Differential Proteome Expression Associated with Urokinase Plasminogen Activator Receptor (uPAR) Suppression in Malignant Epithelial Cancer Rohit G. Saldanha,*,†,‡,§ Ning Xu,§,| Mark P. Molloy,†,| Duncan A. Veal,⊥ and Mark S. Baker*,†,‡,| Department of Chemistry and Biomolecular Sciences and Australian Proteome Analysis Facility, Macquarie University, Sydney, NSW 2109, Australia, and FLUOROtechnics, Sydney NSW 2109, Australia Received May 14, 2008
Dysregulation of the plasminogen activation cascade is a prototypic feature in many malignant epithelial cancers. Principally, this is thought to occur through activation of overexpressed urokinase plasminogen activator (uPA) concomitant with binding to its high specificity cell surface receptor urokinase plasminogen activator receptor (uPAR). Up-regulation of uPA and uPAR in cancer appears to potentiate the malignant phenotype, either (i) directly by triggering plasmin-mediated degradation or activation of uPA’s or plasmin’s proteolytic targets (e.g., extracellular matrix zymogen proteases or nascent growth factors) or indirectly by simultaneously altering a range of downstream functions including signal transduction pathways (Romer, J.; Nielsen, B. S.; Ploug, M. The urokinase receptor as a potential target in cancer therapy Curr. Pharm. Des. 2004, 10 (19), 235976). Because many malignant epithelial cancers express high levels of uPAR, uPA or other components of the plasminogen activation cascade and because these are often associated with poor prognosis, characterizing how uPAR changes the downstream cellular “proteome” is fundamental to understanding any role in cancer. This study describes a carefully designed proteomic study of the effects of antisense uPAR suppression in a previously studied colon cancer cell line (HCT116). The study utilized replicate 2DE gels and two independent gel image analysis software packages to confidently identify 64 proteins whose expression levels changed (by g2 fold) coincident with a moderate (∼40%) suppression of cell-surface uPAR. Not surprisingly, many of the altered proteins have previously been implicated in the regulation of tumor progression (e.g., p53 tumor suppressor protein and c-myc oncogene protein among many others). In addition, through a combination of proteomics and immunological methods, this study demonstrates that stathmin 1R, a cytoskeletal protein implicated in tumor progression, undergoes a basic isoelectric point shift (pI) following uPAR suppression, suggesting that post-translational modification of stathmin occur secondary to uPAR suppression. Overall, these results shed new light on the molecular mechanisms involved in uPAR signaling and how it may promulgate the malignant phenotype. Keywords: UPAR • two-dimensional gel electrophoresis • mass spectrometry • protein expression • colon cancer • metastasis • STMN1
Introduction Colorectal cancer (CRC) is the fourth most common cancer diagnosed worldwide, representing 9.4 and 10.1% of all incident cancers in men and women, respectively. Its worldwide mortality rate is second only to prostate cancer (in men) and breast * To whom correspondence should be addressed. Professor Mark S. Baker, Director, Australian Proteome Analysis Facility, F7B Research Park Drive, Macquarie University, 2109, Australia, telephone (+61 2 9850 6209), fax (+61 2 9850 6200). E mail:
[email protected]. † Department of Chemistry and Biomolecular Sciences, Macquarie University. ‡ Current Address: Pharmacoproteomics Program, Children’s Cancer Institute Australia for Medical Research, Sydney, NSW 2031, Australia. § Both these authors contributed equally to preparation, execution and submission and are considered joint first authors. | Australian Proteome Analysis Facility, Macquarie University. ⊥ FLUOROtechnics.
