Expression of FUS-CHOP fusion protein in immortalized/transformed human mesenchymal stem cells drives mixoid liposarcoma formation

Authors

  • Rene Rodriguez,

    Corresponding author
    1. Hospital Universitario Central de Asturias and Instituto Universitario de Oncología del Principado de Asturias, Oviedo, Spain
    • Correspondence: Pablo Menendez, Ph.D., Josep Carreras Leukemia Research Institute. Facultat de Medicina. University of Barcelona. Carrer de Casanova 143, 08036 Barcelona, Spain. Telephone: +34 655010572; Fax: +34 934651472; e-mail: pmenendez@carrerasresearch.org; or Rene Rodriguez, Ph.D., Hospital Universitario Central de Asturias and Instituto Universitario de Oncología del Principado de Asturias, Laboratorio 2 ORL-IUOPA, C/Celestino Villamil s/n, 33006 Oviedo, Spain. Telephone: +34-985-10-80-00, ext. 38524; Fax: +34-985-10-80-15; e-mail: renerg@ficyt.es

    Search for more papers by this author
  • Juan Tornin,

    1. Hospital Universitario Central de Asturias and Instituto Universitario de Oncología del Principado de Asturias, Oviedo, Spain
    Search for more papers by this author
  • Carlos Suarez,

    1. Hospital Universitario Central de Asturias and Instituto Universitario de Oncología del Principado de Asturias, Oviedo, Spain
    Search for more papers by this author
  • Aurora Astudillo,

    1. Servicio de Anatomía Patológica, Hospital Universitario Central de Asturias, Oviedo, Spain
    Search for more papers by this author
  • Ruth Rubio,

    1. GENYO. Centre for Genomics and Oncological Research, Pfizer, University of Granada, Andalusian Goverment, Granada, Spain
    Search for more papers by this author
  • Carole Yauk,

    1. Environmental Health Science and Research Bureau, Healthy Environments and Consumer Safety Branch, Ottawa, Ontario, Canada
    Search for more papers by this author
  • Andrew Williams,

    1. Environmental Health Science and Research Bureau, Healthy Environments and Consumer Safety Branch, Ottawa, Ontario, Canada
    Search for more papers by this author
  • Michael Rosu-Myles,

    1. Health Canada Centre for Biologics Research, Biologics and Genetic Therapies Directorate, Ottawa, Ontario, Canada
    Search for more papers by this author
  • Juan M. Funes,

    1. UCL Cancer Institute, University College London, London, United Kingdom
    Search for more papers by this author
  • Chris Boshoff,

    1. UCL Cancer Institute, University College London, London, United Kingdom
    Search for more papers by this author
  • Pablo Menendez

    Corresponding author
    1. GENYO. Centre for Genomics and Oncological Research, Pfizer, University of Granada, Andalusian Goverment, Granada, Spain
    2. Josep Carreras Leukemia Research Institute-Cell Therapy Program, University of Barcelona, Barcelona, Spain
    3. Institució Catalana de Reserca i Estudis Avançats (ICREA), Barcelona, Spain
    • Correspondence: Pablo Menendez, Ph.D., Josep Carreras Leukemia Research Institute. Facultat de Medicina. University of Barcelona. Carrer de Casanova 143, 08036 Barcelona, Spain. Telephone: +34 655010572; Fax: +34 934651472; e-mail: pmenendez@carrerasresearch.org; or Rene Rodriguez, Ph.D., Hospital Universitario Central de Asturias and Instituto Universitario de Oncología del Principado de Asturias, Laboratorio 2 ORL-IUOPA, C/Celestino Villamil s/n, 33006 Oviedo, Spain. Telephone: +34-985-10-80-00, ext. 38524; Fax: +34-985-10-80-15; e-mail: renerg@ficyt.es

    Search for more papers by this author

  • Author contributions: J.T. and R.Ru.: performed experiments; C.S., A.A., C.Y., and A.W.: analyzed data and interpreted the results; J.M.F. and C.B.: provided key biological material; P.M.: conceived the study, analyzed the data, interpreted the results, wrote the paper, and financially supported the study; R.Ro.: conceived the study, designed the study, performed experiments, analyzed the data, interpreted the results, wrote the paper, and financially supported the study; M.R.-M.: analyzed data and interpreted the results and financially supported the study. P.M. and R.Ro. share senior authorship.

