Discovery of Novel Bladder Cancer Biomarkers by Comparative Urine

Aug 31, 2010 - University, Taoyuan 333, Taiwan, and Graduate Institute of Medical ... Medical Biotechnology and Laboratory Science, Chang Gung Univers...
0 downloads 0 Views 4MB Size
Download PDF
Discovery of Novel Bladder Cancer Biomarkers by Comparative Urine Proteomics Using iTRAQ Technology Yi-Ting Chen,*,†,‡ Chien-Lun Chen,†,§ Hsiao-Wei Chen,‡ Ting Chung,‡ Chih-Ching Wu,‡,| Chi-De Chen,⊥ Chia-Wei Hsu,⊥ Meng-Chieh Chen,‡ Ke-Hung Tsui,§ Phei-Lang Chang,§ Yu-Sun Chang,‡,⊥ and Jau-Song Yu*,‡,⊥ Molecular Medicine Research Center, Chang Gung University, Taoyuan 333, Taiwan, Chang Gung Bioinformatics Center, Department of Urology, Chang Gung Memorial Hospital, Taoyuan 333, Taiwan, Graduate Institute of Biomedical Sciences, Colleague of Medicine, Chang Gung University, Taoyuan 333, Taiwan, and Graduate Institute of Medical Biotechnology and Department of Medical Biotechnology and Laboratory Science, Chang Gung University, Kwei-San, Taoyuan 333, Taiwan Received June 9, 2010

A urine sample preparation workflow for the iTRAQ (isobaric tag for relative and absolute quantitation) technique was established. The reproducibility of this platform was evaluated and applied to discover proteins with differential levels between pooled urine samples from nontumor controls and three bladder cancer patient subgroups with different grades/stages (a total of 14 controls and 23 cancer cases in two multiplex iTRAQ runs). Combining the results of two independent clinical sample sets, a total of 638 urine proteins were identified. Among them, 55 proteins consistently showed >2-fold differences in both sample sets. Western blot analyses of individual urine samples confirmed that the levels of apolipoprotein A-I (APOA1), apolipoprotein A-II, heparin cofactor 2 precursor and peroxiredoxin-2 were significantly elevated in bladder cancer urine specimens (n ) 25-74). Finally, we quantified APOA1 in a number of urine samples using a commercial ELISA and confirmed again its potential value for diagnosis (n ) 126, 94.6% sensitivity and 92.0% specificity at a cutoff value of 11.16 ng/mL) and early detection (n ) 71, 83.8% sensitivity and 94.0% specificity). Collectively, our results provide the first iTRAQ-based quantitative profile of bladder cancer urine proteins and represent a valuable resource for the discovery of bladder cancer markers. Keywords: urine proteome • iTRAQ • bladder cancer • biomarkers • apolipoprotein • heparin cofactor 2 precursor • peroxiredoxin-2

Introduction Bladder cancer is one of the most common urinary tract carcinomas. The earlier this cancer is found and treated, the better the outcome.1 Cytology, the standard noninvasive detection method for bladder cancer, has low sensitivity and specificity, particularly for low-grade tumors;2 it is often used as an adjunct to cystoscopy, which is an invasive and costly procedure. Several candidate bladder tumor markers have been identified in urine and/or bladder cancer cells for purposes of making an initial diagnosis and monitoring recurrence and treatment response. However, the properties of these candidate markers are not superior to existing detection methods with respect to sensitivity or specificity, or their clinical utility has * To whom correspondence should be addressed. Dr. Jau-Song Yu, Tel: 886-3-2118800ext. 5171. Fax: 886-3-2118891. E-mail: [email protected]. Dr. Yi-Ting Chen, Tel: 886-3-2118800ext. 3558. Fax: 886-3-2118800 ext. 3533. E-mail:[email protected]. † These authors contributed equally to this work. ‡ Molecular Medicine Research Center. § Department of Urology. | Graduate Institute of Medical Biotechnology and Department of Medical Biotechnology and Laboratory Science. ⊥ Colleague of Medicine. 10.1021/pr100576x

 2010 American Chemical Society

not been examined in large patient populations.2-6 Thus, there is a compelling need to develop more reliable bladder tumor markers. Urine is in direct contact with bladder epithelia cells, which may give rise to bladder tumors; proteins released from bladder tumor cells may be enriched in urine samples. One promising approach in the search for useful bladder cancer biomarkers is to study the urine proteome during the occurrence of the disease. Several approaches have been developed for global protein identification of the urine proteome.7-10 For differential urine proteome analysis, techniques such as two-dimensional gel electrophoresis (2-DE), capillary electrophoresis mass spectrometry,11-13 and surface-enhanced laser desorption ionization mass spectrometry (SELDI-MS)14-16 have been applied to analyze the differences in urinary profiles between control and target groups. SDS-PAGE or 2-DE systems have been most widely applied in differential proteomic studies of urine specimens.17-21 A number of inherent limitations of 2DE including the difficulty in separating low-molecular-weight proteins make this technique insufficient for analysis of the urine proteome, which contains many low-molecular-weight proteins as a result of filtration by the kidney. By coupling Journal of Proteome Research 2010, 9, 5803–5815 5803 Published on Web 08/31/2010

research articles

Chen et al.

