Purpose:

Urothelial carcinoma is a malignant cancer with frequent chromosomal aberrations. Here, we investigated the application of a cost-effective, low-coverage whole-genome sequencing technology in detecting all chromosomal aberrations.

Experimental Design:

Patients with urothelial carcinomas and nontumor controls were prospectively recruited in clinical trial NCT03998371. Urine-exfoliated cell DNA was analyzed by Illumina HiSeq XTen, followed by genotyping with a customized bioinformatics workflow named Urine Exfoliated Cells Copy Number Aberration Detector (UroCAD).

Results:

In the discovery phase, urine samples from 126 patients with urothelial carcinomas and 64 nontumor disease samples were analyzed. Frequent chromosome copy-number changes were found in patients with tumor as compared with nontumor controls. A novel diagnosis model, UroCAD, was built by incorporating all the autosomal chromosomal changes. The model reached performance of AUC = 0.92 (95% confidence interval, 89.4%–97.3%). At the optimal cutoff, |Z| ≥ 3.21, the sensitivity, specificity, and accuracy were 82.5%, 96.9%, and 89.0%, respectively. The prediction positivity was found correlated with tumor grade (P = 0.01). In the external validation cohort of 95 participants, the UroCAD assay identified urothelial carcinomas with an overall sensitivity of 80.4%, specificity of 94.9%, and AUC of 0.91. Meanwhile, UroCAD assay outperformed cytology tests with significantly improved sensitivity (80.4% vs. 33.9%; P < 0.001) and comparable specificity (94.9% vs. 100%; P = 0.49).

Conclusions:

UroCAD could be a robust urothelial carcinoma diagnostic method with improved sensitivity and similar specificity as compared with cytology tests. It may be used as a noninvasive approach for diagnosis and recurrence surveillance in urothelial carcinoma prior to the use of cystoscopy, which would largely reduce the burden on patients.

Translational Relevance

Urothelial carcinomas are the fourth most common tumors in men, and urothelial carcinomas harbor frequent copy-number variations (CNVs). CNVs are revealed to be a hallmark of cancer, thus identifying CNVs in urine for noninvasive detection of urothelial carcinomas may be possible. In this study, we developed a novel and optimized low-coverage whole-genome sequencing technique, named Urine Exfoliated Cells Copy Number Aberration Detector (UroCAD), to detect CNVs in urine-exfoliated cells. Moreover, we validated this technique to be a robust noninvasive way to detect urothelial carcinomas in a prospective, double-center, single-blinded clinical trial. We found the UroCAD assay with significantly improved sensitivity and similar specificity compared with urine cytology. Testing the CNVs in urine may be a noninvasive approach for diagnosis and surveillance of urothelial carcinomas in clinical practice, which may largely reduce the burden of cystoscopy on patients with urothelial carcinomas.

Urothelial carcinomas are the fourth most common tumors in men. Urothelial carcinomas arise from the urothelium of urinary tract, which is extended from renal pelvis and ureters (upper urinary tract) to the bladder and urethra (lower urinary tract; ref. 1). Bladder cancer is the most common urinary tract malignancy and accounts for 90%–95% of urothelial carcinomas, while upper tract urothelial carcinomas (UTUCs) account for only 5%–10% of urothelial carcinomas (2). Patients diagnosed with early-stage urothelial carcinomas will undergo lifelong surveillance based on cystoscopy and urine cytology, resulting in urothelial carcinomas costing more per patient lifetime than any other cancers (3). However, cystoscopy is invasive, costly, and may miss flat lesions with only 68%–83% sensitivity (4). Urine cytology is specific, but lacks sensitivity, especially for low-grade urothelial carcinomas, with sensitivity varying from 29% to 84% for different grades of urothelial carcinomas (5). Therefore, there is a clinical need for sensitive, noninvasive, convenient, and affordable techniques to complement current clinical practice of urothelial carcinomas.

As urothelial cells from urinary tracts are in direct contact with urine, numerous attempts have been made to develop noninvasive biomarkers from urine for diagnosis and surveillance of urothelial carcinomas. Emerging biomarkers including urine proteins, circulating tumor DNA, RNAs, extracellular vesicles, and metabolic products have been reported in recent studies (6–11). The FDA has approved six different noninvasive urine-based tests for monitoring bladder cancer recurrence, such as NMP22 and UroVysion, with sensitivity of 30%–100% and specificity of 55%–98% (2, 12). However, none of these markers have been validated for detection of UTUC, and none has been accepted for diagnosis or follow-up of bladder cancer in routine clinical practice guidelines of urothelial carcinomas due to performance inconsistencies, technical issues, and high cost (2, 13).

Genetic alterations, such as mutation and methylation, play a significant role in the development of urothelial carcinomas (14). However, not all bladder tumors harbor common mutations such as FGFR3, TP53, PIK3CA, and MET, and it is also not realistic to include all genetic alterations at nucleotide level in one diagnostic assay. The link between chromosomal abnormalities and cancer was first proposed by Theodor Boveri, more than a hundred years ago (15). Since then, it has been revealed that copy-number variations (CNVs) are nearly ubiquitous in cancer, and are a hallmark of human cancer (15, 16). Chromosomal alterations including deletions on chromosome 3, 8, 9, 11, 13, and 17 have been commonly seen in bladder cancer, and such chromosomal alterations can be measured by karyotyping and FISH (17). Given the prevalence of CNVs in bladder cancer, significant effort has been made to identify specific CNVs in urothelial carcinomas that may contain potential driver genes, or are associated with prognosis as well as diagnostic assay (18–20). The urinary FISH test (UroVysion), which probes alterations in chromosomes 3, 7, 17, and 9p21, is one of the FDA-approved and commercially available urinary biomarkers used to detect urothelial carcinomas, with overall sensitivity of 30%–86% and specificity of 63%–95% (2, 5).

