Soft tissue sarcomas (STS) are malignant tumors of mesenchymal origin. A substantial portion of these tumors exhibits complex karyotypes and lack characterized chromosomal aberrations. Owing to such properties, both histopathologic and molecular classification of these tumors has been a significant challenge. This study examines the protein expression of a large number of human STS, including subtype heterogeneity, using two-dimensional gel proteomics. In addition, detailed proteome profiles of a subset of pleomorphic STS specimens using an in-depth mass-spectrometry approach identified subgroups within the leiomyosarcomas with distinct protein expression patterns. Pathways analysis indicates that key biologic nodes like apoptosis, cytoskeleton remodeling, and telomere regulation are differentially regulated among these subgroups. Finally, investigating the similarities between protein expression of leiomyosarcomas and undifferentiated pleomorphic sarcomas (UPS) revealed similar protein expression profiles for these tumors, in comparison with pleomorphic leiomyosarcomas.
Implications: These results suggest that UPS tumors share a similar lineage as leiomyosarcomas and are likely to originate from different stages of differentiation from mesenchymal stem cells to smooth muscle cells. Mol Cancer Res; 12(12); 1729–39. ©2014 AACR.
Soft tissue sarcomas (STS) are rare tumors of mesenchymal origin that are highly malignant and heterogeneous. Metastases develop in approximately one-third of the patients, most of whom die from the disease (1). Morphologically, more than 50 entities have been described. On the basis of genetic alterations, STS are broadly divided into tumors with specific reciprocal translocations and simple karyotypes, and tumors with complex karyotypes in which specific aberrations have not been recognized (2, 3).
Poorly differentiated sarcomas represent a diagnostic challenge as these tumors lack a typical and easily identifiable phenotype. The undifferentiated diagnostic category is usually considered in four broad groups: tumors with round-cell morphology, spindle-cell morphology, epithelioid morphology, and pleomorphic morphology. The latter group can be simplistically defined as the one in which the cells display marked atypia and should not be taken as synonymous of undifferentiation or biologic aggressiveness. Pleomorphic sarcomas broadly encompass six histotypes (differential diagnoses): pleomorphic liposarcoma, pleomorphic leiomyosarcoma, pleomorphic rhabdomyosarcoma, dedifferentiated liposarcoma, myxofibrosarcoma, and malignant fibrous histiocytoma/undifferentiated pleomorphic sarcoma (MFH/UPS). In most cases, with adequate sampling and complementary diagnostic techniques, for example, immunohistochemistry or electron microscopy, a diagnosis is possible, in spite of only focal evidence of a line of differentiation. However, in 5% to 10% of the cases, no line of differentiation can be identified and these tumors are classified as MFH/UPS, which nowadays essentially represents a diagnosis of exclusion. Until the 1990s, MFH represented the most frequent STS diagnosis, accounting for 40% of all adult mesenchymal malignancies. Despite the heterogeneity of these tumors, currently, there are no alternative courses of treatment for different types of STS used at Lund University Hospital.
Genomic profiling studies by comparative genome hybridization and gene expression analyses have increased our insight into the biology of pleomorphic STS (4–11). Similarities in genomic and gene expression profiles suggest that, in the context of STS of the extremities, tumors that are histopathologically classified as MFH/UPS may be based on genetic alterations corresponding to highly pleomorphic leiomyosarcomas (LMS; refs. 6, 9). Our group has previously performed gene expression analysis on pleomorphic STS, encompassing a series of 40 LMS. Among the LMS, a subgroup of 11 samples clustered together while the remaining 29 LMS showed more heterogeneous patterns of gene expression. The 11-sample cluster showed a high expression of muscle-associated genes. Similar results were reported by van Rijn and colleagues, when analyzing a set of LMS, which could be divided in three subgroups depending on gene expression analysis. The heterogeneity of LMS, morphologically as well as genetically, indicates that further investigations are needed to define subtypes and define molecular classifiers.
Data on the STS proteome are scarce though some protein targets are amenable to clinical diagnostic application through immunohistochemistry (12, 13). mRNA expression levels and corresponding proteins profiles do not necessarily correlate well, which suggests a complex interplay between copy number and protein abundance (14). In addition, factors such as posttranslational modifications, compartmentalization and relative synthesis, and degradation rates play an important role in protein function. To profile protein expression in UPS and LMS, we used nano-flow liquid chromatography coupled with tandem mass spectrometry and performed quantitative protein data analysis to highlight pathways with differential activity to define STS subsets.
