Profilin 1 is a Potential Biomarker for Bladder Cancer Aggressiveness*

Of the most important clinical needs for bladder cancer (BC) management is the identification of biomarkers for disease aggressiveness. Urine is a “gold mine” for biomarker discovery, nevertheless, with multiple proteins being in low amounts, urine proteomics becomes challenging. In the present study we applied a fractionation strategy of urinary proteins based on the use of immobilized metal affinity chromatography for the discovery of biomarkers for aggressive BC. Urine samples from patients with non invasive (two pools) and invasive (two pools) BC were subjected to immobilized metal affinity chromatography fractionation and eluted proteins analyzed by 1D-SDS-PAGE, band excision and liquid chromatography tandem MS. Among the identified proteins, multiple corresponded to proteins with affinity for metals and/or reported to be phosphorylated and included proteins with demonstrated association with BC such as MMP9, fibrinogen forms, and clusterin. In agreement to the immobilized metal affinity chromatography results, aminopeptidase N, profilin 1, and myeloblastin were further found to be differentially expressed in urine from patients with invasive compared with non invasive BC and benign controls, by Western blot or Elisa analysis, nevertheless exhibiting high interindividual variability. By tissue microarray analysis, profilin 1 was found to have a marked decrease of expression in the epithelial cells of the invasive (T2+) versus high risk non invasive (T1G3) tumors with occasional expression in stroma; importantly, this pattern strongly correlated with poor prognosis and increased mortality. The functional relevance of profilin 1 was investigated in the T24 BC cells where blockage of the protein by the use of antibodies resulted in decreased cell motility with concomitant decrease in actin polymerization. Collectively, our study involves the application of a fractionation method of urinary proteins and as one main result of this analysis reveals the association of profilin 1 with BC paving the way for its further investigation in BC stratification.

Of the most important clinical needs for bladder cancer (BC) management is the identification of biomarkers for disease aggressiveness. Urine is a "gold mine" for biomarker discovery, nevertheless, with multiple proteins being in low amounts, urine proteomics becomes challenging. In the present study we applied a fractionation strategy of urinary proteins based on the use of immobilized metal affinity chromatography for the discovery of biomarkers for aggressive BC. Urine samples from patients with non invasive (two pools) and invasive (two pools) BC were subjected to immobilized metal affinity chromatography fractionation and eluted proteins analyzed by 1D-SDS-PAGE, band excision and liquid chromatography tandem MS. Among the identified proteins, multiple corresponded to proteins with affinity for metals and/or reported to be phosphorylated and included proteins with demonstrated association with BC such as MMP9, fibrinogen forms, and clusterin. In agreement to the immobilized metal affinity chromatography results, aminopeptidase N, profilin 1, and myeloblastin were further found to be differentially expressed in urine from patients with invasive compared with non invasive BC and benign controls, by Western blot or Elisa analysis, nevertheless exhibiting high interindividual variability. By tissue microarray analysis, profilin 1 was found to have a marked decrease of expression in the epithelial cells of the invasive (T2؉) versus high risk non invasive (T1G3) tumors with occasional expression in stroma; importantly, this pattern strongly correlated with poor prognosis and in- Bladder cancer (BC) 1 is the second in incidence and mortality cancer of the genitourinary system (1) and estimated to be the ninth most common malignancy (2). It is associated with a high recurrence rate underscoring the need for continuous surveillance following initial treatment. Cystoscopy still remains the gold standard for diagnosis and follow-up monitoring of bladder cancer. However, it is an invasive and unpleasant procedure, rendering particularly the regular surveillance program (e.g. cystoscopy every three months for the first year following initial diagnosis) not well accepted by the patients (3,4). Urine Cytology is a noninvasive current detection tool for BC, suffering however from suboptimal sensitivity, especially for low grade tumors and being subjected to interobserver variability (5). The invasive nature of cystoscopy and the low effectiveness of cytology have prompted the search for novel and better ways to diagnose the disease with special emphasis on the early detection of disease recurrences and/or progression.
Urine is regularly used in clinical practice and yields a wealth of information about the state of an individual's health. Because it can be collected in a noninvasive way it is more accessible than plasma or serum. In addition, there is no need for trained personnel for urine collection. Urine contains cells and cellular debris, inorganic ions (K ϩ , Na ϩ , Cl Ϫ , and Ca ϩ2 ), organic molecules (urea, uric acid, and creatinine) and proteins. If renal function is normal, urinary protein content is less than 150 mg/day. Such an amount is much lower than serum/ plasma protein content but it is still sufficient for proteomics studies, and even more, urinary proteins are in general considered to be more stable than blood proteins, as the bulk of proteolytic events have been completed prior to urine excretion (6,7). Notably, analysis of urine has led to the identification of various putative protein biomarkers for BC such as; Nuclear Matrix Protein 22 (NMP22), telomerase, hyaluronidase, cytokeratins (CK19, CK8, CK18), Bladder Cancer Antigen-4 (BLCA-4), survivin, Matrix Metalloproteinases (MMP-9, MMP-2), and others (3,4). The specific context of use of these biomarkers is still not fully defined, accuracy rates received are often not optimal, underscoring the need for a continuous search for more reliable BC biomarkers for the full spectrum of disease manifestations e.g. primary diagnosis, recurrence, progression (8).
