Comparing SILAC and Two-Dimensional Gel Electrophoresis Image Analysis for Profiling Urokinase Plasminogen Activator Signaling in Ovarian Cancer Cells Pauliina M. Uitto,† Braddon K. Lance,‡ Graham R. Wood,‡ James Sherman,§ Mark S. Baker,†,§ and Mark P. Molloy*,†,§ Department of Chemistry and Biomolecular Sciences, Macquarie University, 2109, Sydney, NSW, Australia, Department of Statistics, Macquarie University, 2109, Sydney, NSW, Australia, and Australian Proteome Analysis Facility Ltd., Macquarie University, 2109, Sydney, NSW, Australia Received November 30, 2006
Two-dimensional gel electrophoresis (2-DE) image analysis is conventionally used for comparative proteomics. However, there are a number of technical difficulties associated with 2-DE protein separation that limit the depth of proteome coverage, and the image analysis steps are typically labor-intensive and low-throughput. Recently, mass spectrometry-based quantitation strategies have been described as alternative differential proteome analysis techniques. In this study, we investigated changes in protein expression using an ovarian cancer cell line, OVMZ6, 24 h post-stimulation with the relatively weak agonist, urokinase-type plasminogen activator (uPA). Quantitative protein profiles were obtained by MALDI-TOF/TOF from stable isotope-labeled cells in culture (SILAC), and these results were compared to the quantitative ratios obtained using 2-DE gel image analysis. MALDI-TOF/TOF mass spectrometry showed that differential quantitation using SILAC was highly reproducible (∼8% coefficient of variation (CV)), and this variance was considerably lower than that achieved using automated 2-DE image analysis strategies (CV ∼25%). Both techniques revealed subtle alterations in cellular protein expression following uPA stimulation. However, due to the lower variances associated with the SILAC technique, smaller changes in expression of uPA-inducible proteins could be found with greater certainty. Keywords: metabolic labeling • SILAC • two-dimensional gel electrophoresis • image analysis • MALDI-TOF/TOF • urokinase plasminogen activator • protein quantitation
Introduction Quantitative measurements of protein expression or abundance have traditionally been carried out using two-dimensional gel electrophoresis (2-DE) image analysis strategies. Several studies have shown that quantitative reproducibility of 2-DE as measured by coefficient of variation (CV) falls within a range of 20-30%.1-3 While the validity of quantitation using 2-D gels is well-documented and in widespread use, a serious drawback is that all computer-assisted gel image analysis packages rely heavily on manual user intervention to achieve accurate difference measurements. Unfortunately, this is a costly, labor-intensive task. The recent introduction of sample multiplexing strategies such as difference in-gel electrophoresis (DIGE),4-6 which permits samples to be mixed and separated on a single 2-DE gel, has significantly shortened image analysis time and decreased quantitative variability that is usually * Corresponding author. Dr. Mark P. Molloy, Australian Proteome Analysis Facility, Level 4, Building F7B, Research Park Drive, Macquarie University, Sydney, NSW 2109, Australia. Phone, +61 2 9850 6218; fax, +61 2 9850 6200; e-mail,
[email protected]. † Department of Chemistry and Biomolecular Sciences, Macquarie University. ‡ Department of Statistics, Macquarie University. § Australian Proteome Analysis Facility Ltd.,, Macquarie University. 10.1021/pr060638v CCC: $37.00
2007 American Chemical Society
observed run-to-run. Nonetheless, this approach is not without drawbacks, as reagent costs are high and concerns remain regarding differential CyDye binding and detection sensitivity when minimal labeling is performed. Furthermore, the DIGE approach offers no advantage in correctly assigning absolute abundance to comigrating proteins. One technique that is gaining prominence in the proteomic community is the metabolic labeling strategy known as SILAC (stable isotope labeling by amino acids in cell culture).7 The technique is applicable to all cell culture systems where defined media allows amino acid substitution with “heavy” variants. This is an elegant approach where the label is incorporated directly into proteins as they are synthesized. Several