ITIH5 and ECRG4 DNA Methylation Biomarker Test (EI-BLA) for Urine-Based Non-Invasive Detection of Bladder Cancer

Bladder cancer is one of the more common malignancies in humans and the most expensive tumor for treating in the Unites States (US) and Europe due to the need for lifelong surveillance. Non-invasive tests approved by the FDA have not been widely adopted in routine diagnosis so far. Therefore, we aimed to characterize the two putative tumor suppressor genes ECRG4 and ITIH5 as novel urinary DNA methylation biomarkers that are suitable for non-invasive detection of bladder cancer. While assessing the analytical performance, a spiking experiment was performed by determining the limit of RT112 tumor cell detection (range: 100–10,000 cells) in the urine of healthy donors in dependency of the processing protocols of the RWTH cBMB. Clinically, urine sediments of 474 patients were analyzed by using quantitative methylation-specific PCR (qMSP) and Methylation Sensitive Restriction Enzyme (MSRE) qPCR techniques. Overall, ECRG4-ITIH5 showed a sensitivity of 64% to 70% with a specificity ranging between 80% and 92%, i.e., discriminating healthy, benign lesions, and/or inflammatory diseases from bladder tumors. When comparing single biomarkers, ECRG4 achieved a sensitivity of 73%, which was increased by combination with the known biomarker candidate NID2 up to 76% at a specificity of 97%. Hence, ITIH5 and, in particular, ECRG4 might be promising candidates for further optimizing current bladder cancer biomarker panels and platforms.


Introduction
Bladder cancer is the most frequent urogenital malignant tumor concerning both sexes worldwide, with an estimated~549,400 new cases and 200,000 deaths in 2018 [1], which causes the highest costs of all cancers per patient [2]. In the European Union (EU) alone, the costs were €4.9 billion, with health care accounting for €2.9 billion in 2012 [3] due to long-term survival with the need for lifelong surveillance by cost-intensive diagnostically tools [2]. Cystoscopy, the "gold standard" for the detection of bladder cancer, is an invasive and time-consuming procedure, achieving an operator-dependent sensitivity and specificity of approximately 90% [4]. In particular, repeating cystoscopy for patients with non-muscle invasive bladder cancer (NMIBC) to determine whether their disease has recurred or progressed to muscle invasive bladder cancer (MIBC) represents a major cost associated with treating bladder cancer patients [5]. Nevertheless, only 10% of haematuria patients are faced with a diagnosis of bladder cancer [6], a fact that did not increase the compliance of undergoing cystoscopy. Complementary to these procedures, the current guidelines recommend completion by non-invasive urine cytology, which, however, is characterized by poor sensitivity varying between 20 to 53% [7]. Additional non-invasive urinary assays have been developed, which could help to minimize the invasive procedure of cystoscopy and reduce its economic burden (for an overview see: [8]). Although such assays have been shown to increase the sensitivity of urine cytology, they have not been widely adopted in routine practice: either they are characterized by cost-intensive performances, like UroVysion [9], or failed as point of-care tests due to limited sensitivity or specificity, such as the NMP22-based "BladderCheck TM Test" [10,11]. Given that, none of the currently available urinary biomarkers that have been approved by the FDA can absolutely be recommended as a stand-alone test to replace cystoscopy in the clinic. Recently, several commercially available tests have been developed with improved sensitivity and specificity by using mRNA (e.g., "Xpert BC" [12]) or protein-based ELISA assay technology (e.g., "UBC" [13]), but these data must be independently be confirmed in further studies. Therefore, it is still of great interest to identify novel tumor biomarkers for urine-based early detection of bladder cancer, which might optimize existing panels and platforms to improve both the initial detection of bladder cancer and detection of its recurrence.
For several decades now, epigenetic alterations are an excellent source of biomarker candidates for cancer detection, diagnosis, and prognosis [14]. In particular, aberrant DNA hypermethylation of putative tumor suppressor genes emerged as a potential biomarker source for assessing early cancer detection, which has recently moved towards clinical practice, for instance, in colorectal cancer [15]. For non-invasive detection of bladder cancer, promising DNA methylation biomarkers have been described in various studies [16], but the FDA has approved none of the presented methylation biomarkers (panels) for routinely diagnostic procedures so far. In the presented study, we focused on two putative tumor suppressor genes in bladder cancer that may also hold a prognostic impact, namely inter-α-trypsin inhibitor heavy chain 5 (ITIH5) and esophageal cancer-related gene 4 (ECRG4 or C2orf40). ITIH5 has previously been shown to be epigenetically silenced in various cancer entities [17][18][19], including bladder cancer [20], where its expression was associated with tumor recurrence of the clinical important group of high-grade pT1 patients. In addition, ITIH5 was characterized as a putative metastasis suppressor gene in breast [21,22] and pancreatic cancers [23]. ECRG4 has also been described to be a candidate tumor suppressor gene that is inactivated by DNA methylation in cancers, like esophageal squamous cell carcinoma [24,25], breast cancer [26], renal cell cancer [27], and colorectal cancer [28,29], but not in bladder cancer so far.
We now provide evidence that ECRG4 and ITIH5 DNA methylation could be useful as urinary biomarkers for non-invasive bladder cancer detection. Biomarkers were assessed by comparing different techniques, i.e., bisulfite-pyrosequencing, qMSP, and MSRE qPCR in comprehensive cohorts of patients with bladder diseases and controls, overall composing 474 urine samples, including a significant number of benign and inflammatory diseases. In particular, we demonstrate strong biomarker performance for ECRG4, which might be a suitable candidate to complete and improve current non-invasive biomarker panels and platforms.

