Investigating Colorimetric Protein Array Assay Schemes for Detection of Recurrence of Bladder Cancer

A colorimetric microarray for the multiplexed detection of recurrence of bladder cancer including protein markers interleukin-8 (IL8), decorin (DCN), and vascular endothelial growth factor (VEGF) was established to enable easy and cheap read-out by a simple office scanner paving the way for quick therapy monitoring at doctors’ offices. The chip is based on the principle of a sandwich immunoassay and was optimized prior to multiplexing using IL8 as a model marker. Six different colorimetric assay formats were evaluated using a detection antibody (dAB) labeled with (I) gold (Au) nanoparticles (NPs), (II) carbon NPs, (III) oxidized carbon NPs, and a biotinylated dAB in combination with (IV) neutravidin–carbon, (V) streptavidin (strp)–gold, and (VI) strp–horseradish peroxidase (HRP). Assay Format (III) worked best for NP-based detection and showed a low background while the enzymatic approach, using 3,3′,5,5′-tetramethylbenzidine (TMB) substrate, led to the most intense signals with good reproducibility. Both assay formats showed consistent spot morphology as well as detection limits lower than 15 ng/L IL8 and were thus applied for the multiplexed detection of IL8, DCN, and VEGF in synthetic urine. Colorimetric detection in urine (1:3) yields reaction signals and measurement ranges well comparable with detection in the assay buffer, as well as excellent data reproducibility as indicated by the coefficient of variation (CV 5–9%).


Introduction
Bladder cancer (BCa) is a serious malignancy of the urinary tract and prominent for its high rate of recurrence concerning 50% of all treated patients. To intervene recurrence of BCa, routine cytology and cystoscopy are done, representing the gold standard. Nevertheless, these methods are expensive, time-consuming, and invasive and lead to urinary infections in up to 16% of patients. Urine cytology has a high specificity (78-100%), although it lacks robust sensitivity (12.2-84.6%), especially for low and intermediate grade tumors, and is operator-dependent [1]. A number of urinary biomarkers (including fluorescence in-situ hybridization (FISH), ImmunoCyt, and nuclear matrix protein 22 (NMP22)) and investigational urine markers [2] also exist, but none of these markers has yet been shown to decrease the need for cystoscopy. Furthermore, biomarkers that aid the clinician in making treatment decisions are still an unmet need. Biomarkers identifying patients most likely to respond to chemotherapy for example would have enormous utility, not only reducing morbidity and improving quality of life but also significant cost savings to health providers. In previous work [3], we compiled a list of relevant bladder cancer protein markers based on extensive literature search (>1000 research articles) and prior defined inclusion and exclusion criteria. Microarray technology, a miniaturized high throughput analysis method, that can measure a high number of biomarkers in parallel was applied as a screening tool to narrow down the biomarker candidates and to evaluate the marker profiles in a clinical

Chip Fabrication
IL8, DCN, VEGF capture antibodies (cAB) were diluted in printing buffer (1 × PBS (pH 7.2)/0.01% sodium deoxycholate) to 0.4 g/L. The antibodies were arrayed on Supernitro slides (ArrayIt ® , Sunnyvale, CA, USA) using the Arrayit Nanoprint contact spotter. The spot-to-spot distance was 500 µm. Each probe was spotted in 9 replicates in 12 identical arrays per slide at a relative humidity of 50%. To allow full immobilization of the probes the slides were kept at 4 • C for at least three days.

Preparation of Immunogold Probes
1 mL of 40 nm gold particles (40 nm, CG-40-100, lot#2457227_40) from Cytodiagnostics (Burlington, VT, Canada) was centrifuged at 4 • C for 15 min using 2500 g and the supernatant was discarded. 100 µL of 0.1 mM 16-MHA in water was added and incubated for 18 h at 22 • C. For EDC activation, the carboxylated gold nanoparticles were centrifuged for 15 min and resuspended in 98 µL water and vortexed. 2 µL 50 mM EDC in water were added and vortexed. The particles were incubated for 30 min at 22 • C. Then they were centrifuged for 15 min and resuspended in 100 µL 0.1 g/L unconjugated anti-IL8 in water. The incubation was carried out at 22 • C for 2 h. After a final centrifugation step the pellet was resuspended in 100 µL TBS (20 mM Tris, 150 mM NaCl, pH 8.0) + 1% BSA.

Oxidation of Carbon Nanoparticles
200 mg of carbon Special Black 100 powder (51 nm) from Orion (Frankfurt, Germany) was dispersed in distilled water for 30 min under sonication. The resulting suspension was then refluxed in 10 mL of nitric acid at 120 • C for 3 h to obtain oxidized carbon nanoparticles (oCNPs) [16]. After centrifugation the solid sample was washed two times with deionized water and then dried at 100 • C for 18 h. Oxidation was proved by elemental analysis (SEM TM3030 with EDX, Hitachi, Feldkirchen, Germany) and Fourier-transform infrared spectra (FTIR) based on carbon black doped potassium bromide (KBr) pellets using a Spectrum 100 FT-IR instrument from PerkinElmer (Waltham, MA, USA) [17].

Antibody Coupling to oCNPs
1% (w/v) suspension of oCNPs was prepared and sonicated for 30 s. Then 100 µL of 0.5 g/L unconjugated anti-IL8, anti DCN or anti VEGF was added dropwise to 150 µL of 5-fold dilution of oCNPs in 5 mM sodium borate coupling buffer (pH 8.8) (0.2% w/v) and allowed to react for 18 h at 4 • C under gentle stirring.

Modification of Carbon NPs with Neutravidin
1% (w/v) suspension of carbon NPs (used as received) was prepared and sonicated for 5 min. Then the 1% suspension was diluted 5-fold in 5 mM borate buffer and again ultrasonicated for 5 min. Next, 0.35 mg neutravidin was added to 1 mL of diluted carbon NP suspension and incubated for 3 h at RT on an over-head shaker (Biorotator, Biospec Products) [18].
Both probes from Sections 2.5 and 2.6 were then washed (15 min, 13,600 g) at least three times with washing buffer (5 mM borate buffer + 1% BSA + 0.002% NaN 3 ). Finally, the supernatant was removed and the pellet resuspended in storage buffer (100 mM borate buffer + 1% BSA + 0.002% NaN 3 ) at a final concentration of 0.2% and stored at 4 • C till usage. Before use in the on-chip assay, the carbon particles were sonicated for 10 s.

Chip Processing
Surface blocking was done for 45 min in 1× PBS (pH 7.2)/0.1% Tween 20 in order to remove any unbound probes and to deactivate reactive surface groups. Slides were washed twice in 1× PBS (pH 7.2) and dried using compressed air. Slides were mounted into a Hybridization Cassette 4 × 16 (ArrayIt Corporation, Sunnyvale, CA, USA). 12 arrays/wells (7 mm × 7 mm) per slide were used. Each array was incubated with 50 µL calibration standard for 2.5 h. Spiked samples were prepared by serial dilution of mixes of analytes in assay buffer or synthetic urine (certified drug free urine (88121-CDF(F)) UTAK Laboratories Inc., Santa Clara, CA, USA) (1:3) (1 part urine, 2 parts assay buffer). Then the slides were washed three times with 1× PBS (pH 7.2)/0.1% Tween 20.

Detection Antibody Labeled with Carbon-and Gold NPs
Slides were incubated for 45 min with 50 µL oCNPs labeled dAB (1:5 diluted in assay buffer) and gold NPs labeled dAB (1:5 diluted in buffer: 5% BSA, 0.15 M NaCl, 0.01 M sodium phosphate (pH 7) and 0.05% Tween). Finally, the slides (treated with oCNPs) were washed twice with 1× PBS (pH 7.2)/0.1% Tween 20, twice in 1× PBS (pH 7.2) and once in distilled water. Slides incubated with gold-labeled dAB were in addition incubated with 0.05 M EDTA for 5 min before silver staining (10 min, Section 2.8). All incubation steps were carried out on the orbital shaker (Stovall, Greensboro, NC, USA) at maximum speed and at room temperature. After washing the slides were dried using compressed air and stored in the dark until scanning.
Slides incubated with strp-gold were washed with 1× PBS (pH 7.2)/0.1% Tween 20, 1× PBS (pH 7.2), distilled water and incubated with 0.05 M EDTA for 5 min before silver staining (10 min, Section 2.8). The slides incubated with strp-HRP were also washed as before and treated with substrate (TMB or ODI) for 15 min. At the end all slides were washed twice with 1× PBS (pH 7.2)/0.1% Tween 20, twice in 1× PBS (pH 7.2) and in distilled water. All incubation steps were carried out on the orbital shaker (Stovall) at maximum speed at room temperature. After washing the slides were dried using compressed air and stored in the dark until scanning.

