Identification of Serum Biomarkers for Colorectal Cancer Metastasis Using a Differential Secretome Approach Hua Xue,†,‡ Bingjian Lu ¨ ,‡,§ Jun Zhang,| Minliang Wu,⊥ Qiong Huang,† Qiang Wu,# Hongqiang Sheng,† Dongdong Wu,† Jianwen Hu,∇ and Maode Lai*,† Department of Pathology & Pathophysiology, Department of Surgical Pathology, Affiliated Women’s Hospital, Department of Clinical Pathology, Affiliated Sir Run Run Shaw Hospital, Department of Clinical Pathology, the Second Affiliated Hospital, School of Medicine, Zhejiang University, Hangzhou, China, Department of Pathology, Anhui Medical University, Hefei, China, and Research Centre for Proteome Analysis, Institute of Biochemistry and Cell Biology, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences, Shanghai, China Received September 30, 2009
Lymph node metastasis is the major concern that causes death in colorectal cancers. However, biomarkers for cancer metastasis are still lacking. In this study, we applied an LC-MS/MS-based labelfree quantitative proteomics approach to compare the differential secretome of a primary cell line SW480 and its lymph node metastatic cell line SW620 from the same colorectal cancer patient. We identified a total of 910 proteins from the conditioned media and 145 differential proteins between SW480 and SW620 (>1.5-fold change). The differential expression pattern of 6 candidate proteins was validated by Western blot analysis. Among them, trefoil factor 3 and growth/differentiation factor 15, two upregulated proteins in SW620, were further analyzed in a large cohort of clinical tissue and serum samples. Sandwich ELISA assay showed that the serum levels of both proteins were significantly higher in lymph node metastatic colorectal cancers. Receiver operating characteristic curve analysis confirmed that serum trefoil factor 3 and growth/differentiation factor 15 could provide a discriminatory diagnostic test for predicting colorectal cancer metastasis. Immunohistochemical analysis also showed that the overexpression of trefoil factor 3 or growth/differentiation factor 15 in colorectal cancer was associated with lymph node metastatic behavior. This study showed an accurate, sensitive, and robust label-free quantitation approach for differential analysis of cancer secretome. The comparison of the cancer secretome in vitro is a feasible strategy to obtain valuable biomarkers for potential clinical application. Both trefoil factor 3 and growth/differentiation factor 15 could serve as potential biomarkers for the prediction of colorectal cancer metastasis. Keywords: biomarker • colorectal cancer • label-free quantitation • metastasis • secretome
Introduction Colorectal cancer (CRC) is one of the leading causes of cancer-related deaths worldwide. In 2008, 150 000 new CRC cases and over 50 000 deaths from CRC were estimated in the United States.1 Surgical resection is the mainstay of CRC treatment. Despite considerable refinement in therapeutic modalities, almost half of the CRC patients after “curative” surgery develop further diseases within 5 yearssmostly metastatic lesions, which inevitably cause fatalities in the future.2-4 * To whom correspondence should be addressed. Maode Lai, Professor, Department of Pathology & Pathophysiology, School of Medicine, Zhejiang University, 188 Yuhangtang Road, Hangzhou, China 310058. E-mail: lmp@ zju.edu.cn. Tel: +86 571-88208197. Fax: +86 571-88208198. † Department of Pathology & Pathophysiology. ‡ These authors contributed equally to this work. § Department of Surgical Pathology, Affiliated Women’s Hospital. | Department of Clinical Pathology, Affiliated Sir Run Run Shaw Hospital. ⊥ Department of Clinical Pathology, the Second Affiliated Hospital. # Department of Pathology, Anhui Medical University. ∇ Shanghai Institutes for Biological Sciences. 10.1021/pr9008817
2010 American Chemical Society
Cancer-specific biomarkers play crucial roles in cancer detection, prediction and intervention. Over the past several decades, enormous efforts have been made to characterize useful biomarkers for CRC.5-8 However, there are no valuable biomarkers for CRC metastasis. Therefore, it is a pressing need to find novel biomarkers that predict the metastatic potential of CRC and serve as prognostic indicators and intervention targets in the future. Cancer proteomics has been showing its increasing importance in biomarker discovery. However, the common cancertissue-based approach is less optimal in cancer detection because many proteins are not necessarily detectable in serum or plasma. Essentially, a highly desirable biomarker for cancer screening and monitoring should be measurable in the body fluid samples. Accordingly, it is plausible to identify cancer biomarkers directly from the blood proteome, but in fact it is currently frustrating. The major technical obstacle comes from the fact that the abundant blood proteins, such as albumin immunoglobulin, fibrinogen, transferrin and lipoproteins, inJournal of Proteome Research 2010, 9, 545–555 545 Published on Web 11/19/2009
research articles
Xue et al.