4792 Journal of Proteome Research 2008, 7, 4792–4806 Published on Web 09/23/2008
cancer (in women).2 Surgical resection of colorectal cancers appears to be the most effective first-line treatment for the disease with roughly 50% of patients diagnosed with CRC responding favorably to surgery. However, infiltration of the tumor beyond the bowel wall and the presence of nodal and/ or liver metastases, negatively affects survival prognosis dramatically. Differential morphological findings at the invasive tumor margin provide important prognostic information about CRC, secondary to tumor staging. An enhancement in the growth capabilities with tumor budding at the tumor-stromal interface, coupled with uncontrolled expression of elements involved in proteolytic activity, provides invading cancer cells with a potential route to escape its primary confines and migrate to secondary locations.3 Proteolytic activity in this regard has long been thought to play a crucial role in CRC metastasis. Proteolytic activity 10.1021/pr800357h CCC: $40.75
2008 American Chemical Society
research articles
Proteomics of uPAR Suppression in Cancer focused on the surface of penetrating tumor cells drive the breakdown of extracellular matrix barriers and the concomitant release of nascent growth factors facilitating the migration of cancer cells and their survival at secondary sites. Of the various “tumor-associated proteases” studied, plasmin and the various components of the plasminogen activation (PA) cascade stand out as a crucial driver of pericellular proteolysis in malignant epithelial cancers. Plasminogen, urokinase plasminogen activator receptor (uPAR), and urokinase plasminogen activator (uPA) have been localized to the epithelial colonic cancer cell surface and the activation of the zymogen precursor plasminogen to the broad-spectrum serine protease, plasmin on the cell surface is principally brought about by ligation of uPA to its high specificity uPAR. Activated plasmin has extremely broad substrate specificity and is also capable of activating matrix metalloproteases and other proteases/peptidases which further facilitate the degradation of ECM barriers in cancer. Regulation of plasmin activity is achieved by the expression of its naturally occurring serpins like R2 antiplasmin along with plasminogen activation inhibitors types 1 and 2 (PAI-1, PAI-2) which by inhibiting uPA prevent activation of plasminogen and subsequent degradation of matrix scaffolds.3 From a functional perspective, uPAR is a pivotal regulator of plasminogen activation both physiologically and pathologically. This 55-60 kDa heavily glycosylated, disulfide linked cell surface receptor binds both single-chain pro-uPA and active two-chain uPA. Since receptor-bound uPA activates plasminogen more effectively (at least 40-fold higher based on the Km of uPA for its substrate), its up-regulation in cancer reflects the acceleration of plasmin dysregulation and subsequent cancer invasion and metastasis. Several experimental and clinical studies have demonstrated the correlation between elevated levels of uPAR (and uPA) at sites of metastases compared to normal colonic tissues demonstrating the relevance of uPAR and its association with invasion and metastasis. Elevated levels of uPA appear to be an important prognostic indicator of epithelial malignancies, including those of the colorectum.3-9 Similarly, a high uPAR concentration in resected CRC is an independent and significant prognostic factor for 5-year overall survival. For example, Suzuki et al. demonstrated that uPAR expression increases during the transition from adenoma to invasive carcinoma in colonic epithelium. These authors also demonstrated that uPAR levels were significantly elevated in Dukes stage B and C cancer compared to Dukes A cancer.10-14 Clinical studies have demonstrated that the levels of soluble uPAR in serum correlates inversely with survival.15,16 However, these statistical representations of uPAR expression levels in malignant cancers do not shed any relevant information on the underlying etiology that would explain how enhanced uPAR levels influence poor survival outcomes in CRC. Recent attempts to understand the underlying mechanisms that regulate uPAR overexpression in malignant cancers and characterization of downstream cellular effects pertinent to invasion as a consequence of uPAR up-regulation have essentially revealed that uncontrolled proteolytic degradation as a result of plasmin activation is only one facet of uPA/uPAR ability to mediate epithelial cancer invasion and metastasis. Structural studies on the uPA-uPAR interaction have indicated that the semicircular arrangement of the three Ly-6 domains of uPAR (allowing uPA engagement within the cavity) can easily permit the association of this receptor with several other neighboring transmembrane proteins.17 By laterally associating
with neighboring transmembrane proteins and subsequent activation of diverse signal transduction pathways, uPAR is capable of transcriptional regulation of gene and protein expression of several other candidates that are relevant to cancer metastasis. Several in vitro and in vivo colon cancer studies have demonstrated the cellular effects of uPAR expression on cancer metastases. Highly invasive colon cancer cell lines generally display >10 fold more uPAR than their normal or poorly invasive counterparts. Moreover, blocking the expression or function of uPAR in these invasive models using antisense methodologies, RNAi silencing or the cultivation of the cells in the presence of neutralizing antibodies or mimetic peptides against the uPA-uPAR interaction reduced the ability of these invasive cancer cells to invade matrix barriers. For example, Wang and colleagues demonstrated that transfection of the highly metastatic HCT 116 colon cancer cell line with expression vectors containing a 5′ cDNA fragment reduced uPA activity by ∼60% and reduced its ability to induce in vivo pulmonary metastasis in nude athymic mice by about 60%. Interesting, however, is the fact that most of these malignant cancers also display an increased expression of PAI-1 that correlates with poor prognosis and survival outcomes. These observations, coupled with evidence from several in vitro and in vivo studies demonstrating that the pro-tumorigenic functions of the uPA-uPAR cascade occur independent of its proteolytic functions, suggest that uPAR is multifaceted in its cellular functions. To extend our knowledge on how uPAR propagates the malignant phenotype, we employed a comprehensive proteomic approach to characterize global protein expression differences occurring as a result of uPAR suppression in HCT116 versus its 5′ antisense variant cell line. By using a well defined experimental design and the sensitivity of multiply replicated two-dimensional electrophoresis (2D-SDS/PAGE), we proceeded to confidently identify proteins concomitantly upregulated and down-regulated as a result of uPAR suppression using sensitive mass spectrometric identification methods. Identification of candidate proteins that exhibited a change in expression profile as a result of over- or under-expression of uPAR may provide new leads into understanding the molecular mechanisms that underlie uPAR’s influence in CRC metastasis and provide for the identification of novel markers of disease velocity and therapeutic targets.
