Lectin affinity chromatography and quantitative proteomic analysis reveal that galectin-3 is associated with metastasis in nasopharyngeal carcinoma

Nasopharyngeal carcinoma (NPC) is a serious cancer in East and Southeast Asia. Patients are often diagnosed at advanced stages, rendering treatment failure due to high potential of metastasis. This study identified lectin-binding glycoproteins with a potential role in NPC metastasis. Cell lysate and culture medium in highly metastatic 5-8F, and lowly-metastatic 6-10B NPC cell lines were fractionated by ConA- and WGA-affinity chromatography, and subjected to GeLC-MS/MS. A total of 232 and 197 proteins were identified in ConA-enriched fraction of 5-8F and 6-10B cell lysates respectively. In WGA-enriched fraction, 65 and 164 proteins were found in 5-8F and 6-10B cell lysates respectively. Proteins identified in culture medium for both cell lines were 223 and 85 for ConA-enriched fraction, and 94 and 124 for WGA-enriched fraction from 5-8F and 6-10B respectively. Differentially expressed proteins were functionally categorized into cell–cell adhesion, extracellular matrix, glycolysis, protein homeostasis and/or glycosylation enzymes, and lipid metabolism. Interestingly, Galectin-3 (Gal-3) was highly expressed in 5-8F cells but was lowly expressed in 6-10B cells. The Gal-3 knockdown in 5-8F cells, Gal-3 overexpression in 6-10B cells and treatment with Gal-3 inhibitor revealed that Gal-3 was responsible for metastatic phenotypes including adhesion, migration and invasion. So Galectin-3 may serve as a potential target for NPC therapeutic interventions.


Results
Phenotypic characterization of highly and lowly metastatic NPC cells. In our study, we used two NPC cell lines including highly tumorigenic and metastatic (5-8F) and highly tumorigenic but not metastatic (6-10B) NPC cells, both of which were derived from the same parental cell line. The results revealed that the growth rates of both cells were similar (Fig. 1A), however, differential metastatic potentials of 5-8F and 6-10B were observed. Based on our data, 5-8F cells exhibited lower adhesive behaviors to adhere to the extracellular matrix (ECM) (Fig. 1B), but higher migrative and invasive capabilities compared to 6-10B cells. (Fig. 1C-F).

Differential glycoproteins in highly and lowly metastatic NPC cells. The experimental workflow
for our study is shown in a schematic representation (Fig. 2). Total proteins from cell lysates and culture media of 5-8F and 6-10B NPC cells were captured in ConA-and WGA-affinity chromatography specific to mannoserich and sialic acid-rich glycans, respectively. The proteins in the eluted fraction were stained with silver stain and periodic acid-Schiff (PAS), which is specific to glycosylated proteins to ensure the same loading amount. To characterize the proteomes, retrieved fractions were separated by SDS-PAGE and digested by trypsin prior to nano ESI MS/MS analysis. The results showed 353 proteins in ConA-enriched fraction in cell lysates (Fig. 3A) and 261 in culture medium (Fig. 3B). There were 76 and 47 common ConA-enriched proteins present in both 5-8F and 6-10B cell lysates and culture medium respectively. For WGA-enriched fractions, a total of 189 and 170 proteins were identified in cell lysates (Fig. 3C) and culture medium (Fig. 3D) respectively. In addition, the overlapping proteins found in both two cell lines were 40 in cell lysates and 48 in culture medium. In cell lysates, there were 319 (74.2%) proteins unique for ConA-enriched fractions and WGA-enriched fractions in 5-8F and 6-10B cells and only 28 (6.5%) lectin-binding proteins identified in all fractions (Fig. 3E). In culture medium, there were 31/368 (8.4%) common lectin-binding proteins identified in all fractions and 285 (77.4%) unique for each fraction in both cell types (Fig. 3F).