The CNVs burden refers to the altered genome as a percentage of genome length, and it is suggested that tumor CNV burden, rather than individual CNVs, may be associated with cancer characteristics (16). Low-coverage whole-genome sequencing (LC-WGS) was first developed in 2014 by Scheinin and colleagues as a simple, cost-effective, and robust technique for the identification of CNVs in tumor samples (21). In this study, we developed a novel and optimized LC-WGS technique called Urine Exfoliated Cells Copy Number Aberration Detector (UroCAD) to detect all the common CNVs in a highly noisy background, which might be caused by exfoliated “noncancerous” epithelial cells and white blood cells in inflammation patients. The aim of this study was to assess the threshold and accuracy of UroCAD for detecting urothelial carcinomas in a double-center, prospective, single-blinded clinical trial.

### Patients' characteristics and ethics statement

Participants were prospectively recruited (clinical trial, NCT03998371) as approved by the Ethics Committee of Changhai Hospital (Shanghai, China, ethical approval No. CHEC2019-134) and Qilu Hospital (Jinan, Shandong, China, ethical approval No. KYLL-2019-2-099), after written informed consents were obtained. Studies were conducted in accordance with the ethical principles in the Declaration of Helsinki.

### Study design

It was a double-center, single-blinded, prospective study. The participants were informed with consent and diagnoses information was obtained before sample collection. Urine for testing was collected before standard-of-care treatments, such as transurethral resection of bladder tumor (TURBT), cystectomy, or other surgeries, were performed. The UroCAD result was not intended to be used in patient management. The investigators of UroCAD were blinded to the patient information of urine samples and urologists were blinded to the test results. In the discovery phase, participants were recruited from Changhai Hospital (Shanghai, China) from May 2019 to November 2019. When the cutoff was established on the basis of the discovery set, participants were recruited from both Changhai Hospital (Shanghai, China) and Qilu Hospital (Jinan, Shandong, China) from March 2020 to May 2020 in the validation phase.

The inclusion criteria included participants suspected with urothelial carcinoma and planned to undergo surgery, participants without any tumor disease and willing to attend the study by providing morning urine, male or female patients ages ≥18 years, and participants who signed informed consent form. Exclusion criteria included age under 18 years, individuals unwilling to sign the consent form or unwilling to provide morning urine for test or unwilling to provide the medical record, individuals unwilling to participate in this trial, and individuals already having indwelling catheter. All patients underwent urine sampling, performed at the timepoint of hospital admission, to determine the chromosomal aberration and/or urine cytology.

### DNA extraction

Each urine sample was collected from the first miction in the morning and before operative treatment was performed. Urine sample was collected using the Cell Preservation Solution Kit (Prophet Genomics) and transported to the central laboratory at room temperature within 72 hours. Urine (10 mL) was centrifuged at 1,000 × g for 10 minutes, and the pellet was retained for DNA extraction according to the manufacturer's instructions (Qiagen). DNA purity and quantity were measured by NanoDrop One (Pelab) and Qubit 4.0 (Thermo Fisher Scientific), respectively.

### LC-WGS

For LC-WGS, libraries were prepared using the Kapa Hyper Plus Kit with custom adapters (Integrated DNA Technologies and Broad Institute), starting with 2–20 ng of genomic DNA input (median, 5 ng), or approximately 1,000–6,000 haploid genome equivalents, used for low-pass whole-genome sequencing. Up to 22 libraries were pooled and sequenced using 150-bp paired-end runs over 1 × lane on a HiSeq XTen (Illumina). Segment copy numbers were derived via customized workflow, named as UroCAD. Sample was excluded if the median absolute deviation of copy ratios (log2 ratio) between adjacent bins and genome wide were higher than 0.38, which suggested poor quality sequence data. The raw sequence and processed data files are available through the National Omics Data Encyclopedia database (https://www.biosino.org/node/search) with accession number OEP001092.

### Voided urine cytology

In the discovery phase, urine cytology was performed for patients with urothelial carcinomas at the discretion of urologists in outpatient clinics. In the validation phase, all participants provided the same morning urine sample for urine cytology and UroCAD analysis. Urine cytology was conducted by the centrifuge and cytospin methods. The prepared slides were then sent to pathologists for analysis by following the standard protocol. The cytology test results were recorded as negative and positive (finding suspicious tumor cells and tumor cells both as positive).

### Statistical analyses and data visualization

Urine sediments cell DNA was extracted and analyzed by Illumina X10. At least 10 million paired reads were collected for each sample. The reads were mapped to human reference genome hg19. Genomic coverage was then counted by using the software samtools mpileup (22). We then calculated average coverage for each 200k bin. Coverages for each bin were then normalized by Z-score by using the formula as below:

where |covereg{e_{raw}}$| is the raw coverage obtained from sequencing and |coverag{e_{controls,\ raw}}$| is the raw coverage from the technique control samples with matched laboratory protocols. Technique control samples was a set of 11 urine sediments cell DNA from healthy volunteers.

Circular binary segmentation algorithm from R package, DNACopy (23), was then used to detect significant genomic breakpoints and copy number–changed genomic segments. R package “DNACopy” was used to analyze copy-number changes. P = 0.05 was considered as statistically significant binary segmentation. Absolute segment value was used for further analysis. The sensitivity and specificity of UroCAD were estimated by ROC curves. For categorical variables, the χ2 test was used as appropriate. All statistical analyses were performed using SPSS17.0. Proportion trend tests were used to analyze the associations between clinicopathologic UroCAD screening positivity and clinicopathologic parameters. Data are reported as means and SDs, medians and interquartile ranges, and HRs or ORs with 95% confidence intervals (CIs), as appropriate. Missing data were removed from the analyses. All analyses were performed with the use of R software, version 3.4.3 (R Foundation for Statistical Computing). Anonymized data and R code used in the statistical analysis will be made available per reasonable request.