Experimental Procedures Materials. All cell culture media used in these experiments including fetal bovine serum (FBS) was purchased from Gibco, Invitrogen, Australia. Sterile culture flasks were purchased from Greiner, Frickenhausen, Germany. Polyclonal rabbit antibodies to p53 and stathmin were purchased from Cell Signaling Technology (Danvers, MA). Polyclonal rabbit antihuman profilin and horeseradish conjugated antirabbit secondary antibody was purchased from Sigma (St. Louis, Missouri). Sodium dodecyl sulfate (SDS), dithiothreitol (DTT) tributyl phosphine (TBP), Urea, Tris, glycine, CHAPS, 20% Bio-Lyte ampholyte 3/10, APS, tetramethyethylenediamine (TEMED), acrylamide, and agarose were obtained from Bio-Rad (Hercules, CA). Methanol, ammonium bicarbonate and acetonitrile were purchased from Crown Scientific, Australia. Deep Purple Total Protein Stain and FLUOROPROFILE protein quantitation kits were purchased directly from FLUOROtechnics, Australia. All Journal of Proteome Research • Vol. 7, No. 11, 2008 4793
research articles other reagents used in the experiments were purchased from Sigma-Aldrich, Australia. Cell Lines and Culture Procedure. The parental human colon cancer cell line HCT-116 (herein designated as HCT116WT) and its uPAR deficient variant, created by stable transfection of a vector expressing 5′ uPAR cDNA in an antisense orientation (herein designated as HCT116ASuPAR) were kind gifts from Dr Wang, St. George Hospital, Sydney, Australia.9 Cell monolayers were cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% FBS, 5 mM glutamine, antibiotics (penicillin, 100 µg/mL and streptomycin, 100 µg/mL) and 10 mM HEPES at 37 °C in the presence of 5% CO2. In addition, the HCT116 ASuPAR cells were cultured in the presence of a selection antibiotic, hygromycin (400 µg/mL).18 On achieving confluency, the adherent cells were washed with PBS followed by serum-free DMEM thrice prior to cell detachment using enzyme-free cell dissociation buffer (Sigma-Aldrich, Missouri, MO). The cells were subsequently washed with serum free DMEM after centrifugation at 300g, and the cell pellet was stored at -80 °C until further use. Sample Solubilisation and Two-Dimensional Electrophoresis. For each experiment, HCT 116 (WT or ASuPAR) cells (5 × 107 cells) were lysed using 4 mL of protein rehydration buffer comprising of 40 mM Tris-HCl, 7 M urea, 2 M thiourea, 4% (w/v) CHAPS, 65 mM DTT, 1% (v/v) 3-10 Biolytes, supplemented with protease inhibitors (1 mM AEBSF, 80 µM aprotinin, 2 mM leupeptin, 4 mM bestatin, 1.5 mM pepstatin A, and 1.4 mM E-64). After incubating on ice for 10 min, the cells were sonicated on ice for 30 s (10-s bursts) and centrifuged at 12000g for 10 min, and the supernatant recovered. Protein concentration for each sample was determined using the FLUOROProfile protein quantitation assay kit as per manufacturer’s recommendations. To ensure appropriate power and confidence with all of our experimental designs, we employed an embedded, hierarchical experimental design comprised of four 2-D gels per sample (i.e., to determine analytical variation), three sample replicates per experiment (i.e., to determine biological variation), and two complete repetitive experiments (i.e., to determine clonal experimental variation) respectively. For isoelectric focusing (IEF), 17 cm, pH 4-7 Readystrips (Bio-Rad, Hercules, CA) were rehydrated overnight in 400 µL of protein rehydration buffer containing 500 µg of protein at room temperature. Isoelectric focusing was performed on a MultiPhor II (Amersham Biosciences) at the following voltage gradients: 100 V, 2 h; 300 V, 3 h; 1000 V, 2 h; 2500 V, 1 h; 5000 V, 16 h; and then maintained at 100 V for an additional 5 h for a total of 95 000 V hours. Following IEF, the strips were equilibrated in equilibration buffer (375 mM Tris-HCl, 6 M urea, 2.5% w/v acrylamide, 2% w/v SDS, 20% v/v glycerol, 5 mM TBP, pH 8.8) for 30 min and subsequently loaded onto 18 cm × 20 cm × 0.15 cm, 8-18% gradient SDS-PAGE gels. The SDS/PAGE gels were run at 5 mA/ gel for 8 h followed by 10 mA/gel for 11 h and then 40 mA/gel until the dye front reached the bottom of the gel after which the gels were maintained at 30 mA/gel for 0.5 h. The gels were subsequently stained with Deep Purple gel stain as per manufacturer’s instructions and proteins visualized using a Typhoon 9410 variable mode image scanner (GE Healthcare, Sweden) with the pixel resolution set at 100 µm. Gel Image Analysis. The digitized gel images were subjected to analysis using two independent and commercially available image analysis software packages namely, (PDQuest; version 4794
Journal of Proteome Research • Vol. 7, No. 11, 2008