applications of this technique have been described and include comparative quantitative proteomics,8,9 protein turnover studies,10 comparative quantitation of phosphoproteins,11-13 and protein-protein interaction studies.14-16 In principle, any essential amino acid can be used as a stable isotope-labeled precursor for SILAC experiments. As most proteomic studies utilize trypsin for proteolytic cleavage, lysine, arginine, and leucine are the most useful amino acids for labeling. Leucine is useful as it is a highly abundant amino acid contained in most tryptic peptides. Because stable isotope Journal of Proteome Research 2007, 6, 2105-2112
2105
Published on Web 05/02/2007
research articles labeling facilitates multiplexing, quantitative differences attributed to run-to-run technical variation are eliminated. For comparative quantitative experiments the practical implication of lower variance is that smaller, statistically significant differences in protein abundance can be obtained from fewer replicates, thereby increasing sample throughput. The mortality of ovarian cancer is closely associated with the ability of cancer cells from the primary lesion to spread (metastasize) to biologically important sites distant to the organ of origin. The serine protease urokinase-type plasminogen activator (uPA) and its receptor (uPAR) have been implicated in the complicated cellular and molecular interplay underlying extracellular matrix degradation, growth factor activation, and signal transduction, all physiological phenomena thought to play key steps in cancer invasion. For example, when uPA binds to uPAR, it activates surface-bound plasminogen into the ECM degrading protease plasmin, which has broad substrate specificity similar to trypsin. uPAR is linked to the cell membrane via a glycosyl-phosphatidylinositol (GPI) anchor, an attachment that is unable to transmit signals because the attachment is devoid of any transmembrane or cytoplasmic “tail” domain interaction directly with signaling proteins found in cytoplasm.17 Intriguingly, it has been reported that binding of uPA to uPAR results in the activation of certain intracellular signaling cascades in various cell types (e.g., breast cancer and fibrosarcoma cells).18-21 One explanation gaining credence in the literature is that uPAR-signaling occurs after the formation of a multi-protein complex that is assembled upon ligand (uPA) binding and that this complex induces multifaceted signaling responses.22,23 In this study, we conducted a comparative proteomic experiment investigating uPA signaling in OVMZ6 ovarian cancer cells at very low (10 nM) physiological levels of agonist. We undertook this challenging experiment as a benchmarking study to compare protein quantitation techniques that employed SILAC and MALDI-TOF/TOF and compared these to the more traditional 2-DE gel based image analysis approach. We demonstrated that quantitation using SILAC and mass spectrometry produced considerably lower quantitative variation compared to traditional 2-DE image analysis strategies, allowing confident detection of subtle changes in cellular protein expression.
Materials and Methods Reagents. Leucine-deficient DMEM was custom-made by SAFC Biosciences (Brooklyn, Australia). 5,5,5-deuterated Lleucine (Leu-d3) was from Cambridge Isotope Laboratories (MA). L-Leucine (Leu-d0) was from Sigma. Dialyzed fetal bovine serum was from Invitrogen. Gel electrophoresis reagents and IPG strips were all from Bio-Rad. Sypro-Ruby was from Molecular Probes. Sequencing grade porcine trypsin was obtained from Promega. R-Cyano-4- hydroxycinnamic acid was from Fluka. C18 tips were purchased from Eppendorf. Acetonitrile and trifluoroacetic acid (TFA) were of HPLC grade. pERK1/2 polyclonal antibody and the HRP-linked anti-rabbit IgG antibody were from Cell Signaling, and chemiluminescent reagent kit ECL Plus was from GE Healthcare. Cell Culture Conditions and Stable Isotope Labeling. OVMZ6 cells were a kind gift from Dr. Ute Reuning (Technical University, Munich, Germany). Cells were grown in DMEM media containing 10% dialyzed FBS, 10 mM HEPES, and 1% penicillin/streptomycin and supplemented with either 100 mg/L of Leu-d0 or Leu-d3. The metabolic label was incorporated 2106
Journal of Proteome Research • Vol. 6, No. 6, 2007
Uitto et al.
