1. Introduction
Bladder cancer (BC), accounting for 90% of urothelial cancer (UC), is the 10th most common cancer worldwide [
1]. There are two major groups of BC patients with distinctive clinical outcome and molecular features: Non-muscle-invasive and muscle-invasive [
2]. Owing to the high recurrence rate of the non-muscle-invasive BC and expensive cystoscopy as a gold-standard diagnostic method, BC is one of the most expensive cancer to treat and manage [
3]. The current guidelines for routine diagnosis and monitoring of BC rely on a sensitive but invasive cystoscopy as the gold standard and urine cytology, which is relatively less sensitive than cystoscopy, especially for low-grade non-muscle-invasive BC [
4]. The invasiveness, risk of complications, and high expense of the cystoscopy technique poses significant challenges to the management of BC, especially in subjects at sufficiently low-risk for BC, for whom invasive testing may be replaced by highly sensitive and specific methods for non-invasive detection of BC. Currently, clinical guidelines for the follow-up of patients with asymptomatic microscopic hematuria (AMH) recommend invasive cystoscopy testing, with only a small fraction of AMH patients developing bladder cancer [
5]. Individuals exposed to BC carcinogens at their workplace and therefore under regular medical surveillance may also benefit from a BC screening strategy using urine biomarkers with higher diagnostic predictive values than the common non-invasive tests (hematuria and urine cytology). Urine-based biomarkers may therefore be useful in these clinical scenarios, thereby contributing to early detection of BC and subsequent monitoring [
6]. However, to date, no reliable non-invasive biomarker has been proposed for routine BC clinical management or for screening populations at risk of BC. Therefore, detecting molecular alterations of the bladder tumor in the urine promises to be an important addition in the liquid biopsy of BC [
7,
8].
Two mutations in the promoter of the
Telomerase reverse transcriptase gene (
TERT), called C228T and C250T, are frequent in various human cancers and particularly in UCs (carcinomas of the bladder and upper urinary tract) where it is seen in 60–85% of cases (74% for non-invasive UCs) [
9,
10]. Should these mutations be detectable in the urine of BC patients as a means of liquid biopsy, they would provide an unprecedented opportunity for a simple diagnostic assay of UCs, which could be used for detection as well as monitoring progression and recurrence. We have recently developed a Next-Generation Sequencing (NGS)-based assay, called UroMuTERT, for the comprehensive detection of BC in urine cell pellet and cell-free DNA samples [
11]. We specifically showed in case-control studies that urinary
TERT promoter mutations have excellent sensitivity and specificity for the detection of BC [
11] and that these mutations could also be detected in urinary DNA of asymptomatic individuals years prior to primary diagnosis of BC with high specificity [
12], highlighting their potential to be used as simple and inexpensive non-invasive biomarkers for early detection of BC. While the NGS assay shows great promise in detecting
TERT promoter mutations in urine samples, the complex laboratory workflow and requirement of extensive bioinformatics skills for data analysis are still considered as a bottleneck for their large-scale clinical application.
In this study, we report the development of sensitive droplet digital PCR assays for detecting TERT promoter mutations in urine samples. We also present for the first time the comparison of ddPCR- and NGS-based assays for the detection of TERT promoter mutations in urine samples from a series of BC patients and controls.
3. Discussion
Owing to the presence of exfoliated cells from the urinary tract epithelium in urine, the latter represents the best reservoir of urinary DNA for the analysis of tumor-derived mutations in BC patients. While liquid biopsy presents an opportunity for non-invasive detection and monitoring of bladder cancer, there are still technological bars to be crossed [
16,
17,
18]. Due to the dilution of tumor-derived cells with normal epithelial cells in the urine of BC patients, the fractional abundance of the tumor cells can be very low, thereby requiring highly sensitive method(s) for the detection of these urinary mutations [
18]. Droplet digital PCR (ddPCR) assays presents some advantages in detecting low-allelic fraction mutations in the urine [
19]. The performance of ddPCR assays in detecting
TERT promoter mutations in urine cfDNA has only been reported in a limited series of UC cases and controls [
17] and more recently in UC cases undergoing monitoring of post-therapy disease [
19]. Few other studies assessed ddPCR assays for detecting
TERT promoter mutations but in different cancer types including melanoma and glioma and in a very small sample set [
20,
21,
22].
In this study, we developed sensitive ddPCR assays for the detection of urinary
TERT promoter mutations in UC and report comparable performance with our previously developed NGS-based UroMuTERT assay [
11].
TERT promoter mutations are the most common somatic mutations in UC and their detection in urine has significant clinical applications [
11,
18,
23]. A study by Kinde et al., has showed that the somatic mutations in the
TERT promoter detected after the transurethral resection of the bladder tumor (TURBT) could be potential biomarkers of recurrent UC [
10]. Another study has shown an increased risk associated with the detection of these mutations and recurrence of pT1 bladder cancer [
24] and a more recent one demonstrated that the post-therapy urinary C228T status in BC cases under surveillance for relapse matched the clinical observations [
19] thereby suggesting a plausible way of non-invasive dynamic follow-up of non-muscle invasive bladder cancer patients. In addition, within a prospective cohort study, urinary
TERT promoter mutations were detected up to 10 years prior to clinical diagnosis of primary bladder cancer and were absent from matched controls [
12]. All these studies provide exciting evidence for the utility of urinary
TERT promoter mutations as non-invasive biomarkers for bladder cancer detection and monitoring.