Silver Staining
For the silver enhancement the excess solution was wiped off, 50 µL/well silver solution were added and incubated for 10 min. Finally, the slides were fixed with 2.5% sodium thiosulfate solution for 3 min and washed with dH 2 O. The slides were dried using compressed air.

Data Analysis
The colorimetric signal intensity was measured using an office scanner (Epson Perfection V330 Photo). Data were analyzed with the Genepix 6.0 software. The mean signal values were calculated from nine background corrected data points. Data points that were out of the mean signal values ± the standard deviation (SD) were excluded in an outlier test. Calibration curves were set up with OriginPro 8 using the logistic fit. The limit of detection (LOD) and dynamic range of the comparative IL8 assays were determined visually.

Colorimetric Assay Schemes
Six colorimetric immunoassay schemes as described in Figure 1a were investigated in terms of sensitivity, reproducibility, signal intensity, and assay time to finally select the most sensitive one for the development of the multiplexed biomarker chip. Nitrocellulose slides were used as the chip platform. For cost reasons, we chose IL8 for evaluation of the proposed assay schemes. IL8 capture antibodies (cAB Slides were treated with 50 µL biotinylated dAB and further incubated with either neutravidincarbon (1:20 dilution in assay buffer), strp-HRP (1:100 in assay buffer) or strp-gold (1:5 in buffer: 5% BSA, 0.15 M NaCl, 0.01 M sodium phosphate dibasic (pH 7.0) and 0.05% Tween 20) for 45 min.
Slides incubated with strp-gold were washed with 1× PBS (pH 7.2)/0.1% Tween 20, 1× PBS (pH 7.2), distilled water and incubated with 0.05 M EDTA for 5 min before silver staining (10 min, Section 2.8). The slides incubated with strp-HRP were also washed as before and treated with substrate (TMB or ODI) for 15 min. At the end all slides were washed twice with 1× PBS (pH 7.2)/0.1% Tween 20, twice in 1× PBS (pH 7.2) and in distilled water. All incubation steps were carried out on the orbital shaker (Stovall) at maximum speed at room temperature. After washing the slides were dried using compressed air and stored in the dark until scanning.

Silver Staining
For the silver enhancement the excess solution was wiped off, 50 µL/well silver solution were added and incubated for 10 min. Finally, the slides were fixed with 2.5% sodium thiosulfate solution for 3 min and washed with dH2O. The slides were dried using compressed air.

Data Analysis
The colorimetric signal intensity was measured using an office scanner (Epson Perfection V330 Photo). Data were analyzed with the Genepix 6.0 software. The mean signal values were calculated from nine background corrected data points. Data points that were out of the mean signal values ± the standard deviation (SD) were excluded in an outlier test. Calibration curves were set up with OriginPro 8 using the logistic fit. The limit of detection (LOD) and dynamic range of the comparative IL8 assays were determined visually.

Colorimetric Assay Schemes
Six colorimetric immunoassay schemes as described in Figure 1a were investigated in terms of sensitivity, reproducibility, signal intensity, and assay time to finally select the most sensitive one for the development of the multiplexed biomarker chip. Nitrocellulose slides were used as the chip platform. For cost reasons, we chose IL8 for evaluation of the proposed assay schemes. IL8 capture antibodies (cAB ) were spotted onto the chip to bind IL8 ( ) present in the sample solution (Step 2). For detection of the binding, the chip was, in a 3rd step, incubated with a detection antibody that was either labeled with NPs (I-III) or biotinylated (IV-VI ). While incubation with an antibody labeled with (I) gold (AuNPs ), (II) carbon (CNPs ), or (III) oxidized carbon nanoparticles (oCNPs ) allows direct detection of the binding, the use of a biotinylated antibody requires additional incubation steps for signal detection. In Format (IV), Step 4, the biotinylated detection antibody reacts with neutravidin adsorbed on carbon nanoparticles ( ); in Format (V), streptavidin is coupled with Au ( ); in Format (VI), streptavidin is linked to HRP ( ). Additionally, Formats (I) and (V) are treated with silver (Ag), while Format (VI) is further processed with different enzyme substrates (Step 5).
AuNPs were modified using 16-mercaptohexadecanoic acid (MHA) (I), and the antibody was bound to the particle via EDC, a zero length crosslinker that covalently couples the carboxylic group of MHA to the amino groups of the antibody. CNPs were either used untreated for antibody immobilization (II) or oxidized (oCNPs) with nitric acid to introduce reactive quinone or conjugated ketone groups [17] for antibody binding (III). The neutravidin/streptavidin-biotin bridge was exploited in Formats (IV) and (V), so neutravidin and streptavidin, respectively, were adsorbed onto the NPs.
Slides incubated with strp-gold were washed with 1× PBS (pH 7.2)/0.1% Tween 20, 1× PBS (pH 7.2), distilled water and incubated with 0.05 M EDTA for 5 min before silver staining (10 min, Section 2.8). The slides incubated with strp-HRP were also washed as before and treated with substrate (TMB or ODI) for 15 min. At the end all slides were washed twice with 1× PBS (pH 7.2)/0.1% Tween 20, twice in 1× PBS (pH 7.2) and in distilled water. All incubation steps were carried out on the orbital shaker (Stovall) at maximum speed at room temperature. After washing the slides were dried using compressed air and stored in the dark until scanning.

Silver Staining
For the silver enhancement the excess solution was wiped off, 50 µL/well silver solution were added and incubated for 10 min. Finally, the slides were fixed with 2.5% sodium thiosulfate solution for 3 min and washed with dH2O. The slides were dried using compressed air.

Data Analysis
The colorimetric signal intensity was measured using an office scanner (Epson Perfection V330 Photo). Data were analyzed with the Genepix 6.0 software. The mean signal values were calculated from nine background corrected data points. Data points that were out of the mean signal values ± the standard deviation (SD) were excluded in an outlier test. Calibration curves were set up with OriginPro 8 using the logistic fit. The limit of detection (LOD) and dynamic range of the comparative IL8 assays were determined visually.

Colorimetric Assay Schemes
Six colorimetric immunoassay schemes as described in Figure 1a were investigated in terms of sensitivity, reproducibility, signal intensity, and assay time to finally select the most sensitive one for the development of the multiplexed biomarker chip. Nitrocellulose slides were used as the chip platform. For cost reasons, we chose IL8 for evaluation of the proposed assay schemes. IL8 capture antibodies (cAB ) were spotted onto the chip to bind IL8 ( ) present in the sample solution (Step 2). For detection of the binding, the chip was, in a 3rd step, incubated with a detection antibody that was either labeled with NPs (I-III) or biotinylated (IV-VI ). While incubation with an antibody labeled with (I) gold (AuNPs ), (II) carbon (CNPs ), or (III) oxidized carbon nanoparticles (oCNPs ) allows direct detection of the binding, the use of a biotinylated antibody requires additional incubation steps for signal detection. In Format (IV), Step 4, the biotinylated detection antibody reacts with neutravidin adsorbed on carbon nanoparticles ( ); in Format (V), streptavidin is coupled with Au ( ); in Format (VI), streptavidin is linked to HRP ( ). Additionally, Formats (I) and (V) are treated with silver (Ag), while Format (VI) is further processed with different enzyme substrates (Step 5).
AuNPs were modified using 16-mercaptohexadecanoic acid (MHA) (I), and the antibody was bound to the particle via EDC, a zero length crosslinker that covalently couples the carboxylic group of MHA to the amino groups of the antibody. CNPs were either used untreated for antibody immobilization (II) or oxidized (oCNPs) with nitric acid to introduce reactive quinone or conjugated ketone groups [17] for antibody binding (III). The neutravidin/streptavidin-biotin bridge was exploited in Formats (IV) and (V), so neutravidin and streptavidin, respectively, were adsorbed onto the NPs.
) present in the sample solution (Step 2). For detection of the binding, the chip was, in a 3rd step, incubated with a detection antibody that was either labeled with NPs (I-III) or biotinylated (IV-VI Slides were treated with 50 µL biotinylated dAB and further incubated with either neutravidincarbon (1:20 dilution in assay buffer), strp-HRP (1:100 in assay buffer) or strp-gold (1:5 in buffer: 5% BSA, 0.15 M NaCl, 0.01 M sodium phosphate dibasic (pH 7.0) and 0.05% Tween 20) for 45 min.
Slides incubated with strp-gold were washed with 1× PBS (pH 7.2)/0.1% Tween 20, 1× PBS (pH 7.2), distilled water and incubated with 0.05 M EDTA for 5 min before silver staining (10 min, Section 2.8). The slides incubated with strp-HRP were also washed as before and treated with substrate (TMB or ODI) for 15 min. At the end all slides were washed twice with 1× PBS (pH 7.2)/0.1% Tween 20, twice in 1× PBS (pH 7.2) and in distilled water. All incubation steps were carried out on the orbital shaker (Stovall) at maximum speed at room temperature. After washing the slides were dried using compressed air and stored in the dark until scanning.

