Identification of Candidate Biomarker Proteins Released by Human Endometrial and Cervical Cancer Cells Using Two-Dimensional Liquid Chromatography/Tandem Mass Spectrometry Hongyan Li,†,‡ Leroi V. DeSouza,†,‡,§ Shaun Ghanny,†,‡ Wei Li,†,‡ Alexander D. Romaschin,|,⊥ Terence J. Colgan,⊥,O and K. W. Michael Siu*,†,‡,§ Department of Biology, Centre for Research in Mass Spectrometry, and Department of Chemistry, York University, 4700 Keele Street, Toronto, Ontario, Canada M3J 1P3, Division of Clinical Biochemistry, St. Michael’s Hospital, 30 Bond Street, Toronto, Ontario, Canada M5B 1W8, Department of Laboratory Medicine and Pathobiology, University of Toronto, Toronto, Ontario, Canada M5G 1L5, and Pathology and Laboratory Medicine, Mount Sinai Hospital, 600 University Avenue, Toronto, Ontario, Canada M5G 1X5 Received February 14, 2007
Candidate biomarker proteins, including chaperonin 10 and pyruvate kinase, previously discovered and identified using mass-tagging reagents with multidimensional liquid chromatography and tandem mass spectrometry (DeSouza, L.; et al. J. Proteome Res. 2005, 4, 377-386) have been identified in serum-free media of cultured endometrial cancer (KLE and HEC-1-A) and cervical cancer (HeLa) cells. These and other cancer-associated proteins were released by the cultured cells within 24 h of growth. A total of 203 proteins from the KLE cells, 86 from HEC-1-A, and 161 from HeLa are reported. Keywords: Endometrial cancer • Cell released proteins • Candidate biomarkers • Two-dimensional liquid chromatography/tandem mass spectrometry
Introduction Cells release proteins to its extracellular space by multiple means, including secretion, ectodomain shedding of proteins, and shedding of membrane-derived vesicles.1,2 Many of these proteins represent main classes of bioactive molecules, including growth factors,3,4 cytokines,5 proteases, protease inhibitors,6,7 transmembrane receptors, and cell adhesion molecules.8,9 The dynamic change and interaction of these proteins with the extracellular-matrix (ECM) molecules constitutes a microenvironment for maintaining cell growth and tissue development in a controlled way.10 Over the past decade, focus on the tumor microenvironment has not only led to a better understanding of tumorigenesis, but even brought an evolution in our thinking of cancer development.11 Tumors are now considered as complex tissues containing transformed cells and co-evolving “normal” neighboring cell types.12 There is increasing evidence for the significant roles played by cell-released proteins, for example, heparin-binding epidermal growth factor-like growth factor,13 macrophage inhibitory cytokine-1,14 and matrix metalloprotease,15 in communications between diseased cells and their surrounding microenvironment, and * To whom correspondence should be addressed: Prof K. W. Michael Siu, Department of Chemistry, York University, 4700 Keele Street, Toronto, Ontario, Canada M2J 1P3. Tel: (416) 650-8021; Fax: (416) 7365936; E-mail:
[email protected]. † Department of Biology, York University. ‡ Centre for Research in Mass Spectrometry, York University. § Department of Chemistry, York University. | St. Michael’s Hospital. ⊥ University of Toronto. O Mount Sinai Hospital. 10.1021/pr0700798 CCC: $37.00
2007 American Chemical Society
consequently on the final pathology. Studies on tumor cellreleased proteins have also shed light on the discovery of biomarkers for early detection of cancer and intervention.11,16 Proteins that are secreted, or shed from cell membranes, have been identified in blood or other bodily fluids of patients afflicted with a variety of cancers. As sampling of bodily fluids is relatively straightforward and is typically of minimal invasiveness; these proteins are potentially useful as biomarkers, if the same proteins are not released (or released to the same extent) from healthy cells. Indeed, blood- or bodily fluid-based assays are the preferred methods for disease diagnosis and prognosis, the blood test of prostate-specific antigen for prostate cancer being the prime example.17 Endometrial cancer (EmCa) is the fourth most-common malignancy in Canadian women with a 2% lifetime risk. At present, there are no biomarkers available for diagnostic testing of this disease. Women with symptoms typical of EmCa, perimenopausal, postmenopausal, or abnormal uterine