Author contributions: R. Rodriguez: designed the study, performed experiments, analyzed the data and interpreted the results, and wrote the paper; R. Rubio, I.G.-A., and G.J.M: performed experiments; C.E.: analyzed data; J.G.-C.: conceived the study and analyzed data; P.M.: conceived the study, analyzed the data and interpreted the results and wrote the paper; P.M. and R. Rodriguez: financially supported the study.
Disclosure of potential conflicts of interest is found at the end of this article.
First published online in STEM CELLSEXPRESS November 23, 2010.
Human sarcomas have been modeled in mice by expression of specific fusion genes in mesenchymal stem cells (MSCs). However, sarcoma models based on human MSCs are still missing. We attempted to develop a model of liposarcoma by expressing FUS (FUsed in Sarcoma; also termed TLS, Translocated in LipoSarcoma)-CHOP (C/EBP HOmologous Protein; also termed DDIT3, DNA Damage-Inducible Transcript 3), a hallmark mixoid liposarcoma-associated fusion oncogene, in wild-type and p53-deficient mouse and human adipose-derived mesenchymal stem/stromal cells (ASCs). FUS-CHOP induced liposarcoma-like tumors when expressed in p53−/− but not in wild-type (wt) mouse ASCs (mASCs). In the absence of FUS-CHOP, p53−/− mASCs forms leiomyosarcoma, indicating that the expression of FUS-CHOP redirects the tumor genesis/phenotype. FUS-CHOP expression in wt mASCs does not initiate sarcomagenesis, indicating that p53 deficiency is required to induce FUS-CHOP-mediated liposarcoma in fat-derived mASCs. In a human setting, p53-deficient human ASCs (hASCs) displayed a higher in vitro growth rate and a more extended lifespan than wt hASCs. However, FUS-CHOP expression did not induce further changes in culture homeostasis nor initiated liposarcoma in either wt or p53-depleted hASCs. These results indicate that FUS-CHOP expression in a p53-deficient background is sufficient to initiate liposarcoma in mouse but not in hASCs, suggesting the need of additional cooperating mutations in hASCs. A microarray gene expression profiling has shed light into the potential deregulated pathways in liposarcoma formation from p53-deficient mASCs expressing FUS-CHOP, which might also function as potential cooperating mutations in the transformation process from hASCs. STEM CELLS 2011; 29:179–192
It is becoming evident that mesenchymal stem cells (MSCs) may be the target cell in which transforming mutations responsible for sarcoma development arise . Several types of sarcomas have been reproduced in vivo on overexpression of specific fusion oncoproteins in bone marrow (BM)-derived mouse MSCs (mMSCs). These models include the recapitulation of Ewing's sarcoma by expression in BM-mMSCs of EWS-FLI-1 [2, 3], mixoid liposarcoma (MLS) by expression of FUS (FUsed in Sarcoma; also termed TLS, Translocated in LipoSarcoma)-CHOP (C/EBP HOmologous Protein; also termed DDIT3, DNA Damage-Inducible Transcript 3) , and alveolar rhabdomyosarcoma by expression of PAX-FKHR . Furthermore, cancer-initiating cells displaying MSC properties have been recently identified in Ewing's Sarcoma . However, the success in reproducing these tumors using human MSCs (hMSCs) has not yet been achieved. The expression of EWS-FLI-1 in hMSCs does not transform but induces a gene expression profile similar to that observed in Ewing's sarcomas .
A liposarcoma classification based on the differentiation stage of the MSCs has been suggested . Among liposarcomas, MLS is the most common, representing 35% of all cases . At the molecular level, MLS is characterized by the presence of the chromosomal translocation t(12;16) (q13;p11) which generates the FUS-CHOP fusion oncogene . This oncogene encompasses the NH2-terminal domain of FUS fused in-frame to the entire coding sequence of CHOP. The NH2-terminal domain of FUS confers the transactivation domain to the fusion protein . CHOP is a member of the C/EBP family of transcription factors, capable of forming heterodimers with and inactivate other (CCAAT/enhancer binding protein) members .
Previous in vitro approaches have shown the transforming effects of FUS-CHOP in fibroblasts , but not in 3T3-L1 preadipocytes, suggesting that the activity of FUS-CHOP may be cell-dependent. The generation of transgenic mice expressing FUS-CHOP under the control of EF1α promoter gave rise to liposarcomas, which resembled their human counterparts . As aforementioned, the expression of FUS-CHOP in BM-mMSCs gave rise to MLS  supporting the idea that liposarcoma develops in mesenchymal progenitors, at least in murine models.
The origin and pathogenesis of many tumors has extensively been studied in vivo using animal models and in vitro using murine cells. However, differences between the mouse and human have left gaps in our understanding. Unfortunately, a human model to reproduce FUS-CHOP-mediated liposarcoma using hMSCs is still missing. Such a human model would be crucial to further investigate not only many aspects associated with the pathogenesis of liposarcoma but also to undertake high-throughput small compounds screening, thus, eliminating confounding in vitro and in vivo effects owing to the differences between murine- and human-based disease models .
Human cells do not transform as easily as mouse cells, suggesting that in addition to the initiating fusion oncogene, secondary transforming hits may be necessary to transform human cells. We and others have demonstrated the relevant role that the deficiency of different cell cycle regulators, especially p53, has on the transformation of mMSCs [16–20]. In fact, p53 mutations and p53 deregulated expression are frequent in liposarcomas [21, 22], supporting p53 as a potential secondary hit for hMSC transformation.