to >99% as judged by mass spectrometry of proteins from cultures grown for 9 days in Leu-d3 media prior to stimulation. To cells of 70% confluency, 10 nM single-chain uPA (a gift from Prof. Douglas Cines, Department of Pathology and Laboratory Medicine, University of Pennsylvania, School of Medicine, Philadelphia, PA) was added for 15 min to the cells labeled with Leu-d3. Control cells of similar confluency were mock-treated with PBS. The cells were rinsed once with PBS to remove unbound uPA and then grown for 24 h. Cells were rinsed twice with 10 mM Tris-HCl, pH 7.4, and 250 mM sucrose prior to collection in 0.5 mL of 2-DE rehydration buffer (7 M urea, 2 M thiourea, 40 mM Tris, 4% (w/v) CHAPS, and 65 mM DTT). Cell lysates were sonicated three times for 10 s and cleared by centrifugation at 14 000 rpm for 10 min at 20 °C. Protein concentration was determined using a Bradford assay. Western Blot Analysis for pERK. Proteins were separated on 10-20% SDS-PAGE gels (Bio-Rad) under reducing conditions and transferred onto a nitrocellulose membrane. The membrane was blocked in 5% skim milk powder and incubated with pMAPK 42/44 antibody at 1:1000 dilution overnight at 4 °C as instructed by manufacturer. HRP-linked secondary antibody was used at 1:2000 dilution and proteins were visualized using chemiluminescence. Two-Dimensional Gel Electrophoresis and Image Analysis. For SILAC experiments, equal quantities of protein (150 µg) from control (Leu-d0) and uPA-stimulated (Leu-d3) cells were mixed and diluted to 200 µL with rehydration buffer (7 M urea, 2 M thiourea, 65 mM DTT, 4% (w/v) CHAPS, 1% (v/v) Biolytes pH 3-10, and a trace of Bromophenol Blue). Triplicate 11 cm pH 4-7 IPG ReadyStrips were focused for 50 kVh. IPGs were equilibrated, reduced, and alkylated (6 M urea, 2% (w/v) SDS, 375 mM Tris, 20% (v/v) glycerol, 5 mM TBP, and 2.5% (w/v) acrylamide) and separated on 10-20% SDS-PAGE Criterion gels (Bio-Rad). For image analysis, the light and heavy samples were run individually in triplicate. Gels were stained with Coomassie Blue G250 for mass spectrometry or with Sypro-Ruby for image analysis. Image analysis was conducted with Progenesis PG240 software (NonLinear Dynamics, Manchester, U.K.) using automated spot detection. Mass Spectrometry for Protein Identification and Quantitation. In total, 150 randomly selected protein spots were excised from the 2D-E gels containing mixed samples of control and uPA-stimulated cells, destained, and digested overnight at 37 °C with 0.1 µg of trypsin/50 mM NH4HCO3, pH 7.8. Peptides were desalted and concentrated using C18 media and deposited onto a 4700 MALDI-TOF/TOF (Applied Biosystems, Foster City, CA) target plate using 8 mg/mL R-cyano-4- hydroxycinnamic acid. A peptide mass fingerprint (PMF) spectrum (m/z range 750-4000 Da) and eight MS/MS spectra (70-3500 Da) were recorded for each digest. Mass spectra were converted into a .PKL file and submitted to Mascot (Matrix Sciences). Spectra were queried against the NCBI human database to obtain protein identifications. Searches were conducted with a mass tolerance of 0.05 Da. Protein identification was regarded as reliable when the Mowse score calculated by Mascot was g60. In addition, the number of leucines detected in the PMF scan was useful for confirmation of the identity of peptides that produced poor MS/MS spectra. The isotope cluster area (ICA) of the precursor ion was measured for each peptide using Data Explorer 4.5 (Applied Biosystems), allowing the Leu-d0/Leu-d3 ratios to be calculated manually. Statistical Analysis. For the image analysis data, the coefficient of variation (standard deviation/mean) of spot volume
Comparing SILAC and 2-DE for Profiling uPA Signaling
Figure 1. uPA phosphorylates ERK1/2 (10 nM) in OVMZ6 cells. Stimulation reaches maximum at 10 min post stimulation. FBS stimulation (10%) for 10 min served as a positive control. The experiment was repeated three times.
was calculated separately for each protein within the control and treatment groups, resulting in two CVs for each protein. For the SILAC data, the coefficient of variation was calculated using the ratios of isotopic cluster area of light and heavy forms of individual peptides, resulting in a single CV of expression ratio for each protein. The significance of differential expression for the SILAC and image analysis data was tested using a paired t test.