Differentially expressed glycoproteins were clustered using DAVID gene functional classification tool. The analysis of the abundance of each functional category revealed substantial differences in all conditions. Here, we focused on the proteins that were involved in cancer metastasis, especially cell-cell adhesion, extracellular matrix, wound healing and secreted proteins (Tables 1, 2, 3 and 4). There were proteins that play roles in cell-cell adhesion found differentially expressed in 5-8F and 6-10B cells such as CD44, MUC13, transferrin receptor and major histocompatibility, which were highly expressed in 5-8F cells, while integrin alpha 3 and 5 were only expressed in 6-10B cells. For secreted proteins from culture medium, soluble proteins such as s100A8 and s100A9, which have been shown to be associated with cancer cell invasion, were only detected in highly metastatic 5-8F cells, but not lowly-metastatic 6-10B cells. Candidate proteins were further selected by considering proteins that exhibited more than 1.5-fold of protein scores corresponding to the EMPAI values. Interestingly, the expression of galectin-3 was found ~ 2.1-fold higher in 5-8 cells compared to 6-10B cells, while galectin-3 binding protein was found only in WGA-enriched fraction in 5-8F, but not 6-10B cells. Based on evidence for its roles in metastatic potential in cancer cells, the availability of specific inhibitors and no previous reports in NPC, galectin-3 was selected for further validations.
Galectin-3 is associated with metastasis in NPC. To validate the differential expression of Galectin-3 in both types of NPC cells, Galectin-3 antibody was then probed to each fraction of proteins. The immunob-Scientific RepoRtS | (2020) 10:16462 | https://doi.org/10.1038/s41598-020-73498-y www.nature.com/scientificreports/ lotting results were in accordance with the protein score obtained from the proteomic analysis (Fig. 4A,B). To further confirm these findings, we subsequently evaluated the expression of Galectin-3 in total whole cell lysates and culture medium from 5-8F and 6-10B cells. The results revealed that 5-8F cells and the medium derived thereof expressed Galectin-3 protein higher than 6-10B cells (Fig. 4C,D). Thus, the findings from both proteomic analyses and in vitro cell cultures were highly in agreement to each other. To further investigate the functional role of Galectin-3 in NPC cells, the siRNA was introduced to the Galectin-3 highly expressed 5-8F cells and the gal-3 expression plasmid was introduced to the Galectin-3 poorly expressed 6-10B cells. The immunoblotting analyses revealed that we successfully generated the gal-3 knockdown 5-8F cells and the Galectin-3 overexpressing 6-10F cells (Fig. 5A,B). The phenotypic characterization on these cells, together with the treatment with modified citrus pectin as a Galectin-3 specific inhibitor was performed.

Discussion
Tumor metastasis is one of the major causes of mortality in cancer patients. The metastatic process involves tumor cell extravasation and subsequent migration and invasion into adjacent tissues, all of which are tightly controlled by cell surface mechanisms 13 . It is evident that aberrations in certain glycosylated proteins play an important role in cancer metastasis 14 . Thus, the discovery of differentially expressed glycoproteins during the metastatic process will help to unravel the fundamental biomolecular activities associated with cancer metastasis and potentially can prompt us to explicate novel therapeutic targets. Recent advancements in MS technology have boosted high-throughput analyses of many proteins and glycoproteins. In this study, two immobilized lectin affinity chromatography (ConA and WGA) normally used to enrich glycoproteins were applied to two different biological systems. These lectins are commonly used for reproducibly enriching glycoproteins from serum or plasma 12 . ConA primarily recognizes alpha-linked mannose and terminal glucose residues, which are a part of the N-glycan core structure, while WGA exhibits a primary affinity to GlcNAc groups and bind to sialic acid 12 . The combination of separation techniques using lectin affinity chromatography and a suitable mass spectrometry in proteomic evaluation would simply provide higher discovery rate of glycoproteins as effectively and reproducibly used for proteomic studies 15 . As NPC occurs in a silent small