Here, we attempted to model liposarcoma by ectopic expression of FUS-CHOP in both wild-type (wt) and p53-deficient mouse and human adipose-derived mesenchymal stem/stromal cells (ASCs). In mouse, we found that FUS-CHOP expression does not transform mouse ASCs (mASCs) on its own. However, the expression of FUS-CHOP in p53−/− mASCs induces liposarcoma. On the other hand, FUS-CHOP was unable to transform either wt or p53-deficient human ASCs (hASCs), therefore, emphasizing a further need for unknown cooperating mutations to induce liposarcoma from hASCs. A microarray gene expression profiling has revealed potential deregulated pathways in liposarcoma formation from p53-deficient mASCs expressing FUS-CHOP, which might also function as potential cooperating mutations in the transformation process from hASCs.
MATERIALS AND METHODS
MSC Sourcing and Culture
Mouse ASC cultures were established from adipose tissue from gonadal, retroperitoneal, and subcutaneous depots of FVB mice as previously described . p53−/− mASCs on a FVB background were previously described . BM-mMSCs were obtained by flushing the femurs and tibias from 8- to 12-week-old FVB mice with phosphate-buffered saline (PBS) supplemented with 2% fetal bovine serum (FBS). Mononuclear cells were then plated at a density of 106 cells per centimeter square in murine mesenchymal medium and supplements (http://www.stemcell.com; Vancouver, Canada, Stem Cells Technologies) and incubated at 37°C in a 5% humidified CO2 atmosphere. After 48 hours, nonadherent cells were discarded and fresh medium was added. When cell culture achieved >85% of density, adherent cells were trypsinized, washed, and replated at a concentration of 5,000 cells per centimeter square.
Human ASCs were obtained from Inbiobank (www.inbiobank.org; San Sebastian, Spain). Two independent cultures established from lipoaspirates from two donors (39- and 45-year-old white females with 29.5 and 32 body mass index, respectively) were used. hASCs were cultured in advanced-Dulbecco's modified Eagle's medium (http://www.invitrogen.com; Carlsbad, CA, Gibco) plus 10% FBS. When cell cultures achieved over 85% of density, adherent cells were trypsinized, washed in PBS, and replated at a concentration of 5 × 103 cells per centimeter square.
Lentiviral Vectors, Viral Production, and ASC Transduction
FUS-CHOP cDNA (exon 7 of FUS is fused to exon 2 of CHOP; kindly provided by Dr. I Sanchez-Garcia, Salamanca, Spain) was inserted into the pRRL-EF1α-PGK-GFP lentiviral expression vector. Viral particles pseudotyped with VSV-G were generated on 293T cells using a standard calcium-phosphate transfection protocol and were concentrated by ultracentrifugation. Murine and human ASCs were infected overnight with concentrated viral particles expressing either pRRL-EF1α-PGK-GFP or pRRL-EF1α-FUS-CHOP-PGK-GFP. The next day, the viral supernatant was removed and transduced cells were washed with MSC media and allowed to expand . In those experiments where the transduction efficiency resulted lower than 85%, the transduced (Green Fluorescent Protein) GFP+ fraction was enriched using a FACSAria cell sorter (www.bd.com; Franklin Lakes, NJ, Becton Dickinson; supporting information Fig. 1 and Fig. 5A). Depletion of p53 was achieved by transduction with lentiviral particles carrying a p53-shRNA expression vector (pLVUH-shp53; Addgene plasmid 11,653 ). In a typical experiment, the efficiency of transduction using these lentiviral particles was of approximately 80% (supporting information Fig. 2).
In Vitro Culture Homeostasis and Differentiation Analyses
Morphology of the different ASC cultures was recorded daily. As for the growth kinetics of the different ASC cultures, cells were counted every 4 days and replated at a density of 3 × 103 cells per centimeter square. Cumulative population doublings were calculated as previously described . Colony-forming assays and replicative senescence was determined as previously described . MSCs differentiation studies were performed by plating either mouse or human ASCs in specific differentiation inductive media for 2 weeks as described . G-banding karyotype analysis was performed as previously described [26, 27]. Fifty metaphases were consistently analyzed for each ASC genotype.
Flow Cytometry and Cell Sorting
The immunophenotype of cultured ASCs was determined by flow cytometry using fluorochrome-conjugated monoclonal antibodies anti-Sca-1, CD11b, CD14, CD29, CD44, and CD45 for mASCs or anti-CD90, CD73, CD105, CD166, CD106, CD45, CD34, HLA-DR, CD19, and CD14 (BD Bioscience, http://www.bdbiosciences.com; Franklin Lakes, NJ) for hASCs as described [19, 23]. 5-bromo-2-deoxyuridine incorporation and total DNA content were measured as described . Cell sorting of GFP+ ASCs was conducted under BSL-2 conditions using a FACSAria cell sorter Becton Dickinson (http://www.bd.com; Franklin Lakes, NJ).
Reverse Transcription Polymerase Chain Reaction
Total RNA extraction and reverse transcription polymerase chain reaction (RT-PCR) were done as previously described . RT-PCR conditions were as follows: cDNA synthesis at 37°C for 2 hours, pre-PCR denaturation at 94°C for 2 minutes followed by 35 cycles of denaturation at 94°C for 30 seconds, annealing at 64°C for 30 seconds, and extension at 72°C for 30 seconds. FUS-CHOP primers (forward: 5′-GACAGCA GAACCAGTACAACAG-3′ and backward: 5′-TGAGTCATT GCCTTTCTCCTTC-3′) amplified a 443 bp sequence spanning the FUS-CHOP fusion breakpoint. GAPDH primers (forward: 5′-GAAGGTGAAGGTCGGAGTC-3′ and backward: 5′-GAAGATGGTGATGGGATTTC -3′) were used as loading control and amplified a sequence of 228 bp.