The existence of several thousand proteins in urine with concentrations spanning multiple orders of magnitude hinders comparative studies of the urinary proteome. To reduce sample complexity and facilitate the detection of low abundance biomarkers for BC aggressiveness, we applied a fractionation strategy based on immobilized metal affinity chromatography (IMAC). IMAC was initially developed for purification of native proteins with an intrinsic affinity to metal ions (9). However, it soon turned out to be an approach with a broad spectrum of applications such as phosphoprotein enrichment and purification of recombinant His-tagged proteins (10). The application of IMAC fractionation in the investigation of the urine proteome has not yet been adequately addressed.
Our results indicate an enrichment for proteases, cell adhesion, cytoskeletal and signaling proteins as well blood proteins with affinity for metals, in urine of BC patients following IMAC fractionation. Among the identified proteins, aminopeptidase N (APN), profilin 1, and myeloblastin were further confirmed to be differentially expressed in invasive BC by Western blot or Elisa analysis of individual urine samples; nevertheless large inter-individual variability at the expression levels of these urinary proteins could be observed, raising concerns as to their clinical potential as urinary BC biomarkers. On the other hand, the association of the actin binding protein profilin 1 with BC was further confirmed by tissue microarray and in vitro blocking experiments which collectively strongly support the potential of this protein as a BC stratification and clinical outcome biomarker.

EXPERIMENTAL PROCEDURES
Urine Sample Collection, Handling, and Storage-Urine samples (random-catch) were collected in Laikon Hospital, Athens, Greece under Institutional Review Board (IRB) approved protocols. All involved subjects had signed informed consent forms kept at the physician's (KS) office, samples were coded and no identifiers were revealed. For the IMAC fractionation, four cancer pools were generated containing respectively: Pool 1: 9 samples; stage Ta, mean age 67 (median 67, range: 52-76); Pool 2: 7 samples, stage Ta, mean age   73 (median 75, range 57-82); Pool 3: 9 samples, grade T2ϩ , mean  age 75 (median 74, range 73-81) and Pool 4: 10 samples, grade T2ϩ,  mean age 62 (median 67, range 27-79). Samples were collected according to the standard protocol for urine collection as defined by the European Kidney and Urine Proteomics (EuroKUP) and Human Kidney and Urine Proteome Project (HKUPP) networks (http:// www.eurokup.org/node/138) including freezing in Ϫ80°C within 3 h of collection and a brief centrifugation step prior to proteomics analysis. Prior to IMAC fractionation individual urine samples were pooled at equal volumes (7-8 ml) and concentrated 40ϫ (2 ml) using an Amicon (3000Da MW cut-off) apparatus under nitrogen pressure (p ϭ 4.5 atm). For investigation of findings on APN, profilin 1, and myeloblastin expression by Western blot and/or Elisa assays a total of 82-108 individual urine samples was analyzed (as described below-Results) corresponding to urine samples from patients with benign diseases of the urogenital tract (BPH, inflammation, lithiasis, and hematuria), Ta (the vast majority being low grade), T1 (low grade and high grade) and T2 (mainly high grade) cancers with similar age and sex distribution.
Immobilized Metal Affinity Chromatography (IMAC)-IMAC uncharged resin (Profinity-BioRad, Hercules, CA) was used to generate a 10 ml chromatography column (BioRad) according to the manufacturer's instructions. In brief, 0.5 ml of resin was applied in the column, washed with 8 ml of deionized water, followed by a wash with 8 ml of binding buffer (50 mM NaH 2 PO 4 , 300 mM NaCl, 1 mM Imidazole, pH 8). Prior to metal ion loading (Ni 2ϩ ), the column was equilibrated with 8 ml of 50 mM CH 3 COONa, 300 mM NaCl, pH 4. A volume of 8 ml 200 mM NiSO 4 , pH 6.8 was added and the column was washed again with 10 ml of deionized water followed by a wash with 8 ml binding buffer (50 mM NaH 2 PO 4 , 300 mM NaCl, 1 mM Imidazole, pH 8). A volume containing 3 mg of total protein per sample (210 to 770 l) was mixed with binding buffer at a ratio of 1:4 respectively and applied to the column. Several fractions of flow through were collected (2 ml/fraction). The column was then washed with 10 ml of wash buffer (50 mM NaH 2 PO 4 , 300 mM NaCl, 5 mM Imidazole, pH 8) during which five fractions were collected (2 ml/fraction) to monitor washing efficiency. Elution was performed by the addition of 4 ml elution buffer (50 mM NaH 2 PO 4 , 300 mM NaCl, 500 mM Imidazole, pH 8) and two elution fractions were collected (2 ml/fraction). These were subsequently dialyzed against 3 L of water (three water changes in 4 h intervals). Following this dialysis step, eluates were dried in speed vac device, resuspended in 40 -50 l Laemmli buffer, subjected to vortex and bath sonication (5min) followed by heating at 92°C for 10min. The samples were then centrifuged at 13,000 rpm for 10 min and stored at Ϫ20°C until use. Monitoring of enrichment efficiency was conducted by analyzing a small aliquot (10 l) of each fraction by one dimensional gel analysis (5% stacking, 10% separating polyacrylamide mini gels) followed by silver staining (11). Prior to MS analysis SDS-PAGE was performed and gels were fixed with 30% methanol, 10% acetic acid for 40min followed by Coomassie Colloidal Blue Stain overnight (12). To estimate the reproducibility of the fractionation scheme, six aliquots of a "cancer pool," corresponding to 3 mg of urinary protein each, was processed through IMAC and eluates were analyzed for their protein content by Bradford assay and one-dimensional (1D) SDS (supplemental Fig. S1).
