Calreticulin as A Novel Potential Metastasis-Associated Protein in Myxoid Liposarcoma, as Revealed by Two-Dimensional Difference Gel Electrophoresis

Myxoid liposarcoma (MLS) is a mesenchymal malignancy. To identify innovate seeds for clinical applications, we examined the proteomes of primary tumor tissues from 10 patients with MLS with different statuses of postoperative metastasis. The protein expression profiles of tumor tissues were created, and proteins with differential expression associated with postoperative metastasis were identified by two-dimensional difference gel electrophoresis (2D-DIGE) and mass spectrometry. The validation was performed using specific antibodies and in vitro analyses. Using 2D-DIGE, we observed 1726 protein species and identified proteins with unique expression levels in metastatic MLS. We focused on the overexpression of calreticulin in metastatic MLS. The higher expression of calreticulin was confirmed by Western blotting, and gene silencing assays demonstrated that reduced expression of calreticulin inhibited cell growth and invasion. Our findings suggested the important roles of calreticulin in MLS metastasis and supported its potential utility as a prognostic biomarker in MLS. Further investigations of the functional properties of calreticulin and other proteins identified in this study will improve our understanding of the biology of MLS and facilitate novel clinical applications.


Introduction
Myxoid liposarcoma (MLS) is a common type of soft-tissue sarcoma, accounting for approximately one-third of all liposarcomas and 10% of all adult soft-tissue sarcomas [1]. Clinical features of MLS are diverse, ranging from localized and curable tumors to metastatic tumors, which can cause tumor-related death. MLS has unique molecular characteristics, including the recurrent translocation t(12:16)(q13:p11) and the corresponding chimeric protein FUS-CHOP (DDIT3, CREBP-homologous protein/DNA damage inducible transcript 3), which is a target of the drug trabectedin [2][3][4][5]. Despite the fact that the unique molecular features are well known in MLS, the origin of MLS remains

Patients
This study included 10 patients with MLS who underwent curative resection from 1996 to 2007 at National Cancer Center Hospital (Tokyo, Japan). The patients in this study had not received neoadjuvant therapy at the time when the tumor tissue was sampled. The clinical and pathological characteristics of the patients in this study are summarized in Table 1. This study was approved by the ethical committees of National Cancer Center Hospital (2004-050), and the informed consent was obtained from all patients (Supplementary Materials).

Protein Expression Profiling
Proteins were extracted from frozen tumor tissues according to our previous report [12]. Briefly, tumor tissues were frozen after the surgical operation and stored until use at −80 degree. The frozen tissues were crushed and powdered in the presence of liquid nitrogen using a Multi-beads shocker (Yasui-kikai, Osaka, Japan). The tissues were then treated with protein lysis buffer (2 M thiourea, 6 M urea, 3% CHAPS, and 1% Triton X-100) for 30 min. Protease inhibitors were not used at any stage. After centrifugation at 15,000 rpm for 30 min, the supernatant was recovered and used as the protein lysate, and the resultant pellet was discarded.
Protein expression profiles were created using two-dimensional difference gel electrophoresis (2D-DIGE) according to our previous report [12]. Briefly, an internal standard sample was created by combining equal amounts of all individual protein samples. The internal standard sample was labeled with Cy3, individual samples were labeled with Cy5 (CyDye DIGE Fluor saturation dye; GE Healthcare, Uppsala, Sweden), and 5 µg of each labeled sample was combined and separated by two-dimensional polyacrylamide gel electrophoresis (2D-PAGE). Separation in the first dimension was achieved using a non-linear IPG dry-strip gel (GE Healthcare, Immobilize DryStrip, pH3-10 NL, 24 cm, GE Healthcare). The second dimension separation was achieved by SDS-PAGE using our original, large format gels (separation distance: 33 cm) [12]. The gels were scanned with laser scanners using appropriate wavelengths (Typhoon Trio; GE Healthcare). To compensate for gel-to-gel variations, Cy5 intensity was standardized against the intensity of Cy3 for all protein spots in an identical gel using image analysis software (Progenesis SameSpot; Nonlinear Dynamics, New Castle, UK). All samples were examined in triplicate, and the averaged standardized spot intensities were used for comparative studies. Proteome data were statistically examined using Expressionist software (Genedata, Basel, Switzerland).