Analytical Performance of ITIH5 and ECRG4 qMSP and Pyrosequencing Biomarker Assays
ITIH5 and ECRG4 have been previously identified as putative class II tumor suppressor genes, which are epigenetically silenced in various tumor entities. In the presented study, we aimed to assess the biomarker quality of both candidates to detect bladder tumors via urine samples. The analytical performance of quantitative methylation-specific PCR (qMSP) and pyrosequencing assays was assessed involving standard biobank processing procedure of the RWTH cBMB biobank prior to assessing the biomarker performance of ECRG4 and ITIH5 by patient materials.
For ECRG4, Figure 1A shows the genomic location of the qMSP and pyrosequencing assays. Both of the assays spanned CpG sites, which were characterized by strong median hypermethylation in bladder tumors within the TCGA data set [30] as compared to normal controls. CpG sites of the ECRG4 target region showed a significant inverse correlation between DNA methylation and ECRG4 mRNA expression ( Figure 1A). Subsequently, a spiking experiment was performed to assess both sensitivity and reproducibility in dependence on a distinct number of RT112 tumor cells (range: 100-10,000 cells), i.e., RT112 bladder cancer cells harboring methylated ECRG4 (see Supplementary Figure S1) and ITIH5 [20] genes were spiked into 20 mL pooled urine of healthy donors (n = 4), respectively. Urine samples that were spiked with RT112 cells were processed according to the standard operating protocol of the RWTH cBMB. Urine pellets were either directly used for DNA extraction (probe set A) or urine sediments were stored according to the RWTH cBMB conditions (−80 • C) for two weeks (probe set B) ( Figure 1B). We found that DNA yield was significantly higher in freshly processed samples when compared to those processed after two weeks of storage at −80 • C (probe set B, Figure 1C). However, this was not associated with significant changes in the detection sensitivity of ECRG4 methylation by both assays, qMSP and pyrosequencing). Overall, the ECRG4 qMSP assay showed the highest sensitivity with a detection limit of 25 tumor cells/ml (equivalent to 89.75 pg tumor DNA), whereas pyrosequencing-based detection of ECRG4 methylation required 45 tumor cells/ml (161.55 pg tumor DNA) urine ( Figure 1D). Furthermore, the ECRG4 qMSP assay exhibited a robust reproducibility when comparing methylation detection rates of the two storage time points, i.e., the qMSP assay convinced with high sensitivity and strong reliability (Spearman r: 0.955, p < 0.001) ( Figure 1E). In contrast, pyrosequencing missed a significant correlation of ECRG4 methylation levels of spiking samples, which were directly used for DNA extraction and those processed after two weeks (Supplementary Figure S2). qMSP and pyrosequencing assays for the detection of the ITIH5 promoter methylation were established similar to ECRG4. In Figure 2A, the relative location of qMSP and pyrosequencing primers are indicated. TCGA BLCA data analyses confirmed differences in median DNA methylation level for the targeted region of the ITIH5 qMSP and pyrosequencing assay, which spanned the CpG site (#10119075), showing an inverse correlation between ITIH5 DNA methylation and corresponding gene expression ( Figure 2A). Of clinical significance, ITIH5 hypermethylation of the CG site #10119075 was associated with shorter overall survival in advanced (pT > 2) bladder tumors ( Figure 2B). In concordance with the results that were observed for ECRG4, pyrosequencing-based detection of ITIH5 methylation failed to perform with a suitable reproducibility when using samples of both storage time points (directly processed vs. two weeks biobank storage). The ITIH5 qMSP assay achieved strong reliability (Spearman correlation: 0.902, p < 0.001, Figure 2C,D) and high sensitivity being characterized by a detection limit that ranged between 25 and 30 tumor cells/mL (89.75 pg/107.70 pg tumor DNA), whereas a robust detection of ITIH5 methylation by pyrosequencing required at least 125 tumor cells/mL (448.75 pg tumor DNA) urine (Supplementary Figure S2). Hence, qMSP assays of both genes were selected for assessment in a clinical cohort setting.