Whole cell extracts were prepared as previously described . Thirty micrograms of protein was resolved on 10% (sodium dodecyl sulfate polyacrylamide gel electrophoresis) gels and blotted onto nitrocellulose membrane (BioRad, http://www.bio-rad.com; Hercules, CA). Proteins were detected using a chemiluminescence detection system (BioRad, http://www.bio-rad.com; Hercules, CA) according to the manufacturer's instructions. Antibodies used were as follows: anti-p53 ([sc-6243 and sc-126], 1:500 dilution), anti-GADD153/CHOP ([sc-7351], 1:200; used for the detection of FUS-CHOP), anti-PPARγ ([sc-7196], 1:500) and anti-C/EBPα ([sc-61], 1:500) from Santa Cruz Biotechnology (http://www.scbt.com; Santa Cruz, CA); anti-FUS/TLS ([A300-292A], 1:7,000) from Bethyl Laboratories (http://www.bethyl.com; Montgomery, TX); anti-p21 (, 1:750) from BD-Pharmingen (http://www.bdbiosciences.com; Franklin Lakes, NJ); and anti-β-actin ([A-1978], 1:20,000) from Sigma (http://www.sigmaaldrich.com; St. Louis, MO).
Anchorage-Independent Cell Growth
Soft agar colony formation assay was carried out using the CytoSelect 96-well cell transformation assay kit (Cell Biolabs Inc., http://www.cellbiolabs.com; San Francisco, CA) as described . For each genotype, two independent experiments were performed in triplicate. Hela cells and the transformed line T-AD-p53-FC #4 were used as positive controls.
In Vivo Tumorogenesis Assays
Nonobese diabetic/severe combined immunodeficient NOD. Cg-Prkdcscid IL2rgtm1Wjl/SzJ (NOD/SCID-IL2R−/−) mice were obtained from Jackson Laboratories (http://jaxmice.jax.org; Bar Harbor, ME). All mice were housed under specific pathogen-free conditions, fed ad libitum, and maintained under veterinary care according to animal welfare facilities guidelines. Eight- to twelve-week-old animals were used . Mice were inoculated subcutaneously with 5 × 106 ASCs (between passage 5 and 10 for mASCs and 5 and 15 for hASCs) or 1 × 106 tumor-derived cells. Animals were sacrificed when tumors reached approximately 10 mm or 4 months after infusion. All animal research protocols were approved by the Animal Research Ethical Committee of the University of Granada prior to the study. On tumor removal, half tumor was mechanically disaggregated to establish ex vivo tumor cell lines as described . The remaining portion of the tumor was used for immunohistopathology analysis.
Tumor samples were fixed in formol, embedded in paraffin, cut into 4-μm sections, and stained with hematoxylin and eosin (H&E). Multiple tumor sections were stained with specific antibodies against S-100 (1:500 dilution), smooth-muscle β-actin (1:50 dilution), desmin (1:100 dilution), myogenin (1:50 dilution; all from DakoCytomation, http://www.dako.com; Glostrup, Denmark), and GFP (1:400 dilution) from Invitrogen (http://www.invitrogen.com; Paisley, UK) as previously described .
Gene Expression Microarray Analysis
ASCs were collected during the exponential cell growth and stabilized in RNA later (Ambion, http://www.ambion.com; Austin, TX) solution until RNA extraction. RNA was isolated using the Agilent Total RNA Isolation Kit (Agilent Technologies, http://www.agilent.com; Palo Alto, CA) and its quality checked in the Agilent 2100 Bioanalyzer. Total RNA samples were labeled with Cy3 using the Quick-Amp Labeling Kit and hybridized with the Gene Expression Hybridization Kit to the Whole Human Genome Microarray (G4112F) or Whole Mouse Genome Microarray (G4122F), following Manufacturer's instructions (Agilent Technologies, CA). Each sample was labeled and hybridized as independent triplicates . Primary data was examined using GeneSpring 11.0 software (Silicon Genetics, Redwood City, CA). Gene expression in the control and experimental groups were compared using Student's t test and a Benjamini Hochberg multitesting correction. Only genes satisfying the threshold of p value < .05 and a fold change expression >2 were included and assigned as significant. Pools of genes that were differentially expressed were clustered according to their expression pattern dynamics into hierarchical tree clustering algorithms using the Pearson's centered correlation distance definition as similarity measure and Centroid's as the linkage rule. Analysis of pathways significantly altered by FUS-CHOP was performed using the Ingenuity Pathway software 8.0 (Ingenuity Systems, Inc., http://www.ingenuity.com; Redwood City, CA). Microarray data has been deposited and is available at Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo/).
FUS-CHOP Induces Liposarcoma When Expressed in p53-Deficient Mouse ASCs
FUS-CHOP cDNA was subcloned into a pRRL-EF1α-PGK-GFP lentivector and GFP- and FUS-CHOP-expressing lentiviral particles were prepared and used to transduce both wt and p53−/− mASC. Transduction efficiency measured as GFP expression was determined by fluorescence microscopy and flow cytometry (supporting information Fig. 1). To ensure, we worked with homogeneous mASC cultures, those cultures displaying transduction efficiencies <85% underwent cell sorting to enrich the GFP+ cell fraction (supporting information Fig. 1). The expression of FUS-CHOP transcript and protein was verified by RT-PCR (Fig. 1A) and Western blot using an anti-GADD153/CHOP antibody, which recognizes a 74-kDa band corresponding to FUS-CHOP and a unspecific band of higher molecular weight (Fig. 1B). p53 absence in p53−/− mASCs was also confirmed by Western blot (Fig. 1B).