area, patients are usually diagnosed at late stages with higher rates of metastasis. Hence, in this study, we performed the comparative proteomics of ConA-and WGA-enriched fractions in highly and lowly metastatic NPC cells in order to retrieve glycoprotein biomarkers responsible for NPC metastatic process. Glycoproteins in cell lysates and culture media were enriched by lectin affinity chromatography. After digestion and LC MS/MS analysis, the enrichment fraction profiling in both types of NPC cells were then elucidated. In cancer cells, glycoproteins with mannose-rich glycan moieties may regulate key cellular processes, which drive oncogenes and corresponding proteins involved in potential cell progression, proliferation and cancer cell transformation 16 , while changes in sialylation may enhance integrin-mediated cellular adhesion or inhibit interactions to integrin-extracellular matrix, leading to potentially facilitating cancer aggressively spread and eventual metastasis 17 . Based on our data, many lectin-specific glycosylated proteins were identified only in highly metastatic 5-8F cells including UDP-glucose:glycoprotein glucosyltransferase 1 (UGGT1), dolichyl-diphosphooligosaccharide protein glycosyltransferase subunit 1 (RPN1), GDP-fucose protein O-fucosyltransferase 2 (POFUT2), and galactosylgalactosylxylosyl protein 3-beta-glucuronosyltransferase 3 (B3GAT1), whereas, only GDP-fucose protein O-fucosyltransferase 1 (POFUT1) was found in lowly metastatic 6-10B cells. It has been well established that glycosylation processing pathways are disturbed during carcinogenesis and tumor progression as resulted from aberrant activity of glycan modification enzymes, affecting various aspects of the biological behaviors of cancer cells including metastatic potential [18][19][20] . In accordance, differential expression of lectin-binding proteins involved in glycosylation processing pathways between highly and lowly metastatic NPC cells supported this notion. It is notable that several proteins which are recognized as metastasis-related molecules were differentially expressed in both NPC cell types. These include adhesion molecules such as L1 cell adhesion molecule (L1CAM), www.nature.com/scientificreports/ activated leukocyte cell adhesion molecule (ALCAM), major histocompatibility complex (MHC), CD44, CD166, integrin, vinculin, melanoma-associated antigen B2 (MAGE-B2), as well as other related proteins including mammalian ependymin-related protein 1 (MERP1), fibulin-7 (FBLN7), glial fibrillary acidic protein (GFAP), hornerin serpin H1 (SERPINH1), calumenin (CALU), aminopeptidase N (CD13), annexin A1 (ANXA1), peroxiredoxin 1, 2, and 4 (PRDX1, 2, 4), alpha-2-HS-glycoprotein (AHSG), cathepsin Z (CTSZ), protein S100 A8 and A9. Among the identified proteins, Galectin-3 was selected for further analysis of the protein levels and the extent of its possible role in NPC metastasis. Galectin-3 belongs to a soluble glycoprotein binding family of immunoregulatory lectins, which exhibits an affinity for β-galactosides via a carbohydrate recognition that is either β-1,3 or β-1,4 linked to N-acetylglucosamine 21 . Although, thus far, there has been no experimental report whether Galectin-3 is glycosylated, however, based on artificial neural networks, the glycosylation has been predicted for this protein. It is also possible that Galectin-3 might bind to other glycoproteins, forming complex molecules, which could be pulled out together during the enrichment process. Galectin-3 can be found in both intracellular and extracellular compartments of the cells. The intracellular Galectin-3 can regulate cell proliferation, differentiation, survival, and cell death via many effector proteins such as K-Ras protein, Akt protein, and annexin VII, while the extracellular Galectin-3 mediates cell adhesion and cell signaling [22][23][24] . The released Galectin-3 can bind to components of extracellular matrix and its membrane counterparts of other cells, which have in a wide range of biological events including cellular homeostasis, immune function, angiogenesis, tumor invasion and metastasis 25 . Galectin-3 has been reported as a diagnostic or prognostic marker linked to metastasis in many types of cancer such as thyroid 26,27 , breast cancer 28 , melanoma 29 , lung cancer 30 , sarcoma 31 , gastric cancer 32 , prostate cancer 33 and oral tongue cancer 34 .