LC-MS/MS Analysis (Liquid Chromatography Coupled to Mass Spectrometry)-The eluate fractions were analyzed by SDS-PAGE as described above, and gels were sliced into 12 bands (shown in Fig. 1). These were further washed with ultra-pure water, and then destained by multiple serial changes of 150 l of destain solution (30% acetonitrile (ACN) v/v, 50 mM NH 4 HCO 3 , pH 8) and dried in a vacuum drier. In gel digestion was performed by incubating the gel bands with 10 l of trypsin solution (10 ng/l Trypsin, 10 mM NH 4 HCO 3 , pH 8) overnight at 37°C, in the dark and with excess humidity. The extracts were pooled, dried again and reconstituted with 30 l LC-MS mobile phase A (3% ACN, 0.1% Formic acid). LC-MS/MS experiments were performed on a quadrupole ion trap mass analyzer (Agilent Technologies, model 6330) fitted to a 1200 nano-HPLC system. The system was equipped with a micro well plate autosampler and a nano-electrospray ionization source. Three l per sample were injected and then separated (flow 150 nL/min) on a 0,075 ϫ 150 mm C18 capillary column (Zorbax C 18 , 300 Å pore, 3,5 m particle size). The nano-LC separation started with an isocratic composition of 100% mobile phase A (3% ACN, 0.1% formic) for 30 min followed by a gradient step for 70 min to 100% mobile phase B (97% ACN, 0.1% Formic acid) and then isocratic 100% B for 20min. The system was reequilibrated for 20 min. The ion trap analysis was performed as follows: scan range was 300 -2200 amu and the four most abundant multiply charged precursors above an absolute intensity threshold of 10 4 during the MS survey scan were selected for fragmentation. Accumulation time for MS/MS was 150 ms. All MS/MS experiments were conducted with Ultra-Scan Resolution with an automated adjustment of the collision energy using the smart fragmentation operation. Precursors were excluded after the acquisition of two spectra and released again after 0.2 min. The MS/MS scan range window was set at 100 -2200 amu.
Raw data were loaded into Mascot Distiller 2.3.2 software and peaklists were generated with settings optimized for the Agilent Ion Trap mass spectrometer. The peaklists were exported in mgf format and submitted to Mascot Server search engine. Protein searches were performed against SwissProt database (Uniprot_sprot, 2010_01) for human taxonomy, calculating 1 maximum miss cleavage for trypsin, oxidation of methionine and deamidation of asparagine or glycine as random modifications. Error tolerances were 1.5 Da and 0.7 Da for precursor and fragment mode respectively. Mascot result files were submitted to Scaffold 3 software (Proteome Science) for validation and meta analysis. For the evaluation of protein hits, a threshold confidence level (according to Scaffold) of 90% was selected with a minimum of two positive peptides per identification. False discovery rate for proteins was 1.2% as calculated automatically by the Scaffold software.
Tissue Samples and Microarrays-Several bladder cancer tissue microarrays were constructed at the Spanish National Cancer Center and used in this study. These arrays included primary urothelial cell carcinomas of the bladder, belonging to patients recruited under Institutional Review Board approved protocols at collaborating institutions. Tumor tissues were embedded in paraffin and five-m sections were stained with hematoxylin and eosin to identify viable, morphologically representative areas of the specimen from which needle core samples were taken, using a precision instrument (Beecher Instruments, Silver Spring, MD). From each specimen, triplicate or quadruplicate cores with diameters of 0.6 mm were punched and arrayed on the recipient paraffin block. Five-m sections of these tissue array blocks were cut and placed on charged polylysinecoated slides and used for immunohistochemistry analysis. We constructed 2 different BC tissue microarrays including a total of 194 bladder tumors: 92 nonmuscle invasive T1G3 and 102 muscle invasive T2ϩ tumors. Tumor stage and grade were defined according to the WHO and TNM system. In patients with invasive BC (pT2, pT3, pT4) lymph nodal and distant metastatic status was known enabling associations with tumor metastases. Additional clinicopathologic and annotated follow-up information of the tumors spotted onto the tissue microarrays allowed the evaluation of associations of protein expression with staging and outcome.
Immunohistochemistry-Protein expression patterns of profilin 1 were assessed at the microanatomical level, using the tissue microarrays described above. Standard avidin-biotin immunoperoxidase procedures were used for immunohistochemistry. Antigen retrieval meth-ods (0.01% citric acid for 15 min under microwave treatment) were utilized prior to an overnight (at 4°C) incubation with the primary antibody (Alexis Biochemicals, rabbit polyclonal profilin 1 Ab at 1:500). Staining conditions were optimized on sections from formalinfixed, paraffin-embedded tissue controls as specified by the Ab manufacturer. The secondary antibody (Vector Laboratories) was a biotinylated goat anti-rabbit antibody (1:1,000 dilution). The absence of primary antibody was used as negative control. Testis was used as positive control. Diaminobenzidine was utilized as the final chromogen and hematoxylin as the nuclear counterstain as previously described (13). Evaluation of the staining pattern was conducted as described below (Statistical analysis).