Figure 1. Schematic representation of the experimental design used to evaluate the reproducibility of the iTRAQ-based quantitative platform for urine proteome analysis. Variations at the digestion/labeling level and the whole-workflow level were evaluated using 115/114 and 117/116 ratios, respectively.

stable isotope labeling (SIL) techniques with liquid chromatography fractionation and nano-LC-MS/MS analysis, the platform enables global protein identification and quantification with high resolution and sensitivity, no molecular weight limitations, and retention of post-translational modification information. However, applications of SIL techniques to the urine proteome have been very limited,22,23 and the suitability of the techniques to biomarker discovery in the urine proteome, which is highly varied due to the complicated source and physiological condition, has not been clearly established. In this study, we established a urine sample preparation workflow for the iTRAQ technique and evaluated its reproducibility. We then applied this platform to the discovery of proteins with differential levels between pooled urine samples from nontumor (NT) controls and three subgroups of bladder cancer patients. Several urine proteins with increased levels were confirmed by Western blot analyses of individual samples and found to be potential markers for early detection, diagnosis, or stage differentiation of bladder cancer. A further analysis of the urinary levels of one of these proteins, APOA1, in a larger number of urine samples by an enzyme-linked immunosorbent assay (ELISA) yielded results that were consistent with the Western blot analysis. To the best of our knowledge, this is the first study to discover disease biomarkers in the human urine proteome using an SILbased proteomic technique systematically.

Experimental Procedures Clinical Specimens. For platform development and reproducibility evaluation, first morning urine samples were collected in the presence of a protease inhibitor cocktail tablet (one tablet per 50 mL urine; Roche, Mannheim, Germany) and sodium azide (1 mM) from 12 volunteers (6 males, 6 females; 27.00 ( 3.07 years old) without a history of urinary disease. For biomarker discovery and validation, first morning urine samples were similarly collected from bladder cancer patients 5804

Journal of Proteome Research • Vol. 9, No. 11, 2010

and NT volunteers. The collected samples were centrifuged at 5000× g for 30 min at 4 °C within 5 h to remove cells and debris, and the clarified supernatants were stored at -20 °C for further processing. The proteomic platformsiTRAQ labeling coupled with LCbased fractionationswas evaluated using 37.5 mL urine from a pooled sample from 12 volunteers. The urine sample was separated into three parts and processed as shown in Figure 1 to evaluate variations at the digestion/labeling level (iTRAQ tag 114/115 ratios) and at the whole-workflow level (iTRAQ tag116/ 117 ratios). For the discovery phase of this study, equal amounts of protein from each individual were pooled into a subgroup. The four subgroups of 4-plex iTRAQ were classified as NT, low grade with early stage (LgEs), high grade with early stage (HgEs), and high grade with advanced stage (HgAs), according to the TNM staging system.24,25 Two independent clinical sample sets were used for biomarker discovery by iTRAQ labeling. Supplemental Table 1 (Supporting Information) lists the diagnosis, sex, and sample number information for the two clinical sample sets. All NT and bladder cancer patients were older than 45 years old. In addition to NT and bladder cancer specimens, urine samples from patients with urinary tract infection (UTI) or hematuria (HU) were included for Western blot analyses of candidate proteins for individual validation. The clinical attributes of these samples are shown in Supplemental Table 2 (Supporting Information). All urine samples were collected at Chang Gung Memorial Hospital, Taoyuan, Taiwan. The study protocol was approved by the Medical Ethics and Human Clinical Trial Committee at Chang Gung Memorial Hospital, Taiwan. Concentration and Desalting of Urine Samples. Urine proteins were enriched using a 10-kDa centrifugal filter as described by the manufacturer (Millipore, Carrigtwohill, Ireland). Briefly, 12.5 mL urine samples were centrifuged at 5000×

