Bone and soft tissue sarcomas are rare malignant tumors [1]. Patients who exhibit a poor response to chemotherapy and develop metastasis continue to have a poor prognosis. Therefore, it is critical to identify proteins associated with tumor malignancy and chemoresistance as predictive biomarkers and novel targets in patients with bone and soft tissue tumors.

The use of high-throughput screening approaches, such as arraybased comparative genomic hybridization analyses and cDNA microarray technology, allows for the screening of several thousand DNA and mRNA sequences and can be used to identify genes relevant to the diagnosis and clinical features of tumors [2-14]. Comprehensive studies have identified several genes that may be involved in the development or progression of tumors, representing candidate biomarkers, and/or drug targets [2-14]. However, DNA sequencing and measurement of the mRNA expression alone cannot be used to detect posttranslational modifications of proteins, such as phosphorylation or glycosylation, or differences in protein stability, factors that play important roles in the malignant behavior of tumor cells [15-18]. Furthermore, many lines of evidence have indicated discordance between the mRNA expression and the protein expression [15-18]. Therefore, proteomic studies are critical tools for understanding the biology of tumors, as well as identifying biomarkers for various cancers. These difficulties undermine the potential advantages of global protein expression studies, an approach known as “proteomics”.

Standard proteomic techniques, such as two-dimensional gel electrophoresis (2DE) and mass spectrometry (MS), have been developed over the past three decades. Since the end of the 1990s, due to the development of high-throughput platforms, proteomics has allowed the simultaneous measurement of multiple protein products and protein modifications. Recently, our studies successfully identified various candidate proteins associated with the differential diagnosis [17,19-21], prognosis [18,21-27] and prediction of the response to chemotherapy [18,23,28] in patients with bone and soft tissue tumors. We also verified the predictive power of these variables using large validation cohorts to develop clinical applications of useful biomarkers. Most of the biomarkers were successfully confirmed; however, the roles of the proteins in the tumors remain unknown and functional analyses of the biomarkers (proteins) have yet to be conducted. Therefore, we performed functional studies of these biomarkers as ongoing proteomic studies.

The following section describes (i) proteomic technologies, (ii) how proteomic approaches have been applied to identify biomarkers in bone and soft tissue tumors, and (iii) our proteomic approaches to conducting functional analyses of biomarkers (proteins), followed by a few examples of our recent proteomic studies.

Proteomics is the large-scale study of proteins, including their structures and functions [29-32]. Unlike studies of a single protein or pathway, proteomic methods enable the researcher to obtain a systematic overview of the profiles of the expressed proteins, which in cases involving tumors, can ultimately improve the diagnosis, prognosis and management of the patient by revealing protein interactions affecting overall tumor progression [29-32]. Technologies used in proteomics research include electrophoresis, mass spectrometric technologies, protein labeling, protein arrays, antibody-based approaches, imaging and bioinformatics technology. In particular, mass spectrometry technologies are now high-throughput, allowing for the rapid and accurate identification of thousands of proteins present within a complex tumor specimen. Therefore, various technologies are now being employed to identify tumor-specific proteins in sarcomas using proteomics technologies. In this section, we briefly describe twodimensional difference gel electrophoresis (2D-DIGE) and GeLC-MS [33,34], as these technologies are the most frequently used methods for obtaining protein expression profiles in our proteomic studies [17-28].

2D-DIGE

We routinely employ 2D-DIGE for biomarker identification using surgical samples [17-28]. 2D-DIGE is an advanced variation of 2D-PAGE (two-dimensional polyacrylamide gel electrophoresis) that has the potential to address many of the drawbacks of classical 2D-PAGE [31,32]. 2D-DIGE is frequently applied in sarcoma proteomics, in which the overall features of the protein expression are correlated with the sarcoma phenotypes to identify the molecular background of cancer biology. 2D-DIGE generates 2,000-5,000 protein spots as quantitative proteomic data [31,32].

In 2D-DIGE, proteins are extracted from surgical samples and all protein samples are labeled with different fluorescent dyes before gel electrophoresis (Figure 1). We create a common internal control sample that includes a mixture of a small portion of all individual samples and label it with a fluorescent dye that differs from the dyes used to label the individual samples. The differently labeled internal control and individual samples are then mixed together and separated according to both the pH and molecular weight ranges using 2D-PAGE. Laser scanning can be used to obtain gel images, because all proteins are labeled with fluorescent dye before gel electrophoresis. These gel images provide data regarding protein spots as protein expression profiles. Protein spots whose intensity statistically differs between the groups examined are identified using software programs in each study [17-28]. Proteins corresponding to the spots of interest are identified using mass spectrometry.