Saldanha et al. 7.3; Bio-Rad, Hercules, CA, USA and Progenesis Discovery, v2005; Nonlinear Dynamics, Newcastle, UK). Analysis of the images using the two software platforms was performed to identify and exclude any bias in protein detection and quantitation that would otherwise be attributed to differences in the algorithm(s) and/or analytical workflows in the various software packages. Initial processing of gel images using PDQuest was performed using the “Automated spot detection and matching wizard” tool as per the manufacturer’s instructions. After grouping the gels within a specified match set and spot detection, the software automatically generated a “match set master gel” image, which essentially is a synthetic gel image comprising of all the spots that were detected within the match set by PDQuest. The filtered gel images were then grouped into either the HCT 116 WT (18 gel images) or HCT116uPAR AS (16 gel images) subsets for differential expression analysis. Spots were assigned a unique identifier number across all the gels by the software to permit accurate comparison between the two groups. This specifically required the spot to be present in no less than three gels/group failing which spots were excluded from further analysis. The density of each spot was normalized to the total density in the gel image and was expressed in parts per million (PPM). In addition, specific protein spots were “landmarked” as indicators of molecular weight (MW) and isoelectric point (pI). These markers were extrapolated by the software to calculate the purported MW and pI of all protein spots within the master gel. Finally, protein spots that demonstrated at least a 2-fold up or down change (∆g2) in expression and that were statistically significant across all gels (p e 0.05 as determined using a Student’s t test) were further subjected to MS identification. Analysis of the same 2DE gel images was also undertaken using the Progenesis Discovery v2005 software package (NonLinear Dynamics, UK). Briefly, equal numbers of gel images were grouped into two categories (n ) 17 for HCT116WT and n ) 17 for HCT116uPAR AS respectively) for comparative analysis. For each group, a synthetic “average” gel which is a statistical combination of all the gels in that particular group was created. Spot detection, filtering, and background subtraction were automatically performed by the software using its unique nonparametric algorithm and extrapolated to all the subgels in the particular group. Additionally, a reference gel was automatically selected from the set of 34 gels to serve as a universal index for spot numbering prior to matching analysis. Spots that did not match were manually warped and updated at the individual, average, and reference gel levels. Protein spots that demonstrated an expression difference (∆g2; as estimated by Students t test, p e 0.05) between HCT116WT vs HCT116ASuPAR were selected for mass spectrometric identification and analysis. SDS/PAGE-Western Blotting Analysis of Cell Lysates. Whole cell lysates (10 µg) were mixed with NUPAGE LDS sample buffer (4 µL) in the absence of reducing agent, loaded onto a 10% NuPAGE Bis-Tris PAGE gels (Invitrogen, Carlsbad, CA) and electrophoresis performed as per manufacturer’s instructions. The gels were then transferred to nitrocellulose membranes (Bio-Rad) and probed with monoclonal antibodies to p53, stathmin or profilin as specified (1:1000 dilution) overnight and proteins detected with a horseradish peroxide conjugated
research articles
Proteomics of uPAR Suppression in Cancer secondary antibodies (1:2000 dilution; Sigma Aldrich). Visualization was performed using enhanced chemiluminesence. Protein Identification by Mass Spectrometry and Database Searching. Protein spots that were differentially expressed as a result of uPAR suppression were excised using the EXQuest spot excision robot (Bio-Rad, Hercules, CA). After reduction and alkylation, the gel plugs were digested overnight at 37 °C using trypsin (30 µL, 15 ng/µL in 25 mM NH4HCO3, pH 7.8). Peptides were extracted from the gel plugs with 1% formic acid/2% acetonitrile and concentrated using C-18 Zip Tips (Millipore, Bedford MA). The digests were spotted on a MALDI target using R-cyano-4-hydroxycinnamic acid (1 µL, 10 mg/mL in 70% acetonitrile, 0.1% trifluoroacetic acid) as matrix. The peptide mixture was analyzed using the 4700 MALDI-TOF/TOF mass spectrometer (Applied Biosystems, Foster City CA) in the positive MS reflector mode. Spectral data from each sample spot was obtained by 4000 laser shots comprised in 20 subspectral accumulations (each consisting of 200 laser shots per sub spectrum) in a mass range of 750-3500 Da. For each precursor MS scan, the top eight peaks with the best signalto-noise (S/N) ratio were automatically selected for MS/MS analysis. Peaks with a S/N ratio of less than 6, local noise window width (m/z) of below 50 and minimum peak width at full width half-maximum (bins) of 2.9 were excluded to improve the sensitivity of MS/MS scans. The interpretation of MS and MS/MS spectra was carried out using the MASCOT search program (www.matrixscience.com, version 1.8) against the NCBI nr protein database (20060505 as of 10th May, 2006). A tolerance of 100 ppm was used for the precursor mass and 0.3 Da for MS/MS fragments. Carboxyamidomethylation of cysteine, oxidation of methionine and phosphorylation of serine, threonine or tyrosine were chosen as variable modification. Proteins with a Mowse score of greater than 63 were considered significant (p e 0.05). Protein spots that were not identified by MALDI-MS/MS were subsequently analyzed by nano LC-MS/MS using a QSTAR XL mass spectrometer (Applied Biosystems, CA) operated in an information-dependent acquisition mode (IDA). Briefly, the peptide digest (15 µL) was injected onto a Michrom Peptide Captrap (Michrom, Auburn, CA) for preconcentration and desalted with 0.1% formic acid at a flow rate of 10 µL/ min. The precolumn was then switched into line with the analytical column containing C18 RP silica (150 µm × 100 mm, Protocol C18, 3 µm, SGE, Australia). Peptides were eluted from the column using a linear solvent gradient from H2O:CH3CN (95:5, + 0.1% formic acid) to H2O:CH3CN (50:50, + 0.1% formic acid) at 500 nL/ min over a 40 min period. The eluted peptides were subjected to positive ion nanoflow electrospray. In the IDA mode, TOF/MS survey scans were acquired (m/z 400-2000, 1.0s), with the four most abundant multiply charged ions (counts >25) in the survey scan sequentially being subjected to MS/MS analysis. MS/MS spectra were accumulated for 1 s (m/z 50-2000). Peak lists for each LC-MS/MS analysis was generated using a Mascot script plug in (mascot.dll; ABI/MDX Sciex analyst) to the Analyst QS software package (version 1.4). Specific criteria used in the generation of the mascot generic files (.mgf) included the analysis of +2, +3, and +4 precursor ions and the exclusion of spectra with less than 10 peaks. All MS/MS peaks below 1% of the overall intensity were removed; the spectra were centroided and deisotoped prior to submission for database searching to MASCOT. The database searches were conducted on the forward and reversed Swiss-Prot database (May 2006) focusing on the Homo
sapiens taxonomy subset. Database analysis was done using the following parameters: selection of trypsin as the cleavage enzyme with a maximum of 1 missed cleavage allowed, detection of monoisotopic masses, peptide precursor mass tolerance of ( 0.5 Da and the MS/MS product ion tolerance of ( 0.8 Da. Carboxyamidomethylation of cysteine and oxidation of methionine were chosen as the variable modifications.
Results The aim of this study was to identify proteins whose differential expression is inherently linked with that of uPAR in malignant CRC cells and to assemble alterations in downstream signal transduction pathways associated with uPAR suppression. This entailed analysis of the cellular proteomic expression profile in a parental (wild type) and uPAR suppressed (∼40%, Supplementary Figure A1 and A2, Supporting Information) colorectal adenocarcinoma cell line, HCT116. This cell line (>15 years study in our hands) has been found to stably and constitutively expresses high levels of uPA and uPAR and that these correlate with the line’s invasive capabilities both in vitro and in vivo.19 We chose to use a conventional 2DE based approach and employed an embedded, hierarchical experimental design comprised of four gels per sample (to determine analytical variation), three replicates per sample (to determine biological variation) and two repetitive experiments (to determine clonal variation) respectively (Figure 1) and we determined the coefficient of variation (CV) and correlation coefficient (r2) therein. The average CV (as determined by PDQuest analysis) between the gel replicates was 27.9 ( 3.4% with an r2 of 0.93. This result is consistent with several other published reports on the degree of variation attributed to 2DE in the hands of experienced users.20,21 To then determine the degree of variation between identical replicate samples which can sometimes be attributed to variability in the steady state growth of cells, we compared the average CV from the replicates of each flask for HCT116WT and HCT116ASuPAR. We determined that the average CV between replicate flasks of HCT116WT was 29.7 ( 2.5% (r2 0.90) and for HCT116ASuPAR was 31.1 ( 0.5% (r2 0.91). This indicates that the proteomic analyses of sample replicates from a cell line which has been derived from a single clone population is highly consistent and that the biological variation due to steady state growth does not considerably alter observed experimental variation in 2DE gels. Finally, we determined the CV between replicate experiments conducted 6 months apart (Figure 1, WT1 vs WT2 and AS1 vs AS2) and found this to be 35.4 ( 1.4% with an r2 of 0.84. Collectively, these results allowed the determination of the maximum degree of variation (over and above the baseline, experimental variation) permissible for this experiment and subsequent confident and statistically powerful analysis of any proteomic changes attributed to uPAR suppression in the HCT116 cancer cell line. Our results were completely consistent with the degree of biological variation attributed to cell culture experiments using 2DE analyses as reported by many other groups.21,22 Taking into account the statistical relationship between variation and sample size reported previously,20 we determined that to account for statistical errors, our experimental design needed to contain a minimum of two replicate experiments with a minimum sample size of 2 flasks per cell line and 2 gels per flask (total of 8 gels/cell line). To undertake an experiment that exceeded these requirements here, we ran a study that comprised two replicate experiments with 3 flasks per cell line Journal of Proteome Research • Vol. 7, No. 11, 2008 4795
research articles
Saldanha et al.