Statistical Analysis-The consensus value of the three or four representative cores from each tumor sample arrayed was used for statistical analyses. The association of the expression of profilin 1 measured by immunohistochemistry on tissue arrays with histopathologic stage and tumor grade was evaluated using the Wilcoxon-Mann-Whitney and Kruskall-Wallis tests. Profilin expression pattern was evaluated as a continuous variable based on the number of cells expressing the protein in the cytoplasm. The intensity of the staining was categorized from negative (-) to low (ϩ), intermediate (ϩϩ), and high (ϩϩϩ). Additionally, it was evaluated whether profilin 1 was observed in the extracellular matrix in the surrounding stroma of epithelial cells. In all cases, immunohistochemistry was evaluated by two independent reviewers (MSC, S.E.) and consensus values were statistically analyzed. It should be noted that the degree of consensus in immunohistochemistry evaluation was high e.g. the different reviewers did not disagree in more than 30% of cells expressing profilin in the cytoplasm. Thus, the mean values of each of the independent observations for each core were estimated and utilized for statistical analyses.
The cut-offs of expression for prognostic evaluation were selected based on the median values of expression among the groups under analyses. The associations of the protein with disease-specific and overall survival and progression were evaluated using the log-rank test in those cases for which follow-up information were available. Disease-specific survival time was defined as the months elapsed between transurethral resection or cystectomy and death as a result of disease (or the last follow-up date). In patients with non-muscle invasive disease, progression time was defined as the months elapsed between transurethral resection until cystoscopy indicating the presence of muscle invasive disease treated by cystectomy (or the last follow-up date). In patients with muscle invasive disease, progression time was defined as the months elapsed between transurethral resection or cystectomy until clinical, pathologic or imaging examinations indicating the presence of metastatic disease (or the last follow-up date). Overall survival time was defined as the months elapsed between transurethral resection or cystectomy and death (or the last follow-up date). Patients who were alive at the last follow-up or lost to follow-up were censored. Survival curves were plotted using the standard Kaplan-Meier methodology. Statistical analyses were performed using the SPSS statistical package (version 18.0).
Cell Culture and Sample Preparation for Western Blot Analysis-T24 cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum (Invitrogen-Invitrogen, Grand Island, New York) at 37°C, 5% CO 2 as previously described (14). When the cells reached a concentration of 10 6 cells per ml, the medium was removed and the cell layer was washed 3 times with phosphate-buffered saline (Invitrogen) and 1 time with Serum and Phenol Red Free Medium (Invitrogen). Serum free medium was added to the cells for an incubation period of 24 h after which the cell layer as well as serum free medium were collected, as previously described (15). In brief, the latter was centrifuged at 1000 ϫ g for 10 min at 4°C to remove dead cells and large debris and incubated with 7.5% trichloroacetic acid, 0.1% N Lauroyl sarcosine at Ϫ20°C overnight. Centrifugation then followed at 10,000 ϫ g for 10 min at 4°C. The supernatant was discarded, and the pellet was washed with ice cold tetrahydrofuran, centrifuged again as previously, and the final pellet was dried in the air and resuspended in isoelectric focusing sample buffer (7 M urea, 2 M thiourea, 4% 3-[(3-cholamidopropyl)dimethylammonio]propanesulfonate, 1% DTE, 2% IPG buffer and 3.6% Protease inhibitors) by 30 min bath sonication (15). Samples were stored at Ϫ20°C until use. For the preparation of cell extracts, cells were collected following trypsinization and cell pellets were washed 3 times in PBS and dissolved in isoelectric focusing sample buffer by bath sonication (14). The suspension was centrifuged at 13,000 rpm for 20 min, supernatants were collected, aliquoted, and stored at Ϫ20°C until used. Protein concentration was estimated by the use of Bradford reagent (Bio Rad).
Transwell Migration Assay-In Vitro Blocking Experiments-Transwell migration assays were performed as previously described (14,16). In brief, T24 cells were cultured for 48 h in Dulbecco's modified Eagle's medium supplemented with 2% fetal bovine serum and then were transferred, at concentration of 5 ϫ 10 4 /100 l, to the 5 m pore size insert of a transwell plate (Corning-Costar, Cambridge, MA). The cells were then allowed to migrate for 6 h across the pore membrane, toward their respective conditioned medium (CM). The nonmigrated cells were removed from the top of the insert with a wet cotton swab. The migrated cells were then fixed on the membrane with 4% paraformaldehyde (Sigma-Aldrich) and stained using the Ral Kit (Ral Reactif, Paris, France), according to the manufacturer's instructions. Photographs from the stained nuclei were taken from a minimum of 10 fields of view (20ϫ) for each membrane using a Leica CTR MIC microscope and then were counted by using Image J software. Migration was quantified by counting the stained nuclei that passed through the pore membrane. Two independent experiments were performed including 2 replicates each. Statistical analysis was performed using Student's t test. For the in vitro blocking experiments, profilin was blocked by incubating T24 cells with an anti-profilin 1 rabbit polyclonal antibody (Santa Cruz, Santa Cruz, CA) at a dilution of 1:10, for 30 min at 4°C. As a control, cells were alternatively incubated with isotype IgG1 Ab (Becton Dickinson) under the same conditions. After blocking procedure, in vitro motility assays were carried out for T24 cells, as described above.
