Aberrantly glycosylated integrin α3β1 is a unique urinary biomarker for the diagnosis of bladder cancer

Study groups

We recruited 2725 participants overall: 408 in the test groups and 2317 in the validation groups (Figure 1). Of all the participants, 1314 had BC, and their demographic characteristics, modes of presentation, tumour stages and tumour grades are summarized in Table 1. The test population was predominantly male (78.8%), and 46.7% were over 63 years of age. Of those with a recorded presentation, 10.3% had visible haematuria, and 89.7% had non-visible haematuria. A total of 72.8% of patients had no recurrence, 14.7% experienced recurrence, and 12.5% were lost to follow-up. A total of 1130 BC patients were enrolled into the validation groups, and their clinicopathological characteristics were well matched to those of the test groups (Table 1); these two groups were well matched for age. Additionally, the patients with non-BC urologic conditions were also matched for age. Cut-off points were set as follows: 65 years of age for RCC, PC and cystitis patients, 47 years of age for nephritis patients, and 57 years of age for prostatitis patients (Supplementary Table 1).

Figure 1. Patient selection process and classification. Overall patient selection process and their classification based on types of diseases.

Urinary AG31 is a sensitive marker for the detection of bladder cancer

To detect AG31 levels in voided urine samples, an ELISA-based assay kit was developed by our laboratory (Supplementary Figure 1A). We generated several monoclonal antibodies that could recognize the AG31 molecule on BC tumours as previously described [17]. For establishment of the ELISA, we screened out an antibody termed BCMab3 as the capture antibody, which could strongly bind to AG31. The BCMab1 antibody also specifically recognizes AG31 on BC tumours [17], and BCMab1-conjugated horseradish peroxidase (HRP) served as the detection antibody. This ELISA kit could be used to accurately measure urinary AG31 levels (Supplementary Figure 1B). Relative light units (RLUs) were used to indicate the test levels of AG31 concentrations in urine.

In the test group, urinary AG31 concentrations in the BC patients were significantly higher than those in the healthy controls (median 6525, interquartile range (IQR) 3600-26,585; mean 54,768, standard definition (SD) 276,559 [BC patients] vs median 576, IQR 378-988; mean 859, SD 819 [healthy controls]; p<0.0001) (Figure 2A, and Supplementary Table 2). Notably, the urinary AG31 levels in patients with RCC, PC, cystitis, nephritis, or prostatitis were comparable to those in healthy individuals (Figure 2A and Supplementary Table 2). The ROC curves were plotted for urinary AG31 in BC patients versus different test groups. For BC patients from all test groups, the area under the curve (AUC) of AG31 was 0.9567 (95% CI 0.9337-0.9797), with a sensitivity of 90.76% and specificity of 91.52% (Figure 1C, and Table 2). The optimum cut-off value was set to 1991 (Supplementary Figure 1C). The four ROC curves of AG31 levels between the BC patients and different test groups showed that the AUCs were greater than 95% (Figure 2C, and Table 2). These results were confirmed by the corresponding validation group tests (Figure 2B, 2D). Predictive values and likelihood ratios for AG31 in the diagnosis of BC are shown in Table 2. Altogether, the urinary AG31 test can distinguish BC patients from patients with other urologic tumours and benign inflammatory diseases.

Figure 2. Urinary AG31 levels are elevated in bladder cancers in the test and validation groups. (A) Urinary AG31 levels for test groups. (B) Urinary AG31 levels for validation groups. Black horizontal lines are means, and error bars are SEs. Urinary AG31 levels were measured with RLU (relative light unit). HC, healthy control; BC, bladder cancer; RCC, renal cell carcinoma; PC, prostate cancer. (C) ROC curves for all patients with bladder cancer versus all controls in the test groups. (D) ROC curves for all patients with bladder cancer versus all controls in the validation groups. Jagged curves denote ROC curves; diagonal lines represent reference lines. HC, healthy control; BC, bladder cancer; RCC, renal cell carcinoma; PC, prostate cancer.

Urinary AG31 levels are well correlated with the clinicopathologic features of bladder cancer

In the test group, the urinary AG31 levels of the BC patients with different disease stages were much higher than those of the healthy controls (Figure 3A, and Supplementary Table 2). Moreover, urinary AG31 levels were significantly increased in patients with high-stage BC. These observations were further confirmed in the validation group (Figure 3B and Supplementary Table 2). Furthermore, the AUCs for the AG31 levels of the BC patients with the different disease stages were greater than 92%, with sensitivities over 89% and specificities over 90% (Supplementary Figure 2, and Supplementary Table 3). Similarly, urinary AG31 levels increased with advancing pathological grade (Figure 3C, D, Supplementary Figure 3, Supplementary Tables 2, 3). Correlation analysis showed that urinary AG31 levels were positively correlated with tumour stage and grade (both p<0.01). Taken together, these data indicate that AG31 levels are well correlated with the clinicopathologic features of BCs.