Figure 1. Experimental design. Graphical representation of the experimental design used in this study. HCT116WT denotes the nontransformed (wild-type) cell line while HCT116 ASuPAR denotes its transformed variant with uPAR suppressed by ∼40%. The experimental hierarchy permits the assessment of variability associated within reproducible experiments (level 1, 2 experiments) due to clonal drifts in cells cultured over long periods of time (∼6 months) or biological variation (level 2, 3 sample replicates) owing to differences in sample growth rates. Experimental variation attributed to gel analysis was assessed by running 4 replicate gels (level 3, 34 gels).
and 4 gels per flask (total of 24 gels/cell line). Gels that were damaged due to fragility or that displayed poor resolution as a result of separation were excluded from analyses, resulting in the inclusion of only the three best gels per flask for image analysis. This allowed us to confidently identify proteins with at least 2-fold change in expression with statistical confidence of 80% power and e0.05 p-value. Proteome Differential Expression between HCT116 WT and HCT116 ASuPAR. To ensure high confidence in the identification of differentially expressed protein spots (due to uPAR suppression), we subjected all 2DE gel images to analyses using two independent, commercially available gel analysis software packages. Analysis of all the gel images using PDQuest entailed extensive manual inspection of all gels and subsequent alignment of overlapping gel spots to the master gel image. PDQuest analysis of HCT116WT gels (n ) 17) identified a total of 826 ( 84 protein spots while the analysis of HCT116 ASuPAR (n ) 16) identified a total of 883 ( 147 protein spots. Using a p value of 0.05
P49321
85.2/4.3
85.2/4.3
6.2
p < 0.01
1.1
p > 0.05
Q9P2 × 0
10.1/5.7
10.1/5.9
4.3
p < 0.01
Out of range
NA
P30041 P15531 Q9UL46
24.9/6.0 20.4/7.1 27.3/5.4
18.0/5.0 18.2/5.8 27.3/5.4
4.1 3.7 3.3
p < 0.01 p < 0.01 p < 0.05
1.7 1.2 1.3
p < 0.01 NA p < 0.05
P05387 P49770
11.7/4.4 39.0/5.8
16.5/ND 34.7/6.1
2.7 2.5
p < 0.01 p < 0.01
1.5 ND
p < 0.05 NA
Q71RC2
80.4/6.2
10.6/5.4
2.4
p < 0.01
1.4
p < 0.05
Q07955
27.6/10.4
29.0/5.4
2.4
p < 0.01
1.4
p > 0.05
Q8TCD5
23.4/6.2
25.0/ND
2.4
p < 0.01
1.4
p > 0.05
P28070
29.2/5.7
24.9/5.7
2.3
p < 0.01
ND
NA
P78318
39.2/5.3
33.9/5.4
2.3
p < 0.05
ND
NA
P49368
60.5/6.1
60.5/6.1
2.2
p < 0.01
1.9
p < 0.01
Q9Y266
38.2/5.3
37.7/5.4
2.1
p < 0.01
ND
NA
P29692
31.0/4.9
32.4/4.9
2.0
p < 0.01
ND
NA
P16949 ???
16.3/5.8 16.3/5.7
17.2/5.7 17.2/5.6
2.0 3.0
p < 0.01 p < 0.01
1.4 1.4
p < 0.05 p < 0.05
Proteins identified as up-regulated by PDQuest alone P38405 44.3/6.2 44.3/6.2 Out of range
p < 0.01
Out of range
p > 0.05
Q14116 Q10713
p < 0.01 p < 0.01
ND 1.1
NA p > 0.05
22.3/4.5 58.2/6.5
18.7/4.3 46.4/6.2
4.1 2.7
P48556
30.0/6.9
28.5/6.4
2.2
p < 0.01
ND
NA
P27797 Q9Y2Z0
48.1/4.29 40.9/5.1
48.1/4.3 31.9/5.3
2.2 2.2
p < 0.01 p < 0.01
1.9 1.6
p < 0.01 p < 0.05
P50453
42.4/5.6
36.4/5.6
2.1
p < 0.01
ND
NA
Journal of Proteome Research • Vol. 7, No. 11, 2008
research articles
Proteomics of uPAR Suppression in Cancer Table 1. Continued Proteins differentially expressed in HCT 116uPAR-ASby g2 fold (p < 0.05) Spot No.
