A multiplex immunoassay for the non-invasive detection of bladder cancer

Under Western Institutional Review Board approval (IRB #Rosser 2014-1), previously collected and banked voided urine samples were available for analysis. Voided urine samples were collected prior to cystoscopy in all BCa subjects and controls and samples were analyzed blinded to the clinical status of the participants, thus the study satisfies both PRoBE and STARD study design [19, 20]. Patients with known renal disease or documented renal insufficiency were excluded from the current study. The study consisted of two independent cohorts (Table 1); cohort #1 consisted of 62 subjects (29 with newly diagnosed BCa and 33 with no previous history of urothelia carcinoma, gross hematuria, active urinary tract infection or urolithiasis, i.e., controls) and cohort #2 consisted of 200 subjects (100 with newly diagnosed BCa and 100 with no previous history of urothelia carcinoma, gross hematuria, active urinary tract infection or urolithiasis, i.e., controls). Controls for the two cohorts consisted of healthy volunteers and individuals with voiding symptoms or microscopic hematuria. All 62 subjects in cohort #1 had their urines analyzed by individual commercial ELISA kits directed towards the ten targets in addition to the MSD multiplex assays in order to compare and contrast the two diagnostic modalities. Cohort #2 was only analyzed by the multiplex assay. Clinical information associated with these urine samples were queried from our database.

Each urine sample was centrifuged at 600×g 4 °C for 5 min. The supernatant was decanted and aliquoted, while the urinary pellet was snap frozen. Both the supernatant and pellet were stored at −80 °C prior to analysis. Aliquots of urine supernatants were thawed and analyzed for protein content using a Pierce 660-nm Protein Assay Kit (Thermo Fisher Scientific Inc., Waltham, MA, USA). Frozen aliquots of urine samples were thawed and protein content was measured using a Pierce 660-nm Protein Assay Kit (Thermo Fisher Scientific Inc., Waltham, MA, USA) and a microplate reader (Synergy HT, BioTek Instruments, Winooski, VT, USA). The relatively constant production of creatinine, a non-enzymatically metabolite of creatine, makes urinary creatinine a useful tool for normalizing the levels of other molecules found in urine [21]. The concentration of urinary creatinine was measured using a commercially available enzymatic assay (Cat # KGE005 R&D Systems Inc., Minneapolis, MN, USA) according to the manufacturer’s instructions. Briefly, urine supernatants were treated with alkaline picrate solution, which when creatinine is present, yields an orange-red color. Intensity at 490 nm corresponds to the concentration of creatinine in the sample. Creatinine concentrations of unknown samples were calculated by comparison to a standard curve.

Levels of human Interleukin 8 (IL8, Cat # ab46032 Abcam), Matrix Metalloproteinase 9 (MMP9, Cat # DMP900 R&D Systems Inc.), Plasminogen Activator Inhibitor 1 (SERPINE1, Cat # EA-0207 Signosis Inc.), Vascular Endothelial Growth Factor A (VEGFA, Cat # 100663 Abcam), Angiogenin (ANG, Cat # CK400 CellSciences), Carbonic Anhydrase 9 (CA9, Cat # DCA900 R&D Systems Inc.), Matrix Metalloproteinase 10 (MMP10, Cat # DMP1000 R&D Systems Inc.), Apolipoprotein E (APOE, Cat # KA 1031 Abnova), Syndecan 1 (SDC1, Cat # ab46506 Abcam) and A1AT (SERPINA1, Cat # ab108799, Abcam) were monitored in urine samples using commercial enzyme-linked immunosorbent assays (ELISA) as previously reported [17]. Calibration curves were prepared using purified standards for each protein assessed. Curve fitting was accomplished by either linear or four-parameter logistic regression following the manufacturer’s instructions.

