Isolation and Identification of Potential Urinary Microparticle Biomarkers of Bladder Cancer David M. Smalley,†,‡,§ Nicholas E. Sheman,‡,| Kristina Nelson,‡,| and Dan Theodorescu*,†,§,⊥ Mellon Medical Biomarker Discovery Laboratory, W. M. Keck Laboratory for Mass Spectrometry, Departments of Urology, Microbiology, and Molecular Physiology and Biological Physics, University of Virginia, Charlottesville, Virginia 22908 Received November 21, 2007
Bladder cancer leads to approximately 13 000 deaths annually in the United States. Early disease is often treated with minimal morbidity and has good prognosis, while the opposite is true for advanced disease. Currently, no tools exist for early detection of this cancer. Microparticles are small, subcellular particles released by essentially all cells upon activation and are known to be produced constitutively by cancer cells. Since most bladder cancers originate in the urothelial cells lining the lumen of the organ, we hypothesize that these cells will release microparticles into the urine. The goal of this study was to identify potential biomarkers in the urinary microparticles of individuals with bladder cancer. Urine microparticles from five healthy individuals and four individuals with bladder cancer were isolated. Samples were delipidated by PAGE and trypsin-digested, peptides were extracted, and the proteome was examined by LC-MS/MS using a Thermo Finnigan LTQ and LTQ-FT ion trap mass spectrometer. Protein identification was determined by SEQUEST and relative quantitation was assessed by comparing spectral counts. Eight proteins were elevated in the microparticles from individuals with bladder cancer. They include five proteins associated with the epidermal growth factor receptor pathway, the alpha subunit of GsGTP binding protein, resistin, and retinoic acid-induced protein 3. Further studies will be needed to validate these potential biomarkers. Keywords: Proteomics • Biomarker Discovery • Exosomes • Ectosomes • Epidermal Growth Factor Pathways • Microparticles • Microvesicles • Bladder Cancer • Urine
Introduction In the United States, urinary bladder cancer is the fourth most common cause of cancer in men and ninth most common cancer in women.1 There are an estimated 61 000 new cases resulting in approximately 13 000 deaths annually.1 If detected when tumor is confined to the mucosa, treatment is associated with a 5-year survival rate of ∼95%.2 On the other hand, invasive and metastatic tumors are associated with a 5-year survival rate of only ∼16%.2 While treatment for bladder cancer has improved over the last several decades, its diagnosis has progressed at a much slower rate. Cystoscopy is still considered the best method to diagnose this malignancy, but is invasive and uncomfortable, and about 10% of all lesions are not detected.3 In addition, the cost associated with this procedure, especially related to monitoring reoccurrence, is the major reason why costs-per* To whom correspondence should be addressed: Dan Theodorescu, MD, Ph.D., P.O. Box 800422, University of Virginia Health Sciences Center, Charlottesville, VA 22908. Tel.: 434-924-0042. Fax: 434-982-3652. E-mail:
[email protected]. † Mellon Medical Biomarker Discovery Laboratory, University of Virginia. ‡ W. M. Keck Laboratory for Mass Spectrometry, University of Virginia. § Department of Urology, University of Virginia. | Department of Microbiology, University of Virginia. ⊥ Department of Molecular Physiology and Biological Physics, University of Virginia.