Clinical Biomarker Performance of the ECRG4-ITIH5 qMSP Test for Accurate Non-Invasive Detection of Bladder Cancer
ECRG4 and ITIH5 performance was initially assessed in a clinical cohort of urine samples (total n = 263) comprising 116 urine samples that were derived from bladder cancer patients. Patients with urological malignancies of other origin (testis, prostate, kidney) as well as benign and inflammatory urological-diseases (prostate hyperplasia, renal stones, chronic cystitis) and healthy (without pathological finding) donors that were included as controls (cohort 1a: benign -inflammatory, cohort 1b: all controls including urological cancers). ECRG4 and ITIH5 methylation for each urine sample is shown as the mean percentage of methylated reference (PMR) in the scatter plots of Figure 3A,B. ECRG4 and ITIH5 methylation were both significantly increased in the urine samples from patients with cancers of the bladder (ECRG4 mean PMR: 5.011, 95% CI: 2.118-7.905; ITIH5 mean PMR: 2.634, 95% CI: 0.730-4.539), the prostate (ECRG4 mean PMR: 1.306, 95% CI: 0.178-2.434; ITIH5 mean PMR: 1.018, 95% CI: 0.441-1.720), and the kidney (ECRG4 mean PMR: 0.985, 95% CI: 0.188-1.781, ITIH5 mean PMR: 1.311, 95% CI: 0.147-2.475) when compared to healthy controls (ECRG4 mean PMR: 0.017, 95% CI: <0.001-0.033, ITIH5 mean PMR: 0.031, 95% CI: 0.009-0.070). In benign and inflammatory diseases, a single statistical outlier was detected, respectively, however, diagnosis had been done in a clinical setting and, thus, a true malignancy in this few cases cannot be completely excluded. No associations were found between both ECRG4 and ITIH5 promoter methylation and clinical-pathological characteristics, including tumor size, histological grade, age at diagnosis, and gender (Supplementary Table S1 and S2). Calculating the optimal cut-off value for a combined ECRG4-ITIH5 (EI-BLA) qMSP application with robust specificity by using ROC statistics (Table 1), we demonstrated significant discrimination of bladder cancer patients from non-malignant controls (control cohort 1a) with a sensitivity of 64.3% and a specificity of 81.5% (AUC: 0.783, 95% CI: 0.716-0.850, p < 0.0001) ( Figure 3C).  Additionally, ECRG4-ITIH5 in combination were able to distinguish bladder cancer patients from patients with neoplasms of other urological origin (control cohort 1b) with similar specificity (81.6%, AUC: 0.695, 95% CI: 0.631-0.760, p < 0.0001) but reduced sensitivity (50.9%) ( Figure 3D). The application of both biomarker candidates in combination achieved the most reliable results as compared with single biomarker performances.
Next, the biomarker quality of ECRG4 and ITIH5 promoter methylation was tested and then compared to a known putative bladder cancer methylation biomarker (NID2) by Methylation Sensitive Restriction Enzyme (MSRE) qPCR at the independent laboratories of Biotype GmbH (Dresden, Germany). The independent urine cohort (overall n = 211) included 130 urines of patients that were diagnosed with primary bladder cancer as well as 81 control urines (benign lesions, inflammatory diseases, and healthy controls). ECRG4-ITIH5 methylation was found to be significantly increased in the urines of bladder cancer patients (mean methylation: 10.31, 95% CI: 6.132-14.50) as compared to all control groups, i.e., benign lesions (mean methylation: 0.143), inflammatory diseases (mean methylation: 2.900), and healthy donors (mean methylation: 0.203) ( Figure 4A). In this independent cohort, a close association of both ECRG4 and ITIH5 with increased tumor size (pTa vs. pT1-4) and age at diagnosis was determined by Fisher's exact test (Tables 2 and 3).   The ROC analyses showed that ECRG4-ITIH5 combination achieved a sensitivity of 71.6% at a specificity of 80.2% (AUC: 0.771, 95% CI: 0.706-0.836) in a cohort that included benign lesions and inflammatory diseases of the urinary tract ( Figure 4B), whereas specificity was increased up to 92% with minimally decreased sensitivity (69%) focusing on healthy controls. When comparing our single biomarkers, i.e., ITIH5 and ECRG4, with the recently proposed biomarker candidate NID2 [31], we revealed a similar biomarker performances of ECRG4 and NID2, achieving a sensitivity of 73.1% (AUC: 0.780, 95% CI: 0.709-0.851) and 75% (AUC: 0.801, 95% CI: 0.729-0.873) at a specificity of 83.1%, respectively (Table 4). According to that, ITIH5 was characterized by reduced sensitivity (56.7%, AUC: 0.674, 95% CI: 0.592-0.755), which did not lead to improved biomarker performance when combining with ECRG4. While considering that, we were able to increase sensitivity for bladder cancer detection up to 75% with improved specificity (85.9%; AUC: 0.807, 95% CI: 0.737-0.877) in a cohort comprising benign lesions and inflammatory diseases of the urinary tract when combining ECRG4 with the known biomarker NID2 ( Figure 4C,D). In comparison to healthy controls, the panel set of ECRG-NID2 achieved 76% sensitivity with 97.3% specificity (AUC: 0.884, 95% CI: 0.831-0.937) ( Table 4).