Phenotypically, both wt and p53−/− mASCs displayed typical MSC phenotype: CD44+, CD29+, Sca1+ (Sca1low in p53−/− mASCs as reported [19, 20]) and absence of the hematopoietic markers CD45, CD14, and CD11b (Fig. 1C). Similarly, both wt and p53−/− mASCs showed identical osteogenic differentiation potential, whereas their adipogenic potential differed as reported , with p53−/− mASCs displaying smaller oil droplets in the cytoplasm indicating a poorer adipogenic differentiation (Fig. 1D). Importantly, the expression of FUS-CHOP did not influence either the phenotype or the differentiation properties of wt or p53−/− mASCs.
To assay the in vivo tumorogenic potential of the different mASCs, NOD/SCID IL2Rγ−/− mice were inoculated subcutaneously with 5 × 106 wt GFP-expressing mASCs (wt-GFP), wt FUS-CHOP (FC)-expressing mASCs (mASC-wt-FC), p53−/− GFP-expressing (mASC-p53-GFP), or p53−/− FC-expressing mASCs (mASC-p53-FC). Opposite to wt-GFP or wt-FC mASCs, both p53-GFP and p53-FC mASCs generated tumors in vivo (Fig. 2A, 2B). Similar to our previous observations , p53−/− mASCs (mASC-p53-GFP) gave rise to leiomyosarcoma with a tumor incidence of 50% and a latency period of 70 days. Interestingly, the expression of FC in p53−/− mASCs (mASC-p53-FC) induced a marked aggressiveness of these cells. On inoculation of mASC-53-FC cells, tumor incidence reached 100% and the latency period was significantly shortened to 37 days (Fig. 2A). Tumors induced by mASC-p53-GFP or mASC-p53-FC displayed relevant differences. Macroscopically, tumors from mASC-p53-FC had a softer texture and displayed large nodules of differentiated fat cells (Fig. 2B). The entire tumors showed GFP fluorescence confirming their ASC origin (Fig. 2B). Likewise, immunohistochemical analysis of serial sections showed a strong GFP staining of tumor cells (supporting information Fig. 3). As we previously described , histology analyses revealed that mASC-p53-GFP-induced leiomyosarcoma is composed of interlacing fascicles of spindle cells positive for desmin and smooth-muscle actin, partially positive for S-100 and negative for myogenin (Fig. 2C). However, tumors derived from mASC-p53-FC cells showed large areas composed of lipogenic cells surrounded by stroma comprising both oval and round-shaped cells (Fig. 2D, panel i). These tumors displayed features of human MLSs including the presence of irregularly sized adipocytes and lipoblasts (Fig. 2D, panel ii; red arrows) and a certain level of capillary vascularization (Fig. 2D, panel iii; black arrows). Tumoral lipogenic areas stained positive for S-100, whereas the larger host-derived mature adipocytes were S-100 negative (Fig. 2D, panel iv). Interestingly, although the expression is weaker than that observed in mASC-p53-GFP tumors, mASC-p53-FC tumors were also partially positive for desmin and SM-actin (Fig. 2D, panels v, vi). Together, the expression of FC in p53−/− mASCs redirects the tumor genesis/phenotype and fate from leiomyosarcoma to a liposarcoma histologically mixed with areas of leiomyosarcoma.
Notably, previous studies reported the development of a mouse model of MLS induced by the sole ectopic expression of FUS-CHOP in wt BM-mMSCs . However, ectopic FUS-CHOP expression in wt mASCs did not generate sarcomagenesis, indicating that different genetic insults (i.e., p53 deficiency) may be required to induce FC-mediated liposarcoma in mASCs but not in BM-mMSCs. To confirm this idea, we expressed FUS-CHOP in BM-mMSCs derived from wt mice, and we addressed their in vivo tumorogenic potential as we previously did for mASCs (supporting information Fig. 4). Opposite to control cells (GFP), FUS-CHOP-expressing BM-mMSCs generated tumors in all injected mice (n = 5). Macroscopically, tumors displayed fat nodules like the tumors from FUS-CHOP-expressing mASCs, although they were much more vascularized. Overall, the tumors are highly vascularized undifferentiated sarcomas; although they also display characteristics typical of MLSs as shown by the presence of atypical multivacoulated lipoblasts and the presence of a plexiform vascular pattern (see supporting information Fig. 4 for a detailed description). These results suggest that BM-mMSCs are more susceptible to FUS-CHOP-induced transformation and sarcomagenesis than mASCs.
Ex Vivo-Established Cell Lines from FUS-CHOP-Expressing Tumors Show MSC Properties In Vitro and Initiate Liposarcoma In Vivo
To gain further insights into these experimentally induced liposarcoma-like tumors, five lines were derived from p53−/− FC-expressing tumors (T-AD-p53-FC #1 to T-AD-p53-FC #5; characterization of two lines is shown in supporting information Fig. 5). These ex vivo-established T-AD-p53-FC-transformed mASC lines retained FUS-CHOP expression as confirmed by RT-PCR (Fig. 3A) and Western blot (Fig. 3B).
To investigate their tumorogenic potential, two T-AD-p53-FC lines were reinoculated into immunodeficient mice. These transformed ASCs forms liposarcoma-like tumors in all transplanted mice (100% tumor penetrance) within a 2- to 4-week period (Fig. 3C). The histology of these secondary tumors was similar to that of the primary tumors (Fig. 3D). Furthermore, cultures derived from these tumors showed the presence of small fat vacuoles in the cytoplasm even in the absence of adipogenic differentiation factors, suggesting the partial commitment to fat lineage of these T-AD-p53-FC tumor cells (Fig. 3D).
