Background
Bladder cancer (BC) is one of the most common cancers in the world. Approximately 90% of patients with bladder cancer present with urothelial carcinoma, which has a high recurrence rate [
1‐
3]. However, the diagnosis and the follow-up monitoring of BC has remained a challenge due to the lack of disease-specific symptoms [
4]. Cystoscopy, although generally accepted as the gold standard for BC detection and surveillance, is an invasive procedure.
Voided urine cytology is a standard non-invasive approach adjunct to cystoscopy. The technique is highly specific (85–100%), but the sensitivity is tumour-grade dependent. Although good sensitivity was demonstrated for detecting high-grade urothelial cancer (80–90%, [
1]), the technique is poor in terms of the detection of low-grade tumours, ranging from only 4–31% detection rates [
4].
A profound number of new urinary biomarkers have been developed by laboratory and clinical investigations, many of which have also been approved by the FDA, such as NMP22, UroVysion® (fluorescence in situ hybridization, FISH) and BTAstat. Although many of these tests exhibit better sensitivity than urine cytology (up to 70–80%), they come with the price of lower specificity (median 70–85%) compared to cytology and therefore need to be further improved for wider application [
4‐
7]. Therefore, a non-invasive, convenient and affordable urine-based test with high sensitivity and specificity is urgently in demand for BC diagnosis and monitoring.
Chromosomal instability is a common feature of tumour cells and has been reported to correlate with the development of bladder cancer [
8]. Chromosomal instability might cause genomic abnormalities, such as alterations in chromosomal numbers and loss and/or gain of DNA in certain chromosomal segments [
9,
10]. It has been recently reported that chromosome instability has the potential to function as a prognostic predictor in non-small-cell lung carcinoma [
11]. In the present study, we developed a next-generation sequencing (NGS)-based evaluation approach, the chromosomal imbalance analysis (CIA), in combination with a previously reported whole genome amplification (WGA) technique of multiple annealing and looping-based amplification cycles (MALBAC) [
12] to assess the chromosomal aberration level of cells in urine, and demonstrated the application of the MALBAC-CIA for BC detection.
Discussion
Chromosomal aberration is a common occurrence in tumours. In addition to aneuploidy, alterations in chromosomal architecture, focal amplifications and deletions are observed in cancer genomes. As a driving factor, the chromosomal abnormality manifests at the earliest stages of tumourigenesis and accumulates throughout subsequent tumour development [
9,
13‐
15]. The urinary FISH test (UroVysion®), which probes alterations in chromosomes 3, 7, 17 and 9p21, is one of the commercially available urinary biomarkers used to detect BC. The sensitivity and specificity have been reported in systematic reviews and meta-analyses to exceed 70% and ~ 80%, respectively, but with a broad range among different studies [
16‐
18]. The methylation patterns of a number of candidate genes have also been explored as potential biomarkers [
16,
19]. More recently, a combination of methylation status of TWIST, ONECUT2, and OTX1 with mutational analyses of FGFR3, TERT, and HRAS has been reported to detect bladder cancer with a sensitivity of 97% and a specificity of 83% [
20,
21]. However, due to the diversity of the tumourigenesis driver mutations and the randomness of the somatic passenger mutations [
22], tremendous genetic heterogeneity is spatially and temporally observed in tumour cells and is expected in different individuals, as is the case in bladder cancer [
23] (Fig.
3). Conceivably, using the manifestation of genomic abnormality/imbalance that comprehensively assesses the variation across the whole genome as an indicator of bladder cancer detection might show superior sensitivity. Previously, different evaluation scores based on whole-genome sequencing were reported in prostate, colorectal and breast cancers for diagnosis and prognosis among limited numbers of patients/controls [
9,
24‐
26]. It is known that the number of exfoliated tumour cells varies in BC patients. This amount may be associated with the size and grade of the tumour. Therefore, an approach capable of detecting a small number of exfoliated tumour cells is in demand for urine-based diagnosis. As a commonly used WGA technique for single cell genome studies, MALBAC facilitates the analysis of trace amounts of starting materials and does not require additional DNA extraction [
12,
27]. MALBAC possesses the advantages of convenience and rapidness compared to a routine library construction process for NGS.
In the present study, we developed a novel strategy based on NGS that incorporates MALBAC and a new chromosomal imbalance evaluation approach, CIA, to assess the aberrant level of the chromosomal genome, and we demonstrated its application in detecting BC for the first time. Approximately 92% of the BC patients were identified as positive in tumour tissues according to the CIA, regardless of the TNM stage and histological grade, indicating its potentially wider utility for diagnosis. Moreover, as demonstrated in Fig.