Discussion
The field of liquid biopsy-based cancer detection systems is rapidly evolving, as novel (epi)genetic biomarkers have been characterized, which can be detected in biological fluids, like blood or urine, offering an easy and non-invasive application for cancer detection, prognosis, and therapy prediction. In colorectal cancer (CRC), for instance, a blood-based screening was realized by targeting Septin 9 (SEPT9) hypermethylation, whose Epi proColon®test has been approved by the FDA in 2016 [32]. In bladder cancer, liquid biopsy still needs improvement [33], as various molecular (epi)genetic biomarker candidates and signatures have been described, but none of those assays are FDA-approved for routine diagnostic so far.
In the current study, we present two novel biomarker candidates, ITIH5 and ECRG4, which show strong potential for improving or even completing existing non-invasive biomarker panels and platforms. Already in the year 2010, Renard and colleagues identified TWIST1 and NID2 as putative biomarker candidates while using qMSP [31]. In the same year, Costa et. al. showed three novel gene loci, i.e., GDF15, TMEFF2, and VIM, whose DNA methylation could be suitable for detecting bladder cancers in urine samples [34]. Meanwhile, many more putative candidates have been presented [16], but most of the studies were characterized by small sample cohorts without taking into account crucial cohorts of non-malignant diseases like chronic inflammation. Hence, only a handful of biomarkers such as TMEFF2, NID2 and TWIST1 meet to some degree the needed requirements and, thus, were independently studied and validated [35]. Therefore, we took great care to implement suitable steps and criteria for biomarker validation from the beginning of the study. In the first step, established biomarker assays were assessed by comparing the reliability of different detection methods (qMSP and pyrosequencing) in dependency on the urine sample processing in vitro. Of importance, we demonstrated that the regular procedure of urine processing analogue to SOPs of the centralized biomaterial bank (cBMB) of the RWTH Aachen University (i.e., storage at −80 • C) did not impair biomarker detection. However, the DNA yield was considerably higher when fresh urine samples were processed. Beyond that, the qMSP technique showed both the highest sensitivity and reproducibility, which is in line with previous studies demonstrating high levels of accuracy and lower rates of false negatives as compared with other techniques [36]. In a second step, we validated the clinical performance of both biomarkers by independent cohorts and laboratories. In this setting, we included high numbers of urological benign and inflammatory diseases as controls that can be endemically and frequently found in larger population groups, thereby reflecting a much more real-world scenario. We achieved a sensitivity ranging between 64 to 72% at a specificity of over 80% by combined performance of the ECRG4-ITIH5 DNA methylation biomarker panel. Importantly, we still detect approximately 50% of bladder tumors with a robust true-negative rate (>80% specificity) by including urological malignancies of other origins supporting a liquid biopsy application in the field of bladder cancer. However, a putative benefit of both biomarker candidates being further useful for future assays dealing with the non-invasive detection of other urothelial malignancies, like prostate or renal cell carcinomas, should not be excluded at this stage. Interestingly, in our independent training cohort from Dresden, ECRG4 reached a sensitivity of over 73% as single biomarker. Hence, a suitable impact was suggested, in particular for ECRG4, for discriminating BPH or urocystitis from bladder tumors, which is comparable with proposed biomarker candidates, like NID2 or TWIST1 [33]. CFTR, SAL3, and TWIST1 have been recently shown to be useful for monitoring bladder cancer in a real clinical scenario, as a sensitivity of 96% was achieved by pyrosequencing in combination with urine cytology-however with low specificity (40%) [35]. In our cohort, the ECRG4-ITIH5 biomarker performance also reached over 90% sensitivity at a specificity of 40%, however, we finally focused on the best panel according to their specificity: combining ECRG4 and NID2 led to an increased sensitivity (76%) at a specificity of 97% when compared to healthy controls, encouraging validation studies of this biomarker setting in the future.