Discovery of Novel Bladder Cancer Biomarkers g for 30 min at 4 °C in the filter tube, and then the tube was refilled with 12.5 mL of 20% acetonitrile/H2O and centrifuged again. This process was repeated once using pure water for desalting purposes. For iTRAQ labeling, samples were subjected to an additional desalting step using 4 mL H2O to prevent against possible interference by metabolites in the labeling reaction. The amount of protein in each concentrated/desalted urine sample was measured using a DC protein assay kit (BioRad, Hercules, CA), and then lyophilized and stored at -80 °C for subsequent processing. The value for total amount (µg) of protein in each individual sample was then transformed into concentration (µg/mL) based on the initial raw urine volume prior to desalting and concentration. iTRAQ Labeling and Fractionation by Strong Cationic Exchange (SCX) and Basic Reverse-Phase Chromatography. Urine proteins from patients with the same histological grades or pathological stages and controls were pooled to minimize individual variation and enhance signals. In this study, hernia patients were chosen as the NT subgroups. Pooled urine protein (100 µg) from each subgroup was processed according to the manufacturer’s protocol for 4-plex iTRAQ (Applied Biosystems, Foster City, CA). Briefly, one unit of iTRAQ reagent (defined as the amount of reagent required to label 100 µg of protein) was thawed and reconstituted in 70 µL ethanol. Pooled urine protein from each subgroup was reduced, cysteineblocked, and digested with trypsin. Tryptic peptides of NT, LgEs, HgEs, and HgAs subgroups were labeled with 114, 115, 116, and 117 iTRAQ tags, respectively, by incubation at room temperature for 1 h. The peptide mixtures were then pooled and dried by vacuum centrifugation. The pooled mixtures of iTRAQ-labeled peptides for platform development and clinical sample set 1 were fractionated by SCX chromatography. The pooled mixtures of iTRAQ-labeled peptides from clinical sample set 2 were fractionated separately by SCX and reverse-phase (RP) chromatography. For SCX chromatography using the Waters 1525 Micro Binary HPLC Pump system, the iTRAQ-labeled peptide mixture was reconstituted and acidified with 0.5 mL buffer A (0.1% formic acid in 25% ACN) and loaded onto a 2.1 × 150 mm BioBasic SCX column containing 5-µm particles and a 300-µm pore size (Thermo Electron, San Jose, CA). The peptides were eluted at a flow rate of 360 µL/min with a gradient of 1% buffer B (400 mM NH4Cl, 0.1% formic acid in 25% ACN) for 5 min, 1-15% buffer B for 15 min, 15-25% buffer B for 10 min, 25-55% buffer B for 10 min, and 55-95% buffer B for 5 min. The system was then maintained in 95% buffer B for 5 min before equilibrating with 1% buffer B for 8 min prior to the next injection. Elution was monitored by measuring absorbance at 220 nm, and fractions were collected every 1 min. The eluted peptides were pooled as 42 fractions and vacuum-dried. For RP chromatography under basic conditions, the iTRAQlabeled peptide mixture was reconstituted in 0.5 mL buffer C (ammonium hydroxide aqueous solution, pH 10) and loaded onto a 4.6 × 150 mm Gemini C18 column containing 3-µm particles and a 160-µm pore size (HPLC Phenomenex, Torrance, CA). The peptides were eluted at a flow rate of 400 µL/min with a gradient of 2% buffer D (ammonium hydroxide in 100% ACN, pH 10) for 5 min, 2-35% buffer D for 40 min, 35-60% buffer D for 5 min, 60-95% buffer D for 3 min. The system was then maintained in 95% buffer D for 3 min before equilibrating in 2% buffer D prior to the next injection. Elution was monitored by measuring absorbance at 220 nm, and fractions were collected every 1 min. After pooling eluted fractions as above

research articles and vacuum drying, samples were ready for nano-ESI-LC-MS/ MS analysis. LC-ESI-MS/MS Analysis by LTQ-Orbitrap Pulsed-Q Dissociation. Each separated peptide fraction was reconstituted in buffer E (0.1% formic acid in H2O), and 2 µg peptides from each fraction were loaded onto a trap column (Zorbax 300SBC18, 0.3 × 5 mm, Agilent Technologies, Wilmington, DE) at a flow rate of 20 µL/min in buffer A, and separated on a resolving 10-cm analytical BioBasic C18 PicoFrit column (inner diameter, 75 µm) with a 15-µm tip (New Objective, Woburn, MA). Peptides were eluted at a flow rate of 0.25 µL/min across the analytical column with a linear gradient of 5-30% buffer E (0.1% formic acid in 99.9% acetonitrile) for 40 min, 30-45% buffer E for 5 min, and 45-95% buffer E for 2 min, and then maintained in 95% buffer E for 4 min. The LC setup was coupled online to a linear ion trap-orbitrap (LTQ-Orbitrap, Thermo Fisher, San Jose, CA) operated using Xcalibur 2.0 software (Thermo Fisher). Intact peptides were detected in the Orbitrap at a resolution of 30 000. Internal calibration was performed using the ion signal of (Si(CH3)2O)6H+ at m/z 445.120025 as a lock mass.26 Peptides were selected for MS/MS using pulsed-Q dissociation (PQD) operating mode with a normalized collision energy setting of 27%; ion fragments were detected in the LTQ. A data-dependent procedure that alternated between one MS scan followed by three MS/MS scans was applied for the three most abundant precursor ions in the MS survey scan. The m/z values selected for MS/MS were dynamically excluded for 180 s. The electrospray voltage applied was 1.8 kV. Both MS and MS/MS spectra were acquired using the 4 microscan with a maximum fill-time of 1000 and 100 ms for MS and MS/MS analysis, respectively. Automatic gain control was used to prevent overfilling of the ion trap; 5 × 104 ions were accumulated in the ion trap for generation of PQD spectra. For MS scans, the m/z scan range was 350 to 2000 Da. Protein Identification and Quantitation by Sequence Database Searching. The resulting MS/MS spectra were searched against the European Bioinformatics Institute (http://www. ebi.ac.uk/) nonredundant International Protein Index (IPI) human sequence database (v3.27, March 2007) containing 67 528 sequences and 28 353 548 residues using the MASCOT engine (Matrix Science, London, U.K.; version 2.2.04) with the Mascot Daemon program (Matrix Science, version 2.2.0). For protein identification, a mass tolerance of 10 ppm was permitted for intact peptide masses and 0.5 Da for PQD fragmented ions, with allowance for two missed cleavages in the trypsin digests, oxidized methionine as a potential variable modification, and iTRAQ (N terminal), iTRAQ (K), and methyl methanethiosulfonate (C) as fixed modifications. The charge states of peptides were set to +2 and +3. Protein identification and quantification were validated using the open source Trans-Proteomic Pipeline (TPP) software (Version 4.0). The MASCOT search resulted in a DAT file for each SCX or RP elution. The MS raw data and the DAT files containing peak list information for identified peptides were then processed and analyzed using the TPP software. The TPP software includes PeptideProphet, a peptide probability score program that aids in the assignment of peptide MS spectrum,27 and ProteinProphet, a program that assigns and groups peptides into a unique protein or a protein family if the peptide is shared among several isoforms.28 ProteinProphet allows filtering of large-scale data sets with assessment of predictable sensitivity and false-positive identification error rates. In this Journal of Proteome Research • Vol. 9, No. 11, 2010 5805

