Proteomic Identification of the Galectin-1-Involved Molecular Pathways in Urinary Bladder Urothelial Carcinoma

Among various heterogeneous types of bladder tumors, urothelial carcinoma is the most prevalent lesion. Some of the urinary bladder urothelial carcinomas (UBUCs) develop local recurrence and may cause distal invasion. Galectin-1 de-regulation significantly affects cell transformation, cell proliferation, angiogenesis, and cell invasiveness. In continuation of our previous investigation on the role of galectin-1 in UBUC tumorigenesis, in this study, proteomics strategies were implemented in order to find more galectin-1-associated signaling pathways. The results of this study showed that galectin-1 knockdown could induce 15 down-regulated proteins and two up-regulated proteins in T24 cells. These de-regulated proteins might participate in lipid/amino acid/energy metabolism, cytoskeleton, cell proliferation, cell-cell interaction, cell apoptosis, metastasis, and protein degradation. The aforementioned dys-regulated proteins were confirmed by western immunoblotting. Proteomics results were further translated to prognostic markers by analyses of biopsy samples. Results of cohort studies demonstrated that over-expressions of glutamine synthetase, alcohol dehydrogenase (NADP+), fatty acid binding protein 4, and toll interacting protein in clinical specimens were all significantly associated with galectin-1 up-regulation. Univariate analyses showed that de-regulations of glutamine synthetase and fatty acid binding protein 4 in clinical samples were respectively linked to disease-specific survival and metastasis-free survival.


Introduction
The surfaces of bladders and ureters are mostly covered by urothelium. Among various heterogeneous types of bladder tumors, urothelial carcinoma is the most prevalent lesion. Urinary bladder urothelial carcinoma (UBUC) is often recognized as non-invasively papillary or superficially invasive tumors [1]. Nevertheless, some of the aforementioned tumors develop local recurrence and may cause distal invasion [1]. UBUC cells are identified as low-grade or less differentiated and high-grade cells.
Galectin-1 is one of the fifteen mammalian proteins belonging to the β-galactoside binding family. It is a 14-kDa monomer with one carbohydrate-binding domain and usually forms a non-covalent signaling pathway [24]. In this study, we exploited proteomics to find more novel molecular pathways evoked by galectin-1 dysregulation in UBUC cells, which might be related to UBUC carcinogenesis. The experimental design for this study is summarized in Figure 1. To obtain the above goal, Lavapurple TM -stained two-dimensional gel electrophoresis  was first carried out to acquire the protein profiles of Sh-Gal-1(+120) (galectin-1 knockdown stable cell line) and Sc-Gal-1(+120) (scrambled control) T24 cells, as described in Materials and Methods. Then, the 2-DE protein pictures of Sh-Gal-1(+120) T24 cells were compared to those of Sc-Gal-1(+120) T24 cells to search for the differentially expressed protein features which were recognized according to the definition, as described in Materials and Methods. To prevent the gel-to-gel variation, 11 replica gel pairs were obtained to find those de-regulated proteins provoked by galectin-1 knockdown. The typical 2-DE gel diagrams of Sh-Gal-1(+120) and Sc-Gal-1(+120) T24 cells are shown in Figure 1 and the other 10 pairs are provided in the supplementary materials (Supplementary Figure S1).  The protein profiles of Sh-Gal-1(+120) T24 cells were very similar to those of Sc-Gal-1(+120) T24 cells, but the amount of some identical proteins in the knockdown cell line was more than twice as high as that in its scrambled counterpart. Fifteen under-expressed and two over-expressed protein spots were found, as indicated by the arrows in Figure 1. Spots 4 (p = 0.07) and 10 (p = 0.05) were near p < 0.05. Spot 7 (p < 0.01) was not listed because it was not identified by liquid chromatography/tandem mass spectrometry (LC-MS/MS) ( Table 1 and Supplementary Table S1). The protein profiles of Sh-Gal-1(+120) T24 cells were very similar to those of Sc-Gal-1(+120) T24 cells, but the amount of some identical proteins in the knockdown cell line was more than twice as high as that in its scrambled counterpart. Fifteen under-expressed and two over-expressed protein spots were found, as indicated by the arrows in Figure 1. Spots 4 (p = 0.07) and 10 (p = 0.05) were near p < 0.05. Spot 7 (p < 0.01) was not listed because it was not identified by liquid chromatography/tandem mass spectrometry (LC-MS/MS) ( Table 1 and Supplementary Table S1).