In view of novel diagnostic platforms, ECRG4 and ITIH5 could also be part of NGS-based gene signatures, which are currently considered to be at the cutting edge of the technical development of future diagnostic applications. In 2017, the multiplex bisulfite NGS-based sequencing concept "UroMark" was described achieving 98% sensitivity and 97% specificity [37]. However, this NGS assay should be confirmed in comprehensive cohorts and the usability of a 150 CpG loci comprising biomarker assay for routinely and cost-effective diagnosis, in particular as a population-based screening tool, in a real-world scenario must be further considered. So far, real-world application of available urinary markers has not reduced any bladder cancer treatment costs, as predicted by decision-analytic economic models [2]. Still, biomarkers are missing, which serve as the basis for decision-making of risk stratification. According to that, DNA methylation of our markers, in particular ITIH5, might hold a prognostic impact, as both candidates have been characterized as putative tumor suppressor genes whose silencing could be triggered by DNA promoter hypermethylation [20]. In 2008, ITIH5 was described to be epigenetically silenced in breast cancer [17] and five years later ITIH5 DNA methylation has been identified as a putative blood-based biomarker for the early detection of breast cancer [38]. Since then functionally studies revealed, for instance, ITIH5 mediated suppression of breast [21,22] and pancreatic cancer metastases [23] in vitro and in vivo. Interestingly, in aggressive mammary cancer cells, ITIH5 triggered an epigenetic reprogramming which was associated with a demethylation of various promoter regions, including that of DAPK1, a tumor suppressor gene and putative blood-based biomarker in several tumor entities [21]. In bladder carcinogenesis, the downregulation of ITIH5 was also associated with worse prognosis while functionally high-grade bladder cancer cells showed reduced growth in vitro after ITIH5 overexpression [20]. Of clinical interest, ITIH5 protein expression was shown to predict tumor relapse of the clinical important subgroup of pT1 high-grade patients [20], of which 30% never displayed recurrence after transurethral resection of the bladder, while a further 30% died due to metastatic disease [39]. In the present study, we now confirmed a putative prognostic impact as ITIH5 promoter hypermethylation was associated with poor patients' outcome in the subgroup of advanced tumors (pT > 2) of the TCGA bladder cancer data set, while increased ITIH5 methylation was further shown to correlate with a higher pT status in our second urine cohort. These findings may support our hypothesis that ITIH5 could be a useful biomarker for risk stratification, helping to monitor patients for the recurrence and/or progression of bladder tumors.
In conclusion, we provide two novel DNA methylation biomarkers for non-invasive detection of bladder carcinomas. As ITIH5 might keep prognostic information for bladder cancer risk stratification, while ECRG4 showed a convincing diagnostic performance, in particular in combination with the known biomarker candidate NID2, both biomarkers, ECRG4 and ITIH5, may be promising candidates to complete and improve current biomarker panels and platforms. For instance, the "Bladder EpiCheck TM " urine assay that combines 15 DNA methylation biomarkers leading to an overall sensitivity of 68.2% and a specificity of 80.0% [40] may benefit from our biomarker candidates to reduce the number of biomarkers while also improving the overall performance. Future studies should be conducted to clarify which of our biomarkers, is suitable for which clinical application, e.g. as a guidance tool for early detection, risk stratification, surveillance, and/or therapeutic management.