Abstract

Increasing evidence supports that mesenchymal stromal/stem cells (MSCs) may represent the target cell for sarcoma development. Although different sarcomas have been modeled in mice upon expression of fusion oncogenes in MSCs, sarcomagenesis has not been successfully modeled in human MSCs (hMSCs). We report that FUS-CHOP, a hallmark fusion gene in mixoid liposarcoma (MLS), has an instructive role in lineage commitment, and its expression in hMSC sequentially immortalized/transformed with up to five oncogenic hits (p53 and Rb deficiency, hTERT over-expression, c-myc stabilization, and H-RASv12 mutation) drives the formation of serially transplantable MLS. This is the first model of sarcoma based on the expression of a sarcoma-associated fusion protein in hMSC, and allowed us to unravel the differentiation processes and signaling pathways altered in the MLS-initiating cells. This study will contribute to test novel therapeutic approaches and constitutes a proof-of-concept to use hMSCs as target cell for modeling other fusion gene-associated human sarcomas. Stem Cells 2013;31:2061–2072

Introduction

It has been recently established that transformed mesenchymal stem cells (MSCs) may act as tumor-initiating cells capable of initiating sarcomagenesis in vivo. Accordingly, many efforts have been undertaken to characterize the transformation process of MSCs, and to prospectively generate models for different sarcomas based on MSCs, which would constitute an unprecedented system to understand the mechanisms underlying sarcomagenesis and to search target-specific therapies [1-3]. Among the different types of sarcomas, those characterized by the presence of tumor-specific fusion oncogenes as a result of chromosomal translocations constitute an active field of research. Several types of tumors resembling human sarcomas have been reproduced in vivo using transgenic mouse models, and also upon the expression of sarcoma-specific fusion proteins in mouse MSC (mMSCs). Specifically, Ewing Sarcoma [4, 5], myxoid liposarcoma [6-9], alveolar rhabdomyosarcoma [10, 11], and synovial sarcoma [12] have been reproduced upon expression in mouse cells of EWS-FLI-1, FUS-CHOP (FC), PAX-FKHR, and SYT-SSX, respectively. However, previous efforts to model sarcomagenesis using human MSC (hMSCs) failed to reproduce the tumor phenotype. For example, the expression of the fusion proteins EWS-FLI-1 or SYT-SSX1 in hMSCs induced a transcriptional expression pattern similar to that observed in Ewing's Sarcomas [13] or synovial sarcoma [14], respectively, but did not cause cell transformation. In fact, no fusion gene-based model of sarcoma has been developed so far using hMSCs as target cells. The “multiple hit” model of cancer suggests that, at least in mesenchymal cancer (leukemias and sarcomas), a fusion oncogene may act as an initiating oncogenic event by blocking a particular differentiation pathway whereas secondary cooperating hits that most likely affect proliferation/apoptosis, may be required to render a fully transformed phenotype [15].

Myxoid liposarcomas (MLS) represent approximately one-third of liposarcomas and account for approximately 10% of all adult soft tissue sarcomas. MLS occurs predominantly in patients aged 30–50 years old, and shows a high tendency to recur locally or to metastasize to other soft tissue locations [16]. A proportion of cases show histological progression to a subclass of MLS displaying round-cell morphology, a feature significantly associated with a poor prognosis. Five-year survival rates vary between 20% and 60%, depending in part on the round-cell histology progression [16]. MLS is characterized by the recurrent translocation t(12;16)(q13;p11), which fuses FUS (FUsed in Sarcoma; also termed TLS, Translocated in LipoSarcoma) to CHOP (C/EBP HOmologous Protein; also termed DDIT3, DNA Damage-Inducible Transcript 3) on chromosome 12. The NH2-terminal domain of FUS confers the transactivation domain to the fusion protein [17]. CHOP is a member of the C/EBP (CCAAT/Enhancer Binding Proteinα) family of transcription factors and heterodimerizes with, and inactivates other C/EBP members [18]. FUS-CHOP (FC) fusion protein is believed to repress the development of adipocytic precursors by repressing PPARγ (Peroxisome Proliferator-Activated Receptor γ) and C/EBP, causing disruption of differentiation and the development of MLS [19].