research articles

Chen et al.

Table 1. Analysis of Urine Samples Using the iTRAQ Platform: Summary of Protein Identification, Quantitation, and Experimental Variationa

Digestion/labeling level (115/114) Whole workflow level (117/116) a

no. of proteins identified

no. of proteins quantified

no. of peptides identified

no. of peptides quantified

578

506

3810

2734

CV (protein)

average ( SD (peptide)

CV (peptide)

1.05 ( 0.14 1.00 ( 0.12

13.7% 12.1%

1.07 ( 0.23 1.01 ( 0.15

21.1% 14.7%

SD, standard deviation; CV, coefficient of variation.

study, we used PeptideProphet and ProteinProphet probability scores g0.95 to ensure an overall false-positive rate less than 0.7%. The ratio of each protein was quantified using the Libra program, a module within the TPP software package that performs quantification on MS/MS spectra that have multiplexed labeled peptides. The minimum intensity threshold of a reporter ion was 20 in the spectrum of a LIBRA peptide. The default parameters of the LIBRA program were used to remove the outlier ratios of peptides quantitation. Each quantified protein contained at least one LIBRA peptide. Information about the PeptideProphet, ProteinProphet, and Libra programs in the TPP software can be obtained from the Seattle Proteome Center at the Institute for Systems Biology (http://www. proteomecenter.org/). Western Blot Analysis. Western blot analyses of target proteins were performed as previously described.29 After desalting and concentrating urine proteins, the amount of protein in each urine sample was measured by the DC protein assay. Urine proteins (100 µg) from individual or pooled samples were resolved on SDS-PAGE gels and transferred electrophoretically onto PVDF membranes (Bio-Rad, Hercules CA). The membranes were blocked for 1 h at room temperature with 5% nonfat dried milk in Tris-buffered saline (J. T. Baker, Phillipsburg, NJ) containing 0.1% Tween 20 (Sigma, St Louis, MO) (TBST). The following antibodies were used for Western blot analysis: antiapolipoprotein A-I (anti-APOA1, 1:500, ab58924, Abcam, U.K.), antiapolipoprotein A-II (anti-APOA2, 1:250, ab54796, Abcam, CB, U.K.), antiheparin cofactor 2 precursor (anti-HCII, 1:2000, MAB0769, Abnova Corp., Taipei, Taiwan), antiperoxiredoxin 2 (anit-PRDX2, 1:5000, AF3489, R&D systems), anti-s100A6 (1:200, AF4584, R&D Systems), and antis100A8 (1:200, AF4570, R&D systems). The membranes were probed with primary antibody followed by horseradish peroxidase-conjugated secondary antibody, and developed using enhanced chemiluminescence detection according to the manufacturer’s instructions (Millipore, Billerica, MA). The relative signal intensity of each target protein detected in the blots was quantified using a computing densitometer (Molecular Dynamics, Sunnyvale, CA). Quantification of APOA1 by ELISA. The levels of APOA1 in diluted urine or desalted/concentrated urine samples were determined using a sandwich ELISA kit (Matritech, Newton, MA). All urine samples were prepared as 2- to 100-fold dilutions in buffer A (PBS containing 0.05% Tween 20 and 0.1% BSA), and then processed according to the instructions provided by the kit manufacturer (Mabtech, Cincinnati, OH). Desalted/ concentrated urine samples or diluted urine were added to Costar clear microplates (Corning, Corning, NY) that had been antibody-coated overnight at 4 °C with the monoclonal antibody HDL 110 (2 µg/mL in PBS, 100 µL/well). The plates were then washed twice with PBS and blocked with 200 µL/well of buffer A at room temperature for 1 h. Purified APOA1 (Matritech, Newton, MA) was used as the calibration standard. Urine samples (100 µL of NT urine at a 1:2 dilution and 100 µL of 5806

average ( SD (protein)