FUS-CHOP Expression in mASCs Alters Signaling Pathways Relevant to Liposarcoma Formation
To gain further insights into the potential molecular and cellular mechanisms underlying the tumor genesis/phenotype redirection induced by FUS-CHOP expression in p53-deficient mASCs, we compared the gene expression profile of mASC-p53-GFP (which form leiomyosarcoma-like tumors) and mASC-p53-FC cells (which form liposarcoma-like tumors) by microarray analysis. We initially generated the lists of genes differentially modulated (p value <0.01; expression more than two fold up or down) between p53-GFP or p53-FC and wt-GFP mASC cultures (supporting information Table 1). A total of 7,603 and 7,170 genes were altered in mASC-p53-GFP and mASC-p53-FC cells, respectively. Most of these genes (5,766) were common in both lists, whereas 1,837 or 1,404 genes were specifically altered in mASC-p53-GFP and mASC-p53-FC cells, respectively (Fig. 4A). Because both mASC-p53-GFP and mASC-p53-FC cells are transformed, the genes commonly altered in both mASC genotypes (5,766) would likely affect important signaling pathways responsible for the tumoral transformation of mASCs. Meanwhile, those genes specifically altered in mASC-p53-FC cells would reflect the genetic program induced by FUS-CHOP in p53-deficient mASCs responsible for tumor phenotype redirection. Therefore, we searched for the signaling pathways associated to these groups of genes using the Ingenuity Pathway Analysis (IPA) software (Fig. 4B, 4C). Many relevant signaling pathways including routes involved in the control of proliferation and/or differentiation of MSCs/ASCs were commonly altered in mASC-p53-GFP and mASC-p53-FC cells. Thus, the disrupted p53 signaling may cooperate with pathways controlled by G-protein-coupled receptors, Wnt, phosphatase and tensin homolog (PTEN), fibroblast growth factor (FGF), and phosphoinositide-3-kinase/v-akt murine thymoma viral oncogene homolog (PI3K/AKT) among others, contributing to tumoral transformation in mASCs (Fig. 4B and supporting information Table 2). On the other hand, the expression of FUS-CHOP induced changes in signaling pathways related with lipid metabolism and/or sarcoma genesis such as the inhibition of the retinoid X receptor (RXR), sphingolipid and fatty acid metabolism, prolactin signaling, and platelet-derived growth factor (PDGF) pathway (Fig. 4C). These changes could explain, at least in part, the FUS-CHOP-induced tumorogenesis redirection toward the development of a liposarcoma phenotype.
FUS-CHOP-mediated deregulation of genes involved in adipogenic differentiation seems to be key for liposarcoma development. We thus analyzed the effect of FUS-CHOP expression on relevant adipogenic regulators [33, 34] by assessing the gene expression changes observed in the microrarray analysis of wt-FC and p53-FC versus wt-GFP mASCs (Fig. 4D). In wt-FC mASCs, FUS-CHOP seems to induce a proadipogenic gene expression profile by upregulating several positive regulators of adipogenesis while inhibiting the negative regulator FoxA2. However, in p53-FC mASCs, there was a more complex regulation of adipogenesis, with several positive and negative regulators being either upregulated or downregulated (Fig. 4D), suggesting an impairment of the adipogenic pathways in these p53-FC-mASCs. Among these adipogenic regulators, the most relevant transcription factors controlling the final stages of adipogenesis, peroxisome proliferator-activated receptor γ (PPARγ) and CCAAT/enhancer binding protein α (C/EBPα), are repressed in liposarcomas of FC-expressing transgenic mice as well as in MLS cell lines . We thus analyzed the protein levels of PPARγ and C/EBPα in cell lines derived from tumors induced by p53-wt (TAD-53-1) and p53-FC mASCs (TAD-FC53-3; Fig. 4E). Similar to previous reports , the expression of PPARγ and especially of C/EBPα was found heavily repressed in the FC-expressing tumors (TAD-FC53) as compared with TAD-53 tumor or white adipose tissue protein extract.
FUS-CHOP Expression Neither Induces Changes in Culture Homeostasis nor Initiates Liposarcoma in Either wt or p53-Depleted hASCs
According to the above liposarcoma-like model based on FUS-CHOP expression coupled to p53 deficiency in mASCs, we next wanted to reproduce this model using hASCs. These fat-derived ASCs were transduced with either GFP- or FC-expressing lentiviruses (Fig. 5A). To ensure, we work with homogeneous hASC cultures, those cultures displaying transduction efficiencies <85% underwent cell sorting to enrich the GFP+ cell fraction (Fig. 5A). Then, p53 expression was depleted in both GFP- and FC-expressing hASCs using lentiviruses expressing a p53-specific shRNA. Four genotypes were generated for downstream experiments: GFP-expressing hASCs (GFP), FUS-CHOP-expressing hASCs (FC), p53-deficient GFP-expressing hASCs (p53-GFP), and p53-deficient FUS-CHOP-expressing hASCs (p53-FC; Fig. 5A). The expression of the FUS-CHOP was confirmed by RT-PCR and WB (Fig. 5B). We also confirmed by western blot (WB) the reduction of p53 protein levels in p53-GFP and p53-FC cells both in unstressed conditions or after the induction of DNA damage by treatment with the topoisomerase I inhibitor campthotecin (CPT; Fig. 5C). The activation of p21 after CPT treatment was also prevented, indicating that p53 depletion is effectively blocking the downstream signaling of this protein (Fig. 5C).