Here, we showed that Galectin-3 was greatly expressed in highly metastatic NPC cells compared to lowly metastatic cells. Furthermore, the downstream investigations including knockdown and overexpression of Galectin-3 in NPC cells as well as the treatment with specific inhibitor revealed that indeed galectin was involved in NPC cell metastatic phenotypes including adhesion, migration and invasion. For cell adhesion, a few studies demonstrated that overexpression of Galectin-3 enhance cell adhesion to ECM 35 . Our study showed that Galectin-3 rendered NPC cells to lose their ability to attach to ECM. These data have been linked to other investigations that various   N, 108 N, 117 N, 152 N, 156 N, 291 N,  372 N, 440 N, 447 N, 449 N, 465 N, 503 N,  546 N, 588 N, 625 N, 652 N, 670 N, 786 N,  823 N, 863 N, 869 N, 901 N, 957 N, 965 N,  1060 N, 1209 ,N 1213 N, 1331 N, 13,777 N, 1386 N, 1467 N, 1499 N, 1741 N, 1740 N, 1767 N, 1773 N, 1821 N, 1901 N, 1945 N, 2184 N, 2225 N www.nature.com/scientificreports/ types of cancer cells which were incubated with Galectin-3 prior to plating, exhibited significantly reduced adhesion to ECM. It has been suggested that this phenomenon occurred as a result of interaction between Galectin-3 and alpha 1 beta 1 integrin 36 . Loss of cell adhesion allows cancer cells to escape from their site of origin and acquire a more motile phenotype. For migration and invasion, similar results have been observed in other cancers. Overexpression of Galectin-3 significantly increased migration and invasion whereas gal-3 knockdown inhibited both processes in oral tongue squamous cell carcinoma 34 . Moreover, an siRNA against gal-3 reduced migration and invasion in tongue cancer cell lines 37 . It has been proposed that Galectin-3 may regulate metastatic phenotypes via the Wnt/β-catenin signaling pathway 35,37 . Indeed, our data pointed to the possible involvement of the active β-catenin and potentially MAPK, AKT and NF-κB pathways. However, further studies are warranted to define the exact roles of these pathways in Galectin-3 mediated metastasis. In agreement with our findings, an immunohistochemical evaluation of 45 undifferentiated NPC tissues revealed that overexpression of Galectin-3 were independently correlated with poor overall survival 38 .
In summary, the current study provides clues for the involvement of a list of lectin-specific glycosylated proteins in NPC metastasis. The data from our findings will provide researchers more understanding about glycoproteins linked to metastasis and may help to develop targeted therapeutic drugs to reduce NPC progression. Galectin-3 has been shown to play a pivotal role in NPC metastasis in vitro. Further investigations including in vivo studies must be performed to determine whether Galectin-3 could be used for therapeutic intervention in human NPC metastasis.  www.nature.com/scientificreports/ 100 U/ml penicillin and 100 µg/ml streptomycin, at 37 °C in a humidified 5% CO 2 incubator. In certain experiments, cells were exposed to modified citrus pectin at the final concentration of 500 µg/ml for 24 h.

Enrichment of lectin-specific glycoproteins in NPC cells.
NPC cells (2 × 10 6 cells) were plated in 10 cm 2 culture dish for 48 h, and then were grown in serum free medium overnight. Cells and culture medium were collected at 72 h, lysed with specific lysis buffer and concentrated by centrifugation evaporator prior to the enrichment. Proteins (80 µg) were retrieved using lectin resin filled spin columns in Qproteome total glycoprotein kit (Qiagen, MD, USA) to capture ConA-enriched and WGA-enriched fractions. Briefly, the spin column was washed with 0.5 ml binding buffer containing 5 µl of protease inhibitor. Proteins were then loaded onto the spin column and incubated for 1 min. The unbound fraction on the column was removed by centrifugation at 100 × g for 2 min. The columns were washed twice with ice-cold binding buffer. The lectin binding proteins were then eluted, and the protein concentrations were measured by BCA method.