Figure 3. Urinary AG31 levels distinguish between stages and grades of bladder cancer. (A, B) Urinary AG31 levels of bladder cancer patients with different disease stages versus healthy controls in the test groups (A) and in the validation groups (B). (C, D) Urinary AG31 levels of bladder cancer patients with different grades versus healthy controls in the test groups (C) and in the validation groups (D). HC, healthy controls; RLU, relative light unit.

Urinary AG31 test is diagnostically accurate for NMIBC patients

Approximately 70% of BC patients with NMIBC (Tis+Ta+T1) will experience one or more recurrences after transurethral resection (TUR), and the BCs 10%-20% of patients will eventually progress to MIBC (3,26). Regarding recurrence and progression, patients with NMIBC are classified as low/intermediate- and high-risk groups (5,27). We next wanted to determine whether the urinary AG31 test could distinguish NMIBC patients. Of all the BC patients recruited, 1058 were classified as having stage Ta or T1 disease. The mean AG31 level of the Ta-stage patients was 13.49 times higher than that of the healthy controls in the test group (Figure 3A and Supplementary Table 2). In contrast, the mean AG31 level of the T1-stage patients was 32.11 times higher than that of the healthy controls. For the patients with NMIBC, ROC curves showed that the AUC of the AG31 levels was 0.9524 (95% CI 0.9248-0.9800), with a sensitivity of 91.04% and specificity of 91.35% in the test group compared with the patients with benign inflammatory diseases and the healthy controls (Figure 4A, and Table 3). When the NMIBC patients were compared only with the healthy controls, the AUC of the AG31 levels was 0.9574 (95% CI 0.9310-0.9838), with a sensitivity of 91.04% and specificity of 91.53% in the test group. Similar results were obtained in the validation groups (Figure 4B, and Table 3). Importantly, the diagnostic accuracy of the AG31 test was not affected by age, sex, or haematuria in BC patients (Supplementary Table 4). Altogether, these findings indicate that the urinary AG31 test is diagnostically accurate for NMIBC patients.

Figure 4. ROC curves for urinary AG31 levels in the detection of NMIBC patients. (A) ROC curves for patients with NMIBC versus the controls in the test groups. (B) ROC curves for patients with NMIBC versus the controls in the validation groups. Jagged curves denote ROC curves; diagonal lines represent reference lines. HC, healthy control; NMIBC, non-muscle-invasive bladder cancer.

The urinary AG31 test is much more sensitive and specific than the NMP22 test

NMP22, a member of the nuclear matrix protein (NMP) family, is much more prevalent in malignant urothelial cells than in normal cells. Given that NMP22 is released into urine upon cell apoptosis, NMP22 is significantly elevated in the urine of BC patients compared to that of healthy individuals. The NMP22 test kit was approved by the U.S. Food and Drug Administration (FDA) for use in surveillance of BC. Given that the NMP22 test was used as a biomarker for the auxiliary diagnosis of BC, we applied the urinary NMP22 test and compared the results to those of the AG31 test. To further assess the diagnostic accuracy of the AG31 assay, we measured AG31 and NMP22 levels in freshly voided urine samples from BC patients (n=53) and healthy individuals (n=54). The AUC of the AG31 levels was 99.65% (95% CI: 0.9906-1.0020), with a sensitivity and specificity for the detection of BC of 90.57% and 98.15%, respectively (Supplementary Figure 4A, 4B). However, the AUC of NMP22 levels was 74.11% (95% CI: 0.6476-0.8346), and the sensitivity and specificity for the detection of BC were 47.17% and 87.04%, respectively. Discordances between the AG31 test and NMP22 test were assessed using the nonparametric McNemar test, and the p value was 0.01. Importantly, the predictive values and likelihood ratios for AG31 were much better than those for NMP22 (Supplementary Figure 4C). In summary, the AG31 test in urine has a better diagnostic accuracy for BC than the NMP22 test.

Primary antibodies

BCMab1 and BCMab3 monoclonal antibodies, which specifically recognize the AG31 epitope on the membranes of BC cells, were generated by our group [17]. BCMab3 was used as the capture antibody, and BCMab1 served as the detecting antibody. BCMab1 was conjugated to HRP by linkage of 2 imine-carbon (1-ethyl-3-(3-dimethylaminopropyl)-carbodiimide, EDC) as described previously [18, 19].

Study population

From January to June 2012, we enrolled consecutive patients with BC from the Peking University People’s Hospital (Beijing, China) and Cancer Institute and Hospital, Chinese Academy of Medical Science (Beijing, China) for constitute our test group. During the same time period, we also recruited consecutive patients with RCC, prostate cancer (PC), cystitis, nephritis, or prostatitis and healthy control subjects from The First Affiliated Hospital of Shenzhen University (Shenzhen, China). From July 2012 to September 2013, validation groups comprising patients with BC, RCC, PC, cystitis, nephritis, or prostatitis and healthy control subjects were recruited from The First Affiliated Hospital of Zhejiang University (Hangzhou, China), The Second Affiliated Hospital of Kunming Medical University (Kunming, China), Renji Hospital Affiliated to Shanghai Jiaotong University School of Medicine (Shanghai, China), and The First Affiliated Hospital of Jilin University (Changchun, China).