Figure 1. Outline of experimental workflow.
evitably mask the less abundant proteins, which are usually potential biomarkers.9,10 Great efforts have been made to remove these abundant proteins before proteomic analysis;11-13 however, most procedures are currently far from perfect due to the inherent low screening efficiency that is associated with nonspecific binding. Hence, scientists and biologists turn to seek other approaches for cancer proteomics. The term “secretome” was first referred to secreted proteins of Bacillus subtilis in a whole genome manner;14 now, in a broader sense, it is to harbor proteins released by a cell, tissue or organism through various mechanisms including classical secretion, nonclassical secretion, and secretion via exosomes.15 More recently, cancer secretome, all proteins released by cancer cells, has been attracting wide attention. These proteins play an important role in many essential physiological and pathophysiological processes, such as cell growth and differentiation, invasion, and metastasis via an endocrine, paracrine or autocrine way.16 More importantly, these cancer secreted proteins or their fragments always enter into body fluids, such as blood or urine, and are probably measurable via noninvasive assays. Thus, cancer secretome emerged to be a promising and reliable source of cancer biomarkers. Initial studies on cancer secretome have successfully identified a rich set of potential biomarkers for cancer detection.17-19 Quantifying changes in protein abundance between samples appear to be a major issue for differential secretome analysis, especially when the high-throughput shot gun liquid chromatography-mass spectrometry (LC-MS) proteomics has become the main technology.20,21 Isotope-labeling methods, such as isotope coded affinity tag (ICAT) and stable isotope labeling by amino acids in cell culture (SILAC), have been commonly used in cancer secretome analysis.17,22 However, these labelbased approaches are costly, time-consuming, and not always feasible as limited by the available tags insufficiently for the simultaneous discrimination of multiple samples.23 Label-free quantitation methods, based on the measurements of mass spectral peak intensities or spectral counts, are devoid of these defects and provide a convenient and reliable way in differential proteomics studies although it was rarely applied in secretome analysis previously.24-26 We herein reported a metastasis-related differential secretome analysis between two classic CRC cell lines from the same 546
Journal of Proteome Research • Vol. 9, No. 1, 2010
patient, the primary SW480 and the lymph metastatic SW620. The workflow and experimental design performed in this study are outlined in Figure 1. We identified 145 metastasis-related differential proteins from CRC secretome by using a label-free quantitative shotgun proteomics strategy. Western-blot analysis validated the differential expression pattern of 6 proteins, including trefoil factor 3 (TFF3), growth/differentiation factor 15 (GDF15), anterior gradient homologue 2 (AGR2), proteinglutamine gamma-glutamyltransferase 2 (TGM2), Lipocalin-2/ neutrophil gelatinase-associated lipocalin (LCN2), and insulinlike growth factor-binding protein 7 (IGFBP7). ELISA and immunohistochemistry demonstrated the potential diagnostic value of TFF3 and GDF15 in predicting CRC metastasis. This study not only presents an accurate, sensitive and robust approach for cancer secretome analysis but also provides potential biomarkers for CRC metastasis prediction.