Journal of Proteome Research • Vol. 9, No. 11, 2010

bladder cancer urine at a 1:100 dilution) were added and incubated at room temperature for 2 h, and the plates were washed five times with buffer B (PBS containing 0.05% Tween 20). Subsequently, biotinylated monoclonal antibody HDL 44 (0.5 µg/mL in buffer A, 100 µL/well) was applied to the wells and the plates were incubated for 1 h. After five washes with buffer B, 100 µL of streptavidin-alkaline phosphatase diluted 1000-fold in buffer A was added to the wells and the plates were incubated for 1 h. After washing five times with buffer B, alkaline phosphatase substrate (pNPP substrate; 1 mg/mL, 100 µL/well, Amresco Inc.) was added to the wells, and color intensity at different times was measured at a wavelength of 405 nm using a SpectraMax M5 microplate reader (Molecular Devices, Sunnyvale, CA). The APOA1 concentration in samples was calculated by reference to a standard curve prepared using purified APOA1. The reference level of APOA1 had been set at 0.2-20 ng/mL by the manufacturer. According to the Western blot data, to avoid the overscale condition in the ELISA measurements, the urines from controls and cancer groups were diluted differently for ELISA analyses. Several dilution ratios were further tested for the cancer cases that were negative at 1:1000. Statistical Analysis. The statistical package SPSS 13.0 (SPSS Inc., Chicago, IL) was used for all analyses. All continuous variables were expressed as means ( SDs. Differences in concentration levels of urine APOA1, APOA2, PRDX2, and HCII between different clinical parameters measured by Western blotting and ELISA were analyzed using the nonparametric Mann-Whitney U test. SPSS software was also used for statistical analyses of receiver operator characteristic (ROC) curves and areas under the curve (AUC). Two-tailed p-values of 0.05 or less were considered significant. ROC curve analysis was applied to detect the optimal cutoff point that yielded the highest total accuracy with respect to discriminating different clinical classifications. The optimal cutoff point was determined using Youden’s index (J), calculated as J ) 1 - (false positive rate + false negative rate) ) 1 - [(1 - sensitivity) + (1 specificity)] ) sensitivity + specificity - 1.30

Results Assessment of the Reproducibility of Quantitative iTRAQ for Urine Proteomics. A model sample, prepared by pooling equal volumes of urine samples from twelve volunteers, was processed as shown in Figure 1 for purposes of developing the basic operation of the platform, with a focus on the reproducibility and compatibility of the established workflow. Table 1 summarizes the protein identification and quantification results. A total of 1533 distinct peptides associated with 578 proteins were identified in this experiment, and 1224 distinct peptides corresponding to 506 proteins were quantified. To assess the variations that occurred at the digestion/labeling and whole-workflow levels, we calculated 115/114 and 117/116 ratios. The average 115/114 and 117/116 ratios were 1.05 and

research articles

Discovery of Novel Bladder Cancer Biomarkers 1.00 with coefficient of variation less than 14%. Protein/peptide identification and quantitation details for this model sample are shown in Supplemental Table 3 (Supporting Information). These results clearly validate the performance and reproducibility of the entire workflow, including the processes for urine protein desalting, concentration, proteolytic digestion, iTRAQ labeling, LC fractionation, MS quantitation, and data analysis. Evaluation of RPLC and SCXLC Strategies for the Separation and Identification of iTRAQ-Labeled Urine Peptides. According to the manufacturer’s protocols, iTRAQ-labeled peptides are typically separated by SCX chromatography before online nanoacidic RP-LC/MS/MS analysis.31-33 To examine the potential utility of RP chromatography in the separation and identification of iTRAQ-labeled urine peptides, we separated iTRAQ-labeled peptides in clinical sample set 2 by off-line basic RPLC before online acidic nano-RPLC-MS/MS. We then compared the protein identification/quantitation results with those obtained from standard off-line SCX chromatography coupled with online acidic nano-RPLC-MS/MS (Supplemental Figure 1, Supporting Information) The result indicates that basic RP chromatography is a more efficient strategy than the regularly used SCX chromatography, allowing identification and quantification of more iTRAQ-labeled peptides in urine samples. Protein Identification and Quantitation in Two Independent Clinical Sample Sets. Several previous studies have reported significant differences in the urine protein composition among individuals with the same diagnostic status.34-37 To minimize the impact of person-to-person variation in urine composition in the biomarker discovery phase, we used two independent sets of biological samples (a total of 14 controls and 23 cancer cases in two multiplex iTRAQ runs, clinical sample set 1 and 2) pooled from different individuals for the iTRAQ experiments. A total of 352 nonredundant proteins (1074 distinct peptides) were identified by combining LC-MS/MS data files from 42 SCX fractions of clinical sample set 1, and 552 nonredundant proteins (1589 distinct peptides) were identified by combining LC-MS/MS data files from 42 SCX and 42 basic RP fractions of clinical sample set 2. Integration of the proteins identified from the two clinical sample sets yielded a total of 638 urine proteins (PeptideProphet and ProteinProphet probability scores g0.95, false positive rate nontumor, (II) bladder cancer e nontumor, (III) nonsignificant difference, and (IV) others. The numbers of quantified protein in each subgroup are summarized in Table 2. Functional analysis revealed no obvious differences in the distributions of global molecular functions between total identified proteins and proteins with differential