Silver Staining
For the silver enhancement the excess solution was wiped off, 50 µL/well silver solution were added and incubated for 10 min. Finally, the slides were fixed with 2.5% sodium thiosulfate solution for 3 min and washed with dH2O. The slides were dried using compressed air.

Data Analysis
The colorimetric signal intensity was measured using an office scanner (Epson Perfection V330 Photo). Data were analyzed with the Genepix 6.0 software. The mean signal values were calculated from nine background corrected data points. Data points that were out of the mean signal values ± the standard deviation (SD) were excluded in an outlier test. Calibration curves were set up with OriginPro 8 using the logistic fit. The limit of detection (LOD) and dynamic range of the comparative IL8 assays were determined visually.

Colorimetric Assay Schemes
Six colorimetric immunoassay schemes as described in Figure 1a were investigated in terms of sensitivity, reproducibility, signal intensity, and assay time to finally select the most sensitive one for the development of the multiplexed biomarker chip. Nitrocellulose slides were used as the chip platform. For cost reasons, we chose IL8 for evaluation of the proposed assay schemes. IL8 capture antibodies (cAB ) were spotted onto the chip to bind IL8 ( ) present in the sample solution (Step 2). For detection of the binding, the chip was, in a 3rd step, incubated with a detection antibody that was either labeled with NPs (I-III) or biotinylated (IV-VI ). While incubation with an antibody labeled with (I) gold (AuNPs ), (II) carbon (CNPs ), or (III) oxidized carbon nanoparticles (oCNPs ) allows direct detection of the binding, the use of a biotinylated antibody requires additional incubation steps for signal detection. In Format (IV), Step 4, the biotinylated detection antibody reacts with neutravidin adsorbed on carbon nanoparticles ( ); in Format (V), streptavidin is coupled with Au ( ); in Format (VI), streptavidin is linked to HRP ( ). Additionally, Formats (I) and (V) are treated with silver (Ag), while Format (VI) is further processed with different enzyme substrates (Step 5).
AuNPs were modified using 16-mercaptohexadecanoic acid (MHA) (I), and the antibody was bound to the particle via EDC, a zero length crosslinker that covalently couples the carboxylic group of MHA to the amino groups of the antibody. CNPs were either used untreated for antibody immobilization (II) or oxidized (oCNPs) with nitric acid to introduce reactive quinone or conjugated ketone groups [17] for antibody binding (III). The neutravidin/streptavidin-biotin bridge was exploited in Formats (IV) and (V), so neutravidin and streptavidin, respectively, were adsorbed onto the NPs.
Slides incubated with strp-gold were washed with 1× PBS (pH 7.2)/0.1% Tween 20, 1× PBS (pH , distilled water and incubated with 0.05 M EDTA for 5 min before silver staining (10 min, Section . The slides incubated with strp-HRP were also washed as before and treated with substrate (TMB DI) for 15 min. At the end all slides were washed twice with 1× PBS (pH 7.2)/0.1% Tween 20, ce in 1× PBS (pH 7.2) and in distilled water. All incubation steps were carried out on the orbital ker (Stovall) at maximum speed at room temperature. After washing the slides were dried using pressed air and stored in the dark until scanning.

Silver Staining
For the silver enhancement the excess solution was wiped off, 50 µL/well silver solution were ed and incubated for 10 min. Finally, the slides were fixed with 2.5% sodium thiosulfate solution 3 min and washed with dH2O. The slides were dried using compressed air.

Data Analysis
The colorimetric signal intensity was measured using an office scanner (Epson Perfection V330 to). Data were analyzed with the Genepix 6.0 software. The mean signal values were calculated nine background corrected data points. Data points that were out of the mean signal values ± standard deviation (SD) were excluded in an outlier test. Calibration curves were set up with ginPro 8 using the logistic fit. The limit of detection (LOD) and dynamic range of the comparative assays were determined visually.

Colorimetric Assay Schemes
Six colorimetric immunoassay schemes as described in Figure 1a were investigated in terms of sitivity, reproducibility, signal intensity, and assay time to finally select the most sensitive one for development of the multiplexed biomarker chip. Nitrocellulose slides were used as the chip form. For cost reasons, we chose IL8 for evaluation of the proposed assay schemes. IL8 capture ibodies (cAB ) were spotted onto the chip to bind IL8 ( ) present in the sample solution p 2). For detection of the binding, the chip was, in a 3rd step, incubated with a detection antibody t was either labeled with NPs (I-III) or biotinylated (IV-VI ). While incubation with an ibody labeled with (I) gold (AuNPs ), (II) carbon (CNPs ), or (III) oxidized carbon oparticles (oCNPs ) allows direct detection of the binding, the use of a biotinylated antibody uires additional incubation steps for signal detection. In Format (IV), Step 4, the biotinylated ection antibody reacts with neutravidin adsorbed on carbon nanoparticles ( ); in Format (V), ptavidin is coupled with Au ( ); in Format (VI), streptavidin is linked to HRP ( ). itionally, Formats (I) and (V) are treated with silver (Ag), while Format (VI) is further processed h different enzyme substrates (Step 5).
AuNPs were modified using 16-mercaptohexadecanoic acid (MHA) (I), and the antibody was nd to the particle via EDC, a zero length crosslinker that covalently couples the carboxylic group HA to the amino groups of the antibody. CNPs were either used untreated for antibody obilization (II) or oxidized (oCNPs) with nitric acid to introduce reactive quinone or conjugated ne groups [17] for antibody binding (III). The neutravidin/streptavidin-biotin bridge was loited in Formats (IV) and (V), so neutravidin and streptavidin, respectively, were adsorbed onto NPs.
), (II) carbon (CNPs ides were treated with 50 µL biotinylated dAB and further incubated with either neutravidin-(1:20 dilution in assay buffer), strp-HRP (1:100 in assay buffer) or strp-gold (1:5 in buffer: A, 0.15 M NaCl, 0.01 M sodium phosphate dibasic (pH 7.0) and 0.05% Tween 20) for 45 min. ides incubated with strp-gold were washed with 1× PBS (pH 7.2)/0.1% Tween 20, 1× PBS (pH stilled water and incubated with 0.05 M EDTA for 5 min before silver staining (10 min, Section e slides incubated with strp-HRP were also washed as before and treated with substrate (TMB I) for 15 min. At the end all slides were washed twice with 1× PBS (pH 7.2)/0.1% Tween 20, n 1× PBS (pH 7.2) and in distilled water. All incubation steps were carried out on the orbital (Stovall) at maximum speed at room temperature. After washing the slides were dried using essed air and stored in the dark until scanning.
ver Staining r the silver enhancement the excess solution was wiped off, 50 µL/well silver solution were and incubated for 10 min. Finally, the slides were fixed with 2.5% sodium thiosulfate solution in and washed with dH2O. The slides were dried using compressed air.
ta Analysis he colorimetric signal intensity was measured using an office scanner (Epson Perfection V330 . Data were analyzed with the Genepix 6.0 software. The mean signal values were calculated ine background corrected data points. Data points that were out of the mean signal values ± ndard deviation (SD) were excluded in an outlier test. Calibration curves were set up with Pro 8 using the logistic fit. The limit of detection (LOD) and dynamic range of the comparative ays were determined visually.