Materials and Methods
Tumor samples and clinical pathologic data
UPS and LMS of the extremities and the trunk wall were selected from the Lund Sarcoma Centre (Lund, Sweden). The Lund University research ethics committee granted ethical permission for the study (LU302-02). Patients with metastases at diagnosis and those who had been treated with neoadjuvant chemo- or radiotherapy were excluded. All tumors were reviewed by two experienced sarcoma pathologists, Dr. Pehr Rissler at Lund University Hospital and Dr. Jonathon Fletcher at Harvard Medical School (Boston, MA), according to the WHO classification (15). Data on histopathologic grade, necrosis, vascular invasion, depth, and size were collected in a standardized manner (Table 1).
LMS diagnosis required the presence of eosinophilic spindle cells with vesicular, blunt-ended, intended, or lobulated nuclei arranged in a fascicular pattern, accounting for 5% to 10% of the surface area examined. Tumors with these characteristics were also required to show unequivocal positivity for smooth-muscle actin (SMA) as well as for desmin and/or h-caldesmon. MFH/UPS were defined as pleomorphic spindle-cell sarcomas without any specific differentiation. These tumors were also negative for melanocytic and hematopoietic markers. Adequate tissue samples and clinical data were available from all cases. Myofibroblastic sarcoma was a pleomorphic spindle-cell sarcoma with either storiform or fascicular growth patterns composed mainly of cells with pale or moderately eosinophilic cytoplasm (without sharply defined cell borders) and pointed, tapering, or somewhat wavy hyperchromatic nuclei. In undoubted cases, there was always widespread positivity for HHF-35 or SMA but not for desmin, and the nuclear features were not convincing for leiomyosarcoma.
Tumor blocks were obtained with 1 cm intervals from the entire specimen and all available blocks were analyzed using 4 μm sections. All tumors were stained at the Department of Pathology, Skåne University Hospital (Lund, Sweden).
Materials and reagents
Chemicals (Urea, CHAPS, Tris, glycerol, magnesium acetate, ammonium bicarbonate, dithiothreitol, iodoacetamide, and bromophenol blue), and the total protein assay kit, micro-Lowry Peterson modification, were purchased from Sigma Aldrich. IPG ampholytes, immobilized pH gradient strips, SDS, and CyDyes were from GE Healthcare. Sequencing-grade modified trypsin was from Promega Biotech AB. 2× Laemmli sample buffer and 12% Criterion TGX precast gels were from Bio-Rad laboratories Inc.. GelCode Blue Stain reagent and Zeba protein desalting spin columns were from Pierce Biotechnology. Formic acid was from JT Baker. High-performance liquid chromatography (HPLC) grade acetonitrile and water were from Fluka analytical (Sigma Aldrich). Ultramicrospin C18 columns were from the Nest Group and the SCX column was from Applied Biosystems. The TMT labeling kit was from Pierce Biotechnology.
Protein extraction from tissue
Tissues were collected after surgery at the orthopedic unit, Lund University Hospital (Lund, Sweden) and made anonymous after informed consent and approval by the ethics committee. One-hundred and thirty nine samples of all types of STS were selected, including 16 UPS and 38 LMS tumors. The resected sample was put on ice and a pathologist first examined all samples to obtain representative, viable, and non-necrotic tumor tissue, which was then snap frozen. Frozen tumor tissue (250 mg) was used for sample preparation for proteomics analysis. The tissue was put in a grinding beaker together with a steel ball and immersed in liquid nitrogen. After this step, the tumor was ground for 30 seconds in a microdismembrator II operating at 60 strokes per second at maximal amplitude, refrozen in liquid nitrogen, and then ground again for 30 seconds.