### Patient characterizations and CNVs in urothelial carcinoma

A total of 238 participants were enrolled from Changhai Hospital (Shanghai, China) in the discovery dataset. A flow diagram summarizing the identified eligible participants is shown in Fig. 1. Detailed demographics and clinical characteristics of included participants are shown in Table 1. There were 33 participants with recurrent bladder cancer and had received TURBT and intravesical instillation with chemotherapy or Bacille Calmette-Guérin. The outcome of DNA quality controls is shown in Supplementary Table S1. Higher DNA concentrations were found in urine samples from urothelial participants as compared with noncancer participants (Table 1).

Figure 1.

Overview of patient selection. CH, Changhai Hospital; FN, false negative; FP, false true positive; QL, Qilu Hospital; TN, true negative; TP, true positive; UC, urothelial carcinoma.

Figure 1.

Overview of patient selection. CH, Changhai Hospital; FN, false negative; FP, false true positive; QL, Qilu Hospital; TN, true negative; TP, true positive; UC, urothelial carcinoma.

Close modal
Table 1.

Patient characteristics of training and validation cohort.

Training cohort (n = 238)Validation cohort (n = 99)
UCControlPUCControlP
Participants excluded  30 18
Participants included for analysis  126 64  56 39
Age Mean (SD) 65.63 (10.61) 51.11 (18.01) <0.001 67.25 (10.41) 56.90 (10.58) <0.001
Gender Male/female 109/17 50/14 0.139 51/5 21/18 <0.001
Smoke Yes/no 84/42 19/45 <0.001 37/19 12/27 0.001
BMI Mean (SD) 24.14 (2.95) 24.34 (2.96) 0.654 24.87 (3.02) 25.00 (2.99) 0.828
DNA conc. Mean (SD) 52.99 (77.42) 27.40 (43.27) 0.015 14.92 (18.33) 18.43 (16.32) 0.34
<1 ng/μL
1–10 ng/μL 30 28  22
≥10 ng/μL 88 27  28 24
Benign urinary diseases in control Urinary stonesa  25   19
Prostateb
Kidneyd
Othere
Healthy volunteers  11
Tumor organ Renal pelvis
Ureter 10
Multiple
Low 31
High 91   45
N.A.
Tumor stage ≥pT2 46   12
<pT2 79   43
N.A.
Training cohort (n = 238)Validation cohort (n = 99)
UCControlPUCControlP
Participants excluded  30 18
Participants included for analysis  126 64  56 39
Age Mean (SD) 65.63 (10.61) 51.11 (18.01) <0.001 67.25 (10.41) 56.90 (10.58) <0.001
Gender Male/female 109/17 50/14 0.139 51/5 21/18 <0.001
Smoke Yes/no 84/42 19/45 <0.001 37/19 12/27 0.001
BMI Mean (SD) 24.14 (2.95) 24.34 (2.96) 0.654 24.87 (3.02) 25.00 (2.99) 0.828
DNA conc. Mean (SD) 52.99 (77.42) 27.40 (43.27) 0.015 14.92 (18.33) 18.43 (16.32) 0.34
<1 ng/μL
1–10 ng/μL 30 28  22
≥10 ng/μL 88 27  28 24
Benign urinary diseases in control Urinary stonesa  25   19
Prostateb
Kidneyd
Othere
Healthy volunteers  11
Tumor organ Renal pelvis
Ureter 10
Multiple
Low 31
High 91   45
N.A.
Tumor stage ≥pT2 46   12
<pT2 79   43
N.A.

Abbreviations: BMI, body mass index; N.A., not available; PUNLMP, papillary urothelial neoplasms of low malignant potential; UC, urothelial carcinoma.

aParticipants with kidney stones, ureter stones, or bladder stones.

bParticipants with benign prostatic hyperplasia

cParticipants with urinary incontinence, cystitis glandularis, interstitial cystitis, bladder diverticulum, papilloma, neurogenic bladder, or infection.

dParticipants with renal cyst, nonfunctional kidney due to stone or stenosis.

eParticipants with varicocele or hydrocele of testis.

A genome-wide overview of CNVs in discovery phase is summarized in Fig. 2. Chromosomal breakpoints were frequently identified on centromeres, resulting in chromosomal arm imbalances (Fig. 2A). No significant CNV was found in healthy controls (Fig. 2B). Frequent genomic aberrations included chr9 loss, chr17p loss, chr7 gain, etc.

Figure 2.

Frequent chromosomal aberrations detected in urine sediments from patients with cancer. A, The common changes included frequent chromosomes gains, 1q, 3q, 5p, 6p, 7, 8q, 10p, 17q, and 20; frequent losses, 2q, 5q, 8p, 9, 11p, 13, 17p, and 18q; and focal amplifications, MYC, CCND1, and E2F3. B, No obvious chromosome changes were found in nontumor controls.

Figure 2.

Frequent chromosomal aberrations detected in urine sediments from patients with cancer. A, The common changes included frequent chromosomes gains, 1q, 3q, 5p, 6p, 7, 8q, 10p, 17q, and 20; frequent losses, 2q, 5q, 8p, 9, 11p, 13, 17p, and 18q; and focal amplifications, MYC, CCND1, and E2F3. B, No obvious chromosome changes were found in nontumor controls.

Close modal

Z-scores for all chromosomal arms except sex chromosomes were calculated by normalizing to health controls, as described in the formula:

where |{V_{tumor}}$| is the normalized sequencing coverage of tumor sample and |{V_{control}}$| is the normalized sequencing coverage of control sample in the same genomic region.

### Copy-number Z-scores showed high concordances in paired tissues and could be reproduced in repeated urine samplings

To validate whether UroCAD assay could serve as liquid biopsy reflecting CNV characteristics of urothelial carcinomas, 20 paired urine samples and tumor tissues had UroCAD analysis. For 16 patients with positive UroCAD assay, we found the copy-number Z-score from urine-exfoliated cells showed high concordance (with correlation coefficient ranging from 0.70 to 0.98) with the ones from paired tumor tissues. While for patients with negative UroCAD assay, the correlation coefficient ranged from −0.31 to 0.40 (Supplementary Table S2).

