A6A⁺ macrophages. MS4A6A⁺ macrophages were suggested to be important immune-related mediators in the progression of NAFLD, PD, and T2DM pathologies.