Molecular Profiling of Human Primary Chondrosarcoma-Derived Spheres Reveals Specific and Target Genes Involved in Multidrug Resistance and Metastasis

Molecular Profiling of Human Primary Chondrosarcoma-Derived Spheres Reveals Specific and Target Genes Involved in Multidrug Resistance and Metastasis Vincenzo Desiderio1,2#, Francesca Paino1#, Angela Nebbioso3, Lucia Altucci3,4, Giuseppe Pirozzi5, Federica Papaccio1, Marcella La Noce1, Alfredo De Rosa6, Gianpaolo Papaccio1*§ and Virginia Tirino1§ 1Department of Experimental Medicine Section of Histology, TERM Laboratory 2nd University of Naples, Naples, Italy 2Department of Otolaryngology: Head and Neck Surgery, University of Michigan Ann Arbor, Michigan, USA 3Department of Biophysics, Biochemistry and General Pathology, 2nd University of Naples, Naples, Italy 4Institute of Genetics and Biophysics (IGB) “Adriano Buzzati Traverso”, Naples, Italy 5Department of Experimental Oncology, National Cancer Institute, Naples, Italy 6Department of Medicine, Surgery and Dentistry, 2nd University of Naples, Naples, Italy #These authors contributed equally to this work §These authors are equal senior authors


Introduction
Chondrosarcoma is a malignant, cartilage-forming bone neoplasm that accounts for approximately 25% of all primary bone tumors. It is the third most common primary malignant bone tumor after myeloma and osteosarcoma. Typically low grade, these neoplasms can arise either de novo or from pre-existing cartilage lesions, such as an osteochondroma or an enchondroma.
It is generally believed that chondrosarcomas are relatively resistant to conventional therapies because of their extracellular matrix composition, low percentage of dividing cells, and poor vascularity [1]. Thus, surgical resection has been the main treatment for over 50 years [2]. However, several novel therapeutic approaches have been evaluated in recent years [3][4][5].
Unfortunately, a minority of patients presents with recurring metastases, and up to 13% of recurrent chondrosarcomas are of a higher grade than the original neoplasm [6]. There is no consensus on prognostic factors that determine which patients have a higher risk of treatment failure and dis¬ease-related death, although several papers have addressed this issue [7,8]. One reason for this may be that most studies were conducted on patients treated over several decades and did not account for the different surgical criteria, indications, and methods applied over the years. Furthermore, most stud¬ies were done with only short follow-up -despite the fact that a high rate of late recurrence and metastases has been reported for chondrosarcoma patients compared with those with other primary bone sarcomas [9] and on patients presenting with rare histopathological sub¬types that have a distinct biologic behavior [1,10] such as dedif¬ferentiated chondrosarcoma, mesenchymal chondrosarcoma, and clear cell chondrosarcoma thus reducing the validity of the results.
There is increasing evidence that many cancers are constituted by a hierarchy of cells, including so-called Cancer Stem Cells (CSCs). CSCs are believed to be progenitor cells from which the tumor is spawned and that may be responsible for relapses and metastases [11]. These cells have a particular ability to grow as spheres when cultured in suspension in serum-free medium supplemented with growth factors and other compounds. This ability was first described for the expansion of neuronal stem cells by Reynolds in 1992 [12].
Sphere culture has been increasingly used as method for enriching stem cells and relies on their property of anchorage-independent growth. Studies have reported the application of sphere culture to isolate, enrich, maintain, or expand potential CSC subpopulations from various types of cancers [13][14][15][16]. Like stem cells, tumor sphereforming cells are capable of proliferation and self-renewal, and possess higher tumorigenicity. Using a sphere formation system, Hansford et al. [17] for the first time successfully expanded tumor cells both from neuroblastomas and metastases of high-risk tumors. The spherederived cells formed metastatic tumors in a murine xenograft model with as few as 10 cells and could also be serially passaged [17].
We have recently reported that CD133 + cells from stabilized osteosarcoma cell lines possess stemness features, such as a potential for differentiation, a high proliferation rate, an ability to form osteospheres and colonies on soft agar, a capacity to exclude the fluorescent DNAbinding dye Hoechst 33342, and the expression of high levels of ABCG2, a multi-drug resistance protein of the ABC transporter family [18]. Moreover, in primary chondrosarcoma and osteosarcoma cell lines from human biopsies, we found CD133 + subpopulations displaying a capacity to grow as spheres: these spheres initiated and sustained tumor growth in immunocompromised mice, expressed stemness genes, such as OCT3/4, NANOG, SOX2, and nestin, and differentiated into mesenchymal lineages, such as osteoblasts and adipocytes [19].
Therefore, because chondrospheres are enriched in cancer stem cells, we decided to compared the molecular expression profile of chondrospheres obtained from a primary chondrosarcoma cell line with the corresponding adherent counterpart. We found that chondrospheres over-express genes involved in stemness, drug resistance, tumorigenesis, cell death, and metastasis. Specifically, CROP1, as well as TAP1 genes, involved in multi-drug resistance were up-regulated on chondrospheres. Again, VEGFA and MALAT1 genes, involved in tumor angiogenesis and metastasis formation, resulted to be over-expressed. Taken together, all these observations suggest that these genes could be potential target genes for the development of specific strategies for chondrosarcoma treatment.