Phalloidin-FITC Uptake, Fluorescence Microscopy and Flow Cytometry Analysis (FACS)-The effects of profilin blockage onto actin polymerization were evaluated by FITC-Phalloidin uptake using fluorescence microscopy and flow cytometry (FACS) analysis. T24 cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, and then were transferred to a 24-well plate at a concentration of 5 ϫ 10 4 cells/well for 24 h. Medium was removed and blocking of profilin was performed as described above. Alternatively, cells were also incubated with isotype IgG1 Ab (Becton Dickinson) under the same conditions. After blocking, cells were washed with phosphate-buffered saline, followed by Phalloidin-FITC (Sigma Aldrich) incubation according to the manufacturer's instructions.
DAPI (Sigma Aldrich) was used for nuclei staining. Fluorescent images were taken from a minimum of 10 fields of view (40ϫ) from each well using a Leica CTR-MIC Fluorescent microscope. FITCphalloidin uptake in T24 cells or profilin blocked T24 cells, was also estimated by flow cytometry analysis (FACS) as described previously (17). Cells were analyzed using a Beckman Coulter Cytomics FC 500 flow cytometer (Beckman Coulter Ltd, Palo Alto, CA, USA).
Western Blot (WB) and Elisa Analysis-Urinary protein precipitation by 7.5% trichloroacetic acid and 0.1% N-lauroylsarcosine sodium salt was performed as previously described (18) and resulting pellets were resuspended in 7 M urea, 2 M thiourea, 4% 3-[(3-cholamidopro-pyl)dimethylammonio]propanesulfonate, 1% DTE. In the cases of myeloblastin WB and profilin 1 Elisa assays, direct analysis of urine provided no/faint or nonspecific signal, regardless the sample protein concentration. The specificity and intensity of the signal increased significantly following analysis of precipitated urinary protein. In the case of APN, no changes in the specificity of the signal or expression trends of the protein could be observed when analyzing starting urine versus concentrated urinary protein (this difference in starting sample requirements between APN, and profilin-myeloblastin also explains the slightly different sample sizes employed in each case and imposed by sample availability). As such, 20 g of total protein from each urine sample (in the case of myeloblastin), total protein from cell extracts and conditioned media, or 15 l of starting urine (in the case of APN) were separated by 12% SDS-PAGE under reducing conditions and electroblotted to Hybond-ECL nitrocellulose membrane (Amersham Biosciences). After blocking with 5% nonfat dried milk in TBST (20 mM Tris/pH 7.6, 137 mM NaCl, 0.1% Tween 20) for 2h at room temperature, membranes were washed with TBST and incubated overnight at 4°C with the primary antibodies, as applicable: rabbit anti-human profilin 1 (Santa Cruz; dilution 1:200), mouse antihuman APN (Santa Cruz; dilution 1:50), mouse monoclonal antimyeloblastin-PR3 (Santa Cruz; dilution 1:200), mouse antihuman tubulin (Sigma Aldrich; dilution 1:2000). Membranes were then washed with TBST and incubated with anti-mouse or anti-rabbit HRP-conjugated secondary antibody (Santa Cruz; dilution 1:5,000, Amersham Biosciences; dilution 1:10,000 respectively) for 2h at room temperature. A final wash with TBST was made and target protein was detected by Enhanced Chemiluminescence (Perkin-Elmer LAS, Inc.) detection system. Films were scanned and images were analyzed using Quantity One software (Bio Rad). Elisa assays for profilin 1 were conducted using the respective kit (Antibodies-online GmbH) according to the manufacturer's instructions. Twenty g of urinary protein prepared following TCA/NLS precipitation as described above, was employed.

Urine Proteins Identified by LC-MS Analysis Following
IMAC Fractionation-To reduce sample complexity a fractionation strategy of urinary proteins based on the use of IMAC was adopted. In brief, four different pooled urine samples from patients with bladder cancer, corresponding to 2 pools from patients with non invasive (Ta) tumors and 2 pools from patients with invasive (T2ϩ) tumors were concentrated by ultrafiltration, and subjected to protein separation, following normalization for protein content, through a Ni 2ϩ IMAC column. Elution of bound proteins was carried out by imidazole. Based on the analysis of one urine sample in six replicate applications through the IMAC column, protein recovery was estimated to be ϳ3.3% with overall reproducible 1D SDS patterns of the eluted fractions (supplemental Fig. S1). Protein identification included protein band excision, tryptic digestion and analysis by LC-MS/MS (Fig. 1). Received identifications following analysis using the Mascot search engine were further evaluated using the Scaffold software and manual analysis of spectra. An additional "in silico" evaluation of findings for their biological significance and relevance was conducted using gene ontology (GO) analysis, text mining as well as search through expression databases such as the Human Protein Atlas database (www.proteinatlas.org). The full list of received identifications with GO data is provided in supplemental Table S1 -IMAC Protein List with the respective detailed peptide and MS identification data provided in supplemental Table S2-IMAC Peptide MS data. As shown, among the identified proteins multiple (about 30%) corresponded to or originated from proteins with affinity for metals. An additional enrichment for proteins expected (according to GO) to be phosphorylated and/or glycosylated could also be observed, as well as, for metal binding blood proteins (such as serotransferrin, hemopexin, hemoglobin, albumin). Among the identified, were proteins with demonstrated function and increased expression in BC such as MMP9 (19 -21) various fibrinogen fragments (22) and clusterin (13).