3, the CNV profile of urine shows characteristics similar to tissues derived from the same patient, suggesting that urine cell pellets are representative of tumours for CNV assessment. The urine CIA also exhibits concordance with tissue CIA, indicating its potential as a non-invasive diagnostic strategy (Table
2). Notably, the performance of urine CIA was superior in the subgroup of patients carrying CIA-positive tumours than in all the patients. Tissue CIA might serve as a prior test to select patients possessing positive CIA scores in primary tumours for the subsequent recurrence surveillance by non-invasive urine CIA. Nonetheless, the heterogeneity in primary and recurrent tumours should also be taken into account, and the feasibility of this method needs to be investigated and validated by prospective studies.
The performance of this technique was also compared with other non-invasive methods. Voided urine cytology is routinely used in the clinic with good specificity (> 90%); however, the sensitivity has been reported to be 30–50% [
4]. CIA showed significantly improved sensitivity in detecting BC patients in this study (80–90% vs. 52.7%), which is also superior to the reported cytology sensitivity in the literature. The FISH (UroVysion®) probes alterations in chromosomes 3, 7, 17 and 9p21 and the sensitivity and specificity have been reported to be ~ 50–80% and 70–85%, respectively [
4,
17,
18]. CIA displayed a superior performance, with both the sensitivity and specificity being ~ 90% in the training and validation groups. Other commercially available markers (such as NMP22, ImmunoCyt and BTA stat) also show unsatisfactory performance [
4,
16].
One major limitation of the currently available urinary biomarkers is the poor sensitivity for early-stage and lower-grade tumours [
1,
4,
18]. However, 30–80% of patients diagnosed with a low-grade Ta/T1 primary tumour undergo recurrence within 5 years [
1‐
3]. In addition, tumours generally are larger or in a more advanced stages at diagnosis than during surveillance. A non-invasive test with high sensitivity, particularly for early-stage and low-grade tumours, is important for the surveillance of patients by reducing the use of invasive tests such as cystoscopy and thereby improving the patient quality of life [
1]. However, it has been reported that cells from men with low-grade BC accumulated fewer CNVs than those from cases with high-grade cancer [
28]. Hurst et al. [
29] also discovered that the more genomically unstable subtype of Ta bladder cancer was distinguished by loss of chromosome 9q, and the other subtype contained no or few copy-number alterations. This outcome might explain the observation that the CIA showed a slightly lower sensitivity in early-stage and low-grade tumours (83–85%) than that of more advanced stage and high-grade tumours (90–100%). However, the observed sensitivity is better than the performance of other commercial markers in the same stage/grade, which have been reported to be less than 80% in most cases [
1,
4,
18]. The sensitivity of CIA was observed to be only 60% in the 5 PUNLMP patients, which might be explained by its low level of malignancy. However, a limited number of patients were included in the present study; therefore, a conclusive statement cannot be made without further validation in a larger set of patients.
The cost of this new technique is estimated to be $200–300 per patient. Although this is relatively costly compared to the methylation and mutation combination assay reported recently ($23) [
21], with the rapid reductions in NGS cost, the CIA assay is expected to become cost effective in the near future. Due to the utility of MALBAC technology, the CIA assay does not require a large amount of DNA, which makes it a more suitable technique for urine-based testing. Compared to the DNA methylation assay, the minimum DNA input of which is approximately 50 ng (the amount present in 8000 cells), the MALBAC assay is applicable to a single cell equivalent amount of DNA [
12,
27].
It has been reported that genetic mutations in certain genes, such as FGFR3, RB1, HRAS, TP53, TSC1, TERT and others, occur in urinary bladder tumours [
30]. A proportion of urothelial tumours also harbour mutations that are potentially therapeutic targets, including the FGFR3, TSC1 and PIK3CA mutations [
31‐
34]. It is quite possible that a combination of the reported CIA and other mutations might improve the detection rate of bladder cancers and might be informative for therapy selection.
Conclusion
Overall, we developed a new strategy based on the chromosomal imbalance/aberration level and demonstrated its application in BC detection for the first time. Good concordance (87.0%) in the assessments obtained from patient tumours and urine was observed. The urine-based evaluation also demonstrated a good performance (accuracy = 89.0%, sensitivity = 83.1%, specificity = 94.5%, NPV = 85.4% and PPV = 93.7%; AUC = 0.917, 95%CI =0.868–0.966, P < 0.001) in the training group, particularly in patients with CIA-positive tumours (accuracy = 92.7%, sensitivity = 89.8%). The performance was also validated in an additional group, with a sensitivity and specificity of ~ 90%. It is conceivable that the present approach might have the potential to be a non-invasive test for BC diagnosis and to be subsequent surveillance prior to cystoscopy use. We envision that prospective cohort studies, with larger samples incorporating both BC patients and a certain percentage of patients with related symptoms and/or signs, will be designed to further validate the feasibility of monitoring bladder cancer patients.
Acknowledgements
The authors would like to thank Dr. Xin Dong, Qinsi Liang and Shujie Ma for their assistance with the data analysis as well as Yunyun Niu and Ting Ma for their help with experimental support and data acquisition.