A panel of commercially available monoclonal antibodies against each of the ten biomarkers was screened pairwise and antibody pairs were selected from an unbiased screen using recombinant protein and normal pooled urine. The final monoclonal antibody pair (capture and detection) was selected based on sensitivity, specificity, physical properties, and recognition of native protein. Detection monoclonal antibodies were labeled with ruthenium (SULFO-TAG™ NHS-Ester), according the manufacturer’ s instructions (Meso Scale Diagnostics, LLC, Rockville, MD, USA). The multiplex assay, using MULTI-SPOT^® plates, is based on a proprietary combination of electrochemiluminescence detection and patterned arrays [22]. Briefly, immobilized capture antibodies were placed in 350–500 μm spots at the bottom of polypropylene 96-well plates to capture target proteins (for details see; http://​www.​mesoscale.​com/​CatalogSystemWeb​/​WebRoot/​). Each spot is coated with a different analyte capture antibody, in this case against IL8, MMP9, MMP10, ANG, APOE, SDC1, A1AT, PAI1, CA9 and VEGFA. The sandwich immunoassay complex that forms generates light when the instrument applies a voltage. Optics and a cooled-CCD camera then collect and quantitatively measure light emitted and processing algorithms convert the data into target concentrations. A seven point standard curve across the 4 log dynamic range of the assays was included in the current assay design.

Initial sample testing noted the need to dilute 7 of the 10 biomarkers fourfold (IL8, MMP9, MMP10, APOE, PAI1, CA9 and VEGFA) and 3 of the 10 biomarkers 200-fold (ANG, SDC1, A1AT) to ensure concentrations were within optimal range on the calibration curve. Thus, we elected to divide the ten biomarkers into two patterned array panels (Panel 1—IL8, MMP9, MMP10, VEGFA, CA9, APOE and PAI1 and Panel 2—A1AT, ANG and SDC1.

Urine samples were handled on ice and diluted with MSD Assay Diluent 37, designed to reduce the effects from heterophilic antibodies and other interferents. Samples and standards (50 μl) were loaded onto the MSD^® plate in duplicate with a multichannel pipettor in order to reduce pipetting error and allowed to incubate for 2 h and then washed out. SULFO-TAG conjugated detection antibody (25 μl) was added to each well and allowed to incubate for 2 h and then washed out. Subsequently, the sandwich immunoassay complex that forms was incubated MSD Read Buffer (150 μl), the electrochemiluminescent substrate, and levels of electrochemiluminescent units were measured on the QuickPlex^® SQ 120 (MSD) instrument. Standard curves were constructed using MSD Discovery Workbench^® 4.0, which allows for the selection of multiple non-linear and linear equations to fit the standard curve. Optimal curve fits were determined by visual graph evaluation and comparison of Akaike’s information criteria (AIC) values [23].

First, urinary concentrations of each of the ten biomarkers were normalized using urinary creatinine concentration. Next, we compared the diagnostic ability associated with the combination of the individual commercial ELISA assays directed at our ten biomarkers versus the diagnostic ability of the multiplex assay, which encompassed these same ten biomarkers. Then we investigated the diagnostic performance of the multiplex assay in an independent cohort. Logistic regression analysis with BCa status (yes vs. no) as the response variable and the ten biomarkers as the explanatory variables was performed. The individual biomarkers were combined into a linear combination with the regression coefficients obtained in logistic regression as the weights, and the linear combination was used as a combined score for the detection of BCa. Cutoff thresholds were identified in cohort #1 and applied to the analysis of cohort #2. Then for a given cutoff threshold, we calculated the sensitivity and specificity of the test. We generated a ROC curve by plotting values for sensitivity against the false-positive rates (1-specificity) for various cutoff thresholds [24]. The relative ability of the combination of biomarkers to indicate BCa was estimated by calculating the area under the ROC curves (AUC), with a higher AUC indicating a stronger predictor. We select the optimal cutoff value defined by the Youden index [25], i.e., the cutoff value that maximizes the sum of the sensitivity and the specificity. We estimated the sensitivity, specificity, PPV and NPV of the combination of biomarkers at the optimal cutoff value. Statistical significance in this study was set at p < 0.05 and all reported p values were 2-sided. All analyses were performed using SAS software version 9.3 (SAS Institute Inc., Cary, NC, USA).