The presence of BC was confirmed by cystoscopy, together with histopathological information obtained after subsequent surgical interventions. Tumours were graded according to the WHO criteria [20, 21], and tumour stages were defined according to the tumour-node-metastasis (TNM) staging system [22]. BC patients in this study were subdivided into five stages (Ta, T1, T2, T3, and T4) and three pathological grades (G1, G2, and G3). For the purpose of this study, we classified TNM stage Tis tumours and Ta tumours including stage Ta, Ta and T1 tumours as NMIBC. The diagnoses of cystitis and nephritis were based on symptoms, physical examination and the results of the urine culture, according to the guidelines of the Infectious Diseases Society of America and the European Society for Microbiology and Infectious Diseases [23]. The diagnosis of prostatitis was based on clinical, laboratory, and imaging evidence (X-ray, ultrasonography, CT, or MRI) [24, 25]. The presence of RCC was confirmed by symptoms, imaging studies (CT or MRI), laboratory data, and renal tumour biopsy. The diagnosis of PC was based on digital rectal examination, the prostate-specific antigen test, imaging evidence (transrectal ultrasonography, CT, MRI, or ECT), prostate biopsy, and pathohistological examination of radical prostatectomy specimens, according to the clinical guidelines of the European Association of Urology. The healthy control subjects were eligible volunteers with no diseases of the urinary system, no viral hepatitis and no malignant diseases. Individuals who had a history of other solid tumours were excluded from the study.

For the test groups, we enrolled patients with BC, RCC, PC, cystitis, nephritis, or prostatitis, along with healthy control subjects, from January to June 2012. For the validation groups, patients with BC, RCC, PC, cystitis, nephritis, or prostatitis and healthy control subjects were recruited from July 2012 to September 2013. These two groups were matched for age (≤63 years versus >63 years) and sex for comparisons of AG31 values. In this study, all urine samples of the patients with BC, RCC, PC, cystitis, nephritis or prostatitis were collected before treatment. Data collection and analysis were performed by three independent researchers (YD, CL, and QZ). Approval for the study was obtained from the institutional ethics review committee of each study centre. Written informed consent was obtained from all participants, according to the committees’ regulations.

Measurement of urine samples for the AG31 test

Each well of a 96-well Nunc-Immuno microtiter plate with a MaxiSorp surface (Nalge Nunc, Penfield, NY, USA) was coated with 100 μL of a 4 μg/mL solution of BCMab3 in 50 mM carbonate buffer (pH 9.5), and plates were incubated at room temperature overnight. The wells were then blocked with 120 μL of 10 mM phosphate buffer (pH 7.4) containing 1% bovine serum albumin and 1% gelatine at 37 °C for 2 h. The plates were washed four times with PBS containing 0.05% Tween-20 and then dried at room temperature. These plates were kept at 4 °C. For the urinary AG31 test, 50 μL of urine samples and 50 μL of HRP-BCMab1 (1 ng/mL) were added to each well and incubated at 37 °C for 1 h. After the plates were completely washed, they were dried at room temperature. A total of 100 μL of freshly prepared substrate solution was added to each well and stirred. Chemiluminescence intensity was measured with a chemiluminescence apparatus (Hamamatsu Photonics, Beijing, China).

NMP22 measurement

A commercial ELISA kit (Alere Scarborough Inc., Scarborough, ME, USA) was used for auxiliary diagnosis of BC as a biomarker. According to the manufacturer’s protocol for the NMP22 test, urinary NMP22 levels were measured for comparison.

Statistical analysis

Statistical analyses were conducted with SPSS for Windows (version 16.0) and GraphPad Prism statistical software (version 5.01). The Mann-Whitney U test was used to assess differences between two independent groups, such as the levels of urinary AG31 between the patients with BC and the healthy controls. The differences in clinicopathological stages or grades were assessed using the Kruskal-Wallis H test, which was used to analyse multiple variables. Receiver operating characteristic (ROC) analysis was used to characterize marker performance, and ROC curves were constructed to assess sensitivity, specificity, and the respective areas under the curve (AUCs) with 95% confidence intervals. In the ROC curves, the true positive rate (sensitivity) is plotted against the false positive rate (1-specificity) for different cut-off points of a parameter. We chose the optimum cut-off value for diagnosis by maximizing the sum of sensitivity and specificity and minimizing the overall error (square root of the sum [1-sensitivity]² + [1-specificity]²) and by minimizing the distance of the cut-off value to the top-left corner of the ROC curve. The correlations between the levels of AG31 in urine and clinicopathological characteristics were analysed using Pearson’s chi-square test (χ²) or Fisher’s exact test. The bivariate correlation and the Spearman correlation coefficient were used to assess the magnitude of the correlation between urinary AG31 levels and tumour stages or grades. Discordances between the AG31 test and NMP22 test were assessed using the nonparametric McNemar test. In all statistical analyses, a p value of 0.05 or less was considered statistically significant.