Protein Identification
Proteins in the protein spots of interest were identified by mass spectrometry according to our previous report [12]. Briefly, 100 µg protein was labeled with Cy3 fluorescent dye (CyDye DIGE Fluor saturation dye; GE Healthcare) and separated by 2D-PAGE as described above. The proteins included in the protein spots were recovered as gel plugs using our original automated spot recovery machine and digested with modified trypsin (Promega, Madison, WI, USA). The trypsin digests were separated using liquid chromatography (Paradigm MS4 dual solvent delivery system; Michrom BioResources, Auburn, CA, USA) and subjected to LTQ linear ion trap mass spectrometry or LTQ Orbitrap XL analysis (ThermoElectron, San Jose, CA, USA) equipped with a nano-electrospray ion source (AMR, Tokyo, Japan). Mascot software (version 2.3.01; Matrix Science, London, UK) was used to search for peptide mass ion peaks against the SWISS-PROT database (Homo sapiens, 471,472 sequences in sprot_57.5 fasta file). Proteins with a Mascot ion score of at least 34 were used for protein identification.

Western Blotting
For separation by molecular weight, protein samples (5 µg) were separated by SDS-PAGE using 10% polyacrylamide gels (ATTO, Tokyo, Japan). For separation according to the isoelectric focusing point and molecular weight, protein samples (50 µg) prepared for 2D Western blotting experiments were separated using IPG dry-strip gels (pI range: 3-10; length: 18 cm; GE Healthcare) for the first dimension and 10% SDS-PAGE (ATTO) for the second dimension. 2D separation was performed as mentioned above. Following electrophoresis, proteins were transferred to polyvinylidene difluoride membranes (Immun-Blot PVDF Membrane for Protein Blotting, Bio-Rad, Hercules, CA) and reacted with anti-calreticulin antibodies (rabbit polyclonal antibody, anti-Calreticulin (#06-661) Upstate, Merck Millipore, Burlington, MA) at a dilution of 1:1000. After incubation with secondary antibodies (ECL Anti-rabbit IgG, Horseradish Peroxidase-Linked Species-Specific F(ab')2 Fragment, GE Healthcare) at a dilution of 1:2000, immunoreactive spots were detected by enhanced chemiluminescence (ECL PRIME; GE Healthcare) and LAS-3000 (FujiFilm, Tokyo, Japan). The intensity of each protein spot was quantified (ImageQuant; GE Healthcare).

Immunohistochemistry
Formalin-fixed paraffin-embedded tissues were sectioned and subjected to immunohistochemical analysis. The sectioned tissues were reacted with antibodies against calreticulin (mouse monoclonal antibody; FMC 75; cat. no. ab22683; Abcam, Cambridge, UK) at a dilution of 1:1000 or Negative Control Mouse IgG1 (cat. no. X0931; Dako) at a dilution of 1:4000. After blocking with skim milk, the sectioned tissues were reacted with EnVision+System-HRP Labelled Polymer Anti-mouse Antibodies (cat. no. K4001; Dako).

Cell Proliferation Assay
2645/94 cells transfected with calreticulin siRNAs or negative control siRNA were seeded in 96-well plates at 5000 cells/well. Cell viability was measured using a Cell Counting Kit-8 (Dojindo, Kumamoto, Japan) according to the manufacturer's instructions. After 2 h, the number of viable cells were determined by measuring the absorbance at 450 nm with a SAFIRE plate reader (TECAN, Männedorf, Switzerland). Cell proliferation was examined every 24-72 h after seeding.

Cell Invasion Assay
Cell invasion was evaluated using BD Biocoat Matrigel Invasion Chambers (BD Biosciences, Bedford, MA, USA). Briefly, 2645/94 cells were transfected with calreticulin siRNA-1, siRNA-2, or siRNA-3 or negative control siRNA. Cells were seeded into the upper chambers at 0.5 × 10 5 cells/well in serum-free RPMI medium, and medium containing 10% fetal bovine serum was added to the lower chamber. Forty-eight hours later, after fixation and staining cells, the numbers of cells on the bottom surface were counted in three separates areas at 100× magnification, and ratios of the number of cells in the Matrigel chamber to the number of cells in the control chamber (% invasion) were calculated.