We previously showed that the expression of FC in transformed leiomyosarcoma-forming p53-deficient adipose tissue-derived mMSCs (ASCs) [20] redirects the tumor genesis/phenotype toward the formation of liposarcoma-like tumors [9]. However, in the human setting, the expression of FC fails to transform either wt or p53-deficient hASCs [9]. Here, we take advantage of an available collection of sequentially mutated human bone marrow MSCs (BM-hMSCs) ranging from wt (no oncogenic hits) to fully transformed hMSCs (targeted with five oncogenic mutations) [21] to address whether expression of FC in BM-hMSCs harboring different oncogenic lesions cooperates to develop a bona fide hMSC-based model reproducing the MLS phenotype. Importantly, most of these oncogenic hits or their downstream signaling pathways are among the few genetic alterations, apart from the FC translocation, described in human MLS to date (see Discussion). Overall, we have developed and characterized a model of MLS based on the expression of FC in transformed hMSCs. This study provides the first human model of sarcomagenesis based on the expression of a fusion oncogene in hMSCs.

Materials and Methods

Generation and Culture of Mutated Human BM-hMSCs

Wild-type or “0-hits” BM-hMSCs (wt-MSCs or MSC-0H) were obtained from Inbiobank (www.inbiobank.org; San Sebastian, Spain) upon signed informed consent. Wt-MSCs depleted of p53 (MSC-1H) (Supporting Information Fig. S1) were generated by transduction with lentiviral particles carrying a p53-shRNA expression vector (pLVUH-shp53; Addgene plasmid 11653; [22]) as previously described [9]. The BM-hMSCs carrying three, four, or five different oncogenic hits (MSC-3H, MSC-4H, and MSC-5H) were developed and characterized elsewhere [21]. Figure 1A summarizes the genotype/oncogenic hits of each hMSC type used in this study. To over-express FC, each type of BM-MSCs (MSC-0H to −5H) was infected overnight with concentrated viral particles expressing either pRRL-EF1α-PGK-GFP (empty vector; green fluorescent protein [GFP]) or pRRL-EF1α-FUS-CHOP-PGK-GFP (FC expressing vector; FC) as previously reported (Supporting Information Fig. S2A) [9]. The transduction efficiency (50%–96%) was analyzed by flow cytometry (% GFP+ cells) (Supporting Information Fig. S2B). All the resulting MSC-GFP and MSC-FC cell lines were cultured and immunophenotypically characterized as previously described [9, 23].

Figure 1.

FC expression in immortalized/transformed hMSCs initiates mixoid liposarcoma-like tumors in vivo. (A): Summary of the BM-hMSCs types used in this study. The oncogenic hits used are indicated. (B): Endpoint RT-PCR confirming the expression of FC at mRNA level in MSCs and human MLS samples (MLS#5 and #6; N and T: normal and tumoral samples obtained from the same patient). GAPDH was used as housekeeping gene. (C): Western blotting showing the expression of FC protein using an anti-GADD153/CHOP antibody. The expression of FUS was used as loading control. (D): RT-qPCR comparing the expression of FC in MSCs with human MLS samples (MLS#1 to #6). mRNA expression was referred to the FC levels in the MLS sample with higher expression (MLS#6: 100%). (E–G): Hematoxylin-eosin staining and GFP immuno-staining detection in tumors developed from MSC-4H-FC (E), MSC-5H-GFP (F), and MSC-5H-FC (G) cells. MSC-4H-FC- and MSC-5H-FC-derived tumors display bona fide histological characteristics of human MLS (myxoid matrix [M]; round-cell areas [R]; lipoblasts [blue arrows]; and plexiform vascular pattern [red arrows]). Original magnification is indicated. Abbreviations: BM-MSC, bone marrow mesenchymal stem cells; GFP, green fluorescent protein; MLS, mixoid liposarcoma.

