A spatial atlas of the healthy human liver from live donors

The new benchmark: A spatially resolved atlas of the healthy human liver

Recent advances in high-resolution analytical methods, particularly in the fields of spatial transcriptomic and proteomic analyses have enabled comprehensive multi-omic analyses of human tissues and substantially advanced our understanding of the pathophysiology of liver diseases (for a recent review, see Suo et al., DOI: 10.1136/gutjnl-2024-332105). However, a limitation exists when comparing pathological changes in disease to the “healthy” liver, as the availability of fully healthy human liver samples is limited Section Head for Translational Systems Immunology in Hepatogastroenterology, University Medical Center Freiburg, Department of Internal Medicine II, Hugstetter Str. 55, 79106 Freiburg, Germanyfor obvious ethical reasons. The commonly used control tissues in many studies are derived from deceased donors or surgical specimens from tumor resections, in which macroscopically “normal” appearing regions are selected; however, these tissues often exhibit postmortem or disease-related changes upon closer analysis. In the present study, the research group led by S. Itzkovitz elegantly resolved this difficulty by performing spatially resolved multi-omic analyses of samples obtained from healthy liver donors during living donor liver transplant (LDLT). Eight liver tissue samples from healthy donors were analyzed using spatially resolved transcriptomics and high-multiplex imaging, complemented by RNA sequencing of isolated cell nuclei. The results were compared with those obtained from macroscopically “healthy” tissue resected from tumor patients, as well as with liver tissue from various animal species. Some expected differences were observed. For example, the livers from living donors examined in the study exhibited significantly different transcriptome profiles from resection specimens of macroscopically normal liver tissue from cancer patients. However, it should be noted that the resection specimens from cancer patients were obtained from a highly heterogeneous group of older patients, some of whom had received prior chemotherapy, among other relevant differences. A better-matched control group would have been desirable here. Nevertheless, the comparison underscores the presence of pathophysiological changes that may exist in macroscopically unremarkable control specimens, with implications for the selection of control groups in liver studies.
The strength of spatially resolved transcriptomics lies primarily in its ability to track gene expression according to its spatial distribution to obtain insights into the tissue microarchitecture. In the liver, the zonation of metabolic processes between the periportal and pericentral regions plays a crucial role in liver function. The authors therefore specifically investigated the zonation of gene expression involved in liver metabolism and compared these findings with various animal models. Interestingly, some distinct differences emerged compared to the commonly used mouse model. For example, human pericentral hepatocytes exhibited high expression of genes involved in gluconeogenesis or the urea cycle, which in mice are expressed in the periportal region. Interestingly, these differences were also observed in large-animal models such as pigs and cows. 
These analyses revealed distinct spatial specialization of certain metabolic processes in human hepatocytes. In contrast, key signaling pathways, such as the role of the WNT signaling pathway in the pericentral zone, were conserved across the species studied. The extent to which other species-specific factors, such as different hormonal signals, play a role here was not investigated further. These analyses do, however, indicate that results obtained in mouse models cannot always be directly transferred to humans, particularly with respect to metabolic processes, and require validation in human model systems whenever possible. 
Finally, the authors investigated the role of changes associated with mild hepatic steatosis. This was possible because, in LDLT, steatosis of up to 20% of the liver is allowed during liver donor selection, and some donors exhibited mild steatosis. Interestingly, these donors already showed an adaptive response in liver metabolism—specifically, increased expression of genes involved in beta-oxidation and altered expression of regulators of lipid droplets and mitochondrial proteins, consistent with a compensatory adaptation of hepatocytes to minimize lipid accumulation.
In summary, the publicly available atlas of the healthy liver (https://shalevapps.weizmann.ac.il/liver_app/) generated in this study, based on LDLT specimen, sets a high standard for the analysis of healthy liver tissue. It is an important new resource for understanding liver physiology and distinguishing it from pathophysiology. The data are particularly relevant for metabolic research in this field and may also be helpful for the development of zonated liver organoid systems.