Identification of a urine metabolite constellation characteristic for kidney allograft rejection

Abstract

Introduction: Allograft rejection is still an important complication after kidney transplantation. Currently, monitoring of these patients mostly relies on the measurement of serum creatinine and clinical evaluation. The gold standard for diagnosing allograft rejection, i.e. performing a renal biopsy is invasive and expensive. So far no adequate biomarkers are available for routine use.

Objectives: We aimed to develop a urine metabolite constellation that is characteristic for acute renal allograft rejection.

Methods: NMR-Spectroscopy was applied to a training cohort of transplant recipients with and without acute rejection.

Results: We obtained a metabolite constellation of four metabolites that shows promising performance to detect renal allograft rejection in the cohorts used (AUC of 0.72 and 0.74, respectively).

Conclusion: A metabolite constellation was defined with the potential for further development of an in-vitro diagnostic test that can support physicians in their clinical assessment of a kidney transplant patient.

Keywords: Diagnostic model; Kidney rejection; Metabolomics; NMR-spectroscopy.

Fig. 1

Overview of our modelling strategy. a Process from binned spectra towards candidate substance set. b Modelling with fitted metabolites towards final candidate models

Fig. 2

Data flow of analyzed samples in the training (a) and test cohort (b). Urine samples with sufficient volume for NMR measurement that passed a visual inspection were measured and the spectra were subjected to automatic quality control. In order to classify the remaining valid spectra as either case or control, they were integrated with the clinical data. In case of the training cohort, samples were filtered according to the biopsy result and the distance between the sample collection and the biopsy. All classified samples were then used for statistical modelling in two phases: first based on binned spectra and then based on quantified metabolites. At this point, two subsets were distinguished: samples from the early phase (i.e. day < 15 after transplantation) and late phase (i.e. day ≥ 15 after transplantation)

Fig. 3

Manual model selection based on AUC comparison in the test set. All graphs show the effect on AUC of different variations of models. a Analysis of effect of adding Lactate. Models substantially improves in performance when lactate was added as another independent variable—all data points are found above the diagonal. b Analysis of effect of exchanging Alanine and Hippurate: on average, AUC was greater, when Alanine was used as an independent variable instead of hippurate—the magenta points are mostly found above the blue points, although the absolute improvement is small. c Analysis of effect of adding DMA. Adding DMA to a model seems to improve performance (data pints above the black line). However the amount of performance gain is negligible at 0.01–0.04 points in the AUC

Fig. 4

Final candidate models. After selecting by model performance and assessing some feature alternatives (e.g. alanine vs. hippurate) we were left with a core model that comprises the features alanine, citrate and lactate. In addition there are a few models that add urea, glucose and/or glucuronate to that core feature set

Fig. 5

Performance in training and test cohort. ROC curve of fitter model for late phase. The blue area indicates 95% confidence intervals of the ROC curves. a Performance in the training set. b Performance in the test set using the strict case/control definition based on biopsy results. c Performance in the test set using the extended case/control definition that additionally includes control samples when no biopsy was performed due to lack of clinical symptoms