Biomarkers for early and late stage chronic allograft nephropathy by proteogenomic profiling of peripheral blood

Abstract

Background: Despite significant improvements in life expectancy of kidney transplant patients due to advances in surgery and immunosuppression, Chronic Allograft Nephropathy (CAN) remains a daunting problem. A complex network of cellular mechanisms in both graft and peripheral immune compartments complicates the non-invasive diagnosis of CAN, which still requires biopsy histology. This is compounded by non-immunological factors contributing to graft injury. There is a pressing need to identify and validate minimally invasive biomarkers for CAN to serve as early predictors of graft loss and as metrics for managing long-term immunosuppression.

Methods: We used DNA microarrays, tandem mass spectroscopy proteomics and bioinformatics to identify genomic and proteomic markers of mild and moderate/severe CAN in peripheral blood of two distinct cohorts (n = 77 total) of kidney transplant patients with biopsy-documented histology.

Findings: Gene expression profiles reveal over 2400 genes for mild CAN, and over 700 for moderate/severe CAN. A consensus analysis reveals 393 (mild) and 63 (moderate/severe) final candidates as CAN markers with predictive accuracy of 80% (mild) and 92% (moderate/severe). Proteomic profiles show over 500 candidates each, for both stages of CAN including 302 proteins unique to mild and 509 unique to moderate/severe CAN.

Conclusions: This study identifies several unique signatures of transcript and protein biomarkers with high predictive accuracies for mild and moderate/severe CAN, the most common cause of late allograft failure. These biomarkers are the necessary first step to a proteogenomic classification of CAN based on peripheral blood profiling and will be the targets of a prospective clinical validation study.

Conflict of interest statement

Competing Interests: The authors have declared that no competing interests exist.

Figures

Figure 1. Pie charts showing the Gene Ontology annotations for both Test Sets for Banff 0 vs. Banff 1 (mild CAN).

Each slice of the pie chart represents the percentage of genes represented by that functional class. A) Test Set 1 (PBL) for Banff 0 vs. Banff 1; B) Test Set 2 (Whole Blood) for Banff 0 vs. Banff 1. The first key point is that there is no difference in the general groups of differentially expressed, functional genes assigned in an unbiased fashion by analysis using Gene Ontology whether we are interrogating the profiles of PBL or whole blood. The second key point is that there are a number of genes representing different pathways connected to immune/inflammatory and tissue injury mechanisms.

Figure 2. Class prediction analysis of Banff 0 vs. Banff 1 (mild CAN) based on Diagonal Linear Discriminant Analysis for the top 50 Banff 0 vs. Banff 1 consensus genes ranked by p values.

A) depicts the Receiver Operating Characteristic (ROC) curves and provides the Sensitivity, Specificity, Positive Predictive Value (PPV) and Negative Predictive Value (NPV); B) depicts the heat map classifying Banff 0 vs. Banff 1 using the top 50 consensus genes where (red) is up-regulated and (green) is down-regulated.

Figure 3. Pie charts showing the Gene Ontology annotations for both Test Sets for Banff 0 vs. Banff 2,3 (moderate to severe CAN).

Each slice of the pie chart represents the percentage of genes represented by that functional class. A) Test Set 1 (PBL) for Banff 0 vs. Banff 2,3; B) Test Set 2 (Whole Blood) for Banff 0 vs. Banff 2,3. The first key point is that there is no difference in the general groups of differentially expressed, functional genes assigned in an unbiased fashion by analysis using Gene Ontology whether we are interrogating the profiles of PBL or whole blood (as was true for the Banff 0 vs. Banff 1 comparisons shown in Figure 1). The second key point is that the number of differentially expressed immune/inflammatory genes is significantly less than observed in mild CAN with many more genes linked to metabolic and other pathways consistent with the hypothesis that early stages of CAN are driven by immune/inflammatory mechanisms and tissue injury but later stages reflect slowly progressive renal dysfunction and fibrosis.

Figure 4. Class prediction analysis of Banff 0 vs. Banff 2,3 (moderate to severe CAN) based on Diagonal Linear Discriminant Analysis for the top 50 Banff 0 vs. Banff 2,3 consensus genes ranked by p values.

A) depicts the Receiver Operating Characteristic (ROC) curves and provides the Sensitivity, Specificity, Positive Predictive Value (PPV) and Negative Predictive Value (NPV); B) depicts the heat map classifying Banff 0 vs. Banff 2,3 using the top 50 consensus genes where (red) is up-regulated and (green) is down-regulated.