lts and Discussion
lorimetric Assay Schemes x colorimetric immunoassay schemes as described in Figure 1a were investigated in terms of vity, reproducibility, signal intensity, and assay time to finally select the most sensitive one for velopment of the multiplexed biomarker chip. Nitrocellulose slides were used as the chip m. For cost reasons, we chose IL8 for evaluation of the proposed assay schemes. IL8 capture dies (cAB ) were spotted onto the chip to bind IL8 ( ) present in the sample solution ). For detection of the binding, the chip was, in a 3rd step, incubated with a detection antibody as either labeled with NPs (I-III) or biotinylated (IV-VI ). While incubation with an dy labeled with (I) gold (AuNPs ), (II) carbon (CNPs ), or (III) oxidized carbon articles (oCNPs ) allows direct detection of the binding, the use of a biotinylated antibody s additional incubation steps for signal detection. In Format (IV), Step 4, the biotinylated on antibody reacts with neutravidin adsorbed on carbon nanoparticles ( ); in Format (V), vidin is coupled with Au ( ); in Format (VI), streptavidin is linked to HRP ( ). onally, Formats (I) and (V) are treated with silver (Ag), while Format (VI) is further processed ifferent enzyme substrates (Step 5). uNPs were modified using 16-mercaptohexadecanoic acid (MHA) (I), and the antibody was to the particle via EDC, a zero length crosslinker that covalently couples the carboxylic group A to the amino groups of the antibody. CNPs were either used untreated for antibody ilization (II) or oxidized (oCNPs) with nitric acid to introduce reactive quinone or conjugated groups [17] for antibody binding (III). The neutravidin/streptavidin-biotin bridge was ted in Formats (IV) and (V), so neutravidin and streptavidin, respectively, were adsorbed onto s.
), or (III) oxidized carbon nanoparticles (oCNPs carbon (1:20 dilution in assay buffer), strp 5% BSA, 0.15 M NaCl, 0.01 M sodium phos Slides incubated with strp-gold were w 7.2), distilled water and incubated with 0.05 2.8). The slides incubated with strp-HRP we or ODI) for 15 min. At the end all slides w twice in 1× PBS (pH 7.2) and in distilled w shaker (Stovall) at maximum speed at room compressed air and stored in the dark until

Silver Staining
For the silver enhancement the excess added and incubated for 10 min. Finally, th for 3 min and washed with dH2O. The slide

Data Analysis
The colorimetric signal intensity was m Photo). Data were analyzed with the Gene from nine background corrected data poin the standard deviation (SD) were excluded OriginPro 8 using the logistic fit. The limit o IL8 assays were determined visually.

Colorimetric Assay Schemes
Six colorimetric immunoassay scheme sensitivity, reproducibility, signal intensity the development of the multiplexed biom platform. For cost reasons, we chose IL8 fo antibodies (cAB ) were spotted onto th (Step 2). For detection of the binding, the ch that was either labeled with NPs (I-III) antibody labeled with (I) gold (AuNPs nanoparticles (oCNPs ) allows direct de requires additional incubation steps for si detection antibody reacts with neutravidin streptavidin is coupled with Au ( ); Additionally, Formats (I) and (V) are treate with different enzyme substrates (Step 5).
AuNPs were modified using 16-merca bound to the particle via EDC, a zero length of MHA to the amino groups of the ant immobilization (II) or oxidized (oCNPs) wi ketone groups [17] for antibody binding exploited in Formats (IV) and (V), so neutra the NPs. des incubated with strp-gold were washed with 1× PBS (pH 7.2)/0.1% Tween 20, 1× PBS (pH tilled water and incubated with 0.05 M EDTA for 5 min before silver staining (10 min, Section e slides incubated with strp-HRP were also washed as before and treated with substrate (TMB ) for 15 min. At the end all slides were washed twice with 1× PBS (pH 7.2)/0.1% Tween 20, 1× PBS (pH 7.2) and in distilled water. All incubation steps were carried out on the orbital (Stovall) at maximum speed at room temperature. After washing the slides were dried using ssed air and stored in the dark until scanning.
er Staining r the silver enhancement the excess solution was wiped off, 50 µL/well silver solution were nd incubated for 10 min. Finally, the slides were fixed with 2.5% sodium thiosulfate solution in and washed with dH2O. The slides were dried using compressed air.
a Analysis e colorimetric signal intensity was measured using an office scanner (Epson Perfection V330 Data were analyzed with the Genepix 6.0 software. The mean signal values were calculated ne background corrected data points. Data points that were out of the mean signal values ± dard deviation (SD) were excluded in an outlier test. Calibration curves were set up with ro 8 using the logistic fit. The limit of detection (LOD) and dynamic range of the comparative ys were determined visually.

