“Background. Late graft failure is a major problem after kidney transplantation (KT) mainly due to chronic renal allograft dysfunction (CRAD). Data about common pathways of kidney fibrogenesis are available, but is still unclear the different susceptibility to fibrosis progression of individual patients. Epigenetic variations might explain this inter-individual variations.

Methods. Donor organ tissue samples were evaluated from 60 deceased donor KT recipients (n=40, were used as training set and n=20, as validation set), and classified as non-Progressors (NP) or Progressors (P) to CRAD based on graft function and histological findings at 2-yrs post-KT. DNA methylation (DNAm) was tested using Infinium 450K methylation assay. The Illumina's GenomeStudio Methylation Analysis Module was used to obtain the β values and detection of p-values for each probe. For each probe, the 'P' and the 'NP' samples were compared using a moderated t-test using the limma Bioconductor package. A penalized logistic regression model was fit to predict P vs. NP using the CpG sites as predictor variables. The final model was selected as that having the minimum AIC. To determine generalization error, N-fold cross-validation (CV) was performed. Differentially methylated CpG sites from the best model were tested in the validation set using Methylight reactions.

Results.1,188 CpG sites were differentially methylated (FDR < 0.05 and |deltaβ|>0.20) between groups (P vs. non-P), mainly affecting genes involved in inflammation, metabolism and oxidative stress). The plots for the top significant CpG sites showed some overlapping between groups. Then, a penalized logistic regression model was fit to predict P vs. NP. The final model included 21 CpG sites with 100% showing prediction accuracy, including CpG sites in the SP1, TNFSF11, and PRDM6 genes . The N-fold cross-validated error rate was 23.1%, with 85% sensitivity, 68.4% specificity, 73.9% positive predictive value, and 81.3% negative predictive value. The area under the curve from the N-fold CV procedure was 0.86, showing a better performance of a combined set of markers in classification of patient groups. The resulting panel validated with similar performance in the validation set.

Conclusions. In addition to identifying CpG sites differentially methylated, we identified a DNAm classifier with good sensitivity and specificity for classifying NP and P to CRAD patients.”

« Back to 2015 American Transplant Congress