Urinary metabolomics for noninvasive detection of borderline and acute T cell-mediated rejection in children after kidney transplantation

Abstract

The goal of this study was to evaluate the utility of urinary metabolomics for noninvasive diagnosis of T cell-mediated rejection (TCMR) in pediatric kidney transplant recipients. Urine samples (n=277) from 57 patients with surveillance or indication kidney biopsies were assayedfor 134 uniquemetabolitesbyquantitative mass spectrometry. Samples without TCMR (n=183) were compared to borderline tubulitis (n=54) and TCMR (n=30). Partial least squares discriminant analysis identified distinct classifiers for TCMR (area under receiver operating characteristic curve [AUC]=0.892; 95% confidence interval [CI] 0.827-0.957) and borderline tubulitis (AUC=0.836; 95%CI 0.781-0.892), respectively. Application of the TCMR classifier to borderline tubulitis samples yielded a discriminant score (-0.47±0.33)midway between TCMR (-0.20±0.34) and No TCMR (-0.80±0.32) (p<0.001 for all comparisons). Discriminant scoring for combined borderline/TCMR versus No TCMR (AUC=0.900; 95% CI 0.859-0.940) applied to a validation cohort robustly distinguished between samples with (-0.08±0.52) and without (-0.65±0.54, p<0.001) borderline/TCMR (p<0.001). The TCMR discriminant score was driven by histological t-score, ctscore, donor-specific antibody and biopsy indication, and was unaffected by renal function, interstitial or microcirculatory inflammation, interstitial fibrosis or pyuria. These preliminary findings suggest that urinary metabolomics is a sensitive, specific and noninvasive tool for TCMR identification that is superior to serum creatinine, with minimal confounding by other allograft injury processes.