lts and Discussion
orimetric Assay Schemes colorimetric immunoassay schemes as described in Figure 1a were investigated in terms of ity, reproducibility, signal intensity, and assay time to finally select the most sensitive one for elopment of the multiplexed biomarker chip. Nitrocellulose slides were used as the chip . For cost reasons, we chose IL8 for evaluation of the proposed assay schemes. IL8 capture ies (cAB ) were spotted onto the chip to bind IL8 ( ) present in the sample solution . For detection of the binding, the chip was, in a 3rd step, incubated with a detection antibody s either labeled with NPs (I-III) or biotinylated (IV-VI ). While incubation with an y labeled with (I) gold (AuNPs ), (II) carbon (CNPs ), or (III) oxidized carbon rticles (oCNPs ) allows direct detection of the binding, the use of a biotinylated antibody s additional incubation steps for signal detection. In Format (IV), Step 4, the biotinylated n antibody reacts with neutravidin adsorbed on carbon nanoparticles ( ); in Format (V), idin is coupled with Au ( ); in Format (VI), streptavidin is linked to HRP ( ). nally, Formats (I) and (V) are treated with silver (Ag), while Format (VI) is further processed fferent enzyme substrates (Step 5). NPs were modified using 16-mercaptohexadecanoic acid (MHA) (I), and the antibody was to the particle via EDC, a zero length crosslinker that covalently couples the carboxylic group A to the amino groups of the antibody. CNPs were either used untreated for antibody ilization (II) or oxidized (oCNPs) with nitric acid to introduce reactive quinone or conjugated groups [17] for antibody binding (III). The neutravidin/streptavidin-biotin bridge was d in Formats (IV) and (V), so neutravidin and streptavidin, respectively, were adsorbed onto .
); in Format (V), streptavidin is coupled with Au ( ated with 50 µL biotinylated dAB and further incubated with either neutravidinon in assay buffer), strp-HRP (1:100 in assay buffer) or strp-gold (1:5 in buffer: Cl, 0.01 M sodium phosphate dibasic (pH 7.0) and 0.05% Tween 20) for 45 min. ed with strp-gold were washed with 1× PBS (pH 7.2)/0.1% Tween 20, 1× PBS (pH and incubated with 0.05 M EDTA for 5 min before silver staining (10 min, Section bated with strp-HRP were also washed as before and treated with substrate (TMB . At the end all slides were washed twice with 1× PBS (pH 7.2)/0.1% Tween 20, 7.2) and in distilled water. All incubation steps were carried out on the orbital maximum speed at room temperature. After washing the slides were dried using stored in the dark until scanning.
enhancement the excess solution was wiped off, 50 µL/well silver solution were ed for 10 min. Finally, the slides were fixed with 2.5% sodium thiosulfate solution ed with dH2O. The slides were dried using compressed air.
ric signal intensity was measured using an office scanner (Epson Perfection V330 analyzed with the Genepix 6.0 software. The mean signal values were calculated und corrected data points. Data points that were out of the mean signal values ± tion (SD) were excluded in an outlier test. Calibration curves were set up with he logistic fit. The limit of detection (LOD) and dynamic range of the comparative termined visually.
cussion say Schemes ic immunoassay schemes as described in Figure 1a were investigated in terms of cibility, signal intensity, and assay time to finally select the most sensitive one for f the multiplexed biomarker chip. Nitrocellulose slides were used as the chip reasons, we chose IL8 for evaluation of the proposed assay schemes. IL8 capture ) were spotted onto the chip to bind IL8 ( ) present in the sample solution ion of the binding, the chip was, in a 3rd step, incubated with a detection antibody beled with NPs (I-III) or biotinylated (IV-VI ). While incubation with an with (I) gold (AuNPs ), (II) carbon (CNPs ), or (III) oxidized carbon Ps ) allows direct detection of the binding, the use of a biotinylated antibody l incubation steps for signal detection. In Format (IV), Step 4, the biotinylated reacts with neutravidin adsorbed on carbon nanoparticles ( ); in Format (V), pled with Au ( ); in Format (VI), streptavidin is linked to HRP ( ). ats (I) and (V) are treated with silver (Ag), while Format (VI) is further processed me substrates (Step 5). modified using 16-mercaptohexadecanoic acid (MHA) (I), and the antibody was le via EDC, a zero length crosslinker that covalently couples the carboxylic group ino groups of the antibody. CNPs were either used untreated for antibody or oxidized (oCNPs) with nitric acid to introduce reactive quinone or conjugated ] for antibody binding (III). The neutravidin/streptavidin-biotin bridge was ts (IV) and (V), so neutravidin and streptavidin, respectively, were adsorbed onto ); in Format (VI), streptavidin is linked to HRP ( ith 50 µL biotinylated dAB and further incubated with either neutravidinssay buffer), strp-HRP (1:100 in assay buffer) or strp-gold (1:5 in buffer: M sodium phosphate dibasic (pH 7.0) and 0.05% Tween 20) for 45 min. strp-gold were washed with 1× PBS (pH 7.2)/0.1% Tween 20, 1× PBS (pH cubated with 0.05 M EDTA for 5 min before silver staining (10 min, Section ith strp-HRP were also washed as before and treated with substrate (TMB e end all slides were washed twice with 1× PBS (pH 7.2)/0.1% Tween 20, nd in distilled water. All incubation steps were carried out on the orbital um speed at room temperature. After washing the slides were dried using in the dark until scanning.
ement the excess solution was wiped off, 50 µL/well silver solution were 0 min. Finally, the slides were fixed with 2.5% sodium thiosulfate solution h dH2O. The slides were dried using compressed air.
al intensity was measured using an office scanner (Epson Perfection V330 d with the Genepix 6.0 software. The mean signal values were calculated rected data points. Data points that were out of the mean signal values ± D) were excluded in an outlier test. Calibration curves were set up with stic fit. The limit of detection (LOD) and dynamic range of the comparative d visually.
mes unoassay schemes as described in Figure 1a were investigated in terms of , signal intensity, and assay time to finally select the most sensitive one for ultiplexed biomarker chip. Nitrocellulose slides were used as the chip , we chose IL8 for evaluation of the proposed assay schemes. IL8 capture re spotted onto the chip to bind IL8 ( ) present in the sample solution e binding, the chip was, in a 3rd step, incubated with a detection antibody ith NPs (I-III) or biotinylated (IV-VI ). While incubation with an ) gold (AuNPs ), (II) carbon (CNPs ), or (III) oxidized carbon ) allows direct detection of the binding, the use of a biotinylated antibody ation steps for signal detection. In Format (IV), Step 4, the biotinylated with neutravidin adsorbed on carbon nanoparticles ( ); in Format (V), ith Au ( ); in Format (VI), streptavidin is linked to HRP ( ). nd (V) are treated with silver (Ag), while Format (VI) is further processed strates (Step 5). d using 16-mercaptohexadecanoic acid (MHA) (I), and the antibody was DC, a zero length crosslinker that covalently couples the carboxylic group roups of the antibody. CNPs were either used untreated for antibody ized (oCNPs) with nitric acid to introduce reactive quinone or conjugated ntibody binding (III). The neutravidin/streptavidin-biotin bridge was and (V), so neutravidin and streptavidin, respectively, were adsorbed onto ). Additionally, Formats (I) and (V) are treated with silver (Ag), while Format (VI) is further processed with different enzyme substrates (Step 5).
AuNPs were modified using 16-mercaptohexadecanoic acid (MHA) (I), and the antibody was bound to the particle via EDC, a zero length crosslinker that covalently couples the carboxylic group of MHA to the amino groups of the antibody. CNPs were either used untreated for antibody immobilization (II) or oxidized (oCNPs) with nitric acid to introduce reactive quinone or conjugated ketone groups [17] for antibody binding (III). The neutravidin/streptavidin-biotin bridge was exploited in Formats (IV) and (V), so neutravidin and streptavidin, respectively, were adsorbed onto the NPs.

Assay Schemes with Nanoparticle-Labeled Detection Antibody
A comparative study was conducted using anti-IL8 labeled with Au and carbon NPs (Figure (I)-(III)) to directly detect IL8 in the concentration range of 1 ng/L to 98.4 µg/L.
Commercially available NPs were utilized to guarantee a high percentage of regularly shap NPs preventing diffuse signals and generating well-confined and consistent spots [13]. Nitrocellulo with a 150 µm nitrosylated polysaccharide layer and a 0.20 µm porosity was applied. In order to wo with colloidal stable [19] and sensitive AuNPs, 40-nm-sized particles were chosen. This is in go agreement with Nath et al. [20], who investigated the effect of particle size on the analytical sensitiv of an optical biosensor for streptavidin-biotin as a model receptor-ligand pair. The authors show that the sensitivity of the sensor increases by a factor of three as size of Au NPs increases in the test range of 13-47 nm.
AuNPs have been coupled with anti-IL8 using MHA and carbodiimide [21]. MHA se assembles on the Au particle forming a monolayer. Its carboxy group is subsequently activated w EDC to covalently bind IL8 through the amino groups present in the amino acids (arginine a lysine). Alternatively, the antibody might only be mixed with the AuNPs and adsorb on the partic via its thiol functions (cysteine). This was realized by Wang et al. [22], who coupled 40 nm A particles to an anti-aflatoxin M1 antibody via adsorption.
In our study, the adsorption of antibodies on plain Au nanoparticles was not successful and t coupling antibody to Au via MHA led to mixed results with sometimes poor signals (Detecti Scheme I) as can be seen from Figure 2. In the case of carbon NPs, brighter signals were obtained w oxidized carbon nanoparticles (oCNPs) (Detection Scheme III) than with carbon NPs that we implemented without further treatment (Detection Scheme II). This is most likely a result of high functionality and hence more antibody bound to oCNPs compared to plain carbon. In fact, FT measurements revealed in the oxidized particles reactive groups, such as quinone or conjugat ketone [17], that are capable of binding with the amino acids of the detection antibody. Furthermo elemental analysis data confirmed oxidation and showed increased oxygen values for oCNPs (9 compared with CNPs (5%). However, in contrast to AuNPs both types of carbon NPs lead to visib spots, proving that Detection Scheme scheme III is more sensitive: the detection limit was 15 ng Figure 1. (a) Schematic presentation of six distinct assay formats for colorimetric detection: (I) cAB + IL8 + gold-labeled dAB, (II) cAB + IL8 + carbon-labeled dAB, (III) cAB + IL8 + oxidized carbon-labeled dAB, (IV) cAB + IL8 + biot. dAB + neutravidin-carbon, (V) cAB + IL8 + biot. dAB + strp-gold, and (VI) cAB + IL8 + biot. dAB + strp-HRP + enzyme substrate. Slides treated via gold nanoparticles were further stained by silver (Ag). (b) Illustration of the antibody (in orange) coupling process using gold (red) and carbon (black) nanoparticles.