Cell Line
The bladder cancer cell line J82 was originally obtained from the American Type Culture Collection (ATCC, Manassas, VA, USA). The urothelial bladder cancer cell line RT112 was used for studies of the analytical performance, a gift from Dr. Alexander Buchner (LMU München, München, Germany). All of the cell lines successfully underwent an identity check (Multiplexion GmbH, Immenstadt, Germany) prior to the experiments.

Urine Samples
In total, 474 urine samples were assessed in this study. The Departments of Urology of the University Hospitals of Aachen, Bonn, and Dresden provided the voided urine samples. The samples that were collected in Aachen were obtained from the RWTH centralized biomaterial bank (RWTH cBMB). The collection of tissue samples was performed within the framework of the Biobank of the Center for Integrated Oncology Köln Bonn. All of the patients gave written consent for asservation and analysis of their samples according to local Institutional Review Board (IRB)-approved protocols of the Medical Faculty of RWTH Aachen University (EK 206/09, 05 Jan 2010), the University of Bonn (EK 205/13, 16 Mar 2013), and the University of Dresden (EK 96032012, 15 Jul 2014). The urine samples derived from patients diagnosed with a primary bladder tumor (n = 246) were used to assess biomarker performance, while samples with a known second malignancy, such as prostate cancer, were excluded from this study. Urines from healthy donors (n = 49) and samples derived from patients with inflammatory (chronic cystitis), benign (benign prostate hyperplasia), and urological malignant diseases of other tissue origin (testicular tumors, prostate cancer, renal cell carcinoma) served as the controls (overall n = 179). For the characteristics of training cohort I (Aachen-Bonn) and training cohort II (Dresden) see Table 5. Unless otherwise stated, 10-20 mL of urines were centrifuged for 10 min. at 2000 × g, washed with PBS and sediments were stored at −80 • C.