The physical components, a library of capture and detection antibodies, and the secondary reagents for the MULTI-ARRAY technology, have undergone extensive optimization by the manufacturers for consistent implementation. Ranges for each analyte assay were evaluated by dilution of standards to determine upper ranges where high-end hook effect and apparent antibody saturation are avoided and lower ranges that are above detection limits. Lower limits of detection (LLOD) were calculated based on 2.5× the standard deviation of the background assessed across the plates run during sample testing. LLOD for each analyte in Panel 1 was 0.081 pg/ml IL8, 38.4 pg/ml MMP9, 3.39 pg/ml MMP10, 0.030 pg/ml VEGFA, 2.13 pg/ml CA9, 236 pg/ml APOE and 1.97 pg/ml PAI1 and in Panel 2 was 17.4 pg/ml A1AT, 0.13 pg/ml ANG and 0.62 pg/ml SDC1, demonstrating sufficient sensitivity to detect proteins present in small amounts in voided urine samples. Intra assay precision was measured with acceptance criteria of a coefficient of variation (% CV) of less than 15.0. Median inter assay variability across all plates was also determined to be less than 15 % CV for each analyte. As the technology is an array, all components were checked for cross reactivity with other components in the antigen and antibody cocktails and confirmed to have less than 0.5 % cross reactivity. Lower limits of quantification (LLOQ) were estimated as being at the concentration with signals least three times over background and the lowest point of the 7-point standard curve where the back-fit regression recovered to within 20 % of the known value and had % CVs of less than 20 %. The dilution linearity of five pooled urine samples was between 80 and 120 % for all analytes at the recommended sample dilution factor, suggesting that there was not any interference due to the matrix.

In order to test the robustness of the multiplex assay, 62 urines samples (29 from BCa subjects) were monitored for the ten biomarkers using both the multiplex assay and the ten commercially available individual assays. Demographics and disease characteristics of cohort #1 are summarized in Table 1. Urinary concentrations of 7 of the 10 biomarkers were significantly elevated in patients with BCa compared to controls in both multiplex assay and individual ELISA assays (Table 2). A combinatorial analysis of all ten biomarkers using optimal cutoff values defined by Youden index calculations resulted in an AUROC (Fig. 1a) of 0.9258 (95 % CI 0.8559–0.9958) in the multiplex assay and 0.9467 (95 % CI 0.8944–0.9990) for the individual ELISA assays (p = 0.625). Thus, there was no difference between the diagnostic performances of the two formats. The multiplex assay achieved an overall sensitivity of 0.79, specificity of 0.97, PPV of 0.96 and NPV of 0.84 for BCa classification. Table 3 provides AUROC for each biomarker and combination of the ten biomarkers tested in the multiplex assay and individual ELISA assay.

In order to validate the diagnostic performance of the assay in a larger cohort, the multiplex assay was applied to an independent cohort of 200 urine samples (100 from BCa subjects). Demographics and disease characteristics of the cohort (cohort #2) are summarized in Table 1. In this validation cohort, urinary concentrations of 9 of the 10 biomarkers were significantly elevated in patients with BCa compared to controls (Table 4). Furthermore, the urinary concentrations of 3 of the 10 biomarkers (IL8, MMP10 and PAI1) were significantly elevated in patients with high-grade BCa compared to low-grade BCa, while 7 of the 10 biomarkers (IL8, MMP9, A1AT, ANG, VEGF, MMP10 and PAI1) were significantly elevated in patients with muscle invasive BCa (MIBC) compared to non-muscle invasive BCa (NMIBC) (Table 4). Based on our prediction rule developed from cohort #1, the multiplex assay achieved an overall diagnostic sensitivity of 0.85 %, specificity of 0.81 %, PPV 0.82 % and NPV 0.84 % for the combination of the ten biomarkers (Fig. 1b). Table 5 provides AUROC and corresponding sensitivity, specificity, PPV and NPV values for all biomarkers tested.