We further validated the consistency of outcomes of UroCAD assay in different conditions. First, we collected urine samples for UroCAD assay from eight participants (four patients with tumor and four controls) at different timepoints, the results supported high concordances in different testing timepoints for the four patients with cancer with correlation coefficients of 0.99, 0.88, 0.99, and 0.98 (see Supplementary Table S3; Supplementary Fig. S1 for data details). Second, we performed dilution test by mixing the urine samples from patients with cancer with the urine samples from nontumor patients with dilution ratios from 1:1 to 1:50. A statistically significant detection (|Z| ≥ 3.21) could still be found at a dilution ratio as low as 1:20. The data suggested UroCAD CNV sequencing assay could detect tumor cells in a urine sample with percentage as low as 5% (Supplementary Table S4). Third, we collected fresh urine and matched urine stored at room temperature for 72 hours. Copy-number Z-score from these samples showed high concordance with matched fresh urine samples (Supplementary Table S5; Supplementary Fig. S2).

### Chromosomal changes Z-scores predict malignancy from nontumor controls

Chromosomal copy gains and losses were also used to predict tumor malignancy. In the discovery phase, AUC for each chromosome arm was found ranging from 0.51 to 0.80 (median AUC = 0.65; Supplementary Table S6). For chromosome aberrations used by UroVysion FISH assay (chr3q gain, 17q gain, 7p gain, and 9p21 loss), the AUCs were 0.69 (95% CI, 0.61–0.76), 0.64 (95% CI, 0.56–0.72), 0.67 (95% CI, 0.60–0.75), and 0.64 (95% CI, 0.56–0.72) for 3q gain, 17q gain, 7p gain, and 9p loss, respectively. We also found chr8q gain, 9q loss, 17p loss, 5q loss, 11p loss, 10q loss, 1q gain, and 8p loss to be more predictive than the ones used by UroVysion FISH assay, with AUCs ranging from 0.721 to 0.797 (Supplementary Table S6).

### The UroCAD diagnostic model by incorporating all chromosomal changes

We, next combined all the autosomal chromosome changes to build diagnostic models for the prediction of urothelial carcinoma. Two diagnostic models were built. Model No. 1 incorporated chr1–22 and model No. 2 used chromosomes from UroVysion FISH assay. A positive diagnosis was defined as |Z| ≥ cutoff. As shown in Fig. 3A, The AUCs were 0.92 (95% CI, 0.89–0.97) and 0.84 (95% CI, 0.81–0.92) for model No. 1 and model No. 2, respectively. Model No. 1 (“all chroms” model) outperformed model No. 2 (“UroVysion chroms”; P < 0.001). The optimal Z-score cutoff, |Z| ≥ 3.21, was found in this study by Youden Index. At this cutoff, “all chroms” model gave a diagnosis with sensitivity of 82.5% and specificity of 96.9% (Table 2). Furthermore, we found the sensitivity of UroCAD was significantly correlated with tumor grade, but not correlated with tumor stage and tumor size (Table 3). As shown in Supplementary Table S7, UroCAD resulted in sensitivity of 65.6% and 87.9% for histologically low-grade and high-grade urothelial carcinomas (P = 0.010), respectively. For low-grade, pTa patients, UroCAD resulted in sensitivity of 60.0% (15/25).

Figure 3.

ROC curve of UroCAD model by incorporating all detectable chromosome aberrations. A, ROC curve of the training cohort. B, ROC curve of the validation cohort. Model 1, all chromosome arms except sex chromosomes and model 2, all chromosome arms from UroVysion FISH (chr 3, 7, 17, and 9p21).

Figure 3.

ROC curve of UroCAD model by incorporating all detectable chromosome aberrations. A, ROC curve of the training cohort. B, ROC curve of the validation cohort. Model 1, all chromosome arms except sex chromosomes and model 2, all chromosome arms from UroVysion FISH (chr 3, 7, 17, and 9p21).

Close modal
Table 2.

The diagnoses performance of the UroCAD model by incorporating all chromosome aberrations.

SensitivitySpecificityPPVNPV
TNFPFNTPTP/(TP+FN)TN/(TN+FP)TP/(TP+FP)TN/(TN+FN)
Model 1 (all chroms) |Z| ≥ 2.4 44 20 11 115 91.3% 68.8% 85.2% 80.0%
|Z| ≥ 3 58 18 108 85.7% 90.6% 94.7% 76.3%
|Z| > = 3.21 62 22 104 82.5% 96.9% 98.1% 73.8%
|Z| ≥ 4 64 34 92 73.0% 100.00% 100.00% 65.3%
Model 2 (UroVysion chroms) |Z| ≥ 2.4 61 40 86 68.2% 95.3% 96.7% 60.4%
(3q, 7p, 7q, 17q, 9p) |Z| ≥ 3 64 49 77 61.1% 100.0% 100.0% 56.6%
|Z| ≥ 3.21 64 51 75 59.5% 100.00% 100.00% 55.6%
|Z| ≥ 4 64 58 68 53.9% 100.00% 100.00% 52.4%
SensitivitySpecificityPPVNPV
TNFPFNTPTP/(TP+FN)TN/(TN+FP)TP/(TP+FP)TN/(TN+FN)
Model 1 (all chroms) |Z| ≥ 2.4 44 20 11 115 91.3% 68.8% 85.2% 80.0%
|Z| ≥ 3 58 18 108 85.7% 90.6% 94.7% 76.3%
|Z| > = 3.21 62 22 104 82.5% 96.9% 98.1% 73.8%
|Z| ≥ 4 64 34 92 73.0% 100.00% 100.00% 65.3%
Model 2 (UroVysion chroms) |Z| ≥ 2.4 61 40 86 68.2% 95.3% 96.7% 60.4%
(3q, 7p, 7q, 17q, 9p) |Z| ≥ 3 64 49 77 61.1% 100.0% 100.0% 56.6%
|Z| ≥ 3.21 64 51 75 59.5% 100.00% 100.00% 55.6%
|Z| ≥ 4 64 58 68 53.9% 100.00% 100.00% 52.4%

Abbreviations: FN, false negative; FP, false true positive; NPV, negative predictive value; PPV, positive predictive value; TN, true negative; TP, true positive.