Protein Expression Profiles of MLS Cells Using 2D-DIGE
Protein expression profiles of primary tumor tissues were created by 2D-DIGE. Figure 1A illustrates the design of experiments using an internal standard sample ( Figure 1A). All gels included the Cy3 image of common internal standard sample. By standardizing the Cy5 intensity of individual protein samples with the Cy3 intensity of corresponding protein spots of an internal standard sample, we compensated for gel-to-gel variations. Using a large-format gel electrophoresis apparatus, we observed 1726 protein species in single 2D-gel images. A typical proteome image is shown in Figure 1B, and its enlarged image is shown in Supplementary Figure S1. Note that all protein spots were well focused and separated from each other in all regions of the gel images. System repeatability was verified by scatter grams using data from triplicate experiments with identical samples ( Figure 1C).
We found that the standardized intensities of at least 96.3% of protein spots scattered within two-fold differences, with a correlation value of more than 0.793. According to the results of scatter gram analysis, we determined that when the spot intensity was different more than two times, that difference is significant in this experiment system.
Proteomes 2019, 7 5 System repeatability was verified by scatter grams using data from triplicate experiments with identical samples ( Figure 1C). We found that the standardized intensities of at least 96.3% of protein spots scattered within two-fold differences, with a correlation value of more than 0.793. According to the results of scatter gram analysis, we determined that when the spot intensity was different more than two times, that difference is significant in this experiment system. Figure 1. Overview of the proteomic study using two-dimensional difference gel electrophoresis (2D-DIGE). (A) The internal control sample and individual sample were labeled with Cy3 and Cy5, respectively, mixed together, and then separated by two-dimensional polyacrylamide gel electrophoresis (2D-PAGE). The gel was scanned using a laser scanner, and the Cy3 and Cy5 images were obtained. To normalize for gel-to-gel variations, the intensity of protein spots on the Cy5 image was normalized to the intensity of corresponding protein spots on the Cy3 image. (B) A typical Cy3 image is shown. Proteins were separated according to their isoelectric point and then their molecular weight. Note that proteins were well separated without remarkable streaking, and the image is not distorted overall. (C) The system repeatability of 2D-DIGE data was evaluated by running identical samples three times.

Comparative Study between Sample Groups with Different Metastatic Statuses
We compare the proteome data obtained by 2D-DIGE between the patients with metastasis (cases 1-5, Table 1) and those without metastasis (cases 6-10, Table 1). The patients who had metastasis after surgical operation received chemotherapy (cases 1-5). One of those five patients had died when we performed this study. In contrast, the patients who did not have metastasis (cases 6-10) did not receive adjuvant chemotherapy, and all of them were alive when this study was conducted. We compared the spot intensity between these two patient groups and identified the protein spots with differential intensity with statistical significance. The proteins were extracted from the identified protein spots and subjected to mass spectrometry for protein identification. The distribution of protein spot intensities is shown as a function of p values and fold differences between samples from patients with or without metastasis ( Figure 2A). The intensities of 149 protein spots were significantly different between the two sample groups (Figure 2A). Unsupervised classification by principal component analysis using all proteome data suggested that the protein samples could be classified according to the metastatic status. Moreover, when the intensities of the selected 149 protein spots were used, the sample groups were further separated ( Figure 2C). These observations suggested that the overall features of the proteome may be associated with the metastatic status and that the selected proteins may represent the characteristics of two sample groups. The intensities of the 149 protein spots in the 10 samples are illustrated as a heat-map ( Figure 2D), and its enlarged image is shown in Supplementary Figure S2. The spot intensities seemed to be homogenous in each sample group, and the intensity difference was obvious between the sample groups. We found that the intensities of 13 and 136 protein spots were higher and lower, respectively, in tumors with metastasis. Among the 149 protein spots with different intensities, proteins corresponding to the 148 protein spots were identified by mass spectrometry.
The results of comparative analyses and protein identification are summarized in Table 2. The protein spots listed in Table 2 had significantly different intensities between metatstatic and non-metastatic MLS.

Calreticulin Expression was Associated with Metastasis
We further focused on the overexpression of calreticulin in MLS. We validated the overexpression of calreticluin in MLS with metastasis by Western blotting. First, we performed SDS-PAGE/Western blotting. However, we did not detect significant differences in intensity associated with metastasis (data not shown), probably because of possible posttranslational modifications or the presence of isoforms without expression difference. Alternatively, the proteins which have an epitope reacting to the anti-calreticulin antibody may exist at the same location of calreticulin in the SDS-PAGE gel. We then performed 2D-PAGE/Western blotting. In 2D-PAGE/Western blotting, calreticulin was observed as a single protein spot, suggesting that there were no obvious post-translational modifications ( Figure 3A). Quantification of Western blotting indicated that the expression level of calreticulin in 2D-PAGE/Western blotting was significantly higher in tumor tissues with metastasis than in those without metastasis (p < 0.05), with average differences of more than two-fold ( Figure 3B and C). These observations were consistent with the results of 2D-DIGE.
Proteomes 2019, 7 11 modifications or the presence of isoforms without expression difference. Alternatively, the proteins which have an epitope reacting to the anti-calreticulin antibody may exist at the same location of calreticulin in the SDS-PAGE gel. We then performed 2D-PAGE/Western blotting. In 2D-PAGE/Western blotting, calreticulin was observed as a single protein spot, suggesting that there were no obvious post-translational modifications ( Figure 3A). Quantification of Western blotting indicated that the expression level of calreticulin in 2D-PAGE/Western blotting was significantly higher in tumor tissues with metastasis than in those without metastasis (p < 0.05), with average differences of more than two-fold ( Figure 3B and C). These observations were consistent with the results of 2D-DIGE.