Materials and Methods Cell Culture. The CRC cell lines SW480 and SW620 were purchased from the American Type Culture Collection (Manassas, VA, USA). Both cell lines were from the same patient. SW480 was derived from the primary site of CRC and SW620 from the recurrent lymph node metastasis. They were maintained in phenol red-free Dulbecco’s Modified Eagle’s Medium (Gibco, Gaithersburg, MD) supplemented with 10% FCS (Gibco) at 37 °C in 5% (v/v) CO2 air atmosphere. For harvesting conditioned media (CM), SW480 and SW620 cells were grown as described above but in the serum free media (SFM). Protein Extraction from the Conditioned Media and the Cell Lysates. SW480 and SW620 cells were grown approximately to 60-70% confluence before being rinsed four times with SFM to remove serum residues in the CM. After 20 h of incubation, the CM containing the secreted proteins was collected and cooled on ice. Floating cells and cellular debris were removed by centrifugation (200 g, 10 min, 4 °C) and the following sterile filtration (pore size: 0.22 µm, Millipore). After the addition of protease inhibitors (Inhibitor cocktail complete, Roche), the CM was concentrated by ultrafiltration using “Amicon Ultra-15” centrifugal filter devices (Millipore) according to the manufacturer’s instructions. Secreted proteins were precipitated by acetone at -20 °C overnight, then collected by centrifugation
A Differential Secretome Analysis (10,000 g, 10 min, 4 °C) and stored at -80 °C. Meanwhile, whole cellular proteins were also isolated by a standard protocol. Briefly, cells on the dishes were washed twice with PBS, and then lysed in buffer consisting of 7 M urea, 2 M thiourea, 4% CHAPS, 65 mM DTT and 0.2% Biolyte (Bio-Rad, Richmond, CA) by sonication on ice. The lysates were centrifuged (10 000g, 1 h, 4 °C) and the supernatants were collected and stored at -80 °C. The protein concentrations were assayed with a standard Bradford protein assay (Bio-Rad, Richmond, CA). For further study, the protein was isolated at least in triplicate from separate cell culture. Trypsin Digestion. Three appropriate volumes of secreted protein aliquots (collected from three separate cell cultures) were combined and dissolved in reducing solution (6 M Urea, 2 M Thiourea, Sigma, St Louis, MO). The supernatant was collected after centrifugation (12 000 g, 30 min, 4 °C). The protein concentration was determined by the Bradford protein assay. 100 µg of protein sample for each fraction was reduced with 10 mM DTT (Sigma, St Louis, MO) at 37 °C for 2.5 h and alkylated with 50 mM iodoacetamide (Sigma, St Louis, MO) at room temperature for 40 min. After diluted in a solution of 50 mM NH4HCO3 (Sigma, St Louis, MO), the protein mixture was digested by sequencing grade modified trypsin (Promega, Madison, WI) using a 1:50 enzyme:protein ratio at 37 °C for 20 h. The tryptic peptide mixture was lyophilized and kept at -80 °C until use. LC-MS/MS. The Ettan MDLC system (GE Healthcare, Piscataway, NJ) was applied for desalting and separation of tryptic peptide mixtures. In this system, samples were desalted on RP trap columns (Zorbax 300 SB C18, Agilent Technologies, Palo Alto, CA), and separated on an RP column (150 µm i.d., 100 mm length, Column Technology Inc., Fremont, CA). Mobile phase A (0.1% formic acid in HPLC-grade water) and the mobile phase B (0.1% formic acid in acetonitrile) were selected. Approximately 20 µg of tryptic peptide mixture was loaded on to the columns and was separated at a flow rate of 2 µL/min by using a linear gradient of 4%-50% B for 120 min. A Finnigan LTQ linear ion trap MS (Thermo Electron, San Jose, CA) equipped with an electrospray interface was connected to the LC setup for eluted peptide detection. Data-dependent MS/ MS spectra were obtained simultaneously. Each scan cycle consisted of one full MS scan in profile mode followed by five MS/MS scans in centroid mode with the following Dynamic Exclusion settings: repeat count 2, repeat duration 30 s, exclusion duration 90 s. Each sample was analyzed in triplicate. Data Analysis and Label-Free Quantitation. MS/MS spectra were automatically searched against the nonredundant International Protein Index (IPI) human protein database (version 3.26, 67 687 entries) using the TurboSEQUEST program in the Bioworks Browser software suite (version 3.1, Thermo Electron, San Jose, CA). The peptides were constrained to be tryptic and up to two missed cleavages were allowed. Carbamidomethylation of cysteines was treated as a fixed modification, whereas oxidation of methionine residues was considered as variable modifications. The mass tolerance allowed for the precursor ions was 3.0 Da and for fragment ions 1.0 Da, respectively. The stringent protein identification criteria were based on Delta Cn (g0.1) and cross-correlation scores (Xcorr, one charge g1.9, two charges g2.2, three charges g3.75). Only proteins identified by at least two unique peptides were reported. BuildSummary, an in-house tool, was used to combine the peptide sequences into proteins and deleted redundant proteins as described by He.27 To determine the false discovery rate (FDR), the data set