bleeding and/or discharge, must undergo an invasive procedure: endometrial biopsy, curettage, and/or hysteroscopy to provide a sample for pathological analysis. Availability of blood-borne EmCa biomarkers would significantly enhance the ability to diagnose the disease, perhaps in an earlier stage, and the effectiveness of treatment. Our earlier work performed with tissue homogenates from EmCa patients, using mass-tagging reagents iTRAQ and cICAT with multidimensional liquid chromatography (LC) and tandem mass spectrometry (MS/MS), resulted in a panel of candidate biomarkers.18 A number of these differentially expressed proteins have recently been verified in an iTRAQ-labeling study involving 40 samples.19 In Journal of Proteome Research 2007, 6, 2615-2622
2615
Published on Web 05/25/2007
research articles addition, the most sensitive and specific of these candidate biomarkers have been independently validated on a second cohort comprising 148 patients, using immunohistochemistry in a tissue-microarray format.20 Here, we report results of an investigation into whether any of the candidate biomarkers discovered and identified in these previous studies can potentially be secreted or shed into the extracellular environment. We expect that proteins isolated from the serum-free media of cultured endometrial cancer and HeLa cell lines would be a first indication of their possible presence in actual bodily fluids.
Materials and Methods Cell Culture and Collection of Culture Media. KLE, HEC1-A, and HeLa cells were obtained from American Type Culture Collection. All cell types are of uterine and epithelial origin: KLE cells were derived from poorly differentiated endometrial carcinoma, HEC-1-A cells from moderately well-differentiated endometrial carcinoma, while HeLa cells were from cervical cancer. KLE and HeLa cells were grown in Dulbecco’s modified Eagle’s medium, while HEC-1-A cells were grown in McCoy’s 5A medium (Wisent Inc.); both media were supplemented with 10% fetal bovine serum (FBS, HyClone) and 1 unit/mL penicillin-streptomycin (Invitrogen, Inc.). Cells were all grown under 37 °C and in a humidified atmosphere with 5% CO2. For conditioned-medium collection, cells were grown to 60-80% confluence in 100 mm dish (SARSTEDT). The culture medium was aspirated, and the plates were rinsed four times with phosphate-buffered saline (Sigma) and once with the corresponding growth medium without FBS. This medium was collected for 0-h control. The cells were then incubated in the appropriate medium free of FBS for 24 h. At the end of the incubation time, the serum-free medium was collected and filtered through a 0.2-µm nylon filter (SARSTEDT) to remove any suspended cells. The medium was then frozen immediately and stored at -80 °C until further processing. Media were collected from 100-mm plates (a total of 50) of each cell type for analysis. Culture-Medium Protein Preparation. Proteins in the culture medium were isolated using 0.02% sodium deoxycholate (Sigma) and 10% trichloroacetic acid (Sigma). Following 2-h precipitation on ice, the samples were centrifuged for 30 min at 11 000g and washed twice with ice-cold acetone. The precipitated proteins were resuspended in 50 mM ammonium bicarbonate. Protein concentrations were determined using the Bradford assay (Bio-Rad). Proteins were digested with trypsin using a modification of a literature method.21 Briefly, the resuspended protein samples were heated to 60 °C for 1 h in the presence of 5 mM dithiothreitol. The samples were allowed to cool to room temperature and then alkylated by incubation with 10 mM iodoacetamide for 1 h in the dark. Sequencinggrade trypsin (Promega, Madison, WI) at 1:20 (w/w) in 50 mM ammonium bicarbonate was then added, and the samples were incubated at 37 °C overnight. The digested samples were then dried in a speed vacuum and resuspended in 10 µL of 0.1% formic acid. LC-MS/MS Analysis. Samples were analyzed by online twodimensional LC-MS/MS. The nanobore LC system and MS/ MS setup used for these analyses have been described previously.18,19 Briefly, the liquid chromatograph was an LC Packings Ultimate (Amsterdam, The Netherlands), and the mass spectrometer was a QSTAR Pulsar-i hybrid quadrupole/time-offlight (TOF) instrument (Applied Biosystems/MDS SCIEX, 2616
Journal of Proteome Research • Vol. 6, No. 7, 2007
Li et al.