MSC Differentiation

MSCs were plated in six-well plates and allowed to grow to confluence. Culture medium was then replaced with specific differentiation inductive media. For adipogenic differentiation, cells were cultured in adipogenic MSC Differentiation BulletKit (Lonza, Basel, Switzerland, http://www.lonza.com) for 7 or 14 days. Differentiated cell cultures were stained with oil red O (Sigma, St. Louis, MO, http://www.sigmaaldrich.com) [24] and the level of differentiation was quantified by extracting the stain with isopropanol and measuring the absorbance at 510 nm. For osteogenic differentiation, cells were cultured in Osteogenic MSC Differentiation BulletKit (Lonza) for 14 days, and the presence of calcium deposits in differentiated cells was checked by Alizarin red S staining (Sigma) [24].

RT-PCR

Total RNA was extracted from undifferentiated or adipogenic differentiated MSC cultures as well as from human MLS tissue samples obtained upon signed informed consent and first-strand cDNA synthesis was performed using the First-Strand cDNA Synthesis Kit (GE Healthcare, Pittsburgh, PA, http://www.gelifesciences.com). The expression of FC and GAPDH was checked by endpoint PCR as previously described [9]. The expression of FC, PPARγ (total and isoform 2), C/EBPα, C/EBPδ, lipoprotein lipase (LPL), MDM2, CDK2, and MET was assessed by Q-PCR using SYBR Green PCR Kit (Qiagen, Valencia, CA, http://www.qiagen.com) [25]. GAPDH was used as a housekeeping gene. The following PCR conditions were used: 5 minutes at 94°C, 35 cycles of 30 seconds at 94°C followed by 50 seconds at 60°C and 50 seconds at 72°C, and a final extension of 10 minutes at 72°C. Primer sequences used are shown in Supporting Information Table S1.

Western Blot

Whole cell protein extraction and Western blot analysis were done as previously described [26]. Antibodies used were as follows: anti-p53 ([sc-126], 1:500 dilution), anti-GADD153/CHOP ([sc-7351], 1:200; used for the detection of FC), anti-PPARγ-2 ([sc-166731], 1:200 dilution), and anti-C/EBPδ ([sc-636], 1:200 dilution) from Santa Cruz Biotechnology (Santa Cruz, CA, http://www.scbt.com); anti-FUS/TLS ([A300-292A], 1:7,000) from Bethyl Laboratories (Montgomery, TX, http://www.bethyl.com); and anti-β-Actin ([A-1978], 1:20,000) (Sigma).

In Vivo Tumorogenesis Assays

NOD/SCID IL2Rγ−/− mice were obtained from Jackson Laboratories (Bar Harbor, ME, http://jaxmice.jax.org). All mice were housed under specific pathogen-free conditions, fed ad libitum, and maintained under veterinary care according to animal welfare facilities guidelines. Animals were used at 8–12 weeks of age. Mice were inoculated subcutaneously with 5 × 106 MSCs or 1 × 106 primary tumor-derived cells. Animals were sacrificed when tumors reached approximately 1 cm3 or 5 months after infusion. All animal research protocols were approved by the Animal Research Ethical Committee of the University of Granada prior to the study. Upon tumor removal, half of the tumor was mechanically disaggregated to establish ex vivo tumor cell lines as described [20]. The remaining portion of the tumor was used for immunohistopathology analysis.

Histological Analysis

Tumor samples were fixed in formol, embedded in paraffin, cut into 4 µm sections, and stained with hematoxylin and eosin. Multiple tumor sections were stained with specific antibodies against GFP (1:400 dilution; Molecular Probes, Eugene, OR, http://www.invitrogen.com) or S-100 (1:500; DakoCytomation, Glostrup, Denmark, http://www.dako.com) as previously described [27, 28].