Results and Discussion Experimental Design and Identification of Proteins. Several different analytical approaches and platforms are used for conducting quantitative proteomic profiling. We have evaluated and compared quantitative profiling data obtained using mass spectrometry with SILAC, to conventional 2-DE image analysis strategies. These approaches were used to quantitate protein changes in an ovarian cancer cell line, OVMZ6, after stimulation with the weak agonist, single chain zymogen urokinase-type plasminogen activator (uPA). uPA (twin chain, active) is classically known for its serine proteolytic activity to produce plasmin which promotes extracellular matrix degradation,24 but recently, the role of uPA in regulating cell signaling through uPA receptor binding has emerged.25-27
research articles For metabolic labeling experiments 5,5,5-deuterated leucine (Leu-d3) replaced L-leucine in culture medium, effectively increasing the mass of each leucine residue by 3 Da compared to control cells cultured in media containing unlabeled Lleucine (Leu-d0). Cells were grown until >99% of the label was incorporated into proteins as determined by mass spectrometry (data not shown). Use of the 4700 MALDI-TOF/TOF which generates singly charged parent ions, provides sufficiently high resolution to distinguish isotope distributions of peptides containing a single Leu-d0 or Leu-d3 amino acid. Leu-d3-labeled cells were stimulated with uPA for 15 min, washed, and cultured for a further 24 h to allow sufficient time for new protein synthesis and protein turnover. Stimulated cells were mixed 1:1 with control cells (Leu-d0-labeled) and separated on triplicate 2-D gels. uPA-induced cell signaling was confirmed by immunoblot for phospho-ERK1/2 (Figure 1). In total, 150 protein spots were randomly selected from each of the triplicate 2-DE gels, excised, digested, and analyzed by MALDI-TOF/TOF to provide protein identification and quantitation. In total, 105 protein spots were identified from either two or three replicate gels. Normalization Strategies for Mass Spectrometry Based Quantitation. Examination of our raw data showed that only 33% of the Leu-d0/Leu-d3 ratios were centered around 1 ((5%), and we attribute this error to the common problem of inaccuracies during sample mixing. We investigated three approaches to normalize the data: (1) single-point normalization using the ratio obtained experimentally for β-actin, (2) threepoint normalization using the mean ratio of a set of three selected ‘house-keeping proteins’, and (3) multi-point normalization using the mean of all 105 observed ratios (see Figure 2). Single-point normalization relies on the assumption that, under the stimulation conditions, β-actin does not change in expression, or changes only minimally to give a Leu-d0/Leu-d3 ratio of 1. β-Actin is commonly used for normalization of
Figure 2. Three different normalization strategies were applied to the data set: (A) ratios of raw data, (B) β-actin was used for the single-point normalization, (C) 78 kDa glucose-regulated protein, R-enolase, and annexin 1 were used for three-point normalization, and (D) all datapoints were used for multi-point normalization. Journal of Proteome Research • Vol. 6, No. 6, 2007 2107
research articles
Uitto et al.
Table 1. Proteins Identified and Quantitated Using 2-D Gel Image Analysis and SILACa
accession number
protein
Q5T937 P09525 P05413 P61978 P26447 P32119 O60664