Confirmation of Proteomics Data Using Cohort Studies of GS, FABP 4, TOLLIP, and AKR1A1
The results from western immunoblotting validated the dys-regulated proteins provoked by galectin-1 knockdown in Sh-Gal-1(+120) T24 cells. Furthermore, we designed cohort studies to confirm the expressed levels of GS, FABP 4, TOLLIP, and AKR1A1 in clinical specimens. Results of cohort studies showed that galectin-1 expression in biopsy samples was significantly correlated with the primary tumor (pT) status, grade, nodal metastasis, vascular invasion, perineural invasion, and mitotic rate. More importantly, the expression levels of GS, FABP 4, TOLLIP, and AKR1A1 in clinical samples were associated with the galectin-1 amount (Table 2), respectively. This finding was consistent with that of proteomics studies.
The results of the univariate log-rank analysis on disease-specific survival (DSS) and metastasis-free survival (MFS) indicated that primary tumor status, histological grade, nodal metastasis, the presence of vascular invasion and perineural invasion, the high mitotic rate, galectin-1 (Figure 3a (Table 3). However, the impacts on both TOLLIP and AKR1A1 abundance were not significant in predicting DSS. In line with our previous results [11] and the above proteomics data, a high galectin-1 expression level was predictive of a shorter metastasis-free survival time ( Figure 3d). The same results were also observed in patients with high amounts of GS (Figure 3e) or the FABP 4 ( Figure 3f) protein.