Assay Schemes with Nanoparticle-Labeled Detection Antibody
A comparative study was conducted using anti-IL8 labeled with Au and carbon NPs (Figure 1, (I)-(III)) to directly detect IL8 in the concentration range of 1 ng/L to 98.4 µg/L.
Commercially available NPs were utilized to guarantee a high percentage of regularly shaped NPs preventing diffuse signals and generating well-confined and consistent spots [13]. Nitrocellulose with a 150 µm nitrosylated polysaccharide layer and a 0.20 µm porosity was applied. In order to work with colloidal stable [19] and sensitive AuNPs, 40-nm-sized particles were chosen. This is in good agreement with Nath et al. [20], who investigated the effect of particle size on the analytical sensitivity of an optical biosensor for streptavidin-biotin as a model receptor-ligand pair. The authors showed that the sensitivity of the sensor increases by a factor of three as size of Au NPs increases in the tested range of 13-47 nm.
AuNPs have been coupled with anti-IL8 using MHA and carbodiimide [21]. MHA self-assembles on the Au particle forming a monolayer. Its carboxy group is subsequently activated with EDC to covalently bind IL8 through the amino groups present in the amino acids (arginine and lysine). Alternatively, the antibody might only be mixed with the AuNPs and adsorb on the particles via its thiol functions (cysteine). This was realized by Wang et al. [22], who coupled 40 nm Au particles to an anti-aflatoxin M1 antibody via adsorption.
In our study, the adsorption of antibodies on plain Au nanoparticles was not successful and the coupling antibody to Au via MHA led to mixed results with sometimes poor signals (Detection Scheme I) as can be seen from Figure 2. In the case of carbon NPs, brighter signals were obtained with oxidized carbon nanoparticles (oCNPs) (Detection Scheme III) than with carbon NPs that were implemented without further treatment (Detection Scheme II). This is most likely a result of higher functionality and hence more antibody bound to oCNPs compared to plain carbon. In fact, FTIR measurements revealed in the oxidized particles reactive groups, such as quinone or conjugated ketone [17], that are capable of binding with the amino acids of the detection antibody. Furthermore, elemental analysis data confirmed oxidation and showed increased oxygen values for oCNPs (9%) compared with CNPs (5%). However, in contrast to AuNPs both types of carbon NPs lead to visible spots, proving that Detection Scheme scheme III is more sensitive: the detection limit was 15 ng/L and improved by a factor of more than 10 compared to Detection Scheme II. Interestingly, Posthuma-Trumpie et al. [13] stated that adsorption is commonly used for carbon NP conjugation because it retains the specificity of proteins, important for antibody-antigen binding, and displays a fast and easy labeling procedure. This might be true for special types of carbon black but can by no way be generalized as there are plenty of different grades of carbon black commercially available made by different manufacturing processes in different sizes and surface areas.
In this comparison, carbon NPs finally yield 10-fold higher assay sensitivity compared to AuNPs labels. The fact that carbon NPs used as labels can be more sensitive than gold NPs was already reported by Gordon et al. [23], who have undertaken a comprehensive survey of literature in PubMed. Thereby, sensitivities in the low pico-molar range were observed for carbon NPs, even by visual inspection. However, we have demonstrated that this is only the case when carbon NPs have been previously treated by nitric acid (oCNPs, Format (III)). and improved by a factor of more than 10 compared to Detection Scheme II. Interestingly, Posthuma-Trumpie et al. [13] stated that adsorption is commonly used for carbon NP conjugation because it retains the specificity of proteins, important for antibody-antigen binding, and displays a fast and easy labeling procedure. This might be true for special types of carbon black but can by no way be generalized as there are plenty of different grades of carbon black commercially available made by different manufacturing processes in different sizes and surface areas.
In this comparison, carbon NPs finally yield 10-fold higher assay sensitivity compared to AuNPs labels. The fact that carbon NPs used as labels can be more sensitive than gold NPs was already reported by Gordon et al. [23], who have undertaken a comprehensive survey of literature in PubMed. Thereby, sensitivities in the low pico-molar range were observed for carbon NPs, even by visual inspection. However, we have demonstrated that this is only the case when carbon NPs have been previously treated by nitric acid (oCNPs, Format (III)).

Evaluation of Enzyme Substrate
Three enzyme substrates either based on TMB (from two different providers) or ODI were evaluated using Assay Format (VI). As shown in Figure 3, significant differences in assay performance were observed. As reported by the manufacturer ODI appears as a green precipitate

Evaluation of Enzyme Substrate
Three enzyme substrates either based on TMB (from two different providers) or ODI were evaluated using Assay Format (VI). As shown in Figure 3, significant differences in assay performance were observed. As reported by the manufacturer ODI appears as a green precipitate after reaction with HRP and normally shows strong adherence on plastic surfaces. TMB precipitates as a dark blue color and is especially suitable for membrane surfaces, such as nitrocellulose. Using the two different TMB substrates, the sensitivities, as defined by the LOD, were different by a factor of 15. The lowest detectable IL8 concentration was 15 ng/L with TMB1 and 405 ng/L with TMB2. This suggests that there are different qualities of enzyme substrate available in the market which all induce different reaction rates depending on the TMB and H2O2 concentration, pH and buffer system. Further influencing factors might be the solubility product of the precipitating agent with the TMB cation and the adhesion of the precipitate on the chip membrane.
Data reproducibility as reflected by the small standard deviations was good with all tested substrates which points to a homogeneous staining pattern and very even distribution of precipitate within the spots. The enzyme substrate ODI was not suitable for this kind of assay, as no calibration curve was obtained. Our experience is actually in good agreement with Josephy et al. [24] who reported poor signal-to-noise ratio for colorimetric assays using ODI as the enzyme substrate arguing that one or more of the dianisidine products formed are unstable. In addition, the authors observed that ODI is short-lived and disappears within a few minutes [24]. This was not the case in our experiments.
In conclusion, TMB1 performed best among the tested enzyme substrates and was therefore used as substrate for HRP in all further investigations.

Detection Schemes Using the Streptavidin-Biotin Bridge
For detection of IL8, samples spiked with IL8 were incubated with the chip and the binding was detected with biotinylated antibodies and made visible in an additional step using Formats (IV)-(VI). To enhance the colorimetric signal of the gold NPs, the Au was stained with silver.
The enzymatic Format (VI) represents by far the most sensitive approach, with an at least 27fold lower LOD compared to Assay Formats (IV) and (V). The good assay performance of the enzymatic approach (VI) is demonstrated in Figure 2, while spots visible with the naked eye are generated at much higher concentrations using Detection Schemes IV and V.
Using Assay Format (V), bound streptavidin-Au was further enhanced with silver. Incubation time was 10 min, which led to 5-fold brighter signals compared to those without silver staining. As also reported by others, incubation time is critical: after 10 min, most silver was deposited; after 25 Using the two different TMB substrates, the sensitivities, as defined by the LOD, were different by a factor of 15. The lowest detectable IL8 concentration was 15 ng/L with TMB1 and 405 ng/L with TMB2. This suggests that there are different qualities of enzyme substrate available in the market which all induce different reaction rates depending on the TMB and H2O2 concentration, pH and buffer system. Further influencing factors might be the solubility product of the precipitating agent with the TMB cation and the adhesion of the precipitate on the chip membrane.
Data reproducibility as reflected by the small standard deviations was good with all tested substrates which points to a homogeneous staining pattern and very even distribution of precipitate within the spots. The enzyme substrate ODI was not suitable for this kind of assay, as no calibration curve was obtained. Our experience is actually in good agreement with Josephy et al. [24] who reported poor signal-to-noise ratio for colorimetric assays using ODI as the enzyme substrate arguing that one or more of the dianisidine products formed are unstable. In addition, the authors observed that ODI is short-lived and disappears within a few minutes [24]. This was not the case in our experiments.
In conclusion, TMB1 performed best among the tested enzyme substrates and was therefore used as substrate for HRP in all further investigations.

