Bladder cancer detection in urine using DNA methylation markers: a technical and prospective preclinical validation

Abstract

Background: The development of accurate urinary biomarkers for non-invasive and cost-effective detection of primary and recurrent bladder tumours is recognized as one of the major clinical needs in bladder cancer diagnostics. The purposes of this study were (1) to validate the results of a previous technical comparison by determining the diagnostic performance of nine methylation markers in urine pellet compared to full void urine, and (2) to validate the diagnostic performance of the optimal marker panel GHSR/MAL from a previous exploratory study in a preclinical setting.

Methods: Urine samples of 108 patients with bladder cancer and 100 age- and gender-matched controls were prospectively collected for methylation analysis. Urinary methylation levels of the markers FAM19A4, GHSR, MAL, miR-129, miR-935, PHACTR3, PRDM14, SST and ZIC1 were determined with quantitative methylation-specific PCR in urine pellet. Area under the curves (AUCs) were determined for individual markers and the marker panel GHSR/MAL. The diagnostic performance of the marker panel GHSR/MAL was evaluated in the total study population and in different subgroups of patients with bladder cancer using the Chi-square test. The diagnostic accuracy was assessed by leave-one-out cross-validation.

Results: All nine urinary methylation markers (FAM19A4, GHSR, MAL, miR-129, miR-935, PHACTR3, PRDM14, SST and ZIC1) showed significantly higher methylation levels in bladder cancer patients than in controls (p < 0.001). Area under the curves (AUCs) of the nine methylation markers tested in urine pellet were similar to AUCs in full void urine of an independent previous cohort. GHSR/MAL reached an AUC of 0.89 (95% confidence interval [CI] 0.84-0.94), at 80% sensitivity and 93% specificity. Sensitivity of GHSR/MAL increased with higher tumour grades, higher tumour stages, in primary vs. recurrent tumours, and in males vs. females.

Conclusions: This technical validation supports the robustness of DNA methylation analysis in urine pellet and full void urine for the non-invasive detection of bladder cancer. Subsequent preclinical validation confirmed the diagnostic potential of GHSR/MAL. These findings underline the diagnostic potential of the marker panel GHSR/MAL for future bladder cancer diagnostics.

Keywords: Biomarkers; DNA methylation; Liquid biopsy; Tumour; Urinary bladder neoplasms; Urine.

Fig. 1

Outline of the study and flowchart for patient inclusion

Fig. 2

Boxplots of the urinary methylation markers FAM19A4, GHSR, MAL, miR-129, miR-935, PHACTR3, PRDM14, SST and ZIC1. The Y-axis displays log2-transformed Ct ratios of the methylation markers, the X-axis divides controls and patients with bladder cancer. Boxes represent medians with 25th and 75th percentiles. Whiskers and outliers are plotted with the Tukey method. p values were calculated with the Mann–Whitney U test and Bonferroni correction (original p value × 9)

Fig. 3

Receiver operating characteristic (ROC) curves of the urinary methylation markers FAM19A4, GHSR, MAL, miR-129, miR-935, PHACTR3, PRDM14, SST and ZIC1. For each marker, the ROC curves for urine pellet in the present study (black line) and for full void urine in the previous exploratory study (grey line) are visualized. The sensitivity is shown at the Y-axis, for 1-specificity at the X-axis. The area under the curve (AUC) of the present study is provided in the lower right corner, with the AUC of the previous exploratory study between parentheses [9]