Gene Expression Profiling Analysis

MSCs were collected in RNA later (Ambion, Austin, TX, http://www.ambion.com) solution until RNA extraction. RNA was isolated using the Agilent Total RNA Isolation Kit (Agilent Technologies, Santa Clara, CA, http://www.agilent.com). Sample purity was checked to ensure that A260/280 ratios for all the RNA samples were ≥2. Only RNA samples with RNA integrity ≥9 were used in the analysis as determined using an Agilent 2100 Bioanalyzer (Agilent Technologies). RNA was amplified using Agilent Fluorescent Linear Amplification kits (Agilent Technologies), and the cRNA of experimental samples was labeled with Cyanine 5-CTP. Universal reference total RNA (Stratagene, Mississauga, ON, Canada, http://www.genomics.agilent.com) was labeled with Cyanine 3-CTP. Samples were hybridized to Whole Human Genome 8 × 60K Microarray (G4851A) and arrays were scanned using an Agilent G2505B scanner (Agilent Technologies). Data were extracted using Feature Extraction software version 10.7.3.1. (Agilent Technologies) and array quality was evaluated using metrics established in Feature Extraction. All preprocessing of the data were conducted using the R software. The median signal intensities for the expression arrays were normalized using the global lowess method using the transform.madata function in the MAANOVA library. Differentially expressed genes were identified using the MAANOVA library using the Fs statistic. The p-values for all the statistical tests were estimated using the permutation method (30,000 permutations with residual shuffling). These p-values were then adjusted for multiple comparisons using the false discovery rate approach. The least-squares means were used to estimate the fold changes for each pairwise comparison. Data from the sample channel only (Cy5) were background subtracted and were normalized using cyclic-lowess normalized followed by collapsing the data to the Gene Symbol using the median. This was done to compare the data from an Affymetrix study (GSE21122) obtained from National Center for Biotechnology Information Gene Expression Omnibus site (http://www.ncbi.nlm.nih.gov/projects/geo). These data were normalized using Robust Multi-array Average using the Affy library in Bioconductor, and then collapsed to the gene symbol using the median. The two datasets were merged together using the gene symbol and a linear model was applied to control for differences between two studies. The adjusted data were then collapsed by tumor type and hierarchical clustering using the dissimilarity matrix based on one minus the Pearson correlation with average linkage was performed. This analysis was conducted using the gene symbols that were significantly expressed in the above analysis. Analysis of canonical pathways and upstream regulators significantly altered by FC was performed with the list of differentially expressed genes using ingenuity pathway analysis (IPA) software 8.0 (Ingenuity Systems, Inc., Redwood City, CA, http://www.ingenuity.com). Microarray data have been deposited and are available in Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo/; GSE48030).

Results

The Expression of FC in Immortalized/Transformed hMSCs Initiates MLS In Vivo

To test whether the expression of FC in immortalized/transformed hMSCs results in liposarcoma formation similar to that observed in the mouse setting [9], we used several sequentially mutated BM-hMSCs ranging from wt (MSC-0H) to fully transformed hMSCs (targeted with up to five oncogenic mutations; MSC-5H) (Supporting Information Table S2) [21]. These oncogenic hits include: (a) p53 inactivation achieved by p53-shRNA or by expression of the E6 antigen of the HPV-16, (b) Rb inactivation induced by expression of the E7 antigen of the HPV-16, (c) ectopic expression of hTERT, (d) introduction of the SV40 small T antigen to inactivate the PPA2 phosphatase leading to stabilization of c-myc, and (e) expression of oncogenic H-RAS (H-RASv-12) (Fig. 1A). Most of these oncogenic lesions or their related signaling pathways are among the few relevant alterations previously described in human MLS (see Discussion). The distinct hMSC types were transduced with GFP- (control) and GFP/FC-expressing lentiviral particles, and the expression of FC transcript and protein was verified by RT-PCR (Fig. 1B) and Western blot (Fig. 1C) using an anti-GADD153/CHOP antibody, which recognizes a 74-kDa band corresponding to FC. There are differences in the level of expression of FC achieved in the different MSCs; however, a range of FC expression is also observed in human MLS samples. The expression of FC in MSC-0H and MSC-1H is in the lower limit of the range of FC expression observed in a panel of human MLS samples (MLS#1 to #6), while the expression of FC in MSC-3H is in the middle of the range, and the expression of FC in MSC-4H and MSC-5H is in the upper limit (Fig. 1D). In any case, the expression of FC did not influence either proliferation or cell death among the distinct MSC types (Supporting Information Fig. S3). Phenotypically, all hMSC types, regardless of the number of oncogenic hits they carry and the FC expression, displayed a typical MSC phenotype: CD90+, CD73+, CD105+, CD166+, CD44+, CD45−, CD34−, CD14−, CD19−, CD106−, and HLA-DR− (Supporting Information Table S3).