Because a limited number of pooled samples was analyzed, prioritization of findings based on statistics would not be reliable; therefore, this (prioritization) was performed taking into consideration the functional annotation of the received identifications. Table I lists some of the received identifications with documented or potential functional relevance to cancer and which were considered to merit further investigation. Among these are proteases (aminopeptidase N, MMP9, dipeptidyl peptidase 4, myeloblastin), protease inhibitors (Inter-alpha-trypsin inhibitor heavy chain H4), proteins involved in cytoskeletal re-arrangements (profilin 1, utrophin, CASC5), cell adhesion (Galectin 3 Binding Protein), signaling molecules (Pro-epidermal growth factor, Kinase suppressor of Ras 2) as well as proteins with documented association to cell proliferation (Antigen ki-67) or apoptosis (clusterin).
Differential Expression and High Interindividual Variability of Urinary APN, Myeloblastin and Profilin 1 in Invasive BC-Among the proteins enriched through IMAC, APN, myeloblastin, and profilin 1 were considered of highest interest and prioritized for further investigation; this decision was based on their functional annotation, as described below, and the lack of reported association with BC and of earlier studies in urine. In brief, APN is an extracellular protease with demonstrated function in tumor invasion for other types of malignancies (23)(24)(25); Myeloblastin is thought to be involved in MMP activation, hence potentially in tumor invasion (26,27) and profilin 1 is involved in actin polymerization and has been implicated in a variety of other malignancies (28 -39).
Western blot and/or Elisa assays of individual urine samples were conducted; received values per protein under study and urine sample are provided in Table II Table II).
In accordance also to the IMAC results, myeloblastin was found to be overexpressed (p Ͻ 0.05) in urine samples corresponding to T2ϩ (n ϭ 25) in comparison to T1 (n ϭ 23) and Ta (n ϭ 26) cancers and benign diseases of the urogenital tract (n ϭ 23) (mean (S.D.): B:7.66 (9.80); Ta: 6.50 (11.52); T1: 3.64 (4.46); T2ϩ: 15.63 (18.92); Fig. 2 and Table II). As evident both APN and myeloblastin showed large variations in the protein levels within each group, potentially reflecting disease heterogeneity and/or presence of confounders affecting the expression of these proteins in urine.

Profilin 1 is Differentially Expressed in Bladder Tumors and Associated With Clinical
Outcome-Besides the analysis of its urinary levels, a detailed characterization of the expression patterns of profilin 1 by means of immunohistochemistry using independent series of bladder tumors contained in several tissue arrays was also conducted. Interestingly, the profilin expression patterns in stroma and epithelial cells were found to be correlated with tumor characteristics and disease outcome (summarized in Table III): In patients with non-invasive disease (n ϭ 92), significant statistical associations were found between expression of profilin 1 in the stroma and increasing tumor size (p ϭ 0.012) as well as the presence of recurrence after three months post-transurethral resection (p ϭ 0.039). Importantly, in patients with noninvasive disease, profilin 1 expression in the stroma was significantly associated with a high recurrence rate (Fig. 4A, log rank, p ϭ 0.032) and also with a poor overall survival (Fig. 4B, log rank, p ϭ 0.005). In patients with invasive disease (n ϭ 102), significant statistical associations were found between an expression of profilin 1 in the stroma and the presence of visceral metastases (Mann-Whitney, p ϭ 0.040). Importantly, low intensity (Ͻϩϩ) of cytoplasmic (epithelial) profilin expression was significantly associated with reduced disease-specific survival (Fig. 4C, log rank, p ϭ 0.014), and reduced overall survival time (Fig. 4D, log rank, p ϭ 0.016). Combining all samples analyzed, both the number of cancer cells expressing profilin 1 and the cytoplasmic intensity in epithelial cells diminished in association with increasing tumor stage (p ϭ 0.002 and p Ͻ 0.0005, respectively). Additionally, patients presenting low number of cancer cells with cytoplasmic profilin expression (Ͻ40%) showed decreased disease-specific survival (log-rank, p ϭ 0.002) and reduced overall survival times (log-rank, p ϭ 0.024). Moreover, the presence of a low intensity of its cytoplasmic expression was significantly associated with increased disease specific mortality (Fig. 4E, logrank, p Ͻ 0.0005) and poor overall survival (Fig. 4F, log-rank, p Ͻ 0.0005) Overall, these observations indicated that loss of intracellular cytoplasmic profilin 1 in epithelial cells and expression of the protein in the stroma is associated with tumor staging and an aggressive phenotype.
Profilin 1 Affects T24 Cell Motility-To obtain initial insight into the role of this protein in BC, in vitro blocking experiments were conducted involving transwell migration assays of T24 cells.