Immunohistochemical Localization of Calreticulin in Tumor Cells
We localized calreticulin in tumor tissues using immunohistochemistry. Using sectioned tumor tissues from all nine patients with MLS in this study, we found that there were no significant differences in calreticulin expression ( Figure S3).

In Vitro Functional Analysis of Calreticulin
The functional significance of calreticluin upregulation in metastatic MLS tissues was examined in cultured MLS cells. Transfection of 2645/94 cells with siRNAs against calreticulin resulted in considerable reduction of calreticulin expression compared with that in cells treated with control siRNA ( Figure 4A). Cell growth assays showed that siRNA silencing of calreticulin resulted in decreased cell proliferation compared with that in control siRNA-transfected cells (p < 0.05, Figure 4B). Moreover, siRNA-mediated silencing of calreticulin significantly upregulated the invasiveness of MLS cells (p < 0.01, Figure 4C,D). These observations suggest that calreticulin promoted tumor progression in MLS cells. blotting confirmed the differential expression of calreticulin. The specific antibody was reacted with the membrane to which two-dimensionally separated protein samples were transferred. (C) The intensity of protein spots on Western blotting was quantified and summarized in the graph.

Immunohistochemical Localization of Calreticulin in Tumor Cells
We localized calreticulin in tumor tissues using immunohistochemistry. Using sectioned tumor tissues from all nine patients with MLS in this study, we found that there were no significant differences in calreticulin expression ( Figure S3).

In Vitro Functional Analysis of Calreticulin
The functional significance of calreticluin upregulation in metastatic MLS tissues was examined in cultured MLS cells. Transfection of 2645/94 cells with siRNAs against calreticulin resulted in considerable reduction of calreticulin expression compared with that in cells treated with control siRNA ( Figure 4A). Cell growth assays showed that siRNA silencing of calreticulin resulted in decreased cell proliferation compared with that in control siRNA-transfected cells (p < 0.05, Figure  4B). Moreover, siRNA-mediated silencing of calreticulin significantly upregulated the invasiveness of MLS cells (p < 0.01, Figure 4C and D). These observations suggest that calreticulin promoted tumor progression in MLS cells.