Patient and generation of a primary cell line
The chondrosarcoma cell line used was obtained in our laboratory from a consenting, 59-year-old male chondrosarcoma patient attending the National Cancer Institute of Naples, Italy. Diagnosis was based on clinical and histological parameters according to WHO classification. The site of tumor involvement was the sternum. The resected tumor measured 18x15x8 cm, was graded G2, focally G3, and was characterized by hypercellularity, the presence of binucleated cells, and elongated, hyperchromatic nuclei. Survival from diagnosis was 8 months. The patient eventually presented extraskeletal metastases and died of chemotherapy-related complications.
The tumor biopsy was dissected, minced, and digested with 1 mg/ ml collagenase I (Invitrogen Life Technologies, San Giuliano Milanese, Milan, Italy) and 1 mg/ml dispase at 37°C overnight with intermittent shaking. Any isolated cells were washed with PBS twice and filtered through a 70 µm filter to generate a stabilized cell line. Cells were cultured in DMEM supplemented with 2 mM glutamine, 100 IU/ml penicillin, 100 µg/ml streptomycin, and 10-or 20-% heat-inactivated FBS (Invitrogen Life Technologies) at 37°C in a humidified atmosphere under 5% CO 2 until an adherent, homogeneous cell population was obtained. The medium was changed twice a week.

Flow cytometry
Chondrospheres and adherent cells (at a density of 3x10 5 cells/ sample) were incubated with fluorescent-labeled monoclonal antibodies or respective isotype controls. The antibodies used were anti-CD133/2-PE (Miltenyi Biotec, Calderara di Reno, Bologna, Italy), anti-OCT4-PE, anti-SOX2-FITC (all purchased from BD Pharmingen, Milan, Italy), and anti-collagen type II (COL2; AbCam, Cambridge, UK). Secondary antibody was anti-rabbit FITC (AbCam). For CD133, the antibody was incubated for 30 min at 4°C in the dark. After washing steps, the cells were analyzed. For intracellular staining of OCT4, SOX2, and COL2, cells were processing using the Fix & Perm Kit (Invitrogen, Milan, Italy) following the manufacturer's guidelines. All data were acquired using FACS Aria II (BD) and analyzed using FCS version 3 software.

Cisplatin treatment
Chondropheres and adherent cells were exposed to different concentrations of cisplatin for 48 h. The concentrations tested were 30-, 50-, 70-, 100-, and 150-µM. After treatment, chondrospheres and adherent cells were analyzed for apoptosis and cell cycle. Cell death was evaluated with flow cytometry using the Annexin/PI kit (Invitrogen) according to the manufacturer's guidelines and the data analyzed with FCS version 3 software. IC50s were calculated by constructing a dose-response curve. Cell cycle was evaluated with PI staining: briefly, adherent cells and condrospheres were harvested in trypsin-EDTA, washed once with PBS, fixed in ice-cold 70% ethanol, incubated with 50 µg/ml PI (Sigma) plus 1 mg/ml RNase for 60 min at 4°C in the dark, and the data analyzed with Mod-Fit 2.0 cell cycle analysis software (Becton-Dickinson).