Discussion
Our previous findings suggest that the galectin-1-increased expression in UBUC is intimately associated with the primary tumor status, grade, and invasion [11]. Univariate and multivariate analyses demonstrate that galectin-1 over-expression in UBUC can also independently predict DSS and MFS. Further signaling pathway investigation indicates that the galectin-1 protein participates in UBUC cell invasion by modulating the MMP9 activity through the Ras-Rac1-MEKK4-JNK-AP1 signaling pathway [24]. To expand our knowledge regarding the role of galectin-1 in UBUC carcinogenesis, in this study, proteomic approaches were exploited to find more key signaling pathways which were provoked by galectin-1 up-regulation in UBUC cells.
In this study, the proteome maps of Sh-Gal-1(+120), a galectin-1 knockdown stable cell line, were compared to those of scrambled control cells, Sc-Gal-1(+120) cells. Our results showed that fifteen under-expressed and two over-expressed proteins were attributed to galectin-1 depletion in T24 cells. These impaired-regulated proteins found by proteomics could be categorized into a wide range of protein types involved in amino acid/energy metabolism (GS), lipid/energy metabolism (FABP 4), cytoskeleton (STMN1, TBCA), cell proliferation (STMN1, PMF1, RNH1), cell-cell interaction (NANS), cell apoptosis (TRXR1), invasion (SNX9), protein degradation (UBE2K, TOLLIP), and glucose metabolism (AKR1A1). SNX9, UBE2K, PMF1, STMN1, and TRXR1 were confirmed by western immunoblotting in the next step, while FABP 4, GS, TOLLIP, and AKR1A1 were validated by western immunoblotting and cohort studies. The aforementioned four proteins are found to be correlated with galectin-1 expression, tumor progression, and DSS/MFS with substantial clinical significance. The FABP 4 protein may participate in the trafficking of lipids to specific compartments. Abnormal FABP 4 functions may disrupt the lipid-mediated biological processes and result in diseases such as diabetes, obesity, and cancer [26]. Recently, Nieman et al. (2011) reported that FABP 4 demonstrates a higher up-regulation level in human omental metastatic ovarian lesions than in primary clinical ovarian tumor specimens and is over-expressed at the adipocyte/cancer cell interface [38]. Their results suggested that adipocytes support the cancer cells to metastasize from ovary to omentum. In omental adipocytes, lipolysis is induced to produce fatty acid and FABP 4 participates in fatty acid trafficking to cancer cells. In tumor cells, fatty acid is broken down by β-oxidation to fuel cancer cells for fast cell proliferation [38]. In contrast to the ovarian cancer, FABP 4 augmented expression is found to be associated with pTa UBUC with subsequent progression [39]. Our present studies suggested that galectin-1 knockdown reduced FABP 4 expression, likely through modulating PPAR-γ. Taken together, galectin-1 might take part in UBUC progression and invasion through regulating PPAR-γ/FABP 4.
Rapidly growing tumor cells would increase nucleotide and protein synthesis. They rely on the continuous provision of amino acids. Glutamine contains the most abundant amino acid in the body and is also the richest source for the nontoxic form of ammonia. Thus, tumor cells are highly dependent upon glutamine. When tumors begin to grow in the body, they elicit the host to mobilize and augment circulating glutamine. In addition, tumor cells use aerobic glycolysis and glutaminolysis (for breaking down glutamine to α-ketoglutarate to enter tricarboxylic acid (TCA) cycle) to quickly obtain the energy and precursor molecules for the synthesis of macromolecules [25]. Glutamine synthetase catalyzes the conversion of glutamate to glutamine and is thus involved in glutamine metabolism. GS has been implied as an early maker for hepatocellular carcinoma [40] and its positive immunostaining is correlated with both specific and overall mortality at the multivariate level [41]. Our data and the above observations all lead to the implication that galectin-1 might be involved in metabolic transformation. TOLLIP and UBE2K are shown to be associated with protein degradation. TOLLIP is originally recognized as an effector in interleukine-1 and Toll-like receptor signaling pathways [36]. However, it is also considered to take part in the trafficking of ubiquitinated proteins [36]. The increased protein turnover contributes to carcinogenesis by giving anti-apoptotic protection to tumor cells and augmented elimination of abnormal proteins [42]. Our present data showed that galectin-1 knockdown reduced TOLLIP expression and galectin-1 expression was closely associated with TOLLIP expression in the findings of biopsy samples. It might be likely that galectin-1 favors UBUC development by increasing the protein turnover. This is the first observation that TOLLIP was correlated with galectin-1 expression and UBUC patient's DSS/MFS. More molecular and functional studies are required to uncover the role of TOLLIP in UBUC carcinogenesis. In this study, we found that AKR1A1 expression was significantly associated with galectin-1 expression. Since the above results were the first findings, further work should be carried out to elucidate the signaling route by which galectin-1 regulates AKR1A1 expression.
Among the confirmed de-regulated proteins by western immunoblotting, STMN1 is a regulatory protein for microtubule assembly, and thus modulates the cytoskeleton [27]. It also participates in cell cycle regulation [29]. Cytoskeleton changes and impaired cell cycle regulation are essential for carcinogenesis. STMN 1 has been shown to be up-regulated in multiple tumors and is significantly linked to a poor survival rate of the patients [43]. Results in this investigation showed that galectin-1 knockdown decreased STMN 1 expression, suggesting that galectin-1 might control UBUC progression, probably through modulating STMN 1 expression. SNX9 is shown to provoke breast cancer metastasis via modulating RhoGTPase [35]. SNX9 can control the activation of RhoA and cdc42GTPase. It also controls breast cancer cell invasion through the RhoA-ROCK pathway and N-WASP. Our previous findings suggested that galectin-1 may promote UBUC invasion via the Ras-Rac1-MEKK4-JNK pathway [24]. The present observation might uncover an alternative pathway for galectin-1 to contribute to UBUC invasion. PMF1 is recognized as one of the proteins participating in polyamine homeostasis. Polyamine homeostasis impacts the cellular transcription rate and the decreased expression of early growth-associated genes including c-jun, c-myc, and c-fos [30]. Besides, Aleman et al. (2008) reported that PMF1 methylation is significantly linked to UBUC progression [44]. Our present data provided the novel observation that galectin-1 could regulate PMF1 expression.
In conclusion, the results of this study implicated that multi-functional galectin-1 contributed to UBUC carcinogenesis, probably through regulating amino acid/lipid/glucose metabolism, cytoskeleton rearrangement, cellular transcription, cell invasion, and protein degradation.