Detection Schemes Using the Streptavidin-Biotin Bridge
For detection of IL8, samples spiked with IL8 were incubated with the chip and the binding was detected with biotinylated antibodies and made visible in an additional step using Formats (IV)-(VI). To enhance the colorimetric signal of the gold NPs, the Au was stained with silver.
The enzymatic Format (VI) represents by far the most sensitive approach, with an at least 27fold lower LOD compared to Assay Formats (IV) and (V). The good assay performance of the enzymatic approach (VI) is demonstrated in Figure 2, while spots visible with the naked eye are generated at much higher concentrations using Detection Schemes IV and V.
Using Assay Format (V), bound streptavidin-Au was further enhanced with silver. Incubation time was 10 min, which led to 5-fold brighter signals compared to those without silver staining. As also reported by others, incubation time is critical: after 10 min, most silver was deposited; after 25 Using the two different TMB substrates, the sensitivities, as defined by the LOD, were different by a factor of 15. The lowest detectable IL8 concentration was 15 ng/L with TMB1 and 405 ng/L with TMB2. This suggests that there are different qualities of enzyme substrate available in the market which all induce different reaction rates depending on the TMB and H2O2 concentration, pH and buffer system. Further influencing factors might be the solubility product of the precipitating agent with the TMB cation and the adhesion of the precipitate on the chip membrane.
Data reproducibility as reflected by the small standard deviations was good with all tested substrates which points to a homogeneous staining pattern and very even distribution of precipitate within the spots. The enzyme substrate ODI was not suitable for this kind of assay, as no calibration curve was obtained. Our experience is actually in good agreement with Josephy et al. [24] who reported poor signal-to-noise ratio for colorimetric assays using ODI as the enzyme substrate arguing that one or more of the dianisidine products formed are unstable. In addition, the authors observed that ODI is short-lived and disappears within a few minutes [24]. This was not the case in our experiments.
In conclusion, TMB1 performed best among the tested enzyme substrates and was therefore used as substrate for HRP in all further investigations.

Detection Schemes Using the Streptavidin-Biotin Bridge
For detection of IL8, samples spiked with IL8 were incubated with the chip and the binding was detected with biotinylated antibodies and made visible in an additional step using Formats (IV)-(VI). To enhance the colorimetric signal of the gold NPs, the Au was stained with silver.
The enzymatic Format (VI) represents by far the most sensitive approach, with an at least 27fold lower LOD compared to Assay Formats (IV) and (V). The good assay performance of the enzymatic approach (VI) is demonstrated in Figure 2, while spots visible with the naked eye are generated at much higher concentrations using Detection Schemes IV and V.
Using Assay Format (V), bound streptavidin-Au was further enhanced with silver. Incubation time was 10 min, which led to 5-fold brighter signals compared to those without silver staining. As also reported by others, incubation time is critical: after 10 min, most silver was deposited; after 25 ), and ODI&#13; ( Using the two different TMB substrates, the sensitivities, as defined by the LOD, were different by a factor of 15. The lowest detectable IL8 concentration was 15 ng/L with TMB1 and 405 ng/L with TMB2. This suggests that there are different qualities of enzyme substrate available in the market which all induce different reaction rates depending on the TMB and H2O2 concentration, pH and buffer system. Further influencing factors might be the solubility product of the precipitating agent with the TMB cation and the adhesion of the precipitate on the chip membrane.
Data reproducibility as reflected by the small standard deviations was good with all tested substrates which points to a homogeneous staining pattern and very even distribution of precipitate within the spots. The enzyme substrate ODI was not suitable for this kind of assay, as no calibration curve was obtained. Our experience is actually in good agreement with Josephy et al. [24] who reported poor signal-to-noise ratio for colorimetric assays using ODI as the enzyme substrate arguing that one or more of the dianisidine products formed are unstable. In addition, the authors observed that ODI is short-lived and disappears within a few minutes [24]. This was not the case in our experiments.
In conclusion, TMB1 performed best among the tested enzyme substrates and was therefore used as substrate for HRP in all further investigations.

Detection Schemes Using the Streptavidin-Biotin Bridge
For detection of IL8, samples spiked with IL8 were incubated with the chip and the binding was detected with biotinylated antibodies and made visible in an additional step using Formats (IV)-(VI). To enhance the colorimetric signal of the gold NPs, the Au was stained with silver.
The enzymatic Format (VI) represents by far the most sensitive approach, with an at least 27fold lower LOD compared to Assay Formats (IV) and (V). The good assay performance of the enzymatic approach (VI) is demonstrated in Figure 2, while spots visible with the naked eye are generated at much higher concentrations using Detection Schemes IV and V.
Using Assay Format (V), bound streptavidin-Au was further enhanced with silver. Incubation time was 10 min, which led to 5-fold brighter signals compared to those without silver staining. As also reported by others, incubation time is critical: after 10 min, most silver was deposited; after 25 ). The&#13; blank (0 ng/L) is represented by 1 ng/L on the log scale.
Using the two different TMB substrates, the sensitivities, as defined by the LOD, were different by a factor of 15. The lowest detectable IL8 concentration was 15 ng/L with TMB1 and 405 ng/L with TMB2. This suggests that there are different qualities of enzyme substrate available in the market which all induce different reaction rates depending on the TMB and H 2 O 2 concentration, pH and buffer system. Further influencing factors might be the solubility product of the precipitating agent with the TMB cation and the adhesion of the precipitate on the chip membrane.
Data reproducibility as reflected by the small standard deviations was good with all tested substrates which points to a homogeneous staining pattern and very even distribution of precipitate within the spots. The enzyme substrate ODI was not suitable for this kind of assay, as no calibration curve was obtained. Our experience is actually in good agreement with Josephy et al. [24] who reported poor signal-to-noise ratio for colorimetric assays using ODI as the enzyme substrate arguing that one or more of the dianisidine products formed are unstable. In addition, the authors observed that ODI is short-lived and disappears within a few minutes [24]. This was not the case in our experiments.
In conclusion, TMB1 performed best among the tested enzyme substrates and was therefore used as substrate for HRP in all further investigations.

Detection Schemes Using the Streptavidin-Biotin Bridge
For detection of IL8, samples spiked with IL8 were incubated with the chip and the binding was detected with biotinylated antibodies and made visible in an additional step using Formats (IV)-(VI). To enhance the colorimetric signal of the gold NPs, the Au was stained with silver.
The enzymatic Format (VI) represents by far the most sensitive approach, with an at least 27-fold lower LOD compared to Assay Formats (IV) and (V). The good assay performance of the enzymatic approach (VI) is demonstrated in Figure 2, while spots visible with the naked eye are generated at much higher concentrations using Detection Schemes IV and V.
Using Assay Format (V), bound streptavidin-Au was further enhanced with silver. Incubation time was 10 min, which led to 5-fold brighter signals compared to those without silver staining. As also reported by others, incubation time is critical: after 10 min, most silver was deposited; after 25 min, the silver enhancement reaction was in saturation [25]. Optimum incubation times reported are in the range of 10-15 min [26,27].

Comparison of the Six Protein Array Assay Schemes
To identify the assay schemes that are both rapid and accurate, we compared the times required to fully complete the assay as well as the assay sensitivity, as summarized in Figures 2 and 4. Assay Schemes (II) and (III) require three steps in total (spotting of the probes, incubation with the sample, and incubation with the labeled detection antibody) and represent the schemes with the shortest assay time (4 h). They are certainly the biomarker assays of choice in time-critical medical situations. Clearly, the assay time was not optimized, especially the incubation times of Steps 2-4 might be further reduced. In addition, Format (III) leads to the lowest LOD being 15 ng/L IL8, as low as Assay Format (VI). From Figure 2, it is obvious that those assay schemes were the only ones that could detect concentrations lower than 1215 ng/L IL8. Nevertheless, concentration-dependent black spots were obtained for all tested assay schemes except for Assay Scheme (I) (the Au nanoparticle-labeled detection antibody).
Results indicate that the choice of detection label strongly influences the assay sensitivity and measurement range of the immunoassay. This adds to the observations made with different types of nitrocellulose platforms (different modifications and different porosities) [28] or generally different types of surface materials [29] and enzyme substrates. Choosing the appropriate label allows the assay sensitivity to be tuned within a range that best fits the detection requirements, for example, to monitor the biomarker over several stages of the disease. In general, no clear trend towards the directly labeled detection antibody or the biotinylated detection antibody was observed. It is the label and its coupling efficiency, not the assay scheme, that governs assay sensitivity. min, the silver enhancement reaction was in saturation [25]. Optimum incubation times reported are in the range of 10-15 min [26,27].