Microarray analyses
Microarray quality control reports generated by the Agilent Feature Extraction software were used to detect hybridization artifacts. Probe level raw intensity were processed using R/BioConductor [20,21] and Limma package.
Background correction was performed using "normexp" limma method and data normalization was carried out in two steps: loess normalization within-arrays to correct systematic dye-bias and quantile normalization between-arrays to detect systematic non-biological bias. Ratios representing the relative target mRNA intensities compared with control RNA probe signals were derived from normalized data. vs adherent cells) were identified using a paired Bayesian T-test [22,23]. For each p-value, the Benjamini-Hochberg procedure was used to calculate the False Discovery Rate (FDR) to avoid the problem of multiple testing. The selected gene lists were obtained using the following thresholds: FDR < 0.05 and abs(ratio) >2. For a more stringent analysis, an FDR < 0.001 and abs(ratio) > 3 were also used.

Differentially expressed genes between conditions (chondrospheres
Data have been deposited in NCBI's Gene Expression Omnibus (GEO) http://www.ncbi.nlm.nih.gov/geo/query/acc. cgi?acc=GSE47823) under the GEO series accession number GSE47823. The relative abundance of "Biological Process" ("Molecular function" and "Cellular Component") Gene Ontology terms in each of the selected lists was analyzed using the Database for Annotation, Visualization, and Integrated Discovery (DAVID) Functional Annotation Clustering tool [24].

Validation of microarray data using quantitative TaqManH real-time PCR (qPCR)
Condrospheres and adherent cells were harvested and total RNA extracted using the Pure Link RNA mini kit (Ambion, Life Technologies) according to the manufacturer's protocol. RNA was quantified by UV-Vis spectroscopy. Extracted RNA was treated with DNase I (Promega) and 1 μg of total RNA was reverse-transcribed using SuperScript® VILO™ master mix (Invitrogen).

Statistical analysis
All the above-described experiments were performed in triplicate. Student t-test was used for statistical evaluation. The level of significance was set at p <0.05.

Cell culture and chondrospheres
After five passages, isolated cells obtained from the chondosarcoma biopsy formed an adherent, homogeneous cell population, characterized mainly by small, chondrosarcoma-like cells ( Figure 1A). When these primary culture cells were grown in condrosphere medium, sphere clusters were clearly observable already after 24 hours ( Figure 1B). After 7 days of culture, chondrospheres were seeded in standard flasks and grown in DMEM with 10% FBS. Polygonal-shaped cells migrated from the spheres within a few hours and adhered to the bottom of the flasks (Figure1C). Single cells from the spheres gave rise to secondary spheres that, in turn, were able to form tertiary spheres, as previously demonstrated by us [19]. Chondrospheres were passaged 20 times during the culture period.

Chondrospheres contain cells expressing stemness markers
The expression levels of stem cell markers connected to self-renewal/ stemness or differentiation capability were evaluated: these were OCT4, SOX2, and CD133 for stemness, and COL2 for differentiation. We found that OCT4-, SOX2-, or CD133-positive cells were significantly more numerous in sphere cultures than in adherent cultures. In contrast, COL2 was significantly less expressed in chondrospheres (Figure1D).

Chondrospheres are resistant to cisplatin treatment
The majority of the cells composing a tumor are killed off by the chemoradiotherapy normally used in standard treatments. However, because of their heightened resistance to damage, CSCs may escape cell death. Therefore, we cultured chondrospheres and adherent cells with different concentrations of cisplatin and then calculated IC50 and analyzed apoptosis and cell cycle. The dose-response curves showed that IC50 for adherent cells was 70 µM, whereas for chondrospheres it was >150 µM (Figure 2A). Moreover, we found that for adherent cells, apoptosis was similar at 30 µM and 50 µM, and then increased from 70 µM to 150 µM; in contrast, necrosis was similar at all concentrations ( Figure 2B). In general, cell viability decreased in a dose-dependent manner for adherent cells. Regarding cell cycle, S-phase block was detectable in adherent cells: the number of cells in G 0 G 1 and G2M phases decreased especially at 100 µM and 150 µM ( Figure 2C). Here, sub-G 0 G 1 picks were evident and attributable to apoptosis or necrosis due to cisplatin treatment. For chondrospheres, sub-G 0 G 1 picks were similar at all concentrations tested and were lower than those of corresponding adherent cells. In addition, untreated chondrospheres had a greater number of cells in G 0 G 1 than corresponding adherent cells, and they remained elevated along the whole series of cisplatin concentrations. Thus, high concentrations of cisplatin induced an S-phase block in adherent cells, but increased the number of chondrosphere cells in G 0 G 1 .
Using Gene Ontology (GO) classification, the identified probe sets were subdivided according to their function into cell motility, regulation of apoptosis, cell cycle regulation, immune response regulation, and stemness ( Figure 3A).