All hASCs genotypes displayed a typical hMSC immunophenotype (supporting information Fig. 6). Likewise, all hASCs genotypes showed similar morphology and adipogenic differentiation potential (Fig. 5D). However, FC-expressing hASCs showed a much robust osteogenic differentiation potential as compared with GFP hASCs (Fig. 5D). Intriguingly, this enhanced osteogenic differentiation potential observed in FC hASCs is lost in either p53-GFP or p53-FC hASCs, suggesting that FUS-CHOP expression boosts osteogenic differentiation in a p53-dependent manner.
Long-term analysis of hASC growth kinetics revealed that p53 depletion robustly enhanced the growth rates of p53-GFP and p53-FC hASC cultures (Fig. 6A), and consequently, these cultures showed a higher proportion of cells in S-phase (Fig. 6B) and an increased clonogenic capacity (Fig. 6C). The expression of FUS-CHOP, however, did not impact these culture homeostasis properties in wt or p53-depleted hASCs. The lifespan of p53-GFP and p53-FC hASC cultures was longer as compared with wt GFP- or FC-expressing hASCs, which reached senescence (measured as β-galactosidase+ staining) about 30 days earlier (Fig. 6A). Additionally, all hASCs remained euploid after long-term in vitro expansion (supporting information Fig. 7)
Finally, we assayed the in vitro and in vivo transformation capacity for the different hASCs genotypes. In contrast to our positive controls, none of the hASCs cultures were capable of generating colonies in the anchorage-independent growth-based in vitro transformation assays (Fig. 6D). Furthermore, none of the hASCs cultures formed tumors on inoculation in NOD/SCID IL2Rγ−/− mice (Fig. 6E). Altogether, these results indicate a differential outcome of FUS-CHOP overexpression in ASC of mouse versus human origin.
Differences in Gene Expression Profiling Between GFP- and FC-Expressing hASCs Suggest that FUS-CHOP Acts As an Overall Transcriptional Repressor
To identify patterns of gene expression, which could help to elucidate the biological effect of FUS-CHOP expression in hASCs, we compared the transcriptional profiles of GFP- versus FC-expressing hASCs by microarray analysis. Data analysis identified 183 genes differentially expressed (p value < .05; expression more than twofold up or down) between GFP- and FC-expressing hASCs (supporting information Table 1). Overall, FUS-CHOP expression functioned as a transcriptional repressor in hASCs because its expression induced the downregulation of 127 differentially expressed genes (70%) and the upregulation of 56 genes (30%; Fig. 7A). This trend was very similar to that observed after FUS-CHOP expression in mASCs (70.5% and 29.5% of downregulated and upregulated genes, respectively).
Analysis of the altered genes using the IPA software revealed that not many signaling pathways and gene families displayed a significantly altered gene expression profile in FC-expressing hASCs (Fig. 7A, 7B). Among the altered pathways endothelin-1, Wnt/β-catenin, and G-protein-coupled receptors signaling were previously reported to be deregulated in cancer [36–38]. Indeed, among the biological/pathological functions identified by IPA software, cancer becomes the most significantly altered after the expression of FUS-CHOP both in human and mouse ASCs (supporting information Table 3). Thus, 46 of the 183 genes (25.1%) modulated by FC in hASCs were previously involved in tumorogenesis (supporting information Table 3).
When the gene expression profile induced by FUS-CHOP was compared between human and mouse ASCs, we found that 17.5% (32 of 183) and 36.8% (174 of 473) of the genes modulated in hASCs-wt-FC versus hASCs-wt-GFP and hASCs-p53-FC versus hASCs-wt-GFP, respectively, were also altered in the matching mASCs. Targets commonly altered by FUS-CHOP expression in both mouse and human ASCs include key regulators of the Wnt signaling pathway (Wnt4, Sfrp1, Sfrp2, and Axin2) and other relevant factors involved in MSCs fate control (Bmp6, Fos, etc.; supporting information Table 4).
Opposite to mASC-p53-FC, hASC-p53-FC cells did not promote tumor formation when injected in immunodeficient mice. In an attempt to elucidate candidate secondary-transforming events, which could be important to drive tumor formation in hASCs, we analyzed the signaling pathways differentially altered in the hASC-p53-FC cells (as compared with hASC-wt-GFP) using the IPA software. The signaling pathways modulated in hASC-p53-FC cells were then compared with those altered in the tumorogenic mASC-p53-FC cells (Fig. 7C, 7D and supporting information Table 2). Among the pathways altered in hASCs-p53-FC cells, there are several routes related to cell cycle checkpoints and apoptosis whose alteration could be relevant in the process of tumoral transformation. On the other hand, mASC-p53-FC cells showed significant changes in many other signaling routes whose deregulation is strongly associated with enhanced tumorogenesis, such as G-protein-coupled receptors, cAMP-mediated factors, PTEN, Wnt/β-catenin, or PI3K/AKT signaling. This robust microarray analysis suggests that some of these pathways may need to be targeted to promote full transformation of FC-expressing human ASCs.
Increasing evidence indicates that MSCs might constitute the target cell for transforming mutations responsible for sarcoma development, suggesting that MSCs/ASCs may become an instrumental tool in studies aimed at dissecting the pathogenesis and cellular origin of sarcomas . It has been demonstrated that BM-mMSCs provided a permissive environment for tumoral transformation and sarcoma development mediated by several sarcoma-associated fusion genes [2–5]. Nevertheless, a human sarcoma model remains to be developed. Here, we have explored the role of FUS-CHOP overexpression on both wt and p53-deficient mouse and human ASCs.