2D DIGE analysis
The protein extracts were thawed on ice in lysis buffer containing 8 mol/L urea, 30 mmol/L Tris, 5 mmol/L magnesium acetate, and 4% CHAPS (pH 8.5). Thirty eight of the samples were extremely viscous and could not be pipetted even after sonication and DNase treatment and were removed from the study. The protein concentrations of the remaining 101 samples were determined using the total protein micro-Lowry assay kit. All 101 samples were labeled with Cy3 and Cy5 independently, according to the manufacturer's protocol. A total of 200 pmol of dye per 25 μg of protein was used. Equal amounts of protein from each sample were mixed to form a pool. The pool was labeled with Cy2. After labeling, the pool was divided into aliquots of 25 μg. The volumes of the labeled samples were adjusted to 20 μL with magnesium lysis buffer, and 20 μl 2× sample buffer [8 mol/L urea, 130 mmol/L dithiothreitol (DTT), 4% CHAPS, 2% IPG ampholytes] was added. A Cy3-labeled sample, a Cy5-labeled sample, and a Cy2-labeled aliquot of the pool were combined and mixed with 330 μL rehydration buffer (8 mol/L urea, 2% CHAPS, 0.002% bromophenol blue, 2.8 mg/mL DTT, 0.5% IPG ampholytes). The samples were left at room temperature for 30 minutes and centrifuged for 30 minutes at 16,200 rpm before loading onto a 24-cm immobilized pH gradient strip (pH 4–7) for 14-hour rehydration.
First-dimension isoelectric focusing was carried out on an Amersham Biosciences IPG-phor with a total focusing time of 67 kVh. Before the second-dimension separation, the strips were incubated in 20 mL equilibration solution [6 mol/L urea, 75 mmol/L Tris (pH 8.8) 30% glycerol (w/v), 2% SDS (w/v), 0.002% bromophenol blue] supplemented with 10 mg/mL DTT for 15 minutes, followed by 15-minute incubation in equilibration buffer supplemented with 25 mg/mL iodoacetamide. The IPG strips were then loaded and run on 12.5% SDS-PAGE for 3 hours at 10 W/gel. The gels were fixed in 30% ethanol and 10% acetic acid for 30 minutes and then kept in water. The gels were scanned with an Amersham Biosciences Typhoon 9400 variable imager. Spot detection and matching was carried out in Progenesis Samespots using on average 10 manual landmarks.
The overall experimental flow design is shown in Supplementary Fig. S5. Twenty UPS and LMS samples were selected on the basis of the results from the two-dimensional (2D)-gel analysis for in depth analysis. Hundred micrograms of protein from each of the selected samples were mixed 1:1 with 2× Laemmli sample buffer, heated for 5 minutes at 95°C, and subsequently run shortly into 12% Criterion TGX precast gels at a constant current of 25 mA/gel at 25°C. The run was stopped as the front entered 3 mm into the resolving gel, so that the entire protein extract was concentrated in the stacking/resolving gel interface. The protein bands were visualized using GelCode Blue Stain reagent, excised and cut into 2 × 2 mm cubes, and in-gel digested essentially according to the method by Wilm and colleagues (16).
Briefly, the gel pieces were de-stained repeatedly in 50% acetonitrile in 50 mmol/L ammonium bicarbonate and 100 mmol/L ammonium bicarbonate and then dehydrated using 100% acetonitrile and dried in a vacuum centrifuge. The disulfide bonds were reduced with 20 mmol/L DTT in 100 mmol/L ammonium bicarbonate for 1 hour at 56°C and subsequently alkylated using 55 mmol/L iodoacetamide in 100 mmol/L ammonium bicarbonate for 45 minutes at room temperature in the dark. After washing with 50% acetonitrile in 50 mmol/L ammonium bicarbonate, the gel pieces were dehydrated in 100% acetonitrile and dried in a vacuum centrifuge. The gel pieces were reswelled on ice for 45 minutes in digestion buffer containing 10 ng/μL sequencing-grade modified trypsin in 50 mmol/L ammonium bicarbonate and then incubated at 37°C overnight. Peptides were extracted by incubating the gel pieces in 5% formic acid in 50% acetonitrile for 15 minutes. The extraction step was repeated three times and the peptide extracts were pooled and dried in a vacuum centrifuge, dissolved in 0.1% formic acid. Peptides were cleaned up using Ultramicrospin C18 columns according to the manufacturer's instructions. The eluted peptides were dried in a vacuum centrifuge, dissolved in 0.1% formic acid, and stored at −20°C until analysis by Reversed Phase (RP)-HPLC tandem mass spectrometry (MS/MS). One microgram of each sample was infused into the mass spectrometer in duplicates for the analysis.