Foster City, CA). The tryptic peptides were first separated in the first dimension using a strong cation exchange (SCX) column (LC Packings: BioX-SCX cartridge, 500 µm × 15 mm). One microliter of sample was loaded onto the SCX column and was eluted in 10 fractions using 10-µL solutions of increasing ammonium acetate concentration (10, 50, 100, 150, 200, 250, 300, 350, and 500 mM and 1 M) directly onto a C18 reversephase precolumn (LC Packings: 300 µm × 5 mm) for subsequent reverse-phase chromatography. Separation was effected by a nonlinear binary gradient: eluent A consisting of 94.9% deionized water, 5.0% acetonitrile, and 0.1% formic acid (pH ≈ 3); and eluent B consisting of 5.0% deionized water, 94.9% acetonitrile, and 0.1% formic acid. During the first 5 min of the LC run, eluent A at a flow rate of 50 µL min-1 was used to load peptides salted out from the SCX column onto the C18 precolumn, after which the SCX column was switched out of line. Desalting continued for 2 additional min. At the seventh minute, the C18 precolumn was switched inline with the reverse-phase analytical column (75 µm × 150 mm packed in house with 3-µm Kromasil C18 beads with 100 Å pores, The Nest Group); separation was performed at 200 nL min-1 using a 90-min binary gradient shown below. Note that the “0” timepoint corresponds to the beginning of elution from the SCX onto the C18 precolumn; the actual time at which the precolumn was brought inline with the analytical column was at the seventh minute.
MS data were acquired in information-dependent acquisition (IDA) mode with Analyst QS 1.1 and Bioanalyst Extension 1.1 software (Applied Biosystems/MDS SCIEX). MS cycles comprised a TOF MS survey scan with an m/z range of 400-1500 Th for 1 s, followed by five product-ion scans with an m/z range of 80-2000 Th for 2 s each. The collision energy (CE) was automatically controlled by the IDA CE Parameters script. Switching criteria were set to ions with m/z g400 and e1500 Th, charge states of 2-4, and abundances of g10 counts. Former target ions were excluded for 30 s, and ions within a 6-Th window were ignored. Additionally, the IDA Extensions II script was set to “no repetition” before dynamic exclusion and to select a precursor ion nearest to a threshold of 10 counts on every fourth cycle. Database Searching and Criteria. LC-MS/MS data were searched using an in-house version of Mascot (Matrix Science, U.K.) against an NCBI nr database (downloaded June 1, 2006 with 3 682 060 sequences) with the taxonomy selected for mammals (which contained 446 729 sequences), and tolerances set for 0.3 Da for peptide matches and 0.2 Da for MS/MS fragment matches. A second search was performed using ProteinPilot software (Applied Biosystems, Foster City, CA) and a Celera human protein database (CDS KBMS 20041109) containing 178 239 protein sequences to verify the results obtained with Mascot. The cutoff for significance used for this search was set for a score of 1.3, which corresponds to a confidence score of 95. These peptides identified were compared with those reported by the Mascot search in order to verify the identifications. Bioinformatics. Identified proteins were analyzed for secreted protein features using the Signal Peptide Predictor (http://www.cbs.dtu.dk/services/SignalP)22 and non-classical and leaderless protein secretion (http://www.cbs.dtu.dk/
Biomarkers from Human Endometrial and Cervical Cancer Cells
research articles
Figure 1. Subcellular locations of proteins released by the endometrial and cervical cancer cell lines. Percentages of the proteins are given. ER, endoplasmic reticulum; ECM, extracellular matrix.