45 46 47 48 49
50
51 52 53 54 55 56 57 58 59 60 61 62 63 64
65
Protein Name
Serine/threonine-protein kinase Nek3 (NEK3) 1,2-dihydroxy-3-keto-5methylthiopentene dioxygenase (MTND) Cytokeratin-18 (KRT18) Heat shock protein 27(HSP27) NADH-ubiquinone oxidoreductase 30 kDa subunit, mitochondrial precursor (NUGM) Dihydrolipoamide dehydrogenase precursor (DLD) Ubiquitin-activating enzyme E1 (UBE1) Transcription elongation factor B, polypeptide 2 (TCEB2) Insulin-degrading enzyme (IDE) Matrin-3 (MATR3) Histone H4 Neurolysin, mitochondrial precursor (NEUL) NADH dehydrogenase (ubiquinone) Fe-S protein 1, 75 kDa, precursor (NDUFS1) D-3-phosphoglycerate dehydrogenase (SERA) L-lactate dehydrogenase B chain (LDHB) 130 kDa leucine-rich protein (LPPRC) Cytokine induced protein 29 kDa (CIP29) spt 62 Cytokine induced protein 29 kDa (CIP29) spt 57 Heat shock protein HSP 90-alpha (HS90A) Nascent polypeptide-associated complex alpha subunit (NACA) Eukaryotic translation elongation factor 1 gamma (EEF1G)
Swiss-Prot ID
b
Theor MW (kDa)/pI-
Fold change (Progenesis)
P-value (Progenesis)
Proteins identified as down-regulated by Progenesis alone P51956 57.7/6.7 18.8/6.1 1.0 p < 0.01
8.9
p < 0.01
Exp MW (kDa)/pI
Fold change (PDQuest)
P-value(PDQuest)-
Q9BV57
21.5/5.4
11.6/5.4
4.4
p > 0.05
3.0
p < 0.01
P05783 P04792 Q13347
47.3/5.3 22.3/7.8 30.2/7.0
38.7/5.4 25.2/6.1 24.9/5.4
1.3 1.8 1.9
p > 0.05 p > 0.05 p < 0.05
4.4 3.1 3.1
p < 0.01 p < 0.01 p < 0.01
P09622
54.1/7.6
38.2/5.3
1.9
p < 0.01
2.6
p < 0.01
Proteins identified as up-regulated by Progenesis alone P22314 117.7/5.6 100.9/5.5 2.3
p > 0.05
4.0
p < 0.01
P22314
13.1/4.7
17.4/4.4
1.3
p < 0.01
3.5
p < 0.01
P14735
117.8/6.3
101.9/6.1
1.1
p > 0.05
3.4
p < 0.05
P43243 P62805 Q9BYT8
94.6/5.9 11.2/11.4 80.6/6.2
105.6/5.8 18.4/5.4 89.6/6.0
1.5 1.4 2.2
p > 0.05 p > 0.05 p > 0.05
3.3 3.3 2.8
p < 0.01 p < 0.01 p < 0.01
P28331
79.4/5.8
90.6/5.5
1.4
p < 0.01
2.8
p < 0.01
O43175
56.5/6.3
46.3/6.1
1.3
p < 0.05
2.7
p < 0.01
P07195
36.5/5.7
32.1/5.9
1.3
p > 0.05
2.4
p < 0.01
P42704
145.1/5.5
107.7/5.8
1.1
p > 0.05
2.3
p < 0.01
P82979
23.5/6.1
29.6/6.4
2.1
p > 0.05
2.2
p < 0.01
P82979
23.5/6.1
26.5/6.5
2.1
p > 0.05
2.4
p < 0.05
P08238
84.5/4.9
96.0/5.5
1.1
p > 0.05
2.2
p < 0.01
Q13765
23.3/4.5
32.3/ND
1.9
p < 0.01
2.1
p < 0.01
P26641
49.8/6.3
41.3/6.1
1.1
p > 0.05
2.0
p < 0.01
a Protein spots that were reported a g2 fold change in their expression levels by PDQuest and/or Progenesis were excised and identified by mass spectrometry. Protein spots are grouped into two categories based on whether they are down-regulated or up-regulated in HCT116 ASuPAR respectively. The proteins are enumerated based on their fold change from highest to lowest. Spot numbers correlate with the annotations in Figures 1 and 2, respectively. Fold change, Mascot score, and peptide and sequence coverage and putative functions are enumerates. b Observed pI and MW has been calculated from PDQuest software. c Out of range indicates spot was absent in replicate gels or was below the detection paramaters set for the software packages.
ligand-receptor engagement between uPA and uPAR.17 The significance of these interactions, especially with respect to cancer metastasis is exemplified by the fact that interfering with the ligand binding properties of uPAR using antibodies or mimetic peptides or reducing the expression levels of uPAR significantly interferes with the proliferatory and invasive propensity of the CRC tumor cells both in vitro and in vivo. However, most analyses to date have been restricted to singular uPAR-protein interactions which provide a rather narrow view of how uPAR functions as part of a multiprotein complex and what net effects suppression exerts on the global phenotype of a cancer. In our recent attempts to overcome these limitations, we reported that uPAR forms a multiprotein complex comprised of a number of important proteins, many of which have independently been strongly implicated in malignant epithelial cancer.17 Incidentally, a number of these proteins had previously also been reported to associate with uPAR in other
cellular models and/or to modulate functions and/or intracellular signal activation in a bidirectional manner.17 However, the intermediate pathways through which uPAR and its lateral associations with transmembrane and/or cytoplasmic proteins modulate cellular function and phenotype in cancer are largely unknown. To extend our understanding on the mechanistic pathways uPAR employs to drive the myriad of downstream cellular changes, we employed sensitive proteomic strategies to identify candidate proteins that change in expression levels when high uPAR levels are suppressed by ∼40% in an invasive colorectal carcinoma cell line model. The rationale was to map multiple proteins and possible pathways that concurrently change when invasive cancer cells lose malignancy as a result of uPAR suppression. The application of 2DE in this analysis was ideal due to its versatility in resolving large numbers of proteins based on their physiochemical properties and the availability of potent software systems that permit high-throughput protein spot alignJournal of Proteome Research • Vol. 7, No. 11, 2008 4799
research articles
Saldanha et al.