Recent findings show that MSCs carrying specific mutations initiate different types of sarcomas depending on the tissue from which those MSCs are sourced. For instance, p53 deficiency in BM-mMSCs forms osteosarcoma , whereas p53 deficiency in fat-derived mASCs induces leiomyosarcoma . Riggi et al.  previously reported the formation of liposarcoma-like tumors when FUS-CHOP was overexpressed in BM-mMSCs. Here, however, we report that the expression of FUS-CHOP in fat-derived mASCs is not sufficient to transform these mASCs or initiate liposarcoma in vivo, suggesting a differential impact of FUS-CHOP in MSCs derived from BM or adipose tissue.
We recently reported that p53−/− mASCs form leiomyosarcoma-like tumors, linking this type of smooth muscle sarcoma to p53 deficiency in ASCs . Intriguingly, expression of FUS-CHOP in p53−/− mASCs induced higher aggressiveness in the tumorogenic properties of mASC-p53-FC as compared with mASC-p53-GFP: tumor incidence reached 100% and the latency period significantly shortened. Importantly, tumors resulting from mASC-p53-FC displayed a phenotypic switch from leiomysarcoma to liposarcoma-like tumors. These tumors displayed features of human MLSs including irregularly sized lipoblasts and numerous dilated capillaries. Our data is further supported by a previous study reporting a switch of tumoral phenotype after FUS-CHOP expression in a human fibrosarcoma cell line . Together, these data suggest that the expression of FUS-CHOP triggers the formation of liposarcoma-like tumors from mASCs, but, in contrast to BM-mMSCs, secondary cooperating oncogenic hits such as p53 deficiency are required for liposarcoma development. Similar results have been reported in a model of alveolar rhabdomyosarcoma based on the expression of PAX-FKHR in mMSCs in which p53 deficiency is needed for tumor development .
Similar to recent studies, which highlight the role that FUS-CHOP plays in the control of important adipogenic regulators, especially PPARγ and C/EBPα , we found a similar downregulation of these transcription factors in those tumors derived from FUS-CHOP-expressing mASC-p53-FC cells. Nevertheless, the gene expression levels of PPARγ are not affected by the expression of FUS-CHOP in mASCs, in line with previous data from BM-mMSCs . Furthermore, the levels of C/EBPα were upregulated by FUS-CHOP in both wt and p53-deficient mASCs (Fig. 4D). Together, these data indicate that FUS-CHOP seems to predispose mASCs to adipogenic differentiation in vitro through the upregulation of C/EBPα, although the in vivo cellular environment seems to play a relevant role in blocking adipogenesis and therefore in the potential development of the malignancy. In line with the partial commitment of FUS-CHOP-expressing mASCs to the adipogenic lineage, an incomplete spontaneous adipogenic differentiation of cells derived from FUS-CHOP-expressing tumors was observed (Fig. 3D). In addition, we observed a consistent loss of Sca-1 expression in p53-deficient mASCs (Fig. 1C) which has been associated with a reduced adipogenic potential of MSCs . Nevertheless, as we previously reported, the loss of Sca-1 in p53-deficient mASCs is not correlated with an enhanced tumorogenic potential .
The molecular basis underlying how FUS-CHOP redirects tumor genesis/phenotype in p53-deficient mASCs could be, at least in part, explained by the signaling pathways specifically altered by the presence of FUS-CHOP in mASC-p53-FC cells (Fig. 4A–4C). Among these pathways, the inhibition of RXR signaling is relevant as this nuclear receptor is an obligate heterodimeric partner for PPARγ necessary to modulate the expression of genes under the control of peroxisome proliferator response elements . Other significantly altered pathways such as prolactin signaling and sphingolipid metabolism were also reported to modulate PPARγ activity and adipogenesis [42, 43]. Notably, other FUS-CHOP-specific altered routes, such as fatty acid metabolism, were also found to be altered in liposarcoma samples . In addition, Riggi et al.  have identified PDGFA as a new relevant target induced by the expression of FC in BM-mMSCs. Importantly, although we have not observed a significant variation in PDGFA, the PDGF signaling was also significantly modulated by the presence of FUS-CHOP in mASC-p53-FC cells (Fig. 4C). Likewise, we also found a relevant overlap of 39 and 112 commonly deregulated genes when the effect of FUS-CHOP expression in BM-mMSCs (data from Riggi et al.) was compared with mASC-wt-FC and mASC-p53-FC cells, respectively (supporting information Table 5). Moreover, Matushansky et al.  have reported 4,027 genes differentially expressed between samples of MLSs and normal fat. We found that 229 and 482 of these genes were also upregulated or downregulated by FUS-CHOP expression in wt and p53-deficient mASCs (supporting information Table 6). In addition, three genes (PEG3, GPX3, and ADH1B) altered in mASC-wt-FC cells and five genes (TRO, SAA1, SOX11, GPX3, and ADH1B) deregulated in mASC-p53-FC cells, were also among the 11 most informative genes for MLS as described by Singer et al. . These gene expression profile comparisons suggest that (a) FUS-CHOP expression in mASCs modify the expression profile toward a liposarcoma-like pattern and (b) that p53 deficiency cooperates in this process.