Figure 1: 2D-DIGE: Proteins extracted from surgical samples. All protein samples are labeled with different fluorescent dyes. The internal control sample, a mixture of a small portion of all individual samples is labeled by Cy3, and the individual samples are labeled by Cy5. The differently labeled samples are then mixed together. The samples are separated according to both the pH and molecular weight ranges. Gel images are then acquired using laser scanning. Finally, interest protein spots selected using data mining are identified in the intact proteins using a mass spectrometer.

GeLC-MS

GeLC-MS involves SDS-PAGE, followed by in-gel tryptic digestion and liquid chromatography-tandem mass spectrometry [33,34]. The technology is a powerful approach for conducting proteomic analyses, and the method directly acquires protein profiles consisting of intact proteins (not protein spots). In our GeLC-MS approaches, the technology identifies 1,500-2,000 protein expressions as semiquantitative proteomic data in one run. We usually employ GeLCMS technology in functional analyses of bone and soft tissue sarcomas.

Using this technique, a protein sample for the analysis is separated using SDS-PAGE, and the entire gel lanes are excised and further subdivided into smaller sections (Figure 2). We usually slice each gel into 24 slices. The proteins in these gel sections are subsequently digested within the gel using trypsin. In addition, the generated peptides are analyzed using an LC-MS experiment to acquire information regarding peptide sequence coverage, and the spectral count values in order to identify proteins present in a particular sample of each study. The database search results for all slices of a biological sample are combined, yielding global protein identification and semiquantification for each sample using the Protomap method [35]. The Protomap method provides a rich set of protein data that reveal global changes in the volume, size, topography and abundance of proteins in complex biological samples.

Figure 2: GeLC-MS: Proteins are extracted from cell lines, including normal cell lines and treated cell lines (for example, those treated with siRNA). All protein samples are separated using SDS-PAGE. The gels are sliced into 24 gels in each lane. The cut gels are digested using in-gel trypsinization. The acquired peptides are analyzed using LC-MS to identify proteins. The results obtained from all slices are combined to generate semiquantified data using the Protomap method [35].

A comparison of the 2D-DIGE and GeLC-MS methods used for our proteomic studies

With respect to the comparison between the 2D-DIGE and GeLCMS methods, there are two important differences: “quantification” and “protein identification”. The 2D-DIGE can provide accurate quantification of protein spots, but the method cannot demonstrate the protein identity directly. Therefore, the protein spots need to be assessed by an additional process to identify the protein names. On the other hand, the GeLC-MS can provide all of protein names directly based on the profiles. However, the GeLC-MS cannot provide accurate quantification because it is only semi-quantitative. Our studies include the discovery of biomarkers and a functional analysis of the findings from the discovery studies. In the discovery study, we usually employ 2D-DIGE to identify novel biomarkers, because we need to obtain the exact expression profiles. In functional studies, we need to know the identity of the most abundant proteins that are related to the protein expression dynamics, including upregulation, downregulation and no change. Therefore, we usually use GeLC-MS for the functional analyses.

Identifying predictive biomarkers and drug targets for tumors is the most important goal of global protein and gene expression studies. Current gene expression profiling technologies have been used to identify upregulated or downregulated genes with prognostic value that can be used to predict the prognosis or chemosensitivity of soft tissue sarcomas [3,4,11-14].

In order to identify useful biomarkers using global protein expression studies, we conduct high-integrity and reliable studies consisting of three sets (Figure 3): (1) a discovery set that attempts to identify candidate biomarkers from the global protein expression profiles of the tissue samples (in our studies, we usually use 2D-DIGE for these analyses); (2) a confirmation set that is used to confirm the protein expression differences identified in the discovery set using other proteomic tools (in our studies, we usually use a Western blot analysis); (3) a validation set that is used to verify the predictive power of a biomarker on a large scale using numerous samples in order to develop biomarkers for clinical application (in our studies, we usually use immunohistochemistry and Western blot analyses).