Table 2. Protein Identification and Quantification Results from Two Independent Clinical Sample Sets by iTRAQ protein number classification

definition

Identified Proteins Quantified Proteins (QP) (I) BCa > NTb

All Ratiosc g 1, and Any Ratiod g 2 All Ratios e1, and Any Ratio e0.5 0.5 < All Ratios >2

(II) BC < NT (III) Non-Significant Change (IV) Others a

(QP)-(I)-(II)-(III) b

clinical set 1

clinical set 2

352 274

552 429

46

81

59

184

78

91

91

73

c

BC: Bladder cancer. NT: Nontumor controls. All Ratios: 115/114, 116/114 and 117/114 ratios. d Any Ratio: 115/114, 116/114 or 117/114 ratios.

levels (Supplemental Figure 5, Supporting Information). Importantly, 22 proteins showed increased levels in both clinical sample sets and levels of 33 proteins were decreased in both sets (Supplemental Table 7, Supporting Information). These proteins represent potential urine bladder cancer biomarkers. Verification of Candidate Biomarkers in Pooled and Individual Urine Samples. On the basis of our iTRAQ results and relevance to human cancer, six candidate biomarkerssAPOA1, APOA2, HCII, PRDX2, S100A6, and S100 A8swere selected for verification in pooled and individual samples. APOA1 and APOA2 were highly elevated in pooled bladder cancer subgroups of clinical sample sets 1 and 2. The fold-difference for APOA2 in urine was particularly dramatic: in clinical sample set 1, APOA2 was ∼125-fold higher in HgAs patients than in NT controls. To the best of our knowledge, this fold-difference value is among the highest observed in complex, real-world biological systems using the iTRAQ technique.22,32,33,38-42 Figure 2 shows the results of MS analyses for quantification (in the low mass reporter ion region) and identification of the six selected candidate biomarkers. In practice, complex matrix contributions, such as coelution of nearly isobaric peptides and nonpeptide contaminants, may significantly limit the dynamic range and accuracy of fold-difference determinations in iTRAQ quantitation.43,44 At first, we applied Western blot analyses to validate the protein levels of those six candidate biomarkers, using 100 µg of pooled urine proteins from subgroups that were originally used for iTRAQ experiments. Figure 3 shows that the quantitative trends of bladder cancer and control subgroups matched well between Western blot analyses and iTRAQ results. We then analyzed APOA1, APOA2, HCII, and PRDX2 levels in 55-64 individual urine samples by Western blotting (Figure 4A-D, upper panels). The detailed information of samples used for iTRAQ, Western blot analyses, and ELISA are listed in Supplemental Table 1 and 2 (Supporting Information). After normalization, the relative protein levels of the four candidate proteins among subgroups were obtained (Figure 4A-D, lower panels and Table 3). APOA1 was barely detectable in 100 µg of total urine protein from the NT subgroups under the conditions used; however, APOA1 levels increased dramatically in all bladder cancer subgroups. The ratios of APOA1 level in LgEs/ NT, HgEs/NT, and HgAs/NT based on constant amounts of protein analyzed were 83.8, 209.4, and 272.9, respectively (Figure 4A, lower left panel). The corresponding values calcuJournal of Proteome Research • Vol. 9, No. 11, 2010 5807

research articles

Chen et al.

Figure 3. Verification of the six selected proteins in pooled urine samples by Western blot analyses. Proteins (100 µg) from pooled urine samples used for the original iTRAQ experiment were used for Western blot analysis (WB). Fold-differences in target protein levels in cancer subgroups relative to the NT group are denoted below each blot.

Figure 2. LC-MS/MS quantification results (in the low mass reporter ion region) and identification of six selected candidate biomarkers: (A) APOA1, (B) APOA2, (C) HCII, (D) PRDX2, (E) S100A6, and (F) S100A8.

lated on the basis of constant urine volume were 207.4, 271.6, and 408.5 (Figure 4A, lower right panel). The other three candidate proteins, APOA2, HCII and PRDX2, were similarly elevated in bladder cancer subgroups based on the amount of total urine protein or urine volume: APOA2, 3.0- to 17.1-fold increase; HCII, 12.7- to 63.1-fold increase; and PRDX2, 2.9- to 47.6-fold increase (Figure 4B-D). To evaluate the interference of bloody urine or other common urology diseases, we also examined the levels of the four candidate proteins in urine samples from 13-22 patients 5808