Table 3.

Clinicopathologic parameters associated with detectable chromosomal changes in urine.

Category++++++P++++++P
Diameter (cm) ≤1 0.165 0.069
(1,3] 12 42  17
>3 36  18
N.A.
T stage Ta, T1, Cis 17 50 0.132 31 0.744
T2–T4 36
N.A.
Low 11 17
High 11 67  36
N.A.
Smoke No 30 0.868 12 0.107
Yes 15 56  29
Recurrent Primary tumor 16 63 0.899 30 0.968
Recurrent tumor 23  11
Category++++++P++++++P
Diameter (cm) ≤1 0.165 0.069
(1,3] 12 42  17
>3 36  18
N.A.
T stage Ta, T1, Cis 17 50 0.132 31 0.744
T2–T4 36
N.A.
Low 11 17
High 11 67  36
N.A.
Smoke No 30 0.868 12 0.107
Yes 15 56  29
Recurrent Primary tumor 16 63 0.899 30 0.968
Recurrent tumor 23  11

Abbreviations: N.A., not available; PUNLMP, papillary urothelial neoplasms of low malignant potential.

aUroCAD strong positive test was marked with “+++,” medium positive test “++,” weak positive “+,” and negative “−.” They were defined as at least three chromosome arms with |Z| ≥ 3.21, two chromosome arms with |Z| ≥ 3.21, one chromosome arm with |Z| ≥ 3.21, and no chromosome arm with |Z| ≥ 3.21, respectively.

### UroCAD assay showed improved sensitivity compared with voided urine cytology tests

In the discovery cohort, the UroCAD diagnosis model found 14 of 14 (100%) cytology-positive tumors. In addition, it also found 30 of 38 (78.9%) cytology-negative tumors. In summary, the UroCAD assay produced 214% more positive findings than cytology (Supplementary Table S8). As shown in Supplementary Table S7, further analyses showed that UroCAD diagnoses model outperformed cytology in both UTUC (100% vs. 21.4%) and bladder cancer (78.8% vs. 28.9%), and high- (87.9% vs. 35.3%) and low-grade (65.6% vs. 13.3%) tumors. The UroCAD model also showed better performance in smaller (diameter less than 3 cm) tumors (78.2% vs. 22.0%).

### External validation proved the diagnostic value of UroCAD assay

To further validate the diagnostic model, we performed single-blinded trial to compare the performance of the UroCAD with urine cytology in an independent external validation cohort. As shown in Fig. 1, 99 participants were enrolled, four participants were excluded with reasons (one participant in control group and one in urothelial carcinoma group were found with incidental prostate cancer through transurethral resection of prostate and cystectomy, respectively. Two participants in the urothelial carcinoma group did not undergo surgery and had no pathologic result). Baseline demographics of participants are shown in Table 1. The UroCAD identified urothelial carcinomas with an overall sensitivity of 80.4%, specificity of 94.9%, and AUC of 0.91 (Fig. 3B). UroCAD resulted in sensitivity of 60.0% and 86.7% for low-grade and high-grade urothelial carcinomas, respectively (Supplementary Table S9). The sensitivity also correlated with the size of tumor, which was 66.7% for tumor ≤1 cm, 72% for tumor between 1 and 3 cm, and 95.5% for >3 cm. In terms participants with low-grade and pTa tumor, the UroCAD detected urothelial carcinomas at a sensitivity of 71.4% (5/7). For UTUC, UroCAD resulted in sensitivity of 63.6% (7/11).

Compared with urine cytology, we found UroCAD had significantly higher sensitivity (80.4% vs. 33.9%; P < 0.001) and comparable specificity (94.9% vs. 100%; P = 0.494). UroCAD diagnosis model detected 17 of 19 (89.5%) cytology-positive tumors and 28 of 37 (75.7%) cytology-negative tumors. For low-grade and pTa tumors, the most frequent challenge confronting cytologists, UroCAD assay had obviously advantages, with sensitivity of 71.4% (5/7) compared with 0% (0/7) for urine cytology. In terms of UTUC also, UroCAD assay showed better performance compared with urine cytology (63.6% vs. 18.2%; Supplementary Table S9).

The symptom of hematuria is strongly correlated with urothelial carcinoma diagnosis (24). Hematuria accounts for 20% of all urological visits, and urological cancer is found to be the cause of up to 20% macroscopic hematuria (25). Thus, the discrimination of these patients with cancer from patients presented with hematuria is of paramount importance. The traditional diagnostic methods of urothelial carcinoma include CT urography (CTU) and cystoscopy (24). Urine-based techniques detecting protein and nucleic acid have been emerging recently. FDA has approved six methods for diagnosis and surveillance of bladder cancer, including NMP22, bladder tumor antigen (BTA), FISH, and fluorescence IHC (ImmunoCyt; ref. 20). According to a meta-analysis, the sensitivity and specificity of quantitative NMP22, qualitative NMP22, qualitative BTA, quantitative BTA, FISH, ImmunoCyt are 69% and 77%, 58% and 88%, 64% and 77%, 65% and 74%, 63% and 87%, and 78% and 85%, respectively (20). However, the clinical application of these methods is still rare due to the low sensitivity, poor accuracy for low-stage and low-grade tumors, high-cost, and ambiguous clinical readout (20, 26). The lack of standardized kits, controlling for patient demographics, and extensive multi-center validation is also an issue (7). Therefore, a noninvasive diagnostic method with high accuracy is warranted.

