Gene expression changes are associated with loss of kidney graft function and interstitial fibrosis and tubular atrophy: diagnosis versus prediction

Abstract

Background: Loss of kidney graft function due to interstitial fibrosis (IF) and tubular atrophy (TA) is the most common cause of kidney allograft loss.

Methods: One hundred one allograft tissues (26 samples with IF/TA, 17 normal allografts, and an independent biopsy group collected at 3 month [n=34] posttransplantation) underwent microarray analysis to identify early detection/diagnostic biomarkers of IF/TA. Profiling of 24 allograft biopsies collected at or after 9-month posttransplantation (range 9-18 months) was used for validation. Three-month posttransplantation biopsies were classified as IF/TA nonprogressors (group 1) or progressors (group 2) using graft function and histology at 9-month posttransplantation.

Results: We identified 2223 differentially expressed probe sets between IF/TA and normal allograft biopsies using a Bonferroni correction. Genes up-regulated in IF/TA were primarily involved in pathways related to T-cell activation, natural killer cell-mediated cytotoxicity, and programmed cell death. A least absolute shrinkage and selection operator model was derived from the differentially expressed probe sets, resulting in a final model that included 10 probe sets and had 100% training set accuracy. The N-fold crossvalidated error was 2.4% (sensitivity 95.8% and specificity 100%). When 3-month biopsies were tested using the model, all the samples were classified as normal. However, evaluating gene expression of the 3-month biopsies and fitting a new penalized model, 100% sensitivity was observed in classifying the samples as group1 or 2. This model was evaluated in the sample set collected at or after 9-month posttransplantation.

Conclusions: An IF/TA gene expression signature was identified, and it was useful for diagnosis but not prediction. However, gene expression profiles at 3 months might predict IF/TA progression.

Figure 1 –. Study design.

The design includes a two step study. In the step 1, a molecular signature of IF/TA samples is established when compared with normal NA biopsies. A volcano plot from the IF/TA vs. NA comparison where the difference between average RMA expression between IF/TA and NA is plotted on the x-axis and –log₁₀(P-value) is plotted on the y-axis. Blue points correspond to probe sets significant using a Bonferroni correction (upper right panel of the Figure). After establishing a the IF/TA signature, a LASSO model was derived from the differentially expressed probesets, resulting in a final model of 10 probesets with 100% training set accuracy. Using N-fold cross-validation, the 41 different LASSO models included an average of 8 probesets (range 2 to 12). The N-fold cross-validated error was 2.4% with only 1 IF/TA sample being misclassified (sensitivity 95.8%, specificity 100%). However, when the final LASSO model was applied to gene expression data from an independently collected set of protocol biopsies (34 biopsies taken at 3-months post-transplantation) to test the predictive ability of the detected gene expression changes, all samples were classified as NA. In a second step, 34 arrays collected at 3 months, were segregated into 2 groups based on eGFR histological findings at 9-months post-transplantation and: Group 1 (G1) had 24 samples with eGFR >45 mL/min and normal histology, and Group 2 (G2) included 10 samples with eGFR ≤45 mL/min and IF/TA (IF (ci≥1) and TA (ct≥1) involves more than 25% of the cortical area. A volcano plot from the G1 vs. G2 comparison where the difference between average RMA expression between G1 and G2 is plotted on the x-axis and –log₁₀(P-value) is plotted on the y-axis. Blue points correspond to probe sets significant using a Bonferroni correction (botton right panel of the Figure). Also, a LASSO model was derived from the differentially expressed probesets, resulting in a final model of 17 probesets with 100% training set accuracy. Finally, the model was tested in an independent set of samples.

Figure 2 –

Heatmap and dendograms after applying hierarchical clustering using Ward’s method with 1-|ρ| as the dissimilarity measure to the 80 probe sets retained from the max-min filter. Cluster 1 includes 15 NA and 1 IF/TA sample; Cluster 2 includes 20 IF/TA and no NA samples; and Cluster 3 includes 2 NA and 4 IF/TA samples. The color bar indicates the sample classification: Blue = NA; Red = IF/TA.

Figure 3 –

Canonical pathways identified from the differentially expressed genes. Each bar represents the percentage of up-regulated (red) or down-regulated (green) or unaffected/undetected (white) genes within the identified pathway. The line represents the –log(p-value).