Journal of Proteome Research • Vol. 9, No. 11, 2010

with UTI and/or HU in parallel. The average levels of APOA1, APOA2 and HCII in UTI/HU patients were both much lower (0.7- to 13.5-fold) compared to bladder cancer groups (3.0- to 408.5-fold) whether based on constant urine protein amounts or urine volume, although their levels were generally slightly higher in UTI/HU patients than in the NT group (Figure 4A-C). The level of PRDX2 in the UTI/HU group was increased to levels comparable to those in bladder cancer subgroups (Figure 4D, and Supplemental Table 8-1 and 8-2, Supporting Information). On the other hand, the urine levels of S100A6 and S100A8 were not significantly difference between NT and bladder cancer subgroups (data not shown). In addition, S100A6 and S100A8 protein levels in the UTI/HU group were even higher than those in bladder cancer subgroups under the assay conditions used (data not shown). Taken together, these observations suggest that APOA1, APOA2 and HCII, but not PRDX2, S100A6 or S100A8, might represent good diseasespecific urine markers for the diagnosis of bladder cancer. Assessment of the Diagnostic Efficacy of Urine APOA1, APOA2, HCII and PRDX2 in Detecting Bladder Cancer. We assessed the diagnostic efficacy of urine APOA1, APOA2, HCII and PRDX2 based on their abilities to differentiate (i) bladder cancer from NT (all bladder cancer subgroups/NT), (ii) early or superficial bladder cancer from NT (LgEs/NT), (iii) early stage tumor from advanced-stage tumor [(LgEs + HgEs)/HgAs], and (iv) low-grade tumor from high-grade tumor [LgEs/(HgEs + HgAs)]. The results of these statistical analyses are summarized in Table 4. All four proteins were present at significantly elevated levels in all bladder cancer subgroups with statistical significance (p-values e0.05) and AUC values >0.70 based on constant protein amounts or constant urine volume (Supplemental Figure 6A-B and 7A-D, Supporting Information), indicating their potential value as novel noninvasive diagnostic markers for bladder cancer. Importantly, Western

Discovery of Novel Bladder Cancer Biomarkers

research articles

Figure 4. Validation of four candidate markers in individual urine samples (17-54 NT controls; 31-75 bladder cancer, including 8 or 19 LgEs, 10-40 HsEs, and 8 or 16 HgAs subgroups; and 13-22 UTI/HU controls) by Western blot analyses. (Upper panels) Equal amounts of proteins prepared from individual urine samples were separated by SDS-PAGE, transferred to PVDF membranes, and probed with antibodies against APOA1 (A), APOA2 (B), HCII (C) or PRDX2 (D); 100 µg urine protein was used in (A), (B) and (D), whereas 50 µg urine protein was analyzed in (C). (Lower panels) Fold-differences between NT, LgEs, HgEs, HsAs and UTI/HU were calculated based on constant total urine protein amounts (left panel) and equal urine volume (right panel). The average fold-difference between each subgroup and the NT subgroup is labeled above each box plot. As an internal standard, a pooled urine sample from five bladder cancer patients was included in each Western blot analysis, and used to normalize the intensity of each target protein detected. The horizontal lines in each box plot denote the 10th, 25th, 50th, 75th and 90th percentiles of the data distribution.

blot analyses revealed that the levels of APOA1 and HCII in urine could be used to discriminate the LgEs subgroup from NT controls based on either constant urine protein amount or constant urine volume (Table 4, Supplemental Figure 6C-D and 7E-F, Supporting Information). The levels of APOA2 and PRDX2 were also found to be statistically different between early and advanced-stage tumors based on constant amount of urine protein and constant urine volume (Table 4, and Supplemental Figure 6G-H and 7G-H, Supporting Information). Furthermore, HCII was able to differentiate LgEs and total bladder cancer groups from UTI/HU based on constant amount of urine protein (Supplemental Table 8-1, Supporting Information). Collectively, these results identified APOA1, APOA2, HCII, and PRDX2 as novel candidate urine markers for

bladder cancer diagnosis; APOA1 and HCII seem to be promising in early detections, and APOA2 and PRDX2 have the potential to serve as tumor-stage discriminators. Measurement of APOA1 Concentrations in Urine Samples by ELISA. We then measured APOA1 concentration in a number of individual urine samples from controls and bladder cancer patients using a commercially available ELISA. Using diluted urine (n ) 126) and desalted/concentrated urine protein (n ) 86) in parallel, the ELISA data reveal that urine APOA1 concentrations are extremely high in all BC subgroups compared with both NT and UTI/HU subgroups (Figure 5 and Table 5). At a cutoff value of 0.12 ng/µg (constant amount of urine protein), APOA1 was able to discriminate all BC subgroups from NT with a sensitivity and specificity of 85.7 and 94.6%, respectively; the corresponding Journal of Proteome Research • Vol. 9, No. 11, 2010 5809

research articles

Chen et al.