Comparison of the Six Protein Array Assay Schemes
To identify the assay schemes that are both rapid and accurate, we compared the times required to fully complete the assay as well as the assay sensitivity, as summarized in Figures 2 and 4. Assay Schemes (II) and (III) require three steps in total (spotting of the probes, incubation with the sample, and incubation with the labeled detection antibody) and represent the schemes with the shortest assay time (4 h). They are certainly the biomarker assays of choice in time-critical medical situations. Clearly, the assay time was not optimized, especially the incubation times of Steps 2-4 might be further reduced. In addition, Format (III) leads to the lowest LOD being 15 ng/L IL8, as low as Assay Format (VI). From Figure 2, it is obvious that those assay schemes were the only ones that could detect concentrations lower than 1215 ng/L IL8. Nevertheless, concentration-dependent black spots were obtained for all tested assay schemes except for Assay Scheme (I) (the Au nanoparticle-labeled detection antibody).
Results indicate that the choice of detection label strongly influences the assay sensitivity and measurement range of the immunoassay. This adds to the observations made with different types of nitrocellulose platforms (different modifications and different porosities) [28] or generally different types of surface materials [29] and enzyme substrates. Choosing the appropriate label allows the assay sensitivity to be tuned within a range that best fits the detection requirements, for example, to monitor the biomarker over several stages of the disease. In general, no clear trend towards the directly labeled detection antibody or the biotinylated detection antibody was observed. It is the label and its coupling efficiency, not the assay scheme, that governs assay sensitivity. Spot morphology is consistent for all assay schemes as indicated by the uniform size (90-120 µm) and appearance of the 3 × 3 replicate spots. The lowest LOD (15 ng/L IL8) was observed for Assay Format (III). Despite excellent sensitivity, a high BG was calculated for the enzymatic approach (Format (VI)), about 3.5-fold higher compared to the oCNPs assay (Format (III)). However, the background signal stays constant in Formats (III) and (VI). The high BG is likely due to nonspecific precipitation of the colored product and has already been previously reported [30]. To overcome this problem, Liu et al. [30] established a washing strategy based on a ring-oven technique integrated on a paper-based immunodevice to measure carcinoembryonic antigen (CEA). By this Spot morphology is consistent for all assay schemes as indicated by the uniform size (90-120 µm) and appearance of the 3 × 3 replicate spots. The lowest LOD (15 ng/L IL8) was observed for Assay Format (III). Despite excellent sensitivity, a high BG was calculated for the enzymatic approach (Format (VI)), about 3.5-fold higher compared to the oCNPs assay (Format (III)). However, the background signal stays constant in Formats (III) and (VI). The high BG is likely due to non-specific precipitation of the colored product and has already been previously reported [30]. To overcome this problem, Liu et al. [30] established a washing strategy based on a ring-oven technique integrated on a paper-based immunodevice to measure carcinoembryonic antigen (CEA). By this method, the authors were able to effectively reduce non-specific binding and improve LOD 10-fold. However, this method requires also additional instrumentation and heating up to 80 • C, which might compromise antibody activity if not properly controlled at the detection zone.
Lönnberg and Carlsson [31] compared the detection ability of using colloidal carbon NPs (at the size of 51 nm) as an antibody label to those obtained with enzymatic labels. A nitrocellulose membrane-based immunochromatographic device and a flatbed scanner as a quantitative test system were used. The scanner detected 0.2-170 amol carbon black/mm 2 with a coefficient of variation lower than 2% and an LOD of 0.02 amol/mm 2 . The detection ability of pure carbon was comparable to the method using horseradish peroxidase (HRP) together with TMB substrate yielding a colored signal of 5 amol HRP/microtiter well. These values are well comparable and about 2 orders of magnitude lower than those of, e.g. polystyrene particles [19]. In contrast to our study, the carbon NPs were adsorbed on the antibodies, because such a process is completed after only 30 min of incubation. In another article by van Dam et al. [32], the sensitivity of carbon NPs conjugates (prepared by adsorption) to detect Schistosomiasis was reported to be equal to that of the corresponding enzyme linked immunosorbent assay (ELISA).
Using the oCNPs (III) assay sensitivity was enhanced by 27 times compared to Scheme (V) using strp-AuNPs (V). This fits well with the results of Linares et al. [33] who demonstrated that carbon black had a remarkably low LOD of 0.01 g/L compared to 0.1 g/L and 1 g/L for silver-coated AuNPs and AuNPs, respectively. Based on the outcome of Sections 3.1-3.4 Assay Scheme (III) using oCNPs and Assay Scheme (VI) using HRP as a label in combination with TMB1 as substrate were further evaluated in a multiplexed chip format. Biomarkers IL8, VEGF, and DCN that were previously confirmed as important markers for detection of recurrence of bladder cancer [1] were spiked into assay buffer and synthetic urine (1:3) and incubated with the chip. The assay format for the three markers was the same as that described for IL8 in Section 2. The respective calibration curves are presented in Figure 5.
As is clear, the dynamic range of the assays is-despite the different signal levels-similar in buffer and in urine (1:3), and interferences by urine (1:3), such as low pH, urea, and other inhibitory components, are negligible. Measurement in pure urine, however, did indeed negatively affect assay sensitivity, and, depending on the protein biomarker, the limit of detection was enhanced by more than 10 times. Matrix effects from urine are reported to be differently pronounced for different assay markers. For example, Goodison et al. [34] evaluated a protein biomarker panel (IL8, MMP9, MMP10, ANG, APOE, SDC1, A1AT, PAI1, CA9, and VEGFA) in voided urine samples using a custom multiplex ELISA assay and divided the biomarkers into two sets, one to be measured with 4-fold diluted and one with 200-fold diluted urine samples. Data reproducibility as defined by the coefficient of variation (CV) was comparable for biomarker analysis in assay buffer and urine (1:3) and as low as 5 to 9%. Clearly, dilution of the urinary samples allows highly sensitive biomarker detection, but at the same time runs the risk to miss extremely low biomarker concentrations. Cut-off values reported in literature for IL8 and VEGF in urine were 25-35 ng/L, and 860 ng/L, respectively [3]. No cut-off value for DCN was found. Such low levels suggest that concentrators or signal enhancement kits shall be used to further improve assay sensitivity and meet the requirements for protein biomarker detection in therapy monitoring. Alternatively, several antibodies and antibody pairs might be tested in terms of sensitivity, specificity, physical properties, and recognition of native protein, and the best one selected for implementation in the multiplex chip [3].

Conclusions
Six protein array assay schemes and three different labels (Au, carbon, enzyme) were evaluated to identify the most sensitive colorimetric assay scheme to be exploited in an on-chip immunoassay for IL8, DCN, and VEGF, three protein markers that have been previously demonstrated to be important for the detection of bladder cancer recurrence. The highest signal intensity and assay sensitivity was observed for the enzymatic Assay Format (VI) with TMB1 as an enzyme substrate and Format (III) using oxidized carbon NPs (oCNPs). Herein, we could show that those schemes perform equally well in synthetic urine (1:3) and assay buffer in a multiplexed arrangement. With this colorimetric chip, a cost-effective, fast (4 h), simple, and efficient tool was established, which has high potential for use in doctors' offices as a non-invasive tool in therapy monitoring.

Acknowledgments:
The authors thank the European Commission for financial support within FP7 Dipromon (development of an integrated protein-and cell-based device for noninvasive diagnostics in the urogenital tract, no. 306157).
Author Contributions: C.P. conceived and designed the experiments; S.G. and S.D. performed the experiments; S.G. and U.S. analyzed the data; C.P. and S.G. wrote the paper.

Conflicts of Interest:
The authors declare no conflict of interest.

Conclusions
Six protein array assay schemes and three different labels (Au, carbon, enzyme) were evaluated to identify the most sensitive colorimetric assay scheme to be exploited in an on-chip immunoassay for IL8, DCN, and VEGF, three protein markers that have been previously demonstrated to be important for the detection of bladder cancer recurrence. The highest signal intensity and assay sensitivity was observed for the enzymatic Assay Format (VI) with TMB1 as an enzyme substrate and Format (III) using oxidized carbon NPs (oCNPs). Herein, we could show that those schemes perform equally well in synthetic urine (1:3) and assay buffer in a multiplexed arrangement. With this colorimetric chip, a cost-effective, fast (4 h), simple, and efficient tool was established, which has high potential for use in doctors' offices as a non-invasive tool in therapy monitoring.