Expression of profilin 1 in this cell line was confirmed by Western blot analysis of total cell extracts and cell line conditioned media preparations. A band at the expected molecular mass was detected in both cell extract and cell media (Fig. 5A). In order to ensure that profilin 1 detected in the latter was not the result of contamination because of cell lysis/cell death, the same membranes were blotted with a tubulinspecific Ab. The results of this analysis indicated a more than 20-fold enrichment of profilin 1 compared with tubulin in the conditioned media indicating that profilin 1 is naturally present in the extracellular medium (Fig. 5A). Blocking was performed by incubating T24 cells in the presence of profilin 1 specific Ab. Effects on cell motility were subsequently investigated employing transwell assays. As shown (Fig. 5B), motility of T24 cells toward their CM decreases when cells were preincubated with profilin-Ab. In contrast, there was no significant effect on T24 cell migration  in the presence of the IgG1 antibody, supporting the specificity of the profilin Ab-observed effect. Profilin plays a key role in the polymerization of actin. To further prove the relevance of the observed phenotype to profilin function, the polymerization of F actin was investigated. Blocking of profilin 1 in T24 cells resulted in a 20% decrease in F actin compared with controls as revealed by phalloidin-FITC uptake and FACS analysis (Fig. 5C). DISCUSSION IMAC is a separation technique that utilizes covalently bound chelating compounds on solid chromatographic supports to entrap metal ions, which themselves serve as affinity ligands for various proteins, making use of coordinative binding of some amino acid residues exposed on the surface. This form of affinity chromatography can be used in cases where rapid purification and substantial purity of the product are necessary, although compared with other affinity separation technologies it is considered moderately specific. However, IMAC holds some benefits such as ligand stability, high protein loading, mild elution conditions, simple regeneration and low cost. Oxidative reductive conditions inside the column, metal-induced cleavage and metal toxicity could be listed as disadvantages of this method (10,40,41).
In the present study we described a fractionation strategy of urinary proteins based on the use of immobilized nickel affinity chromatography. To the best of our knowledge, this is the first time that IMAC was used in cancerous urine samples for the identification of disease biomarkers. However, it should be noted that IMAC in the form of SELDI chromato-

FIG. 4. Survival analyses -Kaplan-Meier curves displaying the associations of profilin 1 expression in epithelial and stroma cells with tumor recurrence and disease outcome (survival).
A, Kaplan-Meier indicating that in patients with non-invasive disease, profilin expression in the stroma was significantly associated with a high recurrence rate (p ϭ 0.032), and B, with a poor overall survival (0.005). C-D, Kaplan-Meier curves indicating that in patients with invasive disease, the presence of low intensity of the cytoplasmic profilin expression was significantly associated with poor survival (C, disease-specific; p ϭ 0.014, and D, overall survival; p ϭ 0.016). E-F, Kaplan-Meier indicating that taking together all tumors under analysis, a low intensity of profilin cytoplasmic expression was significantly associated with poor survival (E, disease-specific (p Ͻ 0.0005), and F, overall survival (p Ͻ 0.0005)). graphic surfaces has been employed: for example, Papale et al. analyzed the impact of several pre-analytical and analytical variables on SELDI-IMAC proteomic profiling of human urine in healthy subjects (42). Cirulli et al. (43) applied IMAC for the detection of free phosphorylated peptides in biological fluids (pathological sera samples, healthy urinary and salivary samples), and Talvas et al. (44) adapted this fractionation scheme as an enrichment method for phosphoproteins, in an approach to identify new targets of leucine deprivation in muscle cells. Levin et al. (45) recently used IMAC as a fractionation technique for blood proteins from a healthy individual whereas various reports exist on the application of IMAC SELDI surfaces for the analysis of serum and plasma protein biomarkers (46 -48). FIG. 5. In vitro effects of profilin 1 on bladder cancer cell motility and actin polymerization. A, Expression of profilin 1 in T24 conditioned media (CM, i) and cell extracts (ii); Profilin 1 found in the CM is not a result of contamination from necrotic cells as shown by respective blots for tubulin (iii, iv); the mean intensity of the tubulin band in CM is 20 times lower than the ones of profilin 1 in CM. The results from the analysis of three different (A-C) cell culture preparations and cell extracts are depicted. B, In vitro blockage of profilin 1 results in decrease of bladder cancer cell motility. i) Diagram showing the number of migrated cells following incubation of T24 cells with specific rabbit anti-human profilin 1 or control Ab. Nontreated cells were also used as control. In all cases 50,000 cells were initially plated, migration was allowed for 6 h toward T24 CM. The nonmigrated cells were then removed from the top of the insert and migrated cells were fixed and stained (Ral Kit). Migration was quantified by counting the nuclei that passed through the filter from a minimum of 10 fields of view (20ϫ) (Leica CTR MIC microscope, Image J software). Two independent experiments were performed including two replicates each. Data are presented as the mean Ϯ S.D. and were analyzed by Student's t test. (ii-iv): Representative optical images of migrated T24 untreated cells (ii), T24 cells incubated with control Ab (iii), and T24 cells following incubation with specific profilin Ab (iv). C, Blocking of profilin 1 results in decrease in polymerized actin. Decreased Phalloidin-FITC uptake from T24 cells after incubation with specific profilin 1 Ab (i) compared with control Ab (ii). Immunofluorescent staining was merged with DAPI nuclear staining. Images were taken from a Leica CTR-MIC Fluorescent microscope (20x). iii) T24 cells incubated with profilin 1 Ab present a decrease in median fluorescence intensity compared with cells incubated with control Ab, as shown by FACS analysis. iv) Representative bar plot denoting a 20% reduction of median fluorescence intensity of T24 cells after incubation with profilin 1 Ab compared with the cells that were incubated with a control Ab.
In our study, pooled urine from patients with non invasive (two pools) and invasive BC cancers (two pools) was employed to allow for fractionation of higher starting protein amounts (3 mg). Pooling automatically decreases the statistical power of observations and cannot reflect existing interindividual variability, as can be readily observed in the cases of urinary APN, myeloblastin and profilin 1. Findings from such comparative analysis of pooled samples should thereby be evaluated with caution, and considered suggestive based on the putative functional rather than statistical associations of the detected proteins with the disease.