Construction of Galectin-1 Knockdown T24 Cells with Short-Hairpin RNA (shRNA)
Galectin-1 knockdown stable clones were established as described previously [24]. In brief, custom-made DNA precursor oligonucleotides of silencing RNA (siRNA) were annealed and the DNA duplex was cloned between BglII and HindIII restriction enzyme sites of the expression vector pSUPER-NEO (Oligoengine Corporate, Seattle, WA, USA), which has a neomycin resistance gene as the selection marker. Two complementary oligonucleotide strands of the siRNA DNA precursor, which would be transcribed into a shRNA processed into a 19-mer small interference RNA (siRNA), were designed with Oligoengine software (version 2.0, Oligoengine Corporate, Seattle, WA, USA). Oligonucleotide DNA sequences are listed in the supplementary File S1. T24 cells were transfected with the siRNA expression plasmid using PolyJet™ transfection reagents (SignaGen Laboratories, Ijamsville, MD, USA) according to the manufacturer's suggestions. Stable knockdown T24 cells were screened in the medium containing 150 µg/mL of G418 antibiotics. T24 cell clones with the best knockdown results were chosen using western immunoblotting and real-time reverse transcriptase polymerase chain reaction (RT-PCR) from 20 cell clones with targeting at +120 nucleotide (nt), named Sh-Gal-1(+120). Control cells were constructed by the stable transfection of T24 cells with the scrambled siRNA DNA precursor and T24 cell clones with the least impacts on galectin-1 protein amount were chosen from 20 cell clones scrambled at +120 nt and recognized as Sc-Gal-1(+120).

Preparation of Protein Lysates for Two-Dimensional Gel Electrophoresis
In this study, 70-80% confluent Sh-Gal-1(+120) and Sc-Gal-1(+120) T24 cells were trypsinized and recovered by centrifugation at 1000× g for 5 min at 25 • C and lysed with lysis solution ( The pellet from the 2D Cleanup kit was first re-suspended in MRB buffer (7M Urea, 4% (v/v) CHAPS) and then incubated at room temperature for 1 h with shaking to dissolve the pellet. Then, the protein concentration of the lysate was quantitated by the Bio-Rad DC protein assay. After measurement, an appropriate amount of thiourea, IPG buffer, and DTT were added into the lysate to re-constitute the rehydration buffer (7 M urea, 2 M thiourea, 0.5% (v/v) immobilized pH gradient (IPG) buffer (pH 4-7), 4% (v/v) CHAPS, 1 mM DTT and 0.1 (w/v) bromophenol blue (aMResco, Solon, OH, USA)) and incubated at room temperature with shaking for another 1 h. Following this, the protein lysates were stored at −80 • C until isoelectric focusing.