2088 Journal of Proteome Research 2008, 7, 2088–2096 Published on Web 03/29/2008
patient expenses for bladder cancer are among the highest of all cancers.4 These factors have led to extensive efforts to develop better biomarkers for this disease. However, to date, none have sufficient specificity and sensitivity to monitor the general population or replace cystoscopy in monitoring reoccurrence.5 One of the most significant obstacles to mass spectrometry based biomarker discovery is detecting proteins present at low levels when other proteins are present at many orders of magnitude higher levels. The most common approach to circumvent this problem is to remove the 6-10 most abundant proteins by affinity chromatography. Then, the sample, normally plasma or urine, is fractionated by various analytical techniques and each fraction is analyzed. In contrast, our approach is to select a subproteome of the sample that has a high likelihood of being informative for disease detection and progression. In the plasma, we have focused our efforts on microparticles, small membrane enclosed subcellular particles less than 1.5 µm.6,7 While a majority of these are ectosomes, arising from the outward blebbing of the plasma membrane, there are also exosomes, which are small membrane vesicles of endocytic origin that are secreted from many cells. In our recent report, we found evidence for both types of microparticles in the plasma.6 10.1021/pr700775x CCC: $40.75
2008 American Chemical Society
Urinary Microparticles in Bladder Cancer Table 1. Clinicopathological Data of the Study Samples patient
age
sex
grade
description
1 2 3 4 5 6
61 55 77 57 59 60
M M M M M F
1/2 3 3 1/2 N/A 3
urothelial papillary carcinoma urothelial papillary carcinoma urothelial carcinoma papillary urothelial carcinoma urothelial carcinoma transitional cell carcinoma
Parallel studies were performed isolating microparticles from urine by Pisitikun and co-workers.8–10 In their first report, they developed methods to isolate and analyze the proteome of these urinary microparticles and determined that a majority of these were exosomes based on the uniformity in size and the orientation of certain plasma membrane proteins.8 In addition, it appears that they originate from the kidney distal tubules. In their next study, they examined the sample storage and processing requirements.9 More recently, they compared the urine exosome proteome in both an animal model of acute kidney injury and from individuals with acute kidney injury and found a potential biomarker for this aliment.10 In the current study, we isolate and examine urine microparticles in an attempt to discovery potential biomarkers for individuals with urothelial bladder cancer. First, we examine the analytical reproducibility of mass spectrometric analysis using a single sample. Next, the technical variation of the sample preparation method which includes isolation of the microparticles, delipidation by PAGE, trypsin digestion and extraction from the gel, is examined. Then we compare the person-to-person variation of the protein composition of urinary microparticles. Finally, we compare the protein composition of microparticles isolated from the urine of individuals with bladder cancer to that of healthy controls.
Experimental Procedures Samples. Random urine was collected from healthy controls and individuals with bladder cancer directly into urine cups containing 4.35 µg of PMSF. Clinical characteristics of individuals with bladder cancer are provided in Table 1. Hematuria was determined using a chemstrip 9 dipstick (Roche Diagnostics, Somerville, NJ) and only urine samples which were negative for hemoglobin were processed except where noted. Typically, 30-50 mL of urine was used for each urine microparticle preparation. After mixing, the urine was transferred into tubes, centrifuged for 250g for 10 min. and supernatants were stored at –80 °C. Human blood was obtained from healthy controls by venipuncture as previously described.7 In one set of experiments, exogenous blood (0, 0.2, 1.0, 5.0, 20, and 100 µL of blood) was mixed with 15 mL of recently collected urine and incubated for 4 h at 37 °C prior to addition of any protease inhibitors. Then, PMSF was added and the samples were processed as described above. Isolation of Urine Microparticles. Urine microparticles were isolated as described8 for urine exosomes with the following modifications. Urine was partially thawed, additional inhibitors were added (33.4 µL of 100 mM NaN3, 50 µL of 10 mM PMSF, and 1 µL of 1 mM leupeptin per mL of urine), and samples were extensively vortexed. Upon complete thawing, the sample was centrifuged at 250g for 10 min and then at 17 000g for 30 min. The supernatant was centrifuged at 200 000g for 60 min in a total of two tubes and the microparticle pellet was