SDS-PAGE and in-gel digestion.
SDS-PAGE and in-gel digestion were performed as described previously with slight modifications 40 . Briefly, protein samples (30 µg) were resuspended in 4 × SDS sample buffer and subjected to 12% SDS-PAGE. SDS gels were stained with Coomassie Brilliant Blue G250, and destained in ultrapure water. For in-gel digestion, each gel lane was cut in pieces and destained in 50% acetonitrile in 50 mM NH 4 HCO 3 until colorless and then newly prepared 10 mM dithiothreitol in 50 mM NH 4 HCO 3 was added to reduce proteins at 60 °C for 15 min. Gel pieces were then cooled down to room temperature and newly prepared 55 mM iodoacetamide in 50 mM NH 4 HCO 3 were added to alkylate the proteins for 30 min in the dark at room temperature. Thereupon, remaining solutions were removed and added with absolute acetonitrile to dehydrate the gel pieces. The gel pieces were let to completely dry. Trypsin enzyme solution (0.01 mg/ml) was directly added to digest the gel pieces and incubated at 37 °C overnight. To extract all peptides by directly adding 50% www.nature.com/scientificreports/ acetonitrile and eventually concentrated using an evaporator. The samples were kept at − 20 °C before to mass spectrometric analysis.
Nano LC-MS/MS and data processing. Nano LC-MS/MS and data processing were performed as described previously with some modifications 41 . Briefly, all tryptic digested samples were resuspended in 0.1% formic acid and then analyzed by micrOTOF-Q II mass spectrometer (Bruker; Bremen, Germany) coupled with an UltiMate 3000 nano-LC system (Dionex; Surrey, UK). The separation flow rate was done at 300 nl/min. HPLC grade mobile phase A (0.1% (v/v) formic acid, 2% (v/v) acetonitrile in water) and HPLC grade mobile phase B (0.1% (v/v) formic acid in acetonitrile) were used to establish 50 min gradient. The gradient initiated with 10 min 2-10% B, followed by 33 min 10-40% B, ramped expeditiously (1 min) 40 -95% B and maintained at 95% B for 1 min. The injection eluent was sprayed and ionized in the nano-electrospray source of mass spectrometer. Data were gained using Hystar software. The MS and MS/MS spectra were collected in mass range of m/z 400-2000 and m/z 50-1500, respectively. The data of mass spectrometric were converted into a mascot generic file (.mgf) using Mascot v.2.3.0 data analysis software version 4.0. (Matrix Science, London, UK) was used to discover data against SWISSPORT human protein database using trypsin enzyme with one possible missing cleavage allowed at 1 and the identified proteins were represented with more than one peptide. The precursor and fragment of mass tolerances ions were set to 1.2 Da and 0.6 Da, respectively. The peptide charge was chosen as 2+, 3+ and 4+ . Methionine oxidations and cysteine carbamidomethylation were set as variable modifications to reduce false positive identification. As a result of the same amount of proteins loaded into SDS-PAGE gel, semi-quantification information of 5-8F and 6-10B were achieved using exponentially modified protein abundance index (emPAI) provided by the Mascot. All protein identification lists were submitted predicted N-Glycosylation sites in human proteins using artificial neural networks using NetNGlyc 1.0 Server.

Western blot analysis.