For Tandem Mass Tag (TMT) labeling, a reference sample was prepared from mixing equal amounts of the 20 selected samples (see above). One-hundred micrograms of protein from each of the 20 samples and four aliquots of 100 μg of protein from the reference sample were run shortly into gels and in-gel digested as described for the label-free analysis. The peptides extracted from the gels were resuspended in 100 μL 200 mmol/L TriEthylAmmoniumBicarbonate (TEAB) and the peptides were labeled with isobaric TMT reagents according to the manufacturer's protocol. The labeled samples were combined into four groups A to D. Each group was composed of five samples labeled with either TMT127–TMT131 together with a reference sample labeled with TMT126.
The four groups were subsequently cleaned and fractionated into eight fractions on a SCX column (ICAT, Strong Cation Exchange Cartridge) at a flow rate of 50 μL/minute. The samples were loaded on the pre-equilibrated cartridge and then eluted in 500 μL fractions by injecting KCl at increasing concentrations (30, 60, 90, 120, 240, 300, 420, 500 mmol/L) in 5 mmol/L KH2PO4, 25% acetonitrile. The volume of the fractions was then reduced to less than 100 μL using a vacuum centrifuge. The fractions were desalted using Ultramicrospin C18 columns. The eluted peptides were dried in the vacuum centrifuge, dissolved in 0.1% formic acid, and subsequently analyzed by RP-HPLC MS/MS.
Mass spectrometry analysis
An ESI-LTQ-Orbitrap XL mass spectrometer (Thermo Electron) interfaced with an Eksigent nanoLC plus HPLC system (Eksigent Technologies) was used for all analyses. Peptides were loaded a constant flow rate of 10 μL/minute onto a precolumn (PepMap 100, C18, 5 μm, 5 mm × 0.3 mm, LC Packings) and subsequently separated on a 10 μm fused silica emitter, 75 μm × 16 cm (PicoTip Emitter, New Objective, Inc.), packed in-house with Reprosil-Pur C18-AQ resin (3 μm Dr. Maisch GmbH). Peptides were eluted with a 150-minute (label-free quantification) or a 90-minute (TMT quantification) linear gradient of 3% to 35% acetonitrile in water, containing 0.1% formic acid, with a flow rate of 300 nL/minute.
The Linear Trap Quadrupole (LTQ)-Orbitrap was operated in a data-dependent mode simultaneously acquiring MS spectra in the Orbitrap (from m/z 400 to 2,000) and MS/MS spectra in the LTQ. For the label-free analysis, four MS/MS spectra were acquired using collision-induced dissociation (CID) in the LTQ and each Orbitrap-MS scan was acquired at 60,000 FWHM nominal resolution settings using the lock mass option (m/z 445.120025) for internal calibration. For the TMT quantification analysis, the instrument selected three precursor ions for sequential fragmentation by CID and higher-energy collisional dissociation (HCD), for analysis in the LTQ and Orbitrap (recorded at a resolution of 15,000), respectively. The normalized collision energy was set to 35% for CID and 45% HCD. The dynamic exclusion list was restricted to 500 entries using a repeat count of two with a repeat duration of 20 seconds and with a maximum retention period 120 seconds. Precursor ion charge state screening was enabled to select for ions with at least two charges and rejecting ions with undetermined charge state.
Raw mass spectrometric data was independently analyzed in Progenesis LC-MS (Nonlinear Dynamics Ltd, version 4.1.4804) and Proteios SE (version 2.19.0; ref. 17) software platforms. In both cases, runs were aligned and peptide identifications were propagated between runs to minimize missing values. MS/MS spectra of ions with charge +2/+3/+4 between 400 and 1,000 m/z and 27 to 62 minutes were filtered and submitted to Mascot for identification using the UniProt database release 2012 filtered for human. Identifications were filtered with an FDR of 0.05 at the peptide and protein levels. Only proteins with at least one unique peptide were kept in the dataset. Qlucore Omics Explorer (Qlucore AB, version 2.3) software was used for statistical analysis of the protein expression profiles. The functional enrichment analysis was carried out with MetaCore (Thomson Reuters, version 6.14) and DAVID (National Institute of Allergy and Infectious Diseases NIH, version 6.7).