Figure 3. Comparison between representative 2DE gel maps of HCT116 and HCT116 ASuPAR using Progenesis. Whole cell lysates (500 µg) from HCT116WT (A) or HCT116 ASuPAR (B) were separated as described earlier. Quantitative analysis of the replicates in the two groups using Progenesis reported a total of 38 proteins that displayed a g2 fold change in expression levels respectively. Of these, 16 protein spots were found to be down-regulated while 22 spots were found to be up-regulated. The protein identities of the spots were confirmed by mass spectrometry and are enumerated in Table 1.
Figure 4. Venn diagram of differentially expressed proteins identified by PDQuest and/or Progenesis. A total of 65 unique proteins were found to change in their expression levels secondary to uPAR suppression in the HCT 116 colorectal cancer cell line. Of these, 17 proteins (∼26.6%) were consistently identified both by PDQuest and Progenesis (seven proteins were upregulated and 10 down-regulated in HCT116ASuPAR). A total of 27 proteins were specific to PDQuest (7 up-regulated and 20 down-regulated in HCT116ASuPAR) and 21 to Progenesis (15 upregulated and 6 down-regulated in HCT116ASuPAR).
ment and quantitative comparisons between replicate gels. Its application in quantitative proteomics has however been hindered due to an inherently high level of analytical variation that occurs during the separation of proteins in the first and second dimension. At an experimental level, this usually translates to misalignment of spots across replicate gels and the subsequent inability to quantitate between the differentially expressed proteins. We addressed this limitation by determining the maximal variability (CV) attributed to our experimental running conditions which included gel to gel, sample to sample and experiment to experiment based variation. This allowed us to account for the experimentally induced distortion between spots (due to analytical and/or biological variation) during image analysis and overcome potential type 1 statistical error (i.e.; false-positives changes).18,21 The overlap between the analytical variation in this study (∼28-30%) and the biological variation between samples (∼26-33%) was encouraging and correlated well with similar studies in the literature. We did find however, that the CV between identical experi4800
Journal of Proteome Research • Vol. 7, No. 11, 2008
ments run at two different time points did slightly increase (35.4 ( 1.4%, r2 ) 0.84) suggesting that clonal drifts in cell lines can contribute to experimental errors. This is crucial since most of the characteristic cell lines models used in proteomic analysis are ones that have been in use for over decades and this could easily influence the interpretation of results derived therein. Differentiating which proteins significantly changed in expression levels between the two cell lines was performed using two independent image analysis software packages to exclude any bias in quantitation attributable to differences in the algorithm(s) and/or analytical workflow in the software packages. Given that the analysis by both software packages were performed on the same gel images and under similar spot selection criteria, we anticipated a large overlap in the proteins identified and their quantitative reports. Surprisingly, only ∼26% (17 of the 65 proteins) of the total spots reported by PDQuest and Progenesis to be differentially expressed were identified by both software packages (Figure 4). The lack of concordance between the two analytical reports could be attributed to the ability of the software to define and warp spot boundaries and align them across replicate gels prior to comparative analysis. The PDQuest analysis of gels in this study entailed considerable manual analysis of each of the gel image to the reference gel prior to comparison between the two subsets while Progenesis analysis of the same subsets was performed as an automated process with minimal user intervention. Although extremely labor intensive, manual 2DE editing permits greater accuracy in delineation of spot boundaries, especially for low abundance or poorly resolved spots. This might also explain why a larger number of spots and variation were identified in the replicate gels identified by Progenesis compared to PDQuest. Additionally, the fundamental differences in the algorithm employed by image analysis programs plays a critical role in defining what spots are differentially expressed. The application of two independent software programs to this analysis represented a robust crossvalidation platform to confidently identify key protein responders to uPAR suppression and also act as a second pass detection step to increase coverage of candidate proteins which might have missed the detection cutoff criteria due to the abovementioned differences in the software packages.
Proteomics of uPAR Suppression in Cancer
research articles
Figure 5. Changes in stathmin expression and isoelectric point observed between HCT116WT and ASuPAR. (A) Analysis of stathmin isoforms 1R and β (spots 29 and 30) in the two cell lines by PDQuest revealed a 2.0 and 3.0 fold decrease in HCT116ASuPAR respectively (p value