research articles was searched against a sequence-reversed decoy IPI human version 3.26 databases using the same search parameters. FDR was calculated as follows: FDR ) Number of false peptides/ (Number of true peptides + Number of false peptides) × 100%. Peptide detection, background subtraction and quantitation were performed on the full scan precursor mass spectra in fully automatic mode using DeCyder MS Differential Analysis Software (version 1.0, GE Healthcare). The relative quantitation analysis consisted of two main procedures. In the first step, the PepDetect module of the software was applied for automated peptide detection, charge state assignments based on resolved isotopic peaks and consistent spacing between consecutive charge states, and quantitation based on MS signal intensities. The second step was the matching of peptides falling within a user-defined mass interval (400-2000 Da) and retention time (120 min) in a quantitative comparison across different signal intensity maps from replicate analyses in the PepMatch module. The intensity distributions for all peptides detected in each sample were employed for normalization. Identification of detected peptides was performed by importing TurboSEQUEST search results back to the Pep-Match module. Default parameters for the software were used throughout the analysis. Bioinformatics Analysis. Cellular localization of identified proteins was further analyzed on the basis of information available from Gene Ontology (GO) (http://www.geneontology. org/) and Human Protein Reference Database (HPRD) (http:// www.hprd.org/). Biological function classifications and signaling pathway analysis were performed with the tools on DAVID Bioinformatics Resources 2008 (http://david.abcc.ncifcrf.gov/) and the Kyoto Encyclopedia of Genes and Genome (KEGG) database (http://www.genome.jp/kegg/pathway.html), respectively. Western Blot Analysis. Secreted or whole cell lytic proteins were dissolved in sample buffer (7 M urea, 2 M thiourea, 4% CHAPS, 0.5%SDS, 65 mM DTT). Equal amounts of protein samples (25-50 µg) were separated on a 10-20% SDS-PAGE gel, transferred to PVDF membranes (Millipore), and then probed with various primary antibodies overnight at 4 °C, followed by fluorescence-labeled secondary antibody (Li-COR, Lincoln, NE) diluted 1:5000 in TBST for 1 h at room temperature. Finally, blots were developed with the Odyssey system (Li-COR, Lincoln, NE). The utilized primary antibodies included rabbit polyclonal anti-TFF3 antibody (catalog number 11810-1AP, Proteintech, Chicago, IL), mouse monoclonal anti-GDF15 antibody (catalog number MAB957, R&D Systems, Minneapolis, MN), rabbit polyclonal anti-AGR2 antibody (catalog number 12275-1-AP, Proteintech, Chicago, IL), mouse monoclonal antiTGM2 antibody (catalog number H00007052-M10, Abnova, Taiwan, China), goat polyclonal anti-LCN2 antibody (catalog number AF1757, R&D Systems, Minneapolis, MN) and mouse anti- IGFBP7 antibody (catalog number MAB1334, R&D Systems, Minneapolis, MN). Serum and Tissue Samples. Preoperative serum samples of 144 sporadic CRC patients (ages 27-88 years; mean 59 years; 67 females, 77 males) were collected from the Second Affiliated Hospital and Sir Run Run Shaw Hospital, Zhejiang University between 2007 and 2008. The serum was obtained by centrifuge at 4 °C in the blood collection tube containing a specific separation gel for biochemical tests (BD Biosciences, San Jose, CA). The serum was kept in -80 °C freezer until use. Additional 69 tissue specimens of sporadic CRC (ages 27-84 years; mean 57 years; 34 female, 35 males) were obtained from the First Journal of Proteome Research • Vol. 9, No. 1, 2010 547
research articles Affiliated Hospital of Anhui Medical University between 2007 and 2008. All patients received no preoperative chemotherapy or radiotherapy. They were diagnosed and staged following the 2000 WHO tumor classification and the 2002 UICC TNM staging system, separately. The control group for ELISA consisted of 156 healthy blood donors (ages 25-81 years; mean 56 years; 71 females, 85 males) from their routine healthy examinations at these hospitals during the same period. The study was approved by the Ethics Board of Biomedicine, Zhejiang University, China. All samples were collected with informed consent. Sandwich ELISA Assay. ELISA plates (Costar; Corning, NY) were coated by adding 0.4 µg of mouse monoclonal anti-TFF3 antibody (catalog number H00007033-M01, Abnova Corporation, Taiwan, China) or 0.2 µg of mouse monoclonal antiGDF15 