Figure 3: Our strategy for conducting proteomic studies using bone and soft tissue sarcomas is herein described. To develop biomarkers (blue arrows), we usually employ a three-step process: (i) 2D-DIGE-based target identification, (ii) confirmation, and (iii) validation. For the functional analyses (yellow arrows), we employ protein-based analyses (proteomic technologies) and DNA- and RNA-based analyses. In this article, we described the protein-based analyses used for the functional studies of the identified biomarkers (proteins).
To develop biomarkers (blue arrows), surgical samples are collected from patients with bone and soft tissue tumors. We organize both the clinical samples and information to establish efficient strategies. Protein expression profiles are generated using 2D-DIGE and analyzed using data mining to identify biomarker candidates. The protein expression levels of the candidates are confirmed using Western blotting analyses, and/or immunohistochemistry. The diagnostic value of the biomarker candidates is verified using additional large variation cohorts. Finally, the validated biomarkers are subjected to novel clinical applications.
In the functional analyses (yellow arrows), we focus on both the interaction proteins and regulated proteins associated with the biomarker proteins as proteomic approaches. The novel findings generated by the functional analyses are verified in validation studies, and/or are used in subsequent studies. Finally, we hope that the novel findings will provide beneficial effects to patients.

With respect to the number of samples included in the discovery set, we usually employ 10 to 20 samples (example 10 vs 10, 7 vs 8, 5 vs 5, and so on) to develop the novel biomarkers. Using a large number of samples may generate abundant protein profiles, and then these results may provide a large amount of information that can be used to choose candidate novel biomarkers. However, we believe that it is critical for the discovery analyses to eliminate noise from samples, even if the sample set will be small. A noisy sample can easily obstruct the identification of novel findings, and provides incorrect results. In our experience, sample sets of 10 to 20 are able to identify novel biomarkers in the bone and soft tissue tumors successfully. Therefore, we believe our strategies regarding the samples are acceptable for sarcoma research.

In this section, we introduce pertinent proteomic studies that have been previously used to identify prognostic biomarkers for GISTs, synovial sarcomas and Ewing’s sarcomas, and chemosensitivity biomarkers for osteosarcomas.

GISTs

Gastrointestinal stromal tumors (GISTs) are the most common mesenchymal tumors of the gastrointestinal tract and are characterized by the expression of the kit oncogene. The tyrosine kinase inhibitor, imatinib, has been proven to be highly effective in treating these tumors [36,37].

In order to identify protein expression profiles that correlate with the prognosis of GISTs, we conducted a quantitative expression study of the intact proteins in GIST samples [24]. We compared the protein expression profiles between a poor prognosis group (eight cases) and a good prognosis group (nine samples). These comparisons identified 43 protein spots with different intensities in the two types of samples. Eight of the 43 protein spots corresponded to pfetin and had higher intensity in the good prognosis group. We confirmed the expression of pfetin using Western blot analyses.

As validation studies, we verified the expression of pfetin in 210 GIST cases using immunohistochemistry. These studies revealed 5-year metastasis-free survival rates of 93.9% and 36.2% for the patients with pfetin-positive and pfetin-negative tumors, respectively (P<0.0001) [24]. Univariate and multivariate analyses demonstrated the pfetin expression to be an independent prognostic factor in patients with GISTs. These results demonstrate that the pfetin expression can be used to correctly distinguish poor prognosis cases from good prognosis cases and suggest that pfetin is a useful biomarker that may contribute to the development of novel therapeutic strategies for treating GIST patients.

Synovial sarcoma

Synovial sarcomas are malignant mesenchymal tumors that are primarily characterized by the presence of a chromosomal translocation, t(X;18)(p11.2;11.2), representing the fusion of the SYT gene with SSX1, SSX2 or SSX4 [1].

In our study, we used a proteomic approach to develop prognostic biomarkers for synovial sarcomas using 2D-DIGE [22]. We used 13 surgical samples (obtained from eight synovial sarcoma patients with a good prognosis and five synovial sarcoma patients with a poor prognosis), and identified 20 protein spots whose intensity statistically differed between the two groups. Mass spectrometric protein identification demonstrated that these 20 spots corresponded to 17 distinct gene products. Three of the 20 spots corresponded to secernin-1 and had higher intensity in the good prognosis group.