As shown, of the proteins enriched following the application of IMAC, several have been implicated in cancer development and/or progression: for example Leucine-rich a2 glycoprotein-1 and ficolin 3 have been recently identified as biomarkers for ovarian cancer (49). Transaldolase has been linked to various cancer types such as breast, colorectal, head and neck cancers and (50,51). Utrophin gene mutations have been observed in breast, melanomas and neuroblastomas suggesting a tumor suppressor role for the protein (52). Altered expression of APN and profilin 1 has been detected in various tumors (23-25, 28 -39). Myeloblastin (also known as proteinase 3) has been considered a leukemia-associated antigen forming the basis for the development of vaccines for hematologic cancers (27,53,54). Antigen ki-67 is a well established proliferation marker having been evaluated through immunohistochemistry in several cancers including BC (55,56). Additionally association of urinary MMP9 (19 -21), clusterin (13,57), and fibrinogen (22), with BC has been documented. Collectively, these results support that the technique can enrich for proteins of relevance to the cancer phenotype and once more reflect the wealth of information contained in urine. Nevertheless, translation of these findings to urinary assays of high specificity and sensitivity in disease detection presents large difficulties related presumably in part to the presence of potential confounders affecting the levels of urinary proteins as well as to the disease vast molecular heterogeneity. This phenomenon on one hand emphasizes on the need to maximize gained prognostic information from excised tissue samples, and at the same time to develop panels of well characterized urinary biomarkers that could more accurately grasp the disease heterogeneity.
As a first report on the association of profilin 1 with BC and investigation of the protein in urine, our studies demonstrated a differential expression of this protein in urine and tissue in invasive compared with non invasive cancers. Even though the former was subject to inter-individual variability underscoring the need for further studies (likely in conjunction with additional markers) targeting better patient stratification, the latter gave a clearer and stronger association with disease outcome. Specifically, expression of profilin 1 in the stroma could be occasionally observed which in the case of non invasive tumors significantly correlated with a high recurrence rate and poor survival whereas in the invasive samples was significantly associated with the presence of visceral metastases. Similarly, associations of cytosolic profilin expression levels with disease-specific and overall survival times could be observed. Notably, none of these associations stratifying tumor aggressiveness can be predicted based on current hematoxylin-eosin evaluations in clinical routine practice pointing to a clear added value of profilin to patient stratification. A potential clinical value of this protein could thereby be in the generation of risk-stratified groups within tumors of similar clinicopathological characteristics, possibly guiding decision-making on treatment selection. Combining profilin 1 with other tissue molecular markers [such as apoptotic, cell cycle, signaling, angiogenic factors; reviewed in (58,59)), in the form of a "prognosticator" panel may be a promising way forward to accurately capture the biological tumor potential and maximize predictive power of clinical samples. Such evaluations will greatly benefit from the development/optimization of automated platforms and software solutions for image capture and analysis.
Profilin 1 belongs to the family of profilins which are small actin binding proteins regulating the dynamics of actin polymerization and cell motility. Besides actin binding, profilins have been shown to interact with other proteins [such as; Wiskott-Aldrich syndrome protein (WASP) and ena/vasodilator-stimulated phosphoprotein (VASP)] and growing evidence further supports a function as hubs controlling interactions related to membrane trafficking and small-GTPase signaling (28,33,35).
In a first attempt to obtain a mechanistic insight on how profilin correlates to aggressiveness, in vitro studies were performed. An inhibitory effect on the motility of T24 cells and a decrease in F (polymerized) actin were observed following cell incubation with profilin Abs, indicating that profilin 1 affects cell motility possibly through actin. Further experiments are needed to investigate this hypothesis and the underlying mechanism. In addition, our in vitro, tissue and urine data collectively strongly support that besides the cytosolic functions of profilin, the protein is also naturally present in the extracellular medium likely secreted via exosomes or a non classical pathway (32,60,61) as it lacks a classical secretory signal (60,61). This evidence in combination with the association of profilin 1 with mechanisms of endocytosis (35,62,63), may also explain the (intracellular) blocking effect of the antibodies on actin polymerization: specifically, our hypothesis is that profilin Ab binds to extracellular profilin, disrupting its interactions with the plasma membrane trafficking molecules and the respective pathways for its endocytosis. Consequently, profilin's entrance to the cell is prevented, affecting to some extent its intracellular function e.g. promotion of actin polymerization. This may also be a mechanism explaining how high stroma profilin promotes aggressiveness onto epithelial cells. Further studies are obviously required to investigate this hypothesis and even more to also recapitulate and investigate how low epithelial profilin levels correlate with the aggressive phenotype.
In conclusion, with the current study, we describe the application of IMAC fractionation in urine toward the identification of potential biomarkers for BC aggressiveness; Among the proteins enriched, urinary aminopeptidase N, myeloblastin and profilin 1 are found to be differentially expressed in invasive compared with non invasive BC cases and benign controls, even though their levels are subject to inter-individual variabilities of yet not defined reason; At the tissue level, a clear added value of profilin 1 to current patient stratification tools is suggested, prompting further validation clinical studies on this protein to define its specific context of use in bladder cancer. Multiple additional research questions and respective avenues open up including, the delineation of cytosolic and potentially autocrine or paracrine functions of profilin 1 in BC epithelial and stromal cells as related to the aggressive phenotype.