DNA Extraction from Urines
The urine sediments of training cohort I (Aachen-Bonn) stored at −80 • C were subjected to DNA extraction by using the ZR Urine DNA Isolation Kit (ZR, Zymo Research, Freiburg, Germany), following the manufacturer's instructions. DNA extraction from the urine sediments of training cohort II (Dresden) stored at −80 • C in RLT buffer was performed by using the QIAamp DNA Mini Kit (Qiagen, Hilden, Germany), according to the manufacturer's instructions. The DNA yield (ng/mL urine) and purity (A 260 /A 280 ) were determined by using the NanoDrop (Thermo Fisher Scientific, Waltham, MA, USA). Only extractions from urines with a minimal total amount of 100 ng genomic DNA and a ratio of ≥1.5 were finally used for qMSP, pyrosequencing, and MSRE qPCR analyses.

DNA Bisulfite Conversion
100 to 250 ng of the genomic DNA (training cohort I) were bisulfite-converted for 14 to 16 h by using the EZ DNA Methylation™ kit (Zymo Research) according to the manufacturer's instructions. Bisulfite-converted DNA was eluted in 20 µL of TRIS-EDTA buffer.

Bisulfite-Pyrosequencing
The pyrosequencing of bisulfite-converted DNA was performed by using the PyroMark PCR Kit, the PyroMark96 ID device, and the PyroGoldSQA reagent Kit (Qiagen), as reported previously [20]. ECRG4 and ITIH5 pyrosequencing assays were designed by using the Pyromark Assay Design Software (Qiagen), and Supplementary Table S3 lists all of the primers. Primers and sequence of interest meet the following criteria: Based on TCGA data analyses, sequences of interest should cover promoter regions that a) are characterized by strong differences in mean DNA methylation between urothelial normal and bladder cancer samples and b) are located in important gene regulatory sequences, i.e., a statistically significant inverse correlation between ECRG4/ITIH5 gene expression and the corresponding DNA methylation had to be observed. The EpiTect®PCR Control DNA Set (Qiagen) was used as the positive controls for unmethylated and methylated DNA in each run.

Quantitative Methylation-Specific PCR (qMSP)
Bisulfite-modified DNA was used as a template for fluorescence-based real-time PCR amplified in an iCycler iQ5 (Biorad, Munich, Germany), as previously described [37] with slight modifications: The designed primers and probes were specific for amplifying bisulfite-converted DNA for the genes of interest (ECRG4 and ITIH5) (for cycle conditions, primer sequences, and annealing temperatures, see Supplementary Table S4). The reference gene GAPDH was used for internal normalization. Eight calibration dilutions of in vitro methylated human leukocyte DNA (0.1%, 1%, 5%, 10%, 20%, 30%, 50%, 100%) and unmethylated sequence (human leukocyte DNA from a healthy donor), as well as multiple water blanks were included in each run. The gene of interest was called methylated if the cycle threshold (Ct) of at least two of three qPCR replicates for each specimen had a value of less than 45 cycles. The amount of methylated DNA (percentage of methylated reference, PMR) at a specific locus was calculated by dividing the GENE/GAPDH ratio of a sample by the GENE/GAPDH ratio of SssI-treated human leukocyte DNA and multiplying by 100, as specified [37]. The primer binding sites of the qMSP assays were located in the same genomic promoter region as covered by pyrosequencing. The efficiencies of real time MSP were calculated according to the equation: E = 10 [−1/slope of calibration dilutions] [41] and the mean efficacy of ECRG4 and ITIH5 qMSP was 76.57% and 77.76%, respectively.