With respect to validation studies, the prognostic performance of secernin-1 was also examined immunohistochemically in 45 synovial sarcoma patients. The 5-year survival rates were 77.6% and 21.8% for the patients with secernin-1-positive and -negative primary tumors, respectively (p<0.01). We concluded that secernin-1 may be used as a biomarker to predict overall and metastasis-free survival in synovial sarcoma patients.

Ewing’s sarcoma

Ewing’s sarcomas are malignant neoplasms of the bone and soft tissue. Ewing’s sarcomas are genetically characterized by the presence of EWS-FLI1 or another related gene fusion, and recent studies suggest that Ewing’s sarcomas may arise from the malignant transformation of mesenchymal, and/or neural crest stem cells [1].

Kikuta et al. [27] reported that the protein expression level of nucleophosmin (NPM1) is correlated with the prognosis of Ewing’s sarcoma.That study investigated the global protein expression profiles of Ewing’s sarcomas using 2D-DIGE and found statistically significant differences in the NPM1 protein expression levels between Ewing’s sarcoma patients with a poor prognosis and those with a good prognosis.

Furthermore, the prognostic performance of nucleophosmin was evaluated immunohistochemically in an additional 34 Ewing’s sarcoma cases. A univariate analysis revealed that the expression of NPM1 was significantly correlated with the overall survival (P<0.01). Additionally, in 29 of the 34 patients with localized disease at diagnosis, the univariate analysis demonstrated that NPM1 positivity was also a strong negative predictor of the overall survival (P<0.01). These results suggest that the expression of NPM1 defines a more aggressive subset of Ewing’s sarcoma patients and is a candidate prognostic marker for Ewing’s sarcoma.

Osteosarcoma

Osteosarcoma is the most common primary malignant bone tumor. It most frequently occurs in the second decade of life, with 60% of patients being under 25 years of age [1]. The response to preoperative chemotherapy provides critical information regarding the patient, and chemosensitive patients are divided into two groups based on the pathological percentage of necrosis [1].

To identify novel biomarkers of the chemosensitivity of osteosarcoma, we employed a proteomic approach (2D-DIGE) [18,23,38]. We generated protein profiles of 12 biopsy samples, including six poor chemosensitivity osteosarcomas and six good chemosensitivity osteosarcomas, according to the Huvos grading system. We compared the expression profiles between the two groups and found 55 spots that corresponded to 38 distinct proteins, including peroxiredoxin 2 (PRDX2). The protein expression of PRDX2 exhibited higher intensity in the poor responder group.

In order to validate the predictive value for chemosensitivity, we conducted a validation study using a Western blot analysis of additional osteosarcoma samples. The validation study also demonstrated that the poor responders had higher PRDX2 expression levels than the good responders. We concluded that PRDX2 is a candidate marker for chemosensitivity in osteosarcoma patients.

We previously reported that our proteomic approaches successfully identified various novel biomarkers for predicting the prognosis and chemosensitivity of bone and soft tissue tumors [17-28]. However, the predictive power of these biomarkers must be confirmed in large validation studies and functional analyses of the biomarkers (proteins) remain to be conducted. Therefore, we continue to research functional analyses of our identified biomarkers using proteomic technologies to identify their functions and roles in tumors. We usually focus on interaction proteins and regulated proteins (Figure 3). Hence, in this section, we describe (i) the identification of interaction proteins, and (ii) the identification of regulated proteins, as well as (iii) demonstrate our the results of our functional analyses of NPM1 in Ewing’s sarcoma using these proteomic technologies.

Identification of interaction proteins

Protein–protein interaction (PPI) networks provide valuable information regarding the understanding of cellular functions and biological processes [39-42]. With the tremendous increase in human protein interaction data, a network approach is used to understand the molecular mechanisms of disease, particularly with regard to cancer phenomena [39-42]. In the setting of cancer, PPI data provide insight into the distinct topological features of cancer genes, cancer classification and cancer-related subnetworks [39-42]. PPI data form signaling nodes and hubs that transmit pathophysiological cues along molecular networks that also provide integrated biological outputs, thereby promoting tumorigenesis and tumor progression, invasion and/or metastasis [39-42]. Therefore, analyses of PPIs are critical for understanding biological processes and developing effective strategies for cancer treatment. In our studies, we focus on the PPIs of the biomarkers identified in our proteomic studies in order to understand the functions of these protein biomarkers.