Chromosomal instability, referred to the ongoing acquisition of genomic alterations ranging from point mutations to gross chromosomal rearrangements, is a hallmark of cancer, which is found in 60%–80% of human cancers, and it positively correlates with high-tumor stage, poor prognosis, metastasis, and therapeutic resistance (15, 27). Here, we investigated the novel LC-WGS–based technology named UroCAD, which is able to detect chromosomal aberrations of the urine-exfoliated cells, as a diagnostic tool for urothelial carcinomas. In the training cohort, the UroCAD model reached performance of AUC = 0.92. With the optimal cutoff |Z| ≥ 3.21, the overall sensitivity, specificity, and accuracy were 82.5%, 96.9%, and 87.4%, respectively. The Z-score of urine sample and the corresponding tumor tissue showed high concordance. Meanwhile, the result was not affected by different timepoints of urine collection, which makes it a practical test in clinical application. Furthermore, the high efficacy of UroCAD test allows 20 times dilution of the urine samples. In the external validation cohort, the UroCAD assay reached a sensitivity of 80.4%, specificity of 94.9%, and AUC of 0.91. Detecting low-grade, Ta lesions is a major challenge in clinical practice, UroCAD reached sensitivity of 60.0% and 71.4% in training and validation cohort, respectively. Hurst and colleagues (28) demonstrated that pTa tumors could be grouped into two major molecular subtypes based on differential CNVs, 55% (78/141) contained no for few CNVs, while 45% (63/141) with more CNVs primarily by high frequency of 9q deletion.

Several researches have investigated the value of detecting chromosomal instability with LC-WGS in either cell-free (cf)DNA or genomic DNA as a noninvasive diagnostic method for bladder cancers. For example, Cheng and colleagues (19) investigated the value of detecting methylation and CNVs of DNA extracted from urine sample of bladder cancer with a LC-WGS technique for diagnosing the disease with sensitivity of 69.6% and specificity of 100% in 85 participants. Ge and colleagues (29) assessed the copy-number profiles of the urine cfDNA by LC-WGS, the sensitivity and specificity of this test were 78.6% and 87.5% in the external validation cohort of 52 participants, respectively. The relatively small sample size and nonprospective study design, without blinding were the main limitations of these studies. However, these researches and this study have shown that detecting CNVs with LC-WGS is a promising noninvasive way for diagnosis of urothelial carcinomas. Other urine tumor DNA assays that measured DNA methylation, mutations, or cfDNA have also been explored for detection and surveillance of bladder cancer, the sensitivity of these biomarkers ranged from 82% to 93%, and specificity ranged from 82% to 97% (8–11, 30).

Cytology tests have been widely used in screening for urothelial carcinomas, which is noninvasive, inexpensive, simple, and valuable for high-grade and flat lesions. But it has a sensitivity varying from 29% to 84% for tumors of different stages and grades (5), and it lacks a clear diagnostic criteria, thus depends on the expertise of the cytopathologist (5, 31). In the validation cohort, our study has shown that UroCAD outperformed cytology by detecting 17 of 19 cytology-positive urothelial carcinomas, as well as 28 cytology-negative patients. We found UroCAD had significantly higher sensitivity (80.4% vs. 33.9%; P < 0.001) and comparable specificity (94.9% vs. 100%; P = 0.494) compared with cytology. UroVysion FISH assay is also currently available in clinic practice. According to previous study, the sensitivity and specificity of FISH test were 63% and 87%, respectively (20). In this study, our assay incorporated all the chromosomal changes, and achieved higher sensitivity and specificity compared with the model that only incorporated chromosomes 3, 7, 17, and 9p21.

Another major merit of our method is the ability of detecting UTUCs. UTUC is an uncommon disease, but has a poor prognosis as a result of the difficulties in detecting the lesions (32). The traditional diagnostic methods include urinary cytology, CTU, and ureteroscopy. But they have shortcomings, such as ureteroscopy is invasive and may cause severe complications as well as lead to intravesical recurrence (IVR) resulting from tumor seeding (33). It was reported that the IVR rate in patients with and without ureteroscopy prior to radical nephroureterectomy range from 39.2% to 60.7% and from 16.7% to 46%, respectively (34). Other methods include UroVysion, ImmunoCyt, NMP22, and BTA, which achieved the sensitivity of 76.7%, 75%, 70.5%, and 50%, respectively (35–38). In our study, we found the sensitivity of UroCAD in detecting UTUC was higher than cytology (100% vs. 21.4% in training cohort and 63.6% vs. 18.2% in validation cohort), which showed the potential to detect UTUC by a noninvasive approach. However, the value of UroCAD in detecting UTUC still needs to be validated in a larger sample cohort.

In summary, UroCAD could be a highly specific, robust, and noninvasive urothelial carcinoma diagnostic method with improved sensitivity and similar specificity as compared with cytology tests. It may be used as a noninvasive approach for diagnosis and recurrence surveillance in urothelial carcinomas prior to the use of cystoscopy, which would largely reduce the burden on patients.

No potential conflicts of interest were disclosed.

S. Zeng: Conceptualization, resources, data curation, supervision, funding acquisition, writing-original draft, writing-review and editing. Y. Ying: Resources, data curation, investigation, writing-original draft, writing-review and editing. N. Xing: Conceptualization, data curation, funding acquisition, investigation. B. Wang: Conceptualization, data curation, software, supervision, methodology. Z. Qian Conceptualization, resources, supervision, funding acquisition, methodology, writing-original draft. Z. Zhou: Data curation. Z. Zhang: Data curation, funding acquisition. W. Xu: Data curation, formal analysis. H. Wang: Data curation, investigation. L. Dai: Data curation, validation. L. Gao: Resources, supervision. T. Zhou: Resources, data curation, funding acquisition. J. Ji: Conceptualization, data curation, supervision, funding acquisition, investigation. C. Xu: Conceptualization, data curation, supervision, funding acquisition, writing-original draft, writing-review and editing.