Lavapurple Staining, Image Analysis, and Statistical Analysis
Image analysis and statistical analysis were carried out as described formerly with some exceptions [45]. LavaPurple staining was carried out following the manufacturer's suggested protocol (Fluorotechnics, Sydney, Australia). In brief, the 2-DE gels were first soaked within solution I ((1% (w/v) citric acid and 15% (v/v) ethanol) for 1.5 h and then incubated in solution II (0.63% (w/v) boric acid, 0.385% (w/v) and 0.5% (v/v) Lavapurple) for 1.5 h in the dark. Lastly, the gels were put in solution III (15% (v/v) ethanol) for 30 min and then in solution I for 30 min. Lavapurple-stained 2-DE gels were scanned using a Typhoon 9400 fluorescence scanner (GE healthcare, Little Chalfont, UK) with a green laser (green laser PMT: 600 volt and emission filter: 580 BP) to obtain the digital images. To find the de-regulated proteins, a total of 11 pairs of well-separated gel images acquired from Sh-Gal-1(+120) and Sc-Gal-1(+120) T24 cells were compared with PDQuest 8.0.1 (BioRad, Hercules, CA, USA) software. Differentially expressed protein spots recognized by computer analysis were further verified by visualization. The intensity (volume) of the spot was quantitated and normalized as a percentage (ppm) of the total intensities of all spots in a gel (total normalized volume). Normalized volumes of individual de-regulated protein spots across all the replica gels of Sh-Gal-1(+120) and Sc-Gal-1(+120) T24 cells were first analyzed by the normal distribution test and then the Student's t-Test (STATISTICA Version 10.0 MR1, StatSoft, Tulsa, OK, USA) was carried out when reaching normal distribution. However, when normal distribution was not obtained, log transformation was carried out, followed by the normal distribution test and then the Student's t-Test. The normalized volume of each spot in Sh-Gal-1(+120) cells was compared to that of the same spot in Sc-Gal-1(+120) cells. In all cases, statistical variance of the Sh-Gal-1(+120): Sc-Gal-1(+120) spot normalized volume ratio within 95% (Student's t-Test; p < 0.05) was deemed as significantly different. Furthermore, the differentially expressed proteins present in at least six out of 11 gel pairs were regarded as galectin-1-associated proteins.

In-Gel Digestion and Protein Identification Analysis Via Liquid Chromatography-Tandem Mass Spectrometry (LC-MS/MS)
Since the ideal identification of Lavapurple stained protein spot was not obtained, the silver stained protein spots of interest were picked up for in-gel digestion. The protocol of silver staining is listed in the supplementary File S1. The in-gel digestion and mass spectrometric protein identification were carried out as described previously [45]. Briefly, the peptides present in protein digest were resolved in the LTQ-Orbitrap hybrid tandem mass spectrometer (ThermoFisher, Waltham, MA, USA) in-line coupled with the Agilent 1200 nanoflow HPLC system mounted with LC Packing C18 PepMap 100 (length: 5 mm; internal diameter: 300 µm; bead size: 5 µm) as the trap column and Agilent ZORBAX XDB-C18 (length: 50 mm; internal diameter: 75 µm; bead size: 3.5 µm) as the resolving column. File Converter in the Xcalibur 2.0SR package (ThermoFisher, Waltham, MA, USA), as well as an in-house program, were exploited to obtain the MS/MS information and the charge and mass of each analyzed peptide were calculated. TurboSequest program (ver. 27, rev. 11; Thermo Fisher Scientific, Waltham, MA, USA) was then implemented to determine the best matched peptides from a non-redundant protein database whose FASTA sequences were downloaded from the National Center for Biotechnology Information (ftp://ftp.ncifcrf.gov/pub/nonredun/) on 12 October 2010 with 541,927 entries. Only the tryptic peptides with ≤2 missed cuts were recorded. During the database search, the mass ranges were 1 and 3.5 m/z for fragment and precursor ions, respectively. The protein identities were only confirmed when there were at least two peptides matched and search results had a high Xcore (i.e., ≥2.0 for doubly charged peptides and ≥3.0 for triply charged ones) and minimal differences between the observed and hypothetical masses (i.e., ∆M < 10 ppm). For each set of MS/MS analyses, 25 fmol of bovine serum albumin (BSA) (aMResco, Solon, OH, USA) in gel was analyzed in parallel for verification of the effectiveness of the entire protein identification procedure, including in-gel digestion, nanoflow HPLC, MS/MS, and informatics analyses. The experimental data were only taken into account when a 10 ppm mass accuracy and over 70% coverage were observed in the co-processed BSA samples.