Results and Discussion
There was little overlap between these tumors and those that were used in the previous genomics study, which limited the value of a correlation analysis between the results from the two studies. Furthermore, the mRNA and proteomic samples from those few tumors that were included in both studies were taken from different regions of the tumor, which further limited the usefulness of a comparison. The gel study was carried out using tumor samples run in duplicate using dye swapping to balance out the different reactivities occasionally seen with the Cy3 and Cy5 dyes.
Unsupervised hierarchical clustering of the 139 tumors, based on 873 spots analyzed in 2D DIGE experiment revealed different protein expression profiles amongst the STS tumors but in general no strong clustering was observed (Supplementary data Fig. S1). The gels were aligned, matched, and analyzed by Progenesis SameSpots and statistical analysis was carried out in SameSpots and also using routines written in R. No significant clustering could be found by either Pearson correlation analysis nor by principal component analysis. A more focused analysis showed several clusters within the LMS samples, but which did not separate LMS samples from MFH/UPS (Fig. 1). On the basis of the hierarchical clustering, we defined three LMS clusters and a forth cluster enriched for myogenic UPS, referred to as LMS_A, LMS_B and LMS_C, and UPS_D. From each cluster, five samples were selected for in depth analysis by HPLC-MS/MS using TMT labeling (Fig. 1).
A total of 1,105 proteins were identified overall, and 778 proteins were quantified in at least one group, using only unique peptides. Measurements within each subgroup were then taken as biologic replicates to allow an intergroup analysis. Principal component analysis of the 778 proteins identified three proteins, vinculin (VINC), COL6A3, and MYH11, that discriminated between the groups in multigroup comparison with R2 ≥ 0.9025. Vinculin is a focal adhesion protein primarily involved in cell–cell and cell–matrix adhesion. Focal adhesion proteins do not only regulate mechanical force transmission, but also indirectly modulate diverse cellular processes such as proliferation, differentiation, apoptosis and cell motility, via integrin signaling (18, 19). Vinculin deregulation has been associated with malignant potential, tumorigenicity, and metastatic potential (20, 21). More specifically, a role for Vinculin in regulation of survival and motility via ERK (22), as well as its indirect control over PTEN activity via regulation of β-catenin–MAGI2 interaction (23) has been reported. The Col6a3 gene codes the alpha-3 chain of collagen type VI, a secreted protein primarily involved in the extracellular matrix (ECM). Col6a3 gene has been shown to undergo tumor-specific splice variation in several different malignancies such as colon, prostate, and pancreatic cancer (24–26). In mesenchymal tumor cells, both vinculin and collagen type VI are known to be regulated (27, 28). Furthermore, COL6A3 overexpression was shown to correlate well with both tumor grade and resistance to the chemotherapy agent cisplatin in ovarian cancer cells, pointing toward a significant ECM remodeling in these tumors (29). Finally, the Myh11 gene codes for myosin heavy chain, smooth muscle isoform. To our knowledge, this particular gene or the protein it encodes has not been associated with pleomorphic STS. Considering that MYH11 is a commonly used marker for smooth-muscle tissue, it is likely that the difference in MYH11 expression we have observed indicates the level of differentiation among the tumors in the study.
Pathway enrichment analysis of this data using MetaCore software identified “glycolysis and gluconeogenesis” pathway significantly enriched in the LMS groups (LMS-A, B, and C), whereas “cytoskeleton remodeling” and “cell adhesion” pathway were significantly enriched in the UPS group (UPS-D).
Label-free quantitative mass spectrometry reveals at least two distinct subgroups within LMS
Given the low overlap of proteins between the groups in the TMT analysis, we further analyzed each tumor sample using a label-free approach with two technical replicates. We identified >2,500 proteins, of which 1632 were quantified with proteotypic peptides following chromatogram alignment using Progenesis LC-MS software. We initially examined whether protein expression profiles allowed the discrimination between LMS and UPS. Two-way comparison (LMS vs. UPS) in Qlucore did not result in a significant separation of samples (not shown), whereas multigroup comparison separated the groups (Fig. 2A). Unsupervised hierarchical clustering based on variables that pass q < 0.05 threshold (Table 2) showed that group-A LMS appear distinct from tumors in LMS-B and LMS-C, whereas group-D containing UPS did not show a distinct pattern (see Fig. 2B) and appeared closer to A than to B and C. This interpretation is consistent with the PCA plot, where UPS group-D samples are spread over the axis of the first principal component between LMS-A, LMS-B, and LMS-C. We therefore suggest that LMS encompass at least two subgroups with respect to protein expression profiles, whereas UPS are spread over the hypothetical space in between these two subgroups, thus suggesting a close relation between UPS and LMS.
