Non-invasive staging of chronic kidney allograft damage using urine metabolomic profiling

Chronic kidney allograft damage is characterized by IFTA and GS. We sought to identify urinary metabolite signatures associated with severity of IFTA and GS in pediatric kidney transplant recipients. Urine samples (n = 396) from 60 pediatric transplant recipients were obtained at the time of kidney biopsy and assayed for 133 metabolites by mass spectrometry. Metabolite profiles were quantified via PLS-DA. We used mixed-effects regression to identify laboratory and clinical predictors of histopathology. Urine samples (n = 174) without rejection or AKI were divided into training/validation sets (75:25%). Metabolite classifiers trained on IFTA severity and %GS showed strong statistical correlation (r = .73, P < .001 and r = .72; P < .001, respectively) and remained significant on the validation sets. Regression analysis identified additional clinical features that improved prediction: months post-transplant (GS, IFTA); and proteinuria, GFR, and age (GS only). Addition of clinical variables improved performance of the %GS classifier (AUC = 0.9; 95% CI = 0.85-0.96) but not for IFTA (AUC = 0.82; 95% CI = 0.71-0.92). Despite the presence of potentially confounding phenotypes, these findings were further validated in samples withheld for rejection or AKI. We identify urine metabolite classifiers for IFTA and GS, which may prove useful for non-invasive assessment of histopathological damage.