Western blot analysis was performed as described previously with slight modifications 42 . Briefly, proteins from cell lysate were lysed with lysis buffer and centrifuged for 15 min at 12,000×g at 4 °C. The cell lysate supernatants were collected, and protein concentration was measured by Bradford assay. All samples from both cell lysates and culture media (~ 30 μg of protein/lane) were separated on 15% SDS-PAGE, and then transferred onto nitrocellulose membranes. Blots were blocked with 1% w/v bovine serum albumin (BSA) for 30 min at room temperature and were probed with anti-galectin-3 monoclonal antibody (1:500) and anti-β actin antibody (1:5,000) at 4 °C overnight. After incubation, the membranes were washed with phosphate saline buffer (PBS) supplemented with 0.1% Tween-20 and were incubated with corresponding secondary antibodies conjugated with horseradish peroxidase (1:1000) for 1 h. Finally, the immunoreactive bands were detected by chemiluminescence (Cell Signaling Technology).
siRNA and plasmid transfection. Gal-3 siRNA was used in 5-8F cell transfection and Gal-3 expression plasmid was used to transfect 6-10B cells. Briefly, cells were seeded in 6 well plates and were incubated for 24 h. The cells were transfected with 40 nM of Gal-3 siRNA to 5-8F cells and 100 ng of Gal-3 expression plasmid to 6-10B cells using lipofectamine. Following 24 h incubation, culture medium was replaced, and the cells were further incubated for 48 h. Gal-3 expression in these cells were verified through Western blotting analysis.
Cell proliferation assay. Cells were seeded in 96-well plates and subsequently cultured in growing medium.
MTT (3-(4,5-dimethythiazol-2-yl)-2,5-diphenyl tetrazolium bromide) solution (0.5 mg/ml) was added to the cells after 24, 48, 72, 96, and 120 h and incubated for 3 h. Following the medium removal, 100 μl of 1:1 dimethyl sulfoxide and ethanol was added to each well to dissolve formazan crystals. The absorbance was measured at 540 nm using microplate reader. The proliferation rates were presented as a percentage of the control.
Cell extracellular matrix adhesion assay. Cells supplemented in serum-free medium with 0.1% BSA were seeded in 24 well plates coated with Matrigel and were incubated for 2 h. Subsequently non-adherent cells were removed gently by washing twice with PBS and adherent cells were subjected to MTT assay. The absorbance of each well was measured at 540 nm. The adhesion index were presented as a relative value compared to the control.
Migration assay. Cells were seeded into 24 well plates for 24 h to reach ~ 100% confluence. The wounds were created by scratching in each confluent monolayer using a pipette tip. Then cells were washed twice with PBS to remove cell debris. After 0 and 24 h, the images of cells were captured by a phase contrast microscope with a 10 × objective lens. The wound enclosure areas were determined using TScratch software to measure the distance traveled during the desired time 43 .
Invasion assay. Invasion assay was performed using Transwell culture plates as described previously with slight adjustments 44 . Briefly, chambers were pre-coated with Matrigel and culture medium for overnight. Cells were then seeded into the upper chamber in 0.5 ml culture medium with 1% FBS. Culture medium supplemented with 10% FBS was added in the lower chamber as a chemoattractant. After incubation for 24 and 48 h, cells migrated to the lower surface of the filters. The chambers were fixed in 0.1% paraformaldehyde followed by 4% paraformaldehyde in PBS for 30 min, and then stained with crystal violet. The invasive cells were expressed as the average number of cells per microscopic field, from at least five fields of view per filter.
Scientific RepoRtS | (2020) 10:16462 | https://doi.org/10.1038/s41598-020-73498-y www.nature.com/scientificreports/ Experimental design and statistical rationale. A pool of proteins from three biological replicates of 5-8F and 6-10B cell lysates and culture medium was processed and subjected to mannose-rich and sialic acidrich lectin columns. All the samples were processed with two technical replicates. For all other experiments, at least 3 independent replicates were performed, and the representative data are shown in Figures. Data were expressed as the mean ± standard error from at least 3 independent experiments. All data were considered significant at P values < 0.05. Statistical analysis was executed using SPSS 13.0 (SPSS, IL, USA) to compare between groups.