We subsequently removed the UPS samples from the statistical analysis, and performed a two-way comparison analysis of samples in LMS group-A versus those in LMS groups B and C. Using the same significance criteria, the number of discriminating variables increased considerably, resulting in 156 proteins (see Fig. 3 and Table 3). We investigated the functional annotations associated with these proteins using the DAVID tool (30) and found out a large group of ribosomal proteins, together with subunits of eukaryotic translation initiation factors (eIF). Consequently, protein biosynthesis and translation were the most significant functional terms, indicating that LMS-A probably has a higher level of protein production compared with LMS-B and LMS-C. However, the distinction between LMS-B and LMS-C is less clear. With the same significance measure (q < 0.05), no protein in our dataset could discriminate between the two groups. Even with less stringent selection criteria, the remaining proteins did not yield a meaningful functional enrichment. We believe this is due to intra-group variability and that it warrants further studies into protein expression of group-B and group-C LMS. It is also worth considering the possibility that the distinction between these two groups is insignificant and that there are essentially two subgroups of LMS and not three.
Pathway analysis of the subgroups
To identify differentially regulated pathways, we have used the list of 156 proteins that discriminate LMS-A from LMS-B and LMS-C. This revealed several significant pathways such as “apoptosis and survival: Granzyme A signaling” (Supplementary Fig. S2), “cytoskeleton remodeling” (Supplementary Fig. S3), as well as “telomere regulation and cellular immortalization” (Supplementary Fig. S4). One of the most significantly differing proteins in Granzyme A signaling pathway was Prelamin-A/C, coded by the LMNA gene (Fig. 4). Lamin family proteins are components of the nuclear lamina, which in turn provides support to the nucleus and are central to maintenance of nuclear integrity. It has been shown in literature that lamins may play an important role in chromatin organization and telomere dynamics (31). Furthermore, it is suggested that Prelamin-A/C accumulation plays a role in smooth-muscle cell senescence by disrupting mitosis and inducing DNA damage in vascular smooth-muscle cells causing genomic instability and premature senescence. In addition, Prelamin-A/C accumulation appears to deregulate the G2–M checkpoint (32).
We also observed differential expression of the Ku70 protein, which is involved in telomere regulation and cellular immortalization, with an average overexpression of 2.5 in LMS-A compared with LMS-B and LMS-C. Ku70 protein is one of the two subunits of the Ku complex in which ATP-dependent helicase is known to play an important role in the nonhomologous end joining pathway as well as telomere regulation (33, 34).
Potential biomarkers for vascular invasion in STS
Vascular invasion is a negative prognostic factor in STS, which motivated assessment of protein profiles in the STS that showed vascular invasion. Two-way comparison in Qlucore Omics Explorer with filtered variables (P < 0.001) that discriminate between the two groups identified three significantly deregulated proteins (Fig. 5); Vinculin (VINC), Intergrin-linked protein kinase (ILK), and Creatine kinase type B. We have shown above that Vinculin is a key player in cell adhesion. In addition, elevated Vinculin levels have been associated with angiogenesis (35, 36). The correlation between metastasis, angiogenesis, and ILK signaling has been addressed in various malignancies (37–40), including osteosarcoma (41) and chondrosarcoma (42). The alpha-parvin (PARVA) protein, which interacts with ILK, is differentially expressed (P = 0.0012). PARVA plays a role in smooth-muscle contraction, as well as in angiogenic sprouting and adhesion between smooth-muscle cells and endothelial cells during vessel development (43–45). Furthermore, the significance of the expression regulation of ILK and its partners was previously studied in chondrosarcomas (46). It is important to note, however, that our sample size is small with respect to tumors showing vascular invasion (n = 4), and thus our dataset is not large enough to draw strong conclusions.