services/SecretomeP-2.0/).23 Signal Peptide Predictor incorporates neural network and hidden Markov model algorithms to detect signal peptides from input protein sequences. SecretomeP utilizes a neural network combining six protein features to predict whether a protein sequence undergoes non-classical secretion or not; these features are the number of atoms, number of positively charged residues, presence of transmembrane helices, presence of low-complexity regions, presence of pro-peptides, and subcellular localization. The Gene Ontology Consortium tool was applied for subcellular and functional annotation analysis.24
Results and Discussion Cell-Culture Condition Optimization. Culture cells are typically grown in serum-supplemented media; however, highabundance serum proteins can interfere with the subsequent detection of secreted proteins by mass spectrometry.25 To circumvent this interference, serum-free media were used in this study. The drawbacks, however, were that cell growth in such media is typically slower and the rate of cell death higher, thus, increasing the chance of cell autolysis and nonspecific release of cytoplasmic proteins. For these reasons, we monitored closely the viability and the death rate of the cells in the serum-free media. To seed the serum-free media with cells of high viability, we grew cells in serum-containing media up to a certain confluence and then switched to serum-free media: HeLa and HEC-1-A cells up to 60% and KLE 75% confluence, before they were washed and inoculated onto the serum-free media. After 24 h, all cell lines reached about 85% confluence, and the media were collected and processed for released
proteins. Harvesting the media after 24 h minimizes the extent of cell stress and autolysis. The strategy of using serum-free medium for the final growth stage and minimizing the duration of this stage is similar to that in studies published within the last year.26-28 Identification of Proteins Released by the Three Cancer Cell Lines. For every cell line, we analyzed the proteomic profiles of the serum-free media for the 0-h and 24-h incubations. Bovine proteins appearing in both profiles, including bovine albumin and γ-globin, were considered as media proteins and were removed from the lists. This resulted in lists of nonredundant proteins totaling 160 for the HeLa, 198 for the KLE, and 87 for the HEC-1-A cell lines. These are given in Supplementary Tables 1s-3s in Supporting Information. Sixtyfive percent of the proteins were identified with two or more peptides. The MS/MS spectra of proteins identified with single peptides were verified by manual inspection and were accepted only when a series of a minimum of four b- or y-type ions was matched with a Mascot score >35. For these proteins, the sequences of the peptides identified are also given in Supplementary Tables 1s-3s in Supporting Information. The extent of overlap between the peptides identified in the Mascot and those identified in the ProteinPilot search is summarized in Supplementary Table 4s in Supporting Information. The number of unique peptides reported with ProteinPilot that are not observed with Mascot appears to be high as the former includes non-tryptic variations/degradation products of tryptic peptides. Thirty-five percent of the identified proteins contain a predicted signal-peptide sequence. As proteins can be exported into the extracellular matrix without a classical N-terminal Journal of Proteome Research • Vol. 6, No. 7, 2007 2617
research articles
Li et al.
Figure 3. Overlaps of proteins identified in the culture media of KLE, HEC-1-A, and HeLa cells.
Figure 2. (a) Molecular weight distributions and (b) isoelectrical point distributions of identified proteins released by the endometrial and cervical cancer cells.
signal peptide, we analyzed the identified proteins also for prediction of nonclassical and leaderless secretion. This revealed that 59% of the proteins could be secreted extracellularly by nonclassical secretory pathways. Categorizing the proteins according to their subcellular locations using GoMiner tool (Figure 1) showed that on average 27% of identified proteins were known to be secretory or extracellular, and 13% were membrane-associated. These percentages compare favorably with those reported in other studies on secretory proteins,26-28 and are consistent with our expectation of locating secreted and released proteins in the cell-culture media. The identified proteins of the three investigated cell lines showed similar molecular-weight distributions. Fifty percent of the proteins are 20-40 kDa (Figure 2a). Eighty-six percent of the identified proteins from the HEC-1-A and HeLa cells had molecular weights 100 kDa) identified from the KLE cells, resulting in 70% of the identified proteins having molecular weights