Discussion
Profiling of the genes relevant to metastasis is the first step to understanding the mechanisms of malignant behaviors in tumors. Although the cancer biomarker discovery has been performed in many cancers using proteomics, there was no report about metastasis biomarkers of MLS until our study. Here, using a set of tumor tissue samples from the patients with different metastasis status, we examined the proteome and identified differentially expressed proteins. Those proteins are potentially biomarker candidates, and we focused on one of the identified proteins, calreticulin.
Calreticulin is a Ca 2+ -binding chaperon protein predominantly located in the endoplasmic reticulum [14]. Higher levels of calreticulin have been reported in various types of cancers [15] and are associated with clinical stages and metastasis in breast cancer [16], gastric cancer [17], prostate cancer [18], and ovarian cancer [19]. Moreover, the overexpression of calreticulin is relevant to shorter survival in pancreatic cancer [20] and esophageal cancer [21]. In bladder urothelial cancer, calreticulin in urine is a candidate biomarker [22]. These reports have suggested that calreticulin plays important roles in carcinogenesis and cancer progression. The possible mechanisms through which calreticulin mediates metastasis have been demonstrated in previous reports. The interaction of calreticulin with integrin α2β1 was observed in cells with active integrin [23], and calreticulin regulates integrin activation by modifying the α1,2-linkaged glycomic status of β1-integrin [24]. Calreticuln affects cell adhesion and migration by altering the components of the extracellular matrix, including fibronectin [25]. Proteins involved in the regulation of focal contacts by calreticulin include the calmodulin/calmodulin-dependent kinase II pathway [26] and vinculin [27]. Calreticulin also functions as an angiogenesis factor. Calreticulin positively regulates vascularization through the activation of vascular endothelial growth factor (VEGF) in gastric cancer [17]. Interestingly, vasostatin, which is a fragment of calreticulin, is considered an anti-angiogenic factor and inhibits VEGF-induced enthothelial cell proliferation [28]. These reports suggested that various mechanisms may explain the association of calreticulin with metastasis in various cancers.
Aberrant expression of carleticulin has been reported in sarcomas. For example, Hisaoka et al. investigated the expression of calreticulin in lipogenic tumors and other sarcomas [29]. They found that calreticulin was frequently overexpressed in dedifferentiated areas of dedifferentiated liposarcomas compared with that in normal surrounding tissue. Moreover, they demonstrated the various expression levels of calreticulin in MLS cases, without describing the association of calreticulin with metastasis. In our study, differences in calraticulin levels between metastatic and nonmetastatic MLS were detected by 2D-DIGE and 2D/Western blotting, but not by immunohistochemistry. The differences in methods may cause the apparent discrepancy. Indeed, protein detection can often vary depending on the method used. 2D-DIGE using saturation dyes detects proteins with cysteine residues after the proteins are highly separated by two-dimensional gel electrophoresis. Thus, proteins that are not labeled by fluorescent dyes, not separated well by electrophoresis, or not expressed at sufficiently high levels may not be identified by 2D-DIGE. In SDS-PAGE/Western blotting, the proteins are heated, separated according to their molecular weight, and transferred to membranes. Proteins that do not exhibit antigenic sequences on the membranes will not be detected. In 2D/Western blotting, proteins are separated by two-dimensional gel electrophoresis and transferred to membranes. Moreover, proteins showing antigenic sequences under these conditions may be detected by antibodies. In immunohistochemistry, the proteins are fixed with formalin and embedded in paraffin. After the fixed tissues are sliced, the proteins are detected by antibodies. Thus, proteins having sequences modified by formalin may react differently with antibodies. Western blotting is best suited for research in the laboratory because it is possible to obtain information regarding protein structure from electrophoretic separation. In contrast, immunohistochemical detection is best suited for confirming the localization of proteins and to perform multi-institutional validation studies.
We found that the knockdown of calreticulin inhibited proliferation and invasion in MLS cells. These observations were concordant with previous reports in other malignancies. In oral cancer cells, depletion of calreticulin was found to result in cell cycle arrest at the G 0 /G 1 phase, suppressing growth rates, colony-formation capacity, and anchorage-independent growth [30]. Additionally, knockdown of calreticulin suppresses the growth, adhesion, and migratory ability of the cells and blocks in vivo metastasis in bladder cancer [31]. In this study, we firstly confirmed the possible contribution of calreticulin to the malignant characters of MLS. These observations suggest that cancer cells with different histological and genetic backgrounds may share some common mechanisms for proliferation, metastasis and invasion.
The limitation of this study is firstly the small number of cases. Further studies are needed to assess the future clinical applications of MLS overexpression. MLS is a rare malignancy, and the frozen tissues are not routinely stored in the hospitals. The results of this study should be validated in the newly enrolled cases, and for this sake, we need to conduct a multi-institutional study. Secondly, we need to elucidate the mechanisms underlying the correlation between higher expression of calreticulin and metastasis in MLS. Thirdly, although we demonstrated the utility of the proteomic approach using 2D-DIGE to find proteins associated with metastasis in MLS, the total number of proteins observed in this study was limited. The pros and cons of 2D-DIGE for biomarker discovery were previously discussed [32], and it may be worth challenging other proteomics modalities such as label-free mass spectrometry and protein arrays. The other protein extraction and fraction methods are worth considering too. The proteome data are always biased by the technical characters of proteomics modalities, and a perfect proteome map is not available. However, as long as the same protocol is used for the comparative study, the results are reliable. Finally, although there are many proteins identified in this study as metastasis-associated ones, we focused only on calreticulin and did not study the rest of the identified proteins. The proteins identified but not examined in this study are also worth investigating.
We found that the expression of calreticulin was significantly correlated with metastasis. If the correlation is established after an extensive validation study, we will be able to predict the metastasis using the surgically resected tumor tissue and optimize the therapeutic strategy.

Conclusions
In this study, we found an association between the higher expression of calreticulin and metastasis in MLS. Further studies are worth conducting to reveal the possible clinical applications of our findings.