Various NGS-based assays have been developed over time for the detection of
TERT promoter mutations in urine [
11,
18]. However, due to the involvement of sophisticated laboratory workflows, extensive bioinformatics post-processing and analysis of the data, the NGS-based assays are not broadly implemented in the clinic [
25]. On the other hand, the advantages of the ddPCR assays are its superior sensitivity, fast processing time, independence from extensive bioinformatics post-processing, and analysis of the data [
26,
27]. In our study, the LOD tests using different amounts of serially diluted cell line template DNA showed that our assays could detect as low as 0.18% for the C250T and the C228T mutant DNA. Our results suggest that the LOD for these mutations using ddPCR assays is also influenced by the amount of input template DNA. The LOD proved to be lower when the amount of template DNA was increased from 5 ng to 40 ng DNA, confirming the findings of a recent study [
28]. This LOD is slightly better than that of our previously developed UroMuTERT assay for which the limits of detection were 0.5% and 0.8% for the C250T and C228T mutations, respectively [
11]. A recent study on ddPCR assays for detecting
TERT promoter mutations in melanoma has shown a similar LOD using
TERT mutated melanoma cell lines [
22]. When applied to urine samples of UC cases, the lowest
TERT promoter MAF detected by ddPCR assays from 10 ng US cfDNA (
n = 172) or UP cellDNA (
n = 222) were equivalent to the LOD measured with cell-lines, demonstrating the transferability of the thresholds to such sample types. While the LOD of standard ddPCR assays for other common somatic mutations such as BRAF V600E could be as low as 0.001% [
29], the high GC content and repetitive sequences of the
TERT promoter region makes it more difficult to get a clear separation of mutated droplets against the background noise [
30].
We also observed that, for the same samples, the MAFs tended to be higher with the UroMuTERT assay than with the ddPCR assays. We postulated that the first PCR-based step of the UroMuTERT assay may preferentially amplify the mutated
TERT allele as a result of incomplete denaturation of extremely high GC-rich heteroduplex fragments, thus resulting in higher MAFs than the ones observed with ddPCR assays, which is based on absolute quantification of the target and should therefore represent a more precise measurement of the MAFs than the ones measured by UroMuTERT. Furthermore, the urinary sample type may matter when comparing the MAF between the two methods and may explain the difference in the R
2 for US cfDNA and UP cellDNA. Few more outliers of the regression were observed for US cfDNA than for UP cellDNA. This may be explained by the differences in the distribution of US mutated and wild-type cfDNA size profiles between samples and the differences in amplicon size between UroMuTERT (147bp) and ddPCR (64bp for C228T/A and 88bp for C250T and CC242-243TT) assays; the latter favoring the amplification of smaller thus highly degraded urinary cfDNA forms, which have been shown to be predominant in such sample [
31]. Altogether, this may partly explain the minor discrepancies of low-level urinary
TERT promoter mutations observed between the two methods. Indeed, in the urine pellet DNA (UP DNA) from the BC cases in the IPO-PORTO series, the ddPCR assays failed to replicate positive mutations calls for two samples analyzed with the UroMuTERT assay, but enabled the mutation detection in two additional BC cases initially negative with the UroMuTERT assay, keeping the overall sensitivity to the similar levels. For the DIAGURO series, the sensitivity of the ddPCR assays was slightly lower (by around 3%) compared to the UroMuTERT assays. However, this decreased sensitivity of the ddPCR assays could be due to the fact that three out of 12 BC cases without mutation after UroMuTERT could not be tested by ddPCR assays. While the 100% specificity observed in the IPO-PORTO series was found for both the UroMuTERT and ddPCR assays, the specificity of the ddPCR assays in the DIAGURO controls was lowered compared to UroMuTERT (specificity of 92.4% versus 95.6%, respectively). While applying a more strict interpretation of the Cohen’s kappa coefficient developed by McHugh to better reflect the level of agreement between two methods in health-related studies [
32], an almost perfect agreement was still observed between ddPCR assays an UroMuTERT for both US cfDNA (
n = 164; kappa coefficient of 0.91) and UP cellDNA (
n = 280; kappa coefficient of 0.94). This was confirmed in a large independent series of serial urine samples from follow-up BC cases, in which, when stratified by MAF, the agreement remains almost perfect for samples with MAFs > 2% (kappa coefficient of 0.99) but dropped to moderate for samples with MAF < 1% and to minimal for the samples with the lowest MAFs (< 0.5%), indicating higher rates of false positive or false negative calls in samples with very-low MAFs. Finally, there was also an almost perfect agreement between UroMuTERT and ddPCR assays for the detection of the mutations in the tumors, further attesting to the robustness of the two methods. In addition, we observed a high concordance of mutational status between tumors and matched urine samples in DIAGURO BC cases with analytical sensitivities of 98.6% for UroMuTERT and 97.2% for ddPCR assays and therefore only a small fraction of false negatives, likely reflecting in such cases the existence of minor tumoral
TERT mutated clones (MAF < 1%) undetectable in urine samples. Bladder tumors are characterized by intra- and inter- tumoral heterogeneity both at the molecular and cellular levels [
33,
34]. The co-existence of multiple sub-clonal populations, including multiple
TERT mutated clones has been also reported in bladder tumors [
11,
24]. As urine represents of ‘screenshot’ of the bladder tumor heterogeneity, tumor-derived mutations may be detected in urine that would not necessarily be captured by a unique piece of biopsy due to the heterogenous nature of this cancer type. This was observed in three out of ten DIAGURO cases with wild-type biopsy, highlighting the superiority of urine over biopsy in reflecting mutational burden of heterogenous bladder tumors. This observation further supports the importance of implementing testing for urinary
TERT promoter mutations for the detection of BC with well validated methodologies.
In conclusion, the two methods developed at our laboratory (ddPCR and UroMuTERT) for detecting urinary TERT promoter mutations are comparable. The high sensitivity and specificity of the ddPCR urinary TERT promoter mutations assays in the two independent cohorts investigated in this study provides evidence that these tests could be easily implemented into the clinic for the non-invasive detection of bladder cancer.