Identification of regulated proteins

The protein profiles regulated by biomarker proteins provide critical information for understanding the functions of the biomarker proteins [43]. These protein lists have the potential to offer important clues for understanding tumor biology and may include candidates for biomarkers and therapeutic targets. In our studies, we routinely use proteomic approaches to identify proteins regulated by the biomarker proteins using a transfection system. The cell lines are treated by either introducing genes encoding the biomarker proteins into the cells without an expression of the proteins (gain-of-protein effect), or removing the biomarker protein expression from the cell lines constantly expressing the proteins using RNAi (loss-of-protein effect). These analyses can be used to identify candidates for regulatory proteins of the biomarker proteins. This approach can also be used to provide critical information for understanding the functions and roles of these biomarkers in tumors.

Functional analyses of NPM1 in Ewing’s sarcoma

We previously reported NPM1 to be a predictive biomarker for the prognosis of Ewing’s sarcoma in patients identified using proteomic [27]. NPM1 is a ubiquitously expressed protein belonging to the nucleoplasmin family of nuclear chaperones, and a highly conserved nucleocytoplasmic shuttling protein that shows restricted nucleolar localization [44-48]. NPM1 is frequently translocated or mutated in hematological malignancies, and mutations of the NPM1 gene leading to aberrant cytoplasmic dislocation of nucleophosmin (NPMc+) occur in approximately one-third of acute myeloid leukemia patients, who exhibit distinct biological and clinical features [44-48]. Although one article revealed a list of interaction proteins with NPM1 in Ewing’s sarcoma, the functions of NPM1 in Ewing’s sarcoma still remain unknown [49]. Therefore, we used proteomic approaches that consisted of the identification of both interaction proteins and regulated proteins associated with NPM1 proteins.

In the PPI analyses, we performed immunoprecipitation (IP) assays using two Ewing’s sarcoma cell lines (SKES1 and CHP100) and NPM1 antibodies to identify the expression profiles of interaction proteins physiologically associated with NPM1 (Figure 4). Proteins extracted from Ewing’s sarcoma cell lines were immunoprecipitated using either NPM1 antibodies or IgG antibodies (control). The IP samples were separated using SDS-PAGE and the gel images were compared between the NPM1 IP samples and the control samples. We found 20 bands with significantly different densities between the two groups (Table 1). The bands were treated with in-gel digestion, and the proteins were identified using MS spectrometry (Table 1). The proteins interacting with NPM1 are shown in Table 1.

Figure 4: Identification of interaction proteins associated with NPM1: Immunoprecipitation (IP) was performed using two Ewing’s cell lines (SKES1 and CHP100) and antibodies (NPM1 and IgG (Santa Cruz, TX)). The IP samples were separated using SDS-PAGE, and the gel images were compared between the NPM1 samples and IgG samples in each cell line. In this study, we identified 20 bands with significantly different densities in the two cell lines. We then identified the proteins included in each band using a mass spectrometer. The identified proteins are listed in Table 1.

Table 1: Protein list of interaction proteins associated with NPM1.

To identify protein expression profiles regulated by NPM1, we employed siRNA knockdown and GeLC-MS in four Ewing’s sarcoma cell lines (A673, TC71, SKES1 and CHP100), using NPM1 siRNA (Figure 5). The cell lines were transfected with either NPM1 siRNA or control siRNA and harvested after 72 hours. Proteins extracted from the cell lines were analyzed using GeLC-MS. We compared the acquired proteomic profiles between the control group and the siRNA group to calculate the semiquantitative expressions. The comparisons identified approximately 1,500 proteins that exhibited upregulation, downregulation or no changes in each of the four cell lines (Figure 5 and Table 2). We analyzed the four profiles to identify commonly regulated proteins in the four cell lines and found 36 upregulated and 18 downregulated commonly regulated proteins (Figure 5 and Table 3). The regulated proteins are shown in Table 3.