Methylation Sensitive Restriction Enzyme qPCR (MSRE) qPCR
Isolated genomic DNA (125 ng) was used for double restriction digest. Methylation-sensitive restriction enzymes AciI and HpaII (New England Biolabs, NEB, MA, USA) were selected based on their capacity to distinguish methylated from unmethylated DNA sequences. Two independent digestion reactions (test reaction and control) were prepared for each patient DNA. Restriction digest was performed within a total volume of 25 µL in CutSmart Buffer (NEB) for 1 h at 37 • C and followed by heat inactivation for 20 min. at 80 • C. The control samples were treated in the same way but without the addition of the enzymes, 50% glycerol was added instead. Finally, DNA digest was diluted with 1x TE buffer before MSRE qPCR. The designed primers and probes for MSRE qPCR are specific for amplifying unrestricted DNA for the genes of interest (ECRG4, ITIH5, and NID2).
MSRE qPCRs were carried out while using the Roche LightCycler 480 II Real-Time PCR detection system. Mono color hydrolysis probe detection (FAM) was used. All of the samples were done in duplicate in 25 µL reactions containing 5 µL Reaction Mix B (Biotype GmbH, Dresden), 3 U Multi Taq 2 (Biotype), 1.5 µL primers and probes (5 µM each), nuclease-free water (Biotype), and 2 µL of digested DNA (2.5 ng/µL, test or control template). For ECRG4 and NID2, the addition of Combinatorial Enhancer Solution (1× CES, [42]) was necessary due to the very high GC content of the amplified region.
For cycle conditions, primer sequences and annealing temperatures, see Supplementary Table  S5. Ct values were analyzed while using LightCycler 480 Software (Hoffmann-La Roche AG, Basel, Switzerland).
Undetected Ct values were normalized to 47 for calculations. The methylation level of the amplified region was calculated by using the following equation: percent methylation = 100 × 2 −∆Ct , where ∆Ct is the average Ct value from the test reaction minus the average Ct values from the control reaction. Methylation values exceeding 100% were set to 100%.

Analytical Assay Performances
RT112 wildtype bladder cancer cells harboring a methylated ECRG4 and ITIH5 promoter were cultured for two weeks. After cell counting RT112 cells were spiked into pooled urine of healthy donors (n = 4). Serial dilutions (10 to 10.000) of RT112 cells were added to 20 mL pooled urine, respectively, which was subsequently processed at the RWTH cBMB laboratories according to its standard operating protocol. Afterwards, urine pellets were either directly used for DNA extraction (probe set A) or urine sediments were stored according to the RWTH cBMB conditions by using two-dimensional (2D) barcoded LVL tubes (LVL technologies, Crailsheim, Germany) at −80 • C for two weeks (probe set B). Pooled urines without any spiked RT112 cells served as the control for normalization and threshold calculation by defining methylation cut-offs. Next, DNA was bisulfite-treated, as mentioned earlier, and the ECRG4 and ITIH5 qMSP as well as pyrosequencing assays were performed for ECRG4 and ITIH5, respectively. The gene of interest was called methylated if the PMR (qMSP) or mean percent of CpG methylation (pyrosequencing) stably exceed the background noise and certainly maintained this threshold (=cut-off).

TCGA BLCA Data Set
Infinium HumanMethylation450 BeadChip data (level 2) and RNASeqV2 data (level 3) of the tumor and normal tissue samples were obtained from the TCGA data portal [30] and analyzed, as previously described [43].

Statistical Data Acquisition
Two-sided p-values that were less than 0.05 were considered to be significant. The non-parametric Mann-Whitney U-test was applied in order to compare two groups, whereas, in the case of more than two groups, the Dunn's multiple comparison test was used. Correlation analysis was performed by calculating a non-parametric Spearman's rank correlation coefficient. Statistical associations between clinico-pathological parameters and DNA methylation of ITIH5 and ECRG4 were determined by Fisher's exact test by using SPSS software version 25.0 (SPSS Inc., Chicago, IL, USA). Survival curves for overall survival (OS) were calculated using the Kaplan-Meier method with log-rank statistics. OS was measured from surgery until death and it was censored for patients alive without evidence of death at the last follow-up date. The receiver operating characteristics (ROC) curves and AUC values were calculated to assess the biomarker performance of ECRG4 and ITIH5 methylation similar to our previous report [43]. The ROC curves of combined biomarkers were based on the binary logistic regression model using the probability as test variable.