antibody (catalog number MAB957, R&D Systems, Minneapolis, MN) in 100 µL of 15 mmol/L sodium carbonate, 35 mmol/L sodium bicarbonate, pH 9.6, to each well. The plates were incubated at 4 °C overnight before emptying the wells and washing three times with PBST (0.05% v/v Tween 20 in PBS, pH 7.4). Coated plates were then blocked by incubation with 300 mL of 3% (w/v) BSA in PBST at room temperature for 1 h. TFF3 full-length recombinant protein (catalog number H00007033-P01, Abnova Corporation, Taiwan, China) and GDF15 full-length recombinant protein (catalog number 957GD-025, R&D Systems, Minneapolis, MN) were diluted with 1% (w/v) BSA in PBST to obtain calibrators for the standard curve preparation, and the diluents were used as zero calibrator. After one wash, 100 µL calibrators or human sera (1:10-1:25) in diluents were added and incubated for 2 h at room temperature. After five washes, 40 ng of unconjugated polyclonal rabbit anti-TFF3 antibody (catalog number 11810-1-AP, Proteintech, Chicago, IL) or 20 ng of biotinylated polyclonal goat anti-GDF15 antibody (catalog number BAF940, R&D Systems, Minneapolis, MN) was added to each well, and the plates were incubated for 2 h at room temperature then washed five times. Onehundred microliters per well of Streptavidin HRP (Boster Corporation, Wuhan, China, dilution 1:10 000) or Streptavidin HRP-conjugated goat antirabbit IgG (catalog number ZB-2301, Zhongshan, Beijing, China, dilution 1:5000) was added and incubated for 30 or 60 min at room temperature for GDF15 or TTF3, respectively, then followed by five washes. The color reaction was developed by the addition of 100 µL Substrate Solution (R&D Systems, Minneapolis, MN) to each well at room temperature for approximately 30 min. The reaction was terminated with 100 µL/well of 2 M H2SO4. Absorbance was measured at 450 nm using an automatic plate reader (Spectramax 340 PC, Molecular Devices Corporation, Sunnyvale, CA). Each sample was examined at least in duplicate. Immunohistochemistry. Paraffin-embedded tissue sections (4 µm) were stained with monoclonal mouse anti-TFF3 antibody (catalog number H00007033-M01, Abnova Corporation, Taiwan, China) and polyclonal rabbit anti-GDF15 antibody (catalog number ab14586, Abcam, Cambridge, MA). Briefly, after deparaffinization and dehydration, sections were subjected to antigen-retrieval treatment in a microwave oven at 95 °C for 20 min in 10 mM citrate buffer (pH 6.0). Endogenous peroxidase activity was inactivated with fresh 0.3% hydrogen peroxide in PBS (100 mmol/L; pH 7.4) for 15 min at room temperature. Slides were then incubated with TFF3 (1:300) or GDF15 (1:100) overnight at 4 °C followed by the EnVision detection system (DAKO Corporation, Carpinteria, CA). Diaminobenzidine (DAKO Corporation, Carpinteria, CA) was used as 548
Journal of Proteome Research • Vol. 9, No. 1, 2010
Xue et al. chromogen and slides were counterstained in hematoxylin. For negative controls, the primary antibodies were replaced with PBS. A definitely brown precipitate indicated positive immunoreactivity. Scoring of TFF3 and GDF15 expression was carried out as suggested by Remmele and Stegner28 with minor modification. The positive cells were scored as: 0 for 20-fold changes). In all 145 candidate proteins, the number of peptides used for quantifying each protein varied between 1 and 10. Among these, 41 proteins were comparatively quantified on the basis of change levels of three or more peptides that varied similarly. From these data, it is possible to indicate the confidence of the quantitative approach. As shown in Supplementary Table S2 (Supporting Information), the average coefficient of variation of the fold changes for peptides from these proteins was 21% (range 3.2-48.7%), indicating a reasonable reproducibility of the quantitative data. The current set of differential proteins contained up to 95 proteins that were quantified with only one peptide by the DeCyder software. The large number of these single-peptide quantified proteins might be associated with a bit lower quantitative alterations (60 Site Colon Rectum Histological grading Low-grade High-grade TNM stage I/II III/IV Distant metastasis No Yes a
number
GDF15
level (ng/mL)
p-value
level (pg/mL)
p-value
77 67
1888 ( 1467 1968 ( 1823
0.469
2123 ( 1123 1924 ( 1120
0.250
74 70
2117 ( 1705 1722 ( 1548
0.418
2034 ( 1012 2027 ( 1236
0.973
78 66
2071 ( 1629 1753 ( 1642
0.059
2031 ( 1162 2029 ( 1082
0.896
115 29
1742 ( 1534 2652 ( 1848
0.005
1810 ( 1014 2905 ( 1120
immunostaining score
>p-value
35 34
3 (1-6) 3 (0-6)
0.666
8 (6-12) 8 (6-12)
0.418
36 33
3 (0-6) 4 (3-6)
0.209
8 (6-12) 8 (6-12)
0.670
33 36
4 (3-8) 2 (0-4)
0.007
8 (6-12) 8 (6-12)
0.599
43 26
3 (1-6) 3 (0-8)
0.806
8 (6-12) 8 (6-12)
0.333
55 14
3 (1-6) 2 (0-12)
0.476
8 (6-12) 10 (6-12)
0.322
31 38
3 (0-6) 4 (1-7)
0.157
6 (6-8) 9 (8-12)