We thank Yan-Lei Chu, the secretary of the Department of Urology in Changhai Hospital, for her assistance in collecting and sending urine samples used in the study. This research was financed by grants from Shanghai Sailing Program (18YF1422700), Discipline development plan in Changhai Hospital (2019YXK041), National Natural Science Foundation of China (81772720, 81572509, and 81802515), Technology Innovation Action Project of Science and Technology Commission of Shanghai City (16411969700), Shanghai Key Laboratory of Cell Engineering (14DZ2272300), the Project of Excellent Academic Leader of Science and Technology Committee of Shanghai City (17XD1404900), National Science and Technology Major Project (2017ZX09304030), Shanghai Clinical Medical Center of Urological Diseases Program (2017ZZ01005), and Department of Science &Technology of Jinan city (201805030).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Siegel
RL
,
Miller
KD
,
Jemal
A
.
Cancer statistics, 2019
.
CA Cancer J Clin
2019
;
69
:
7
34
.
2.
Babjuk
M
,
Burger
M
,
Compérat
EM
,
Gontero
P
,
Mostafid
AH
,
Palou
J
, et al
European Association of Urology guidelines on non-muscle-invasive bladder cancer (TaT1 and carcinoma in situ) - 2019 update
.
Eur Urol
2019
;
76
:
639
57
.
3.
Avritscher
EBC
,
Cooksley
CD
,
Grossman
HB
,
Sabichi
AL
,
Hamblin
L
,
Dinney
CP
, et al
Clinical model of lifetime cost of treating bladder cancer and associated complications
.
Urology
2006
;
68
:
549
53
.
4.
Helenius
M
,
Brekkan
E
,
Dahlman
P
,
Lönnemark
M
,
Magnusson
A
.
Bladder cancer detection in patients with gross haematuria: computed tomography urography with enhancement-triggered scan versus flexible cystoscopy
.
Scand J Urol
2015
;
49
:
377
81
.
5.
Dimashkieh
H
,
Wolff
DJ
,
Smith
TM
,
Houser
PM
,
Nietert
PJ
,
Yang
JJCc
.
Evaluation of UroVysion and cytology for bladder cancer detection: a study of 1835 paired urine samples with clinical and histologic correlation
.
Cancer Cytopathol
2013
;
121
:
591
7
.
6.
Miyake
M
,
Owari
T
,
Hori
S
,
Nakai
Y
,
Fujimoto
K
.
Emerging biomarkers for the diagnosis and monitoring of urothelial carcinoma
.
Res Rep Urol
2018
;
10
:
251
61
.
7.
Chakraborty
A
,
Dasari
S
,
Long
W
,
Mohan
C
.
Urine protein biomarkers for the detection, surveillance, and treatment response prediction of bladder cancer
.
Am J Cancer Res
2019
;
9
:
1104
17
.
8.
Kessel
KEMv
,
Beukers
W
,
Lurkin
I
,
AZ-vd
,
Keur
KAvd
,
Boormans
JL
, et al
Validation of a DNA methylation-mutation urine assay to select patients with hematuria for cystoscopy
.
J Urol
2017
;
197
:
590
5
.
9.
Christensen
E
,
Birkenkamp-Demtröder
K
,
Nordentoft
I
,
Høyer
S
,
van der Keur
K
,
van Kessel
K
, et al
Liquid biopsy analysis of FGFR3 and PIK3CA hotspot mutations for disease surveillance in bladder cancer
.
Eur Urol
2017
;
71
:
961
9
.
10.
Dudley
JC
,
Schroers-Martin
J
,
Lazzareschi
DV
,
Shi
WY
,
Chen
SB
,
Esfahani
MS
, et al
Detection and surveillance of bladder cancer using urine tumor DNA
.
Cancer Discov
2019
;
9
:
500
9
.
11.
Beukers
W
,
Kandimalla
R
,
van Houwelingen
D
,
Kovacic
H
,
Chin
JF
,
Lingsma
HF
, et al
The use of molecular analyses in voided urine for the assessment of patients with hematuria
.
PLoS One
2013
;
8
:
e77657
.
12.
Chou
R
,
Gore
JL
,
Buckley
D
,
Fu
R
,
Gustafson
K
,
Griffin
JC
, et al
Urinary biomarkers for diagnosis of bladder cancer: a systematic review and meta-analysis
.
Ann Intern Med
2015
;
163
:
922
31
.
13.
Springer
SU
,
Chen
C-H
,
Rodriguez Pena
MDC
,
Li
L
,
Douville
C
,
Wang
Y
, et al
Non-invasive detection of urothelial cancer through the analysis of driver gene mutations and aneuploidy
.
Elife
2018
;
7
:
e32143
.
14.
Larsen
KL
,
Lind
EG
,
Guldberg
P
,
Dahl
C
.
DNA-methylation-based detection of urological cancer in urine: overview of biomarkers and considerations on biomarker design, source of DNA, and detection technologies
.
Int J Mol Sci
2019
;
20
:
2657
.
15.
Bakhoum
SF
,
Cantley
LC
.
The multifaceted role of chromosomal instability in cancer and its microenvironment
.
Cell
2018
;
174
:
1347
60
.
16.
Hieronymus
H
,
Murali
R
,
Tin
A
,
K
,
Abida
W
,
Moller
H
, et al
Tumor copy number alteration burden is a pan-cancer prognostic factor associated with recurrence and death
.
eLife
2018
;
7
:
e37294
.
17.
N
,
Mathew
B
,
Jatawa
S
,
Tiwari
A
.
Genetic instability in urinary bladder cancer: an evolving hallmark
.
2013
;
59
:
284
8
.
18.
Lindquist
KJ
,
Sanford
T
,
Friedlander
TW
