Non-Invasive Diagnosis for Acute Rejection Using Urinary mRNA Signature Reflecting Allograft Status in Kidney Transplantation

Abstract

Urine has been regarded as a good resource based on the assumption that urine can directly reflect the state of the allograft or ongoing injury in kidney transplantation. Previous studies, suggesting the usefulness of urinary mRNA as a biomarker of acute rejection, imply that urinary mRNA mirrors the transcriptional activity of the kidneys. We selected 14 data-driven candidate genes through a meta-analysis and measured the candidate genes using quantitative PCR without pre-amplification in the cross-sectional specimens from Korean kidney transplant patients. Expression of 9/14 genes (CXCL9, CD3ϵ, IP-10, LCK, C1QB, PSMB9, Tim-3, Foxp3, and FAM26F) was significantly different between acute rejection and stable graft function with normal pathology and long-term graft survival in 103 training samples. CXCL9 was also distinctly expressed in allografts with acute rejection in in situ hybridization analysis. This result, consistent with the qPCR result, implies that urinary mRNA could reflect the magnitude of allograft injury. We developed an AR prediction model with the urinary mRNAs by a binary logistic regression and the AUC of the model was 0.89 in the training set. The model was validated in 391 independent samples, and the AUC value yielded 0.84 with a fixed manner. In addition, the decision curve analysis indicated a range of reasonable threshold probabilities for biopsy. Therefore, we suggest the urine mRNA signature could be used as a non-invasive monitoring tool of acute rejection for clinical application and could help determine whether to perform a biopsy in a recipient with increased creatinine.

Keywords: acute rejection; kidney; non-invasive diagnosis; transplantation; urinary mRNA.

Figure 1

Workflow of biomarker discovery and validation for the diagnosis of AR in urine pellets. After a meta-analysis and review of the published literature to obtain candidate genes for diagnosis of AR, we measured the levels of 103 training samples composed of stable (STA, n=45), and acute rejection (AR, n=58) using absolute qPCR methods. We developed a signature by combining significantly different transcripts to distinguish AR from STA. We validated significant mRNAs and tested the performance of the diagnostic signature in 391 independent samples including STA (n=153) composed of NP (n=96) and long-term graft survival (LTGS, n=57), AR (n=68) composed of acute TCMR (n=38), acute ABMR (n=11), and chronic active ABMR (n=19), borderline changes (BC, n=58), BKVN (n=15), and other graft injuries (OGIs, n=97) including acute tubular necrosis (ATN, n=30), calcineurin inhibitor (CNI, n=28) toxicity, glomerulonephritis (GN, n=27), and interstitial fibrosis/tubular atrophy (IF/TA, n=12). We also performed a decision curve analysis to determine the value of the signature in predicting AR.

Figure 2

The expression levels of each mRNA between STA (n=45) and AR (n=58) were analyzed using absolute quantitative qPCR without pre-amplification. Each mRNA level was log10-transformed after each mRNA copy number was normalized with 18S rRNA copies (x10-6) in the QC-passed samples (STA, n=40; AR, n=44). (A) The levels of CXCL9, IP-10, C1QB, PSMB9, LCK, CD3e, Foxp3, FAM26F, and Tim-3 mRNAs were significantly elevated in AR compared to STA, and for OX40, ISG20, vWF, IDO1, and PTPRC mRNAs, there was no difference. In the 18s rRNA used as an endogenous control, there was no difference between AR and STA. P values by the non-parametric Mann-Whitney test were expressed as the mean ± SE. NS: not significant, *P < 0.05, **P < 0.01 and ***P < 0.001 versus STA. Although LCK, Foxp3, and FAM26F mRNAs were statistically significant, these mRNAs were not detected in more than 10% of the QC-passed samples. Therefore, we excluded these mRNAs for further analysis. (B) CXCL9 mRNA expression in kidney biopsy tissues of NP, acute TCMR and acute ABMR groups was examined by ISH (original magnification x400). CXCL9 was distinctly expressed in the damaged tubules in kidney allografts of acute TCMR and predominantly in the peritubular capillary area in ABMR groups (black arrows). Scale bars: 50 μm.

Figure 3

ROC curve analysis of the signature to distinguish AR from STA in the training set. The graphs of receiver operating characteristic (ROC) curves and the predicted probability of AR for the six-genes signature show discrimination of AR from STA. (A) The area under the curve (AUC) value of the six-gene model is 0.89 (95% CI, 0.82-0.96; P<0.001). (B) In the graph of the predicted probability of AR with the cutoff point (0.40889), the signature yielded 86% accuracy, 91% sensitivity, 80% specificity, 83% positive predictive value (PPV), and 89% negative predictive value (NPV).

Figure 4

The prediction performance of the signature in the validation set. (A) ROC curve of the AR predicted probability for distinguishing AR (n=57) from STA (n=122) in the QC-passed samples shows the AUC of 0.84 (95% CI, 0.78-0.90; P<0.001). (B) The box plot shows the AR predicted probability of the signature with the fixed cut-off point (0.40889) in AR (n=57), BC (n=42), and STA (n=122), and the signature yielded 77% accuracy, 70% sensitivity, 80% specificity, 63% positive predictive value (PPV), and 85% negative predictive value (NPV). The horizontal line within each box represents the median, the bottom and top of each box represents the 25th and 75th percentile value, and the I bar represents the 5th and 95th percentile value. The plus symbol represents the mean, and the dots indicate outliers. (C) ROC curve of the predicted probability for distinguishing AR from No-AR (STA + OGIs, n=207) in the QC-passed samples is shown. The AUC value of the signature was 0.78 (95% CI, 0.72-0.85; P<0.001). (D) The box plot shows the AR predicted probability of the signature with the fixed cut-off point (0.40889) in AR (n=57) and No-AR (n=207), and the signature yielded 72% accuracy, 70% sensitivity, 72% specificity, 41% positive predictive value (PPV), and 90% negative predictive value (NPV).

Figure 5

Decision curve to evaluate the clinical benefit of the signature to distinguish AR from STA. We performed decision curve analysis to assess the clinical benefit of the diagnostic signature using the predicted probability for each patient in independent samples and determined whether the signature can help avoid the number of unnecessary biopsies for the diagnosis of AR. In the decision curve, the y-axis represents the net benefit ((true-positive count/n)-(false-positive count/n)x[p_t /(1-p_t )]), where the true-positive count is the number of patients with AR, the false-positive count is the number of patients with STA, n is the total number of patients, and p_t is the threshold probability. Here, p_t /(1- pt) is the ratio of the harm caused by a false positive compared to that caused by a false negative. The blue line represents the net benefit of the diagnostic signature. The red line represents the net benefit of the biopsy strategy in all patients. The black line, which represents no net benefit, is the strategy without biopsy.