Thioredoxin Annexin A4 Fatty acid binding protein 3, heart Heterogeneous nuclear ribonucleoprotein K S100 calcium-binding protein A4 Peroxiredoxin 2 Mannose 6 phosphate receptor binding protein 1 Q9H2Y2 Inositol 1-phosphate synthase P23381 Tryptophanyl-tRNA synthetase Q53G38 Keratin, type I cytoskeletal 18 Q14697 Neutral alpha-glucosidase AB [Precursor] P30101 Protein disulfide-isomerase A3 [Precursor] P07741 Adenine phosphoribosyltransferase P30040 Endoplasmic reticulum protein ERp29 [Precursor] NP_006434 Putative c-Myc responsive isoform 1 P63104 14-3-3 protein zeta/delta Q15102 Platelet-activating factor acetylhydrolase IB gamma subunit Q14554 Protein disulfide-isomerase A5 [Precursor] Q9ULG2 KIAA1258 protein [Fragment] P28066 Proteasome subunit alpha 5 Q13177 Serine/threonine-protein kinase PAK2 Q2TB58 Huntingtin interactin protein 1 Q5SP17 Heat shock 70 kDa protein 1A P48643 T-complex protein 1, epsilon subunit P06733 Alpha Enolase P14625 Endoplasmin [Precursor] Q6IPD2 Ribosomal protein SA, 67 kDa P27797 Calreticulin [precursor] P61158 Actin-like protein 3 Q5T5V2 Peroxiredoxin 3 Q9UMX0 Ubiquilin P62258 14-3-3 protein epsilon Q9Y2T3 Guanine deaminase Q9SJO6 Nudix hydrolase motif 5 P23526 Adenosylhomocysteinase P30041 Peroxiredoxin 6 Q53YD7 Eukaryotic translation elongation factor 1 gamma P10809 60 kDa heat shock protein, mitochondrial [Precursor] Q5TA03 Glutathione S-transferase omega 1 P60900 Proteasome subunit alpha type 6 P37235 Hippocalcin-like protein 1 P11021 78 kDa glucose-regulated protein [Precursor] P35232 Prohibitin P36957 Dihydrolipoyllysine-residue succinyltransferase component of 2-oxoglutarate dehydrogenase complex, mitochondrial [Precursor] Q969T9 WW domain-binding protein 2 P61088 Ubiquitin-conjugating enzyme E2 N P43487 Ran-specific GTPase-activating protein P60709 Beta actin P12004 Proliferating cell nuclear antigen Q02790 FK506- binding protein 4 P12429 Annexin A3
fold fold t test change no. of Mascot change peptides CV% SILAC image score SILACb SILAC (d0/d3) analysis usedc
CV 2D (%) (d0)
CV 2D (%) (d3)
t test image t test SILAC: analysis image (d0/d3) analysis
215 253 199 246 184 400 269
0.79 0.81 0.81 0.84 0.84 0.85 0.86
6 7 9 10 12 8 7
6% 7% 8% 10% 4% 3% 9%
0.021 0.009 0.016 0.002 0.036 0.017 0.005
0.91 0.87 0.84 0.92 0.36 0.95 0.97
37% 38% 41% 18% 8% 30% 38%
13% 39% 14% 24% 5% 36% 19%
0.765 0.545 0.339 0.745 0.003 0.906 0.939
0.493 0.503 0.611 0.554 0.001 0.475 0.547
209 342 470 230 245 320 140
0.86 0.87 0.87 0.87 0.87 0.87 0.88
10 10 12 11 8 8 10
7% 13% 4% 12% 11% 9% 13%
0.013 0.007 0.008 0.009 0.030 0.041 0.005
1.44 0.83 1.69 0.90 0.78 1.07 1.06
16% 25% 24% 19% 15% 33% 29%
5% 15% 19% 9% 26% 25% 22%
0.101 0.228 0.169 0.123 0.255 0.856 0.798
0.046 0.588 0.138 0.210 0.994 0.375 0.486
105 400 206
0.88 0.88 0.89
7 8 13
6% 4% 14%
0.002 0.033 0.004
0.83 1.13 1.48
37% 25% 10% 20% 16% 24%
0.645 0.257 0.111
0.746 0.183 0.195
357 263 344 294 199 464 316 280 250 555 394 365 154 167 357 462 353 312 490 356
0.89 0.89 0.90 0.91 0.91 0.93 0.93 0.94 0.94 0.94 0.94 0.94 0.94 0.94 0.94 0.94 0.94 0.95 0.95 0.95
11 6 7 7 6 12 10 10 5 13 10 11 11 5 10 10 12 13 13 10
6% 16% 10% 8% 6% 21% 12% 6% 4% 10% 5% 11% 7% 10% 5% 8% 15% 8% 4% 9%
0.011 0.031 0.017 0.008 0.017 0.027 0.029 0.002 0.004 0.006 0.011 0.012 0.031 0.035 0.043 0.045 0.047 0.004 0.010 0.012
0.80 1.10 1.23 0.92 0.83 1.00 1.31 0.93 1.17 1.02 1.09 1.03 1.15 0.86 1.02 1.77 1.14 0.96 1.20 0.58
31% 30% 27% 17% 30% 26% 10% 30% 11% 11% 28% 16% 36% 31% 33% 21% 41% 29% 30% 36%
25% 15% 14% 18% 36% 10% 13% 18% 47% 19% 21% 21% 40% 20% 23% 20% 13% 19% 10% 16%
0.488 0.591 0.412 0.523 0.664 0.981 0.004 0.825 0.621 0.861 0.482 0.841 0.686 0.607 0.963 0.170 0.729 0.511 0.458 0.022
0.736 0.376 0.260 0.546 0.765 0.235 0.017 0.708 0.286 0.136 0.088 0.370 0.595 0.956 0.540 0.198 0.432 0.490 0.346 0.056
414
0.95
10
9%
0.031
0.99
36% 22%
0.984
0.500
258 497 167 661 325 395
0.96 0.97 0.97 0.97 0.98 0.98
8 12 8 12 11 10
7% 7% 8% 9% 8% 10%
0.030 0.009 0.038 0.033 0.001 0.043
1.12 0.89 1.02 0.96 1.40 0.79
33% 42% 37% 31% 25% 23%
34% 8% 31% 29% 32% 26%
0.661 0.721 0.969 0.889 0.394 0.467
0.490 0.939 0.542 0.685 0.249 0.177
193 197 195 389 292 161 308
0.99 0.99 1.01 1.01 1.02 1.03 1.04
9 10 10 7 13 5 9
7% 7% 5% 4% 5% 9% 4%
0.018 0.029 0.004 0.013 0.009 0.035 0.040
1.60 1.23 1.24 1.45 1.49 1.01 1.27
28% 30% 37% 30% 27% 34% 33%
44% 20% 25% 18% 21% 14% 31%
0.332 0.535 0.623 0.327 0.234 0.968 0.475
0.272 0.342 0.352 0.306 0.229 0.762 0.409
a Only proteins that have p-values