To assay the in vivo tumorogenic potential, NOD/SCID IL2Rγ−/− mice were inoculated subcutaneously with hMSCs of the different genotypes. Independent of FC expression, MSC-0H, MSC-1H, and MSC-3H were unable to initiate tumorogenesis in vivo. Similarly, MSC-4H-GFP did not developed tumors either (Table 1). However, FC expression in MSC-4H (MSC-4H-FC) initiated tumor formation with 85% penetrance, and a latency of 138 ± 32 days. Histological analysis of the MSC-4H-FC-derived tumors classified these tumors as MLS displaying large areas of small spindle/round cells dispersed within a myxoid matrix and showing the characteristic plexiform vascular pattern and some degree of fatty differentiation including the presence of atypical lipoblasts. Some of these tumors also presented areas of closely packed high-grade small blue cells, characteristic of the round-cell subtype of MLS (Fig. 1E; Table 1). These experimentally induced MLS closely resembled human MLS (Supporting Information Fig. S4) and stained positive for S-100, a marker reported to be positive in most cases of MLS [29] (Supporting Information Fig. S5). Interestingly, both MSC-5H-GFP and MSC-5H-FC cells developed aggressive tumors in vivo with 100% penetrance, and very short latency (25 days) (Table 1). Histological analysis classified the MSC-5H-GFP-derived tumors as undifferentiated tumors resembling spindle cell sarcomas (UPCS) (Fig. 1F; Table 1). However, the tumors derived from MSC-5H-FC cells showed together with undifferentiated zones, large areas displaying the main features observed in human MLS (Fig. 1G; Table 1, Supporting Information Figs. S4, S5). In all cases, tumor areas stained positive for GFP, confirming the hMSC origin (Fig. 1E, 1G).

Table 1. In vivo tumor formation ability
MSC typeTumors/mice (penetrance)LatencyaHistological analysis
  1. a

    Mean of days (± SD) needed to observe an approximate tumor volume of 1 cm3.

0H-GFP0/4 (0%)
0H-FC0/4 (0%)
1H-GFP0/4 (0%)
1H-FC0/4 (0%)
3H-GFP0/7 (0%)
3H-FC0/7 (0%)
4H-GFP0/7 (0%)
4H-FC6/7 (85%)138 ± 32Myxoid liposarcoma
5H-GFP7/7 (100%)25 ± 2UPCS
5H-FC7/7 (100%)25 ± 2Myxoid liposarcoma + UPCS

We next investigated whether genes previously associated to human MLS such as MDM2, CDK2, or MET [8, 30-32] were also upregulated in FC-expressing hMSCs. MDM2 was only slightly upregulated in MSC-4H-FC, MSC-5H-FC as well as in human MLS samples (MLS#5), and CDK2 and MET showed variable levels of overexpression in MSC-1H-FC to MSC-5H-FC and MLS#5 (Supporting Information Fig. S6).

Together, these data indicate that FC expression in immortalized, nontransformed MSC-4H (p53 and Rb inactivation, hTERT expression, and c-myc stabilization) originates tumors in vivo resembling the histological and molecular features of human MLS, whereas FC expression in transformed MSC-5H (MSC-4H plus H-RASV12) is capable of redirecting tumor genesis/phenotype from a UPCS to a MLS phenotype, also recapituling the histological/molecular features of human MLS.