Table 3. Average Fold-Differences in the Urine Levels of the Four Target Proteins in Bladder Cancer and UTI/Hematuria Subgroups Compared to the NT Group Measured by Western blot analyses based on constant amount of total urine protein constant total urine protein amount

NT (N ) 17-19)

APOA1 APOA2 HCII PRDX2

1 ( 0.4 1 ( 0.7 1 ( 1.1 1 ( 0.9

LgEs (N ) 8 or 9)

: : : :

83.8 ( 186.5 3.0 ( 5.0 23.6 ( 82.4 2.9 ( 5.8

HgEs (N ) 10 or 15)

: : : :

209.4 ( 252.9 7.0 ( 11.7 12.7 ( 17.8 3.6 ( 7.2

HgAs (N ) 7 or 8)

: : : :

UTI/Hematuria (N ) 13)

272.9 ( 252.1 7.1 ( 6.0 18.0 ( 25.3 12.6 ( 23.1

5.3 ( 4.4 0.7 ( 0.8 4.7 ( 9.1 3.3 ( 3.2

: : : :

Measured by Western blot analyses based on constant urine volume constant urine volume

NT (N ) 17-19)

APOA1 APOA2 HCII PRDX2

1 ( 1.4 1 ( 2.1 1 ( 1.7 1 ( 2.6

LgEs (N ) 8 or 9)

: : : :

207.4 ( 406.0 8.1 ( 20.2 63.1 ( 221.9 9.2 ( 24.7

HgEs (N ) 10 or 15)

: : : :

271.6 ( 338.0 10.2 ( 18.9 30.7 ( 56.0 4.7 ( 10.0

HgAs (N ) 7 or 8)

: : : :

UTI/Hematuria (N ) 13)

408.5 ( 408.2 17.1 ( 19.9 59.4 ( 100.3 47.6 ( 112.6

: : : :

7.4 ( 8.9 1.8 ( 3.9 13.5 ( 38.8 5.3 ( 7.8

Table 4. Summary of Statistical Analyses of the Efficacy of APOA1, APOA2, HCII and PRDX2 Urine Levels in Detecting Bladder Cancera Measured by Western blotting based on constant amount of total urine protein of individual samples (NT vs all BC) protein name

APOA1 APOA2 HCII PRDX2

p-value for diagnosis

< 0.001 (n ) 48) 0.014 (n ) 123) < 0.001 (n ) 129) 0.70) are highlighted in bold.

values based on constant urine volume (cutoff, 11.16 ng/mL) were 100% and 92.0% (Table 5B and Figure 5). At a cutoff value of 0.07 ng/µg (constant amount of urine protein), APOA1 could differentiate the LgEs subgroup from NT controls with a sensitivity and specificity of 100.0 and 83.8%, respectively; the corresponding values based on constant urine volume (cutoff, 16.30 ng/mL) were 100 and 94.0% (Table 5B and Figure 5). Our findings indicate that APOA1 might represent a novel urine marker for early detection of bladder cancer. Although these early results are promising, additional studies using a larger set of urine specimens will be required to confirm the utility of APOA1 in urine as a bladder cancer biomarker. Correlation between Urine APOA1/APOA2 Concentrations and Urine Blood Cells Test. Since the candidate markers described above are also presented in serum, parallel trace 5810

LgEs vs (HgEs + HgAs)

(NT vs LgEs) p-value for early detection

Journal of Proteome Research • Vol. 9, No. 11, 2010

hematuria measurements in urine samples may be one way to overcome the serum contamination problem during biomarker validation. The urine red blood cells (RBC) test, which is the most common way to measure trace hematuria degree in urine, measures the number of red blood cells in a urine sample. The RBC analysis provides strong evidence that the APOA1 or APOA2 levels in urine were not correlated with the degree of hematuria in urine (n ) 102, Supplemental Figure 8, Supporting Information). The elevated urine APOA1/APOAA2 levels could not be predicted by the RBC test in urine.

Discussion In this study, we examined the reproducibility of an LCbased urine proteomic platform that couples multidimensional

research articles

Discovery of Novel Bladder Cancer Biomarkers

Figure 5. ELISA-based detection of APOA1 in individual urine samples and its efficacy in detecting bladder cancer. (Upper panel) Folddifferences between NT, LgEs, HgEs, HsAs, and UTI/HU were detected in desalted/concentrated urine protein (A, n ) 104) and dilute urine specimens (B, n ) 151) of individual samples. The absolute quantitation values are labeled above each box plot. The horizontal lines in each box plot denote the 10th, 25th, 50th, 75th and 90th percentiles of the data distribution. (Middle and lower panel) ROC curve analyses of the ability of APOA1 to discriminate all bladder cancer subgroups from NT (middle) or the LgEs subgroup from NT (lower).

Table 5. Summary of Statistical Analyses of the Efficacy of APOA1 Urine Levels, Measured by ELISA, in Detecting Bladder Cancera Absolute concentration of urine APOA1 measured by ELISA using the appropriate amount of concentrated urine protein or 100 µL of diluted urine from each individual

APOA1 (ng/µg) APOA1 (ng/mL)

NT (N ) 37) 0.1 ( 0.2 NT (N ) 50) 5.0 ( 14.2

LgEs (N ) 14) 1.2 ( 1.8 LgEs (N ) 21) 624.1 ( 1433.9

HgEs (N ) 24) 2.1 ( 2.6 HgEs (N ) 39) 1631.0 ( 7362.0

HgAs (N ) 11) 3.2 ( 3.5 HgAs (N ) 16) 920.9 ( 1347.3

UTI/Hematuria (N ) 18) 0.7 ( 0.8 UTI/Hematuria (N ) 25) 269.7 ( 532.3

p-values, AUC values of individual ROC curves, sensitivity and specificity (NT vs all BC)

protein name

p-value for diagnosis

APOA1