It has been reported that the expression of FUS-CHOP  or EWS1-FLI1  caused cell death in several human cell lines although p53 or p16 gene depletion increased cell survival . We thus attempted to create a model of human liposarcoma by expressing human FUS-CHOP in either wt or p53-deficient hASCs. Based on the role that the deficiency of p53 plays in the transformation of mASCs  and in the fact that the expression of this cell cycle regulator is frequently altered in soft tissue sarcomas, we hypothesized that p53 protein depletion would be a secondary cooperating hit candidate tissue sarcomas [21, 22]. In contrast to previous studies on human cell lines , FUS-CHOP expression resulted nontoxic in wt or p53-depleted primary hASCs, suggesting that hASCs provide a permissive environment for the expression of this fusion protein. p53 depletion resulted on a much higher growth rate, colony-forming ability, and increased lifespan. However, FUS-CHOP expression neither induced further changes in culture homeostasis nor initiated liposarcoma in either wt or p53-depleted hASCs. Similar to our data, the expression of EWS-FLI-1 was able to transform BM-mMSC  but unable to transform BM-hMSCs although it induced a gene expression profile resembling features of Ewing's Sarcomas . Moreover, mMSCs but not hMSCs generated osteosarcoma-like lesions in the lung following systemic injection in mice . These data confirm that mMSCs are much more susceptible to transformation and sarcoma development  than hMSCs, which seem to require additional cooperating mutations to constitute a useful cellular model for sarcomagenesis. It is plausible that several cooperating oncogenic hits are needed to induce lineage-specific transformation of hMSCs .
Gene expression profiling showed that FUS-CHOP functions as a transcriptional repressor in hASCs. This effect was previously observed in the human fibrosarcoma cell line HT1080 . FUS-CHOP expression results in constitutive expression of CHOP mediated by the FUS promoter, whereas in normal cells, the transcription of CHOP is tightly regulated and generally repressed [51, 52]. Both CHOP and FUS-CHOP form dimmers and inactivate many members of the C/EBP family of transcription factors [12, 35], explaining, in part, the transcriptional repression observed in FUS-CHOP-expressing hASCs.
Among the signaling pathways displaying a significantly deregulated gene expression profile in FUS-CHOP-expressing hASCs, Wnt signaling functions as an important regulator of MSC fate. Activation of Wnt signaling stimulates osteogenesis and inhibits adipogenesis . Wnt signaling also controls the growth and transformation of MSCs. Thus, the expression of the Wnt signaling inhibitor Dkk-1 in hMSCs induced the formation of undifferentiated pleiomorphic sarcomas in vivo  and, likewise, this inhibitor is overexpressed in human osteosarcoma . In addition, Dkk-1 is required for BM-MSCs entry into the cell cycle . FUS-CHOP expression in hASCs seems to disrupt the Wnt/β-catenin pathway because several inducers and repressors genes of the pathway were modulated by FUS-CHOP expression. Importantly, the expression of many of these Wnt signaling members (Wnt4, Axin2, Sfrp1, and Sfrp2) is also modulated by FC in mASCs. Other relevant pathways altered by FUS-CHOP in hASCs include the Endothelin axis and the G-protein-coupled receptor signaling. The Endothelin signaling has been implicated in cancer development and deregulation of this pathway has been found in several types of sarcomas . Likewise, the G-protein-coupled receptors play a pivotal role in a wide array of important signaling pathways and many of these receptors have been implicated in tumorogenesis .
Overall, the adipogenic program of hASCs does not seem highly altered by the presence of FUS-CHOP. Among the most relevant proadipogenic factors only Klf-4 was downregulated by FUS-CHOP. Likewise, the overall overlap between genes modulated in hASC-wt-FC cells (25 genes) and hASC-p53-FC cells (37 genes) and those genes differentially expressed between MLSs and normal fat  is more modest than that observed for mASCs (supporting information Table 6). This reduced gene expression deregulation induced by FUS-CHOP expression in hASCs may explain the failure of the hASC-p53-FC cell to produce tumors when injected into immunodeficient mice. On careful comparison of the gene expression profiles of both mASCs and hASCs, several genes specifically modulated in mASC-p53-FC cells came out as potential additional oncogenic insults which may cooperate with FUS-CHOP and p53 deficiency in the transformation process of hASCs. For instance, based on its previously reported link with cancer development , the role of the PTEN/PI3K/AKT pathway in the transformation of hASCs should be carefully addressed in future studies.
Together, our results suggest a differential outcome of FUS-CHOP overexpression in mouse versus human ASCs and indicate that FC expression in a p53-deficient background is sufficient to induce liposarcoma-like tumors in mouse but not in human ASCs, suggesting the need of still uncharacterized additional cooperating mutations in hASCs.
We attempted to model liposarcoma by expression of FUS-CHOP in wild type and p53-deficient mouse and human adipose-derived mesenchymal stem/stromal cells (m/hASCs). The expression of FUS-CHOP triggers the formation of liposarcoma-like tumors from mASCs but, in contrast to bone marrow-derived mMSCs, secondary cooperating hits such as p53 deficiency are required for liposarcoma development. In the human setting, FUS-CHOP was unable to transform either wild type or p53-deficient hASCs, thus emphasizing the need for further cooperating mutations. A microarray gene expression profiling has revealed potential deregulated pathways in liposarcoma formation from p53-deficient mASCs expressing FUS-CHOP.
We thank Enrique de Alava (CIC, Salamanca), Francisco Nogales, and Francisco O'Valle (University of Granada) for their assistance on the pathology diagnosis, Purificación Catalina and Paola Leone for assistance with G-banding assays, and Dr. Patrick Aebischer for the pLVUHshp53 plasmid. This work was supported by the CSJA (0030/2006 to P.M. and 0108/2007 to R. Rodriguez) and CICE (P08-CTS-3678 to P.M.) de la Junta de Andalucía, The FIS/FEDER (PI070026 and PI100449 to P.M.), and the MICINN to P.M. (PLE-2009-0111). R. Rodriguez is supported by a Fellowship from the AECC.
DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
The authors indicate no potential conflicts of interest.