Multidrug resistance:
Interestingly, all tested genes involved in multidrug resistance were up-regulated in chondrospheres vs adherent cells ( Figure 3B) with an FDR<0.05 and fold change=2. ABCC2 and TAP1 belong to the superfamily of ATP-binding cassette (ABC) transporters. ABC proteins transport various molecules across extra-and intra-cellular membranes. TAP1 encodes for a protein belonging to the MDR/TAP subfamily and involved in the pumping of degraded cytosolic peptides across the endoplasmic reticulum into the membrane-bound compartment where class I molecules assemble. ABCC2 and TAP1 were up-regulated with fold changes of 1.49 and 1.65, respectively. Of note, CROP1 -encoding a protein associated with cisplatin resistance -was up-regulated in chondrospheres ( Figure  3B). Up-regulation of CROP1 was confirmed by qPCR ( Figure 4). With an FDR<0.001 and fold change=3, only TAP1 was found up-regulated (Supplementary Table 2).

Cell death and regulation of the cell cycle:
For apoptosis and cell cycle regulation, 258 genes were analyzed: a total of 102 cell death genes and 156 cell cycle genes were differentially expressed with an FDR<0.05 and fold change=2 ( Figure 3B). In particular, 57 genes were involved in apoptosis. Interestingly, all genes tested were implicated in negative regulation of programmed cell death. Of these, 35 (61.4%) were upregulated and 22 (38.6%) were down-regulated, demonstrating that antiapoptotic mechanisms were activated in chondrospheres, as reported in Figure 3B (apoptosis panel). 37 genes were found involved in apoptosis with an FDR<0.001 and fold change=3: 64.8% were up-regulated and 35.2% were down-regulated, as reported in Supplementary Table 3. On the whole, cell death genes were repressed in chondrospheres.
For cell cycle, 156 gene differences were statistically significant with an FDR<0.05 and fold change=2. Of these, 31 genes were up-regulated (19.8%) and 125 genes were down-regulated (80.2%) ( Figure 3B). In particular, most of the down-regulated genes were involved in cell cycle promotion. Aurora kinase B was down-regulated, negatively mediating G2-to-M phase transition and cytokinesis. In addition, CDC2, CDC20, CDC25B, and CDC45L were down-regulated, inhibiting both G2-to-M phase and G1-to-S transition and mitosis. CDC6, which is expressed in all proliferating cells but not in quiescent or differentiated cells, was also down-regulated, confirming again the quiescence of chondrosphere cells. Other down-regulated genes included CDCA2, CDCA3, CDCA5, CDCA7, and CDCA8, which are implicated in regulating chromatin structure during mitosis. Up-regulation of p21 was confirmed by qPCR ( Figure 4).

Immune response and cell growth:
For immune response modulation, 64 genes were found with an FDR<0.05 and fold change=2 ( Figure 3B): 42 (65.6%) genes were up-regulated with a fold change ranging from 1.04 to 2.4, and 22 (34.4%) genes were down-regulated with a fold change between −1.0 and −1.4. In particular, genes codifying for major histocompatibility complex (HLA) class I types A, B, C, E, F, G, and H were up-regulated. Overall, genes involved in the innate immune response and in positive regulation of the immune response showed a balance between up-and down-regulation. For genes involved in negative regulation of the immune response, 6 genes were up-regulated and 2 down-regulated ( Figure 3B).
With an FDR<0.001 and fold change=3, 24 genes resulted to be differentially expressed. In particular, 87.5% were up-regulated and 12.5% down-regulated. Genes codifying for HLA class I proteins were up-regulated (Supplementary Table 7). Three genes involved in negative regulation were up-regulated (Supplementary Table 8).
With an FDR<0.001 and fold change=3, 21 genes resulted to be differentially expressed. In particular, 52.4% were up-regulated, while 47.6% were down-regulated (Supplementary Table 9). Taking into consideration all the different functions and different conditions, cell growth was negatively regulated. HLA class I was validated by qPCR ( Figure 4).

