Evaluation of gene panel mRNAs in urine samples of kidney transplant recipients as a non-invasive tool of graft function

Abstract

Non-invasive monitoring may be useful after kidney transplantation (KT), particularly for predicting acute rejection (AR). It is less clear whether chronic allograft nephropathy (CAN) is also associated with changes in urine cells. To identify non-invasive markers of allograft function in kidney transplant patients (KTP), mRNA levels of AGT, TGF-beta1, EGFR, IFN-gamma, TSP-1, and IL-10 in urine (Ur) samples were studied using QRT-PCR. Ninety-five KTP and 111 Ur samples were evaluated. Patients (Pts) were divided as, within six months (N = 31), and with more than six months post-KT (N = 64). KTP with more than six months post-KT were classified as KTP with stable kidney function (SKF) (N = 32), KTP with SKF (creatinine < 2 mg/dL) and proteinuria > 500 mg/24 h (N = 18), and KTP with biopsy proven CAN (N = 14). F-test was used to test for equality of variances between groups. IL-10 mRNA was decreased in Ur samples from KTP with less than six months post-KT (P = 0.005). For KTR groups with more than six months post-KT, AGT and EGFR mRNA were statistically different among KTP with SKF, KTP with SKF and proteinuria, and CAN Pts (P = 0.003, and P = 0.01), with KTP with SKF having higher mean expression. TSP-1 mRNA levels also were significantly different among these three groups (P = 0.04), with higher expression observed in CAN Pts. Using the random forest algorithm, AGT, EGFR, and TGF-beta1 were identified as predictors of CAN, SKF, SKF with proteinuria. A characteristic pattern of mRNA levels in the different KTP groups was observed indicating that the mRNA levels in Ur cells might reflect allograft function.

Figures

Figure 1

A-Two variable importance measures for each predictor covariate from the random forest classification predicting 2 RT-PCR expression for EGFR, AGT, IL10, and TGF-β1 displayed by < six months versus ≥ six months post-transplant.

Figure 2

Receiver operating characteristic curves (ROC) for the better predictors. For each gene ROC analysis predicting

Figure 3

For ≥ six months samples, two variable importance measures for each predictor covariate from the random forest classification predicting CAN, normal, or proteinuria: the mean decrease in accuracy (left panel) and mean decrease in Gini impurity (right panel). For ≥ six-month samples, box plots of the log2 RT-PCR expression for EGFR, AGT, and TGF-β 1 displayed by CAN, stable kidney function, and stable kidney function with proteinuria. CAN: patients with biopsy proven CAN, Normal: kidney transplant patients with stable kidney function, Proteinuria: kidney transplant patients with stable kidney function (creatinine levels < 2 mg/dL) and proteinuria.

Figure 4

Receiver operating characteristic curves (ROC) for the better predictors. For each studied gene a ROC analysis predicting CAN versus SKF was performed. Afterward, the area under the ROC curve was estimated, A- TGF-β1 (AUC = 0.84); B- EGFR (AUC = 0.71).