research articles resuspeneded in a total of 6 mL prior to its final centrifugation. These urine microparticle pellets were frozen at –80 °C. SDS-PAGE Electrophoresis, Peptide Extraction, and Western Blotting. Isolated urine microparticles were resuspended in a minimal volume of PBS (phosphate buffered saline, pH 7.4) and a small aliquot was taken for protein analysis using the Bio-Rad Protein Assay (Bio-Rad Laboratories, Hercules, CA). An aliquot of the sample was mixed with one-sixth the volume of SDS-PAGE loading buffer (0.5 M Tris, pH 6.8, 10% SDS, 38% glycerol, and 0.1% bromophenol blue), heated to 95 °C for 5 min, allowed to cool to room temperature, and centrifuged for 2 min at 14 000 rpm prior to loading onto the gel. For examination by silver stain, equivalent amounts of protein (1-5 µg) was applied per lane and the gels were stained as described.11 For MS analysis, 10 µg of sample was applied per lane. Microparticle proteins were electrophoresed approximately 1 cm into a 7.5% acrylamide SDS-PAGE using a Minigel system (Bio-Rad, Hercules, CA) at 150 V. The acrylamide gel section containing the proteins was cut out and placed in fixative (50% methanol, 12% acetic acid, and 0.05% formalin) for 2 h. The in-gel tryptic digestion of the lanes and the peptide extraction were performed as described.6 The extracted peptide solutions were lyophilized and reconstituted to 20 µL with 0.1% acetic acid for mass spectrometry analysis. Western analysis was preformed as previously described6 with the following modification. Only 2 µg of urine MPs were applied per lane. Primary antibodies against epidermal growth factor receptor kinase substrate 8-like protein 2 (EPS8L2, Abnova Corporation, Taiwan) and mucin-4 (MUC4, Abnova Corporation) were used. Liquid Chromatography/Mass Spectrometry (LC/MS), and Protein Identification. Samples (2 µg of protein digest) were loaded onto a 360 µm o.d. × 75 µm i.d. microcapillary fused silica precolumn packed with irregular 5–20 µm C18 resin. After sample loading, the precolumn was washed with 0.1% acetic acid for 15 min to remove any buffer salts or gel contaminants. The precolumn was then connected to a 360 µm o.d. × 50 µm i.d. analytical column packed with regular 5 µm C18 resin constructed with an integrated electrospray emitter tip. For preliminary studies examining run-to-run, preparation variation, and the effects of blood on the urine microparticle proteomes, the samples were gradient-eluted at a flow rate of 60 nL/min with an 1100 series binary HPLC solvent delivery system (Agilent, Palo Alto, CA) directly through an electrospray ionization source interfaced to an LTQ ion trap mass spectrometer (Thermo Electron Corp, San Jose, CA). The HPLC gradient used 100% A at 0 min, 5% B at 5 min, 50% B at 220 min, 100% B at 240 min, and 100% A at 280 min (solvent A ) 0.1 M acetic acid, solvent B ) 70% acetonitrile in 0.1 M acetic acid). The LTQ mass spectrometer was operated in the datadependent mode in which first an initial MS scan recorded the mass to charge (m/z) ratios of ions over the mass range 300–2000 Da, and then the 10 most abundant ions were automatically selected for subsequent collision-activated dissociation. All MS/MS data were searched against a human protein database downloaded from the European Bioinformatics Institute (www. ebi.ac.uk) on April 18, 2007 using the SEQUEST program (Thermo Electron Corp.). For unlabeled peptides, a static modification of 57.02150 Da for cysteine residues, variable modifications of 15.9949 and 14.01550 for methionine and cysteine, respectively, and a mass tolerance of 0.5 Da were used. Peptide identifications were made based on fully tryptic peptides, using a first-pass filtering of standard criteria as previously described,12 including cross-correlation Journal of Proteome Research • Vol. 7, No. 5, 2008 2089
research articles
Smalley et al.
values g2.0 (+1 charge), 2.2 (+2 charge) and 3.3 (+3 charge). Protein assignments required at least 2 MS/MS spectra that passed the above criteria. Manual validation of at least one MS/ MS spectrum-peptide sequence match per protein was performed for all proteins that were determined to be differentially expressed. For studies focusing on the person-to-person variation, and analysis of urine microparticles from individuals with bladder cancer, analysis was performed identical to above with the following modifications. The LC-MS system consisted of a Surveryer binary HPLC solvent delivery system (Agilent, Palo Alto, CA) directly through an electrospray ionization source interfaced to a Thermo Electron LTQ-FT ion trap mass spectrometer (Thermo Electron Corp, San Jose, CA). The parent mass tolerance was set to 10 ppm. The two-sample t test was performed to examine the significant difference in protein spectral counts between urine microparticles from healthy controls and from individuals with bladder cancer. P < 0.05 was used to determine differentially expressed proteins.