Discussion
Cancers are composed of heterogeneous cell populations and are defined on the basis of three main characteristics: 1) a selective tumorigenic capacity; 2) self-renewal and differentiation, i.e., the ability to sustain growth of heterogeneous cancer tissues; and 3) expression of specific surface markers, allowing reproducible selection [30]. CSCs can be selected from tumors through their ability to grow as spheres [31]. Weiss and colleagues [32] showed in neurospheres that between 4% and 20% of cells were stem cells, while the others were progenitor cells in different phases of differentiation [32,33]. Stem/progenitor cells were also enriched from mammary cell populations on the basis of their ability to grow as spheres (mammospheres) [34]. Subsequently, sphere culture techniques have been applied to cell populations from a variety of cancers, such as brain cancers, breast cancers, lung cancers, and melanomas, with successful enrichment of cells with CSC features [35][36][37][38].
The present study confirms and expands the recent observations showing that chondrosarcomas are a source of cancer stem cells [19,39] and demonstrates that chondrospheres may constitute an excellent model for the study of tumoral heterogeneity and multidrug resistance. The molecular pathways controlling genesis and growth of chondrosarcoma have not been defined until now. By analyzing the global pattern of gene expression, we define a signature that is distinctive for sphere vs adherent cell cultures of human chondrosarcoma. In fact, our microarray analysis revealed that adherent cells and chondrospheres differ substantially in their pattern of gene expression. First, we studied the expression of a well-defined set of known stemness markers, such as OCT4, NANOG, and SOX2 [25], by FACS, qRT-PCR, and microarray and found, as expected, that these genes are upregulated in chondrospheres. Also, genes such as LIF, NOG, and FZD7 resulted to be up-regulated. These genes contribute to and support selfrenewal signaling of human embryonic stem cells [26][27][28]. Secondly, we compared the global pattern of gene expression of chondrospheres vs adherent cells. We used two analysis settings: 1) FDR<0.05 with fold change=2 and 2) FDR<0.001 with fold change=3. With the first, we identified a set of 629 up-regulated and 776 down-regulated genes, whereas with second, a set of 251 up-regulated and 302 down-regulated genes was identified in condrospheres.
Three main results are noteworthy. First, multidrug resistance markers -CROP1 above all -are differentially up-regulated in chondrospheres, a finding that is associated with increased cisplatin resistance of sphere cells compared with their adherent counterparts. CROP1 encodes for the cisplatin resistance-associated over-expressed protein (CROP). The N-terminal half of CROP contains cysteine/ histidine motifs and leucine zipper-like repeats, whereas the C-terminal half is rich in arginine and glutamate residues (RE domain) as well as arginine and serine residues (RS domain). This protein localizes with a speckled pattern to the nucleus, and could be involved in the formation of the splicesome via the RE and RS domains. Two alternatively spliced transcript variants encoding the same protein have been found for this gene, conferring cisplatin resistance [40]. Although CROP1 resulted to be up-regulated with an FDR<0.05 and fold change=2, we consider this finding very important. In fact, because CROP1 is involved in splicing processes, even a small variation may lead to a remarkable biological effect. Thus, CROP1 probably plays a key role in addition to other multidrug resistance factors -such as ABCC1, BCAR3, TAP1, and NBR1 -in the notorious resistance to chemoradiotherapy of chondrosarcoma [1,41].
Although the expression of ABC transporters as well as CROP could render chondrospheres resistant to drugs, it is not the sole determinant of resistance, as both quiescence of the cells and the reluctance to undergo apoptosis could be equally or even more important. Generally regarded as quiescent and non-dividing [42], stem cells would be expected to be inherently refractory to drugs that target either the cell cycle or rapidly dividing cells. In line with this, we found that there was an inhibition of apoptosis-and cell cycle-related genes in chondrospheres, confirming the hypothesis that quiescence and negative regulation of cell cycle and apoptosis, in addition to overexpression of multidrug resistance genes, may be involved in the resistance to treatment that is typical of chondrosarcoma.