,
Paris
PL
,
Porten
SP
.
Copy number gains at chr3p25 and chr11p11 are associated with lymph node involvement and survival in muscle-invasive bladder tumors
.
PLoS One
2017
;
12
:
e0187975
.
19.
Cheng
THT
,
Jiang
P
,
Teoh
JYC
,
Heung
MMS
,
Tam
JCW
,
Sun
X
, et al
Noninvasive detection of bladder cancer by shallow-depth genome-wide bisulfite sequencing of urinary cell-free DNA for methylation and copy number profiling
.
Clin Chem
2020
;
65
:
927
36
.
20.
Abbosh
PH
,
McConkey
DJ
,
Plimack
ER
.
Targeting signaling transduction pathways in bladder cancer
.
Curr Oncol Rep
2015
;
17
:
58
.
21.
Scheinin
I
,
Sie
D
,
Bengtsson
H
,
Van De Wiel
MA
,
Olshen
AB
,
Van Thuijl
HF
, et al
DNA copy number analysis of fresh and formalin-fixed specimens by shallow whole-genome sequencing with identification and exclusion of problematic regions in the genome assembly
.
Genome Res
2014
;
24
:
2022
32
.
22.
Li
H
,
Handsaker
B
,
Wysoker
A
,
Fennell
T
,
Ruan
J
,
Homer
N
, et al
The sequence Alignment/Map format and SAMtools
.
Bioinformatics
2009
;
25
:
2078
9
.
23.
Seshan
VE
,
Olshen
A
.
DNAcopy: DNA copy number data analysis. R package version 1.62.0
.
Available from:
https://bioconductor.org/packages/release/bioc/html/DNAcopy.html.
24.
Kamat
AM
,
Hahn
NM
,
Efstathiou
JA
,
Lerner
SP
,
Malmstrom
PU
,
Choi
W
, et al
.
Lancet
2016
;
388
:
2796
810
.
25.
Mariani
AJ
,
Mariani
MC
,
Macchioni
C
,
Stams
UK
,
Hariharan
A
,
Moriera
A
.
The significance of adult hematuria: 1,000 hematuria evaluations including a risk-benefit and cost-effectiveness analysis
.
J Urol
1989
;
141
:
350
5
.
26.
Chen
A
,
Fu
G
,
Xu
Z
,
Sun
Y
,
Chen
X
,
Cheng
KS
, et al
Detection of urothelial bladder carcinoma via microfluidic immunoassay and single-cell DNA copy-number alteration analysis of captured urinary-exfoliated tumor cells
.
Cancer Res
2018
;
78
:
4073
85
.
27.
Sansregret
L
,
Vanhaesebroeck
B
,
Swanton
C
.
Determinants and clinical implications of chromosomal instability in cancer
.
Nat Rev Clin Oncol
2018
;
15
:
139
50
.
28.
Hurst
CD
,
Alder
O
,
Platt
FM
,
Droop
A
,
LF
,
Burns
JE
, et al
Genomic subtypes of non-invasive bladder cancer with distinct metabolic profile and female gender bias in KDM6A mutation frequency
.
Cancer Cell
2017
;
32
:
701
15
.
29.
Ge
G
,
Peng
D
,
Guan
B
,
Zhou
Y
,
Gong
Y
,
Shi
Y
, et al
Urothelial carcinoma detection based on copy number profiles of urinary cell-free DNA by shallow whole-genome sequencing
.
Clin Chem
2019
;
66
:
188
98
.
30.
Birkenkamp-Demtröder
K
,
Nordentoft
I
,
Christensen
E
,
Høyer
S
,
Reinert
T
,
Vang
S
, et al
Genomic alterations in liquid biopsies from patients with bladder cancer
.
Eur Urol
2016
;
70
:
75
82
.
31.
Black
PC
,
Brown
GA
,
Dinney
CP
.
Molecular markers of urothelial cancer and their use in the monitoring of superficial urothelial cancer
.
J Clin Oncol
2006
;
24
:
5528
35
.
32.
Mian
C
,
Mazzoleni
G
,
Vikoler
S
,
Martini
T
,
Knuchel-Clark
R
,
Zaak
D
, et al
Fluorescence in situ hybridisation in the diagnosis of upper urinary tract tumours
.
Eur Urol
2010
;
58
:
288
92
.
33.
Fojecki
G
,
Magnusson
A
,
Traxer
O
,
Baard
J
,
Osther
PJS
,
Jaremko
G
, et al
Consultation on UTUC, Stockholm 2018 aspects of diagnosis of upper tract urothelial carcinoma
.
World J Urol
2019
;
37
:
2271
8
.
34.
Marchioni
M
,
Primiceri
G
,
Cindolo
L
,
Hampton
LJ
,
Grob
MB
,
Guruli
G
, et al
Impact of diagnostic ureteroscopy on intravesical recurrence in patients undergoing radical nephroureterectomy for upper tract urothelial cancer: a systematic review and meta-analysis
.
BJU Int
2017
;
120
:
313
9
.
35.
Lodde
M
,
Mian
C
,
Wiener
H
,
Haitel
A
,
Pycha
A
,
Marberger
M
.
Detection of upper urinary tract transitional cell carcinoma with ImmunoCyt: a preliminary report
.
Urology
2001
;
58
:
362
6
.
36.
Marín-Aguilera
M
,
Mengual
L
,
Ribal
MJ
,
Musquera
M
,
Ars
E
,
Villavicencio
H
, et al
Utility of fluorescence in situ hybridization as a non-invasive technique in the diagnosis of upper urinary tract urothelial carcinoma
.
Eur Urol
2007
;
51
:
409
15
.
37.
Jovanovic
M
,
Soldatovic
I
,
Janjic
A
,
Vuksanovic
A
,
Dzamic
Z
,
Acimovic
M
, et al
Diagnostic value of the nuclear matrix protein 22 test and urine cytology in upper tract urothelial tumors
.
Urol Int
2011
;
87
:
134
7
.
38.
Siemens
DR
,
Morales
A
,
Johnston
B
,
Emerson
L
.
A comparative analysis of rapid urine tests for the diagnosis of upper urinary tract malignancy
.
Can J Urol
2003
;
10
:
1754
8
.