Liver sinusoidal scores: a new tool for the diagnosis and prognosis of chronic liver disease

Background and Aims: Sinusoidal cells drive the progression of chronic liver disease (CLD). Therefore, novel biomarkers for CLD should reflect the state of sinusoidal cells. However, the liver is mainly constituted from hepatocytes (80-90% of cells), while sinusoidal cells only represent the 10% of the cellular populations. Therefore, determination of biomarkers from plasma or bulk transcriptomics underrepresents smaller sinusoidal cell subpopulations, such as liver sinusoidal endothelial cells (LSECs), hepatic stellate cells (HSCs), and macrophages (MPs). On the other hand, recent studies performing single-cell sequencing allow for finer phenotype determination, but these analyses are too expensive for routine implementation in patient care. The aim of this study is to propose biologically-relevant markers of liver sinusoidal phenotype for the diagnosis and prognosis of CLDs. Methods: We estimated the relevant sinusoidal cell subpopulations in an internal cohort of 11 controls and 12 cirrhotic livers by gene deconvolution. Then, we reanalyzed published single-cell sequencing data in order to find a gene signature specific to these subpopulations. We combined the expression of these sinusoidal genes into 3 sinusoidal scores (endothelial, mesenchymal and macrophage) correlating with patient phenotypes. The ability of these scores to classify patients and predict fibrosis progression, regression and clinical decompensation was assessed in an internal cohort of 113 patients and further validated in 3 external cohorts for a total of 1008 patients with F3-F4 CLD (Simtuzumab, ATLAS & STELLAR trials). Results: We obtained a sinusoidal gene signature which was more specific for than traditional markers of LSEC capillarization, HSC activation and macrophage polarization. The three sinusoidal scores generated with these gene signatures were significantly higher in those patients with worse clinical parameters in our internal cohort and in the external validation cohort. The sinusoidal scores at baseline were significantly higher in patients that progressed from F3 to F4 fibrosis stage after 96 weeks of follow-up compared to those that stayed at F3. Similarly, the baseline sinusoidal scores were significantly lower in those patients that regressed from F4 to F3 compared to those that stayed at F4. Finally, the scores were able to predict decompensation in the following 96 weeks of follow-up (AUROC = 0.98) in the validation cohort and correlated with conventional parameters for evaluation of clinical decompensation (HVPG, Child-Pugh and MELD). Conclusions: This sinusoidal gene panel could aid in the diagnosis and prognosis of patients with CLD and predict future complications of cirrhosis. This technique can be easily assessed by RT-PCR and could be a useful tool for clinical decision making in the future, possibly aiding in assessment of drug response or in choosing the most relevant cell target for therapy.