Case Selection
For immunohistochemistry, the Institutional Review Board (IRB) of Chi Mei Medical Center admitted the retrospective retrieval of 295 available primary UBUC blocks (Identification code: IRB10102-004, approved by IRB of Chi Mei Medical Center, project started from 14 Feb. 2012), which were adopted from the patients undergoing surgical resection with curative intent between January 1998 and May 2004. However, those who received the palliative resection were excluded. Patients with validated or suspicious lymph node metastasis underwent regional lymph node dissection. Cisplatin-based postoperative adjuvant chemotherapy was carried out in patients with pT3-pT4 status or nodal linkage. The UBUC histologic diagnosis was validated in all specimens according to the recent World Health Organization classification. Histologic grading was identified based upon Edmonson-Steiner criteria, while tumor staging was recognized according to the seventh edition of the American Joint Committee on Cancer system. Medical charts were reviewed for each patient to ensure the accuracy of other related clinicopathologic data. Follow-up information was available in all cases with a median period of 42 months (ranging 3-176 months).

Immunohistochemistry (IHC) and Statistical Analysis
IHC was conducted on representative tissue sections dissected from formalin-fixed, paraffin-embedded tumor blocks at a 3-µm thickness. Xylene was first used to deparaffinize the slides. Then, the slides were rehydrated with ethanol and heated using a microwave in a 10 mM citrate buffer (pH 6) for 7 min to retrieve the antigen epitopes. A total of 3% H 2 O 2 was used to quench endogenous peroxidase. After antigen retrieval, slides were washed in Tris-buffered saline for 15 min and then hybridized with a primary antibody again for 1 h, followed by antibody detection using a ChemMate EnVision TM kit (K5001; DAKO, Glostrup, Denmark). Two expert pathologists (CF Li and HL He) blinded to clinicopathological information and patient outcomes interpreted the immunostainings. The immune-intensities of membranous and cytosolic staining of Galectin 1, FABP 4, GS, TOLLIP, and AKRAK1 were respectively recorded by H score. Scoring of the immunoreactivity was measured based on a composite of both the percentage and intensity of the cytoplasm of positively stained tumor cells to generate an H score, using the equation: H score = ∑ P i (I + 1) where i is the intensity of the stained tumor cells (0 to 3+), and Pi is the percentage of stained tumor cells for each intensity varying from 0% to 100%. The specimens that showed an H-score above the median value were defined as high expression, while the cases below the median value were defined as low expression. The median H-score of all specimens was regarded as the cutoff point to divide high and low expression.
Statistical analyses were carried out with SPSS 14.0. The immuno-expressions of Galectin 1 (Invitrogen, Carlsbad, CA, USA), FABP 4 (Epitoimcs, Burlingame, CA, USA), GS (BD Transduction Laboratories™-BD Bioscience, San Jose, CA, USA), TOLLIP (Novus Biologicals, Littleton Colorado, USA), and AKRAK1 (Abcam, Cambridge, MA, USA) were compared by various parameters with the chi-square test. The end points evaluated in the whole cohort were DSS and MFS measured based on the date of UBUC operation to disease-related mortality and metastasis. In stepwise forward fashion, parameters at univariate p < 0.05 were basically analyzed for their relative prognostic importance using the Cox regression model. Of all analyses, the two-sided test of significance with p < 0.05 was considered significant.

Conclusions
The results of this study implicated that multi-functional galectin-1 contributed to UBUC carcinogenesis, probably through regulating amino acid/lipid/glucose metabolism, cytoskeleton rearrangement, cellular transcription, cell invasion, and protein degradation.

Conflicts of Interest:
The authors declare no conflict of interest.