Figure 5: Identification of proteins regulated by NPM1: In order to identify proteins regulated by the NPM1 expression, we performed siRNA knockdown and GeLC-MS analyses in four Ewing’s sarcoma cell lines (A673, TC71, SKES1 and CHP100). The four Ewing’s sarcoma cell lines were treated with either NPM1 siRNA (SASI_ Hs01_00214118; SIGMA-ALDRICH) or negative control siRNA (SIGMA-ALDRICH). A Western blot analysis confirmed that the cells treated with NPM1 siRNA exhibited a significant decrease in the NPM1 expression compared to the controls. These protein samples were then analyzed using GeLC-MS to obtain their protein profiles, and the acquired data were calculated as semiquantitative expressions (control vs siRNA). Approximately 1,500 proteins were identified in each cell line (Table 2). Finally, 36 upregulated proteins and 18 downregulated proteins were identified as common proteins in the four cell lines (Table 3).

Table 2: Number of regulated proteins.

In order to further understand the biological processes and networks and determine whether the proteins were direct or indirect proteins, we routinely employed network analyses using the Ingenuity Pathways Analysis (IPA) system (Ingenuity Systems, Inc, CA, USA) (Figure 6). In this study, we performed network analyses using each PPI profile (Table 1) and regulated protein profile (Table 3 and Figure 7). In both independent analyses using each set of data, the network analyses identified the MYC pathway as playing a critical functional role as an upstream regulator of NPM1 in Ewing’s sarcoma (Table 4 and Figure 7). Additionally, in order to confirm the relationships between MYC and NPM1, we conducted siRNA assays of the Ewing’s sarcoma cell lines using MYC siRNA and verified the protein expressions of both MYC and NPM1 in the cells using Western blotting. The results revealed that silencing MYC in parallel inhibited the NPM1 expression, indicating that MYC is an upstream regulator of NPM1 in Ewing’s sarcoma . We believe that the findings obtained in the functional analyses will contribute to improving understanding of the relationship between NPM1 and malignant behavior in Ewing’s sarcoma and lead to the development of novel therapeutic strategies.

Figure 6: Identification of protein networks: To identify networks and upstream proteins, we routinely employed the Ingenuity Pathways Analysis (IPA) system. We analyzed these networks using either the interaction protein profiles (Table 1), or regulated protein profiles independently (Table 3). The results of these analyses are demonstrated in Table 4. We found that both pathway lists included the MYC pathway as an upstream protein. We conducted a confirmation study and successfully confirmed MYC to be an upstream regulator of NPM1 in Ewing’s sarcoma (data not shown).

Figure 7: (A) The Ingenuity Pathways Analysis (IPA) system demonstrated pathways based on the interaction proteins with NPM1 (Table 1 and Table 4A). (B) The IPA system demonstrated pathways based on the proteins regulated by NPM1 (Table 3 and Table 4B).

Table 3: List of proteins regulated by NPM1 suppression.

A: Interaction proteins associated with NPM1

B: Proteins regulated by NPM1

Table 4: Upstream regulators.

Our proteomic studies of soft tissue sarcomas identified various candidate biomarkers relevant to the prognosis and chemosensitivity of tumors [17-28]. These proteomic studies successfully verified the value of the biomarkers in validation sets using immunohistochemistry. We believe that these proteins are potentially useful biomarkers for various clinical applications. However, although we identified useful biomarkers in our proteomic studies, the functions of the biomarker proteins in tumors remain unknown. Therefore, we conducted functional studies in order to identify the roles and functions of these proteins in the tumors. In particular, we employed proteomic technologies as a tool for conducting functional studies, which revealed novel findings. These results indicate that our proteomic approaches used to perform functional analyses are efficient. Therefore, we should continue these studies in order to further understand these functions. Proteomic analyses are more directly linked to aberrant tumor phenotypes; therefore, there are limitations in our approaches to revealing all processes of molecular biology. In fact, in comparison to cDNA microarray analyses (50,000 probe sets), the sensitivity of the current 2D-DIGE analysis (5,000 spots) remains unsatisfactory. Therefore, these technologies, including CGH arrays, cDNA microarrays, whole genome sequences and proteomic techniques should be used in combination to overcome their individual disadvantages. We believe that hybrid comprehensive studies consisting of genomics, transcriptomics and proteomics will provide important, novel clues for understanding the biology of tumors and identifying biomarkers and therapeutic targets.

This work was supported by a grant from the Japan Society for the Promotion of Science (JSPS) and science Grants-in- Aid for Young Scientists B (No- 25861342 to YS) and Scientific Research (No. 23590434 to TS).

The corresponding author declares that there are no conflicts of interest.