Our analysis has shown that LMS, in agreement with previous gene expression studies, could be divided into three subgroups, with distinct proteomic profiles. One of the subgroups, referred to as group-A LMS, displayed a strikingly distinct protein profile and was enriched for ribosomal proteins as well as eIFs. These findings, at a proteomic level, are in line with previous expression studies that confirm LMS complexity. In general, the protein profiles do not correlate well with the gene expression profiles due to temporal differences in expression and turnover. In our study, we found that several proteins, including ARPC3, UB2D3, DDX17, and PHB2, all upregulated in group-A LMS, as well as LMNA, which is downregulated in group-A LMS, can be used to discriminate group-A LMS from other types of LMS. Pathway analysis of our dataset indicated that apoptosis and survival (particularly Granzyme A signaling) cytoskeleton remodeling and telomere maintenance/regulation to be likely candidates for differential regulation between the LMS subgroups.
Our results also indicate that expression regulation of several key proteins, in particular VINC and COL6A3, might play important roles in the progression of pleomorphic STS. Specifically their role in the ECM and cell adhesion warrants future studies on expression levels of these proteins, and their associates, in tumor invasiveness and metastatic potential. Differential MYH11 expression, on the other hand, could be useful in determination of the differentiation state of the tumors.
We have also addressed the relationship and similarity between UPS and LMS. We have previously performed a comparison of UPS/LMS at a genetic/gene expression level, suggesting that UPS could indeed correspond to highly pleomorphic LMS. In this study, UPS appeared to spread out in the PCA analysis and cluster in between the LMS subgroups in unsupervised clustering. This reinforces the concept that UPS and LMS may represent a common lineage at different differentiation stages. Mesenchymal stem cells (MSC) have been shown to be the likely origin of MFH/UPS tumors, and the role of Wnt signaling in commitment to differentiation has been highlighted (8). Furthermore, a recent study has shown that both types of tumors (UPS and LMS) can originate from the same Pten/p53-inactivated murine model, which also points toward shared lineage between these histotypes (47). In this light, tumors classified as MFH/UPS originate from different time points along smooth-muscle differentiation of hMSCs, and represent different steady (or semi-steady) states of gene expression based on other relevant mutations that might have been accumulated or inherited. Thus, we conclude that this discovery-phase experiment using first 2D-PAGE, then two independent LC-MS analyses, label-free and isotopic labeling has provided compelling evidence that there are indeed subtypes to be found in the UPS and LMS sarcomas. The distinct lack of clustering in the majority of tumor types indicates the extent of the heterogeneity in this cancer type and the need for individual analyses of each patient tumor to define the type of aberration that has led to development of the tumor and possibly indicate how it should be treated. This discovery-phase experiment has opened the way for a future independent validation using new tumor sets. The proteins we have identified here as differentially expressed among the subgroups may play a relevant biologic role to be addressed in further studies.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: U. Kirik, M. Nilbert, P. James, A. Carneiro
Development of methodology: M. Jonsson, A. Carneiro
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K. Hansson, M. Jonsson, M. Nilbert, A. Carneiro
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): U. Kirik, M. Krogh, A. Carneiro
Writing, review, and/or revision of the manuscript: U. Kirik, M. Nilbert, P. James, A. Carneiro
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): U. Kirik, M. Nilbert, A. Carneiro
Study supervision: M. Nilbert, P. James, A. Carneiro
This project was supported by grants from Swedish National Research Council (VR-NT; to P. James), the Foundation for Strategic Research (SSF; P. James) (Strategic Centre for Translational Cancer Research–CREATE Health), the Knut and Alice Wallenberg foundation (to P. James) and Vinnova (to P. James) as well as Swedish Cancer Fund (M. Nilbert), the Nilsson Cancer Fund (to M. Nilbert), and Yngre ALF 2013–2015 (to A. Carneiro).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The authors thank Josefin Fernebro, Emelie Styring, Anders Rydholm, Fredrik Vult von Steyern, and Pehr Rissler for contributing with tumor samples and clinical pathologic data, as well as Liselotte Andersson and Fredrik Levander for their help and support throughout the project. The authors also thank the PRIDE team for their help in making the dataset available.
The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium (http://proteomecentral.proteomexchange.org) via the PRIDE partner repository with the dataset identifiers PXD000616 and PXD000617.
Note: Supplementary data for this article are available at Molecular Cancer Research Online (http://mcr.aacrjournals.org/).
- Received February 7, 2014.
- Revision received July 2, 2014.
- Accepted July 6, 2014.
- ©2014 American Association for Cancer Research.