Results Reproducibility of Spectral Count Analysis of Urine Microparticle Samples. Spectral count analysis is one method used to estimate the relative abundance of a protein in complex protein mixture using label-free liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis. This method sums the total number of MS/MS spectra over certain quality criteria for peptides of a given protein during a LC-MS/MS experiment and has good linear correlation with the protein abundance.13 To determine if the variation in the results of this approach is due to variability in sample chromatography and/or peak selection for fragmentation by the mass spectrometer, we analyzed one preparation of urine microparticles from a healthy individual in triplicate and compared the spectral counts. Three analyses of the peptides of a tryptic digest of proteins from a urine microparticle preparation identified 141 proteins with a minimum spectral count of 2. Eighty-three percent (83%) of the proteins were identified in at least 2 of the 3 analyses and 56% of the proteins were identified in three (Figure 1A). Therefore, if a peptide is detected in one analysis, it has a good likelihood of being detected in a second if the overall spectral count was over two. Technical Variation in the Preparation of Urinary Microparticles. To determine if the variation in a protein’s detection is due to small differences in the preparation of urine microparticles, pooled urine from 4 healthy individuals was separated into six equal fractions and urine microparticles were isolated from these aliquots. The amount of protein isolated from each aliquot was consistent (7.56 ( 0.79 µg). A fraction of these isolated microparticles was separated by PAGE and visualized by silver staining. No significant difference in the protein composition was evident by this method (Figure 1B). Another fraction of four of the samples was delipidated by PAGE and trypsin-digested, and the peptides were extracted and analyzed twice by LC-MS/MS. One hundred and seventy proteins were detected with an overall spectral count of at least 2. If a protein was detected with a spectral count of at least 2 in any give sample, it had 90.8% chance of being detected in each of the other urine microparticle preparation (Table 2). If the protein was detected with a spectral count of at least 4, it had a 97.8%. These results suggest that the microparticle preparations are highly reproducible. In addition, it indicates that if a given protein is detected with a spectral count difference of 4 or greater in one sample (2 LC-MS/MS runs 2090
Journal of Proteome Research • Vol. 7, No. 5, 2008
Figure 1. Technical and biological reproducibility of sample preparation and MS analysis. (A) Reproducibility of MS analysis. Peptides from the trypsin digests of proteins from human urine microparticles from one healthy individual, analyzed in triplicate by LC-MS/MS. A total of 141 were identified with a total spectral count of two or more. Venn diagram of the number of proteins identified in each of the three LC-MS/MS analyses. (B) Technical reproducibility of urine microparticle preparations. Urine from four healthy individuals was pooled and divided into six equal aliquots. Urine microparticles were isolated from each aliquot; the proteins were separated by PAGE and visualized by silver staining. (C) Individual-to-individual variation in the urine microparticles proteome. Urine microparticles were isolated from six healthy individual. The proteins were separated by PAGE and visualized by silver staining. While a majority of the proteins appear similar, there are noticeable differences, some of which are denoted by arrows.
each) and not detected in another, it is likely to be qualitatively different in the two samples. Person-to-Person Variability in the Urinary Microparticle Proteome. Urine samples were obtained from five healthy individuals. Urine microparticles were isolated from each sample and examined by 1D PAGE and silver staining. Noticeable differences were found between the protein profiles of these samples (Figure 1C). The microparticles were delipidated, trypsin-digested, and analyzed in duplicate by LC-MS/MS. Comparable number of spectra identifying proteins were identified for each sample (2006 ( 638, see Supplemental Table 1 in Supporting Information). Percentage of total spectral counts for the 20 most abundant proteins are shown in Figure
research articles
Urinary Microparticles in Bladder Cancer Table 2. Reproducibility of Spectral Count Analysis of Proteins Isolated from Four Different Urine Microparticle Preparationsa For Peptides Identified by 2 or More MS/MS in Sample:
Percentage of peptides also identified in Sample:
A B C D
A
B
100 93 94 98
93 100 95 97 Aver
C
90 89 100 95
D
83 88 77 100
90.85
For Peptides Identified by 4 or More Ms/Ms in Sample:
Percentage of peptides also identified in Sample:
A B C D
A
B
100 97 100 98
100 100 98 98 Aver
C
100 97 100 100
D
93 96 95 100
97.85
a Urine from four healthy individual was pooled and then divided into aliquots, and microparticles were isolated from each fraction. Proteins from four of these fractions were delipidated by PAGE and digested using trypsin, and the peptides were extracted from the gel and analyzed twice by LC-MS/MS. If a given protein was detected in a given sample with a spectral count of at least 2 (above) and 4 (below), the likelihood that it will be detected in the other samples is high. For example, of the proteins detected in sample A (with a spectral count of 2 or more), 93% of these proteins were detected in sample B, 94% of these proteins were detected in sample C, and 98% of these proteins were detected in sample D.
2. There appears to be dramatic differences in protein content between these samples. For example, from one healthy control, uromodulin accounted for 45% of the total spectral count, while it was