The second prominent result concerns two of the most important traits of cancer: angiogenesis and the development of metastases. Chondrospheres had differentially up-regulated VEGFA, VWCE, ANG, and MALAT1, factors involved in tumor angiogenesis and metastasis. Tumor vascularity and the factors that stimulate angiogenesis have been shown to be correlated with biologic aggressiveness, high pathological grade, and poor survival [43]. In fact, once a tumor grows beyond several millimeters in diameter, angiogenesis must occur to support further growth. Moreover, the first step in metastasis is for the tumor cell to gain access to the circulation. Neo-formed vessels increase permeability and allow tumor cells to egress into the circulation and initiate metastasis. In particular, VEGF-A is a basic, homodimeric glycoprotein of approximately 45,000 Da that is expressed in four isoforms via alternative exon splicing of the gene [44]. It is the most commonly studied angiogenic cytokine, the primary inducer of angiogenesis in multiple types of carcinomas, and has been strongly correlated with biologic aggressiveness, metastasis, and prognosis [45]. On the other hand, MALAT1 regulates both the expression of metastasis-associated genes and cell motility via transcriptional and/ or post-transcriptional regulation of motility-related genes [46][47][48][49]. It is up-regulated in several solid tumors and is a highly abundant nucleusrestricted RNA that localizes to nuclear speckles, a sub-nuclear domain suggested to coordinate RNA polymerase II transcription, pre-mRNA splicing, and mRNA export. MALAT1 interacts with several pre-mRNA splicing factors, including serine arginine dipeptide-containing SR family splicing factors, thereby influencing alternative splicing of pre-mRNAs [50,51]. Through such a mechanism, cells could alter the local concentration of a particular splicing factor upon a specific external signal or during specific stages of the cell cycle [51].
Finally, a result that deserves to be discussed regards immune response modulation. Overall, we found a balance between up-and down-regulated genes in chondrosphere cells; however, negative modulation of the immune response seemed to be favored. Notably, HLA class I genes appeared to be up-regulated. Little is known concerning the antigen processing and presentation machinery in CSCs and normal stem cells, let alone peptide-HLA complexes unique to stem cells that may serve an immunological target. Usually, loss of HLA class I antigens appears to be a significant mechanism by which tumor cells escape specific immune attack and cause problems in the design of antitumor immunotherapy. The loss of HLA class I antigens on tumor cells has been reported in several human tumors, and the loss of HLA class I molecules has been discussed in the context of tumor aggressiveness, such as differentiation of histology, invasiveness, and metastatic potential [52]. Our result seems to be in conflict with these data. Despite this, Aptsiauri et al. [53] showed that if apparent tumor cells expressed HLA class I, various types of HLA class I alterations were found in malignancies and in the molecular mechanisms that underlie these defects. In this context, the HLA class I molecules preserved in breast cancers may exhibit altered expression and dysfunction as antigen presentation molecules [54]. There are several mechanisms proposed to explain abnormal HLA class I phenotypes: 1) impaired transcriptional activity of genes codifying for HLA class I; 2) deregulation of antigen-processing machinery components responsible for functional HLA class I expression, such as β-2 microglobulin and transporters associated with antigen processing; and 3) degradation of HLA class I proteins by proteasomes [55].
In fact, HLA-G, which is considered to be one of the major factors in the immune mechanism preventing the rejection of the semi-allogenic fetus in the mother, suppresses the proliferation of CD4 + T lymphocytes in vitro, induces apoptosis in activated CD8 + T lymphocytes, and inhibits the cytotoxicity of natural killer cells [56]. Therefore, although there was an up-regulation of HLA genes involved in the immune response, in chondrosphere cells, other mechanisms must be considered.
In conclusion, chondrospheres express stemness-, invasion-, multidrug resistance-and metastasis-related genes and are highly resistant to cisplatin treatment. Our study highlights that either CROP1 and TAP1 genes, involved in multidrug resistance, or VEGFA and MALAT1 genes, involved in angiogenesis and metastasis, are strongly up-regulated, and could be considered potential therapeutic targets for the treatment of chondrosarcoma. In this context, effects of targeting VEGFA and MALAT1 genes, have been described in in vivo models [57,58]. Phase I studies have reported a response of chondrosarcomas