Defining an NK Cell–enriched Rejection-like Phenotype in Liver Transplant Biopsies From the INTERLIVER Study

Background. Initial analysis of liver transplant biopsies in the INTERLIVER study (ClinicalTrials.gov; unique identifier NCT03193151) using rejection-associated transcripts failed to find an antibody-mediated rejection state (ie, rich in natural killer [NK] cells and with interferon-gamma effects). We recently developed an optimization strategy in lung transplants that isolated an NK cell–enriched rejection-like (NKRL) state that was molecularly distinct from T cell–mediated rejection (TCMR). Here we apply the same strategy to a liver transplant biopsy population. Methods. We used this strategy to search for a molecular NKRL state in 765 consented liver transplant biopsies collected at participating international centers for gold-standard histology and molecular assessment by genome-wide microarrays. Validation through a training set-test set approach of an optimized selection of variables as inputs into unsupervised rejection classification identified an NKRL state in livers. Results. The full model classified 765 biopsies into the following molecular phenotypes, characterized by their gene expression: no-rejection 54%, TCMR 16%, NKRL 13%, and injury 16%. Top TCMR transcripts were expressed in effector T cells; top NKRL transcripts were almost exclusively expressed in NK cells; and both had increased interferon-γ-inducible transcripts, which were more pronounced in TCMR. Most TCMR biopsies had significant parenchymal injury, molecular fibrosis, and abnormal biochemistry. NKRL biopsies had no excess of injury, fibrosis, or biochemistry abnormalities. Conclusions. Optimized rejection algorithms indicate that some liver transplants manifest an NKRL state that is well tolerated in the short term postbiopsy and with minimal injury and relatively normal biochemistry, while also underscoring the potential of TCMR to produce extensive parenchymal injury.