Primary FC-Expressing MLS Are Transplantable into Secondary NOD/SCID Mice

To determine whether the experimentally induced MLS can be serially transplanted, primary tumors formed from MSC-5H-FC or MSC-4H-FC cells as well as from MSC-5H-GFP cells were mechanically disaggregated and placed back in MSC culture conditions to establish immortalized cell lines (T-5H-GFP#1 to #3, T-5H-FC#1 to #3, and T-4H-FC#1 and #2). These ex vivo-established cell lines retained FC expression (Fig. 2A, 2B) at levels comparable to human MLS samples (Fig. 2C) and displayed identical morphology to the parental MSC-4H and MSC-5H cells. Upon inoculation into secondary recipients, all the ex vivo established cell lines were able to generate tumors in vivo with a 100% penetrance and with a short latency period of 15 days for T-5H cells and 38 days for T-4H cells (Fig. 2D). Histopathological analysis of the secondary tumors generated from T-5H-FC and T-4H-FC lines revealed that they retained the main features of human MLS observed in the primary tumors including the presence of large areas of mixoid matrix containing atypical lipoblasts and a plexiform vascular pattern together with varying amounts of the round-cell histological subtype (Fig. 2E, 2F). Additionally, hierarchical clustering of the gene expression profiling (GEP) of T-5H-GFP and T-5H-FC cell lines (Fig. 2G) together with a previously published human sarcoma GEP datasets [33] showed that T-5H-FC cells cluster more closely with human MLS tumors. The in vivo development of serially transplantable tumors with a phenotype and transcriptome similar to that of primary human MLS demonstrates that the expression of FC in immortalized (MSC-4H) or transformed (MSC-5H) hMSCs provides a bona fide model of human MLS.

Figure 2.

Immortalized FC-expressing cell lines ex vivo-derived from primary tumors reinitiate MLS in secondary recipients. Several cell lines were ex vivo-derived from primary tumors developed from MSC-5H-GFP (T-5H-GFP#1 to #3), MSC-5H-FC (T-5H-FC#1 to #3), and MSC-4H-FC (T-4H-FC#1 and #2) cells. (A): Endpoint RT-PCR confirming the expression of FC at mRNA level in mesenchymal stem cells (MSCs) and human MLS samples (MLS#5 and #6; N and T: normal and tumoral samples obtained from the same patient). GAPDH was used as housekeeping gene. (B): Western blotting showing the expression of FC protein in T-5H-FC and T-4H-FC cell lines. The expression of FUS was used as loading control. (C): RT-qPCR comparing the expression of FC in MSCs with human MLS samples (MLS#1 to #6). mRNA expression was referred to the FC levels in the MLS sample with higher expression (MLS#6: 100%). (D): Summary table indicating tumor incidence, latency, and histological analysis of tumors developed upon secondary inoculation in NOD/SCID mice of the indicated cell lines. (E, F): Hematoxylin-eosin staining of tumors developed from the T-4H-FC (E) and the T-5H-FC cells (F). Similar to primary tumors, second-round tumors display the main characteristic of myxoid liposarcomas (myxoid matrix [M]; round-cell areas [R]; lipoblasts [blue arrows]; and plexiform vascular pattern [red arrows]). (G): Hierarchical clustering of the gene expression data of T-5H-GFP (combination of T-5H-GFP#1 and #2) and T-5H-FC (combination of T-5H-FC#1 and #2) cell lines and the human sarcoma gene expression dataset GSE 21122 [33]. T-5H-FC cells cluster closer to human primary MLS as compared to T-5H-GFP cells. Abbreviations: GFP, green fluorescent protein; MLS, mixoid liposarcoma; UPCS, undifferentiated spindle cell sarcoma.

FC Expression and the Accumulation of Oncogenic Lesions Block the Adipogenic Differentiation Potential of hMSCs

A blockage of the adipogenic differentiation pathway appears to be a hallmark of liposarcoma development [34]. We investigated how the expression of FC and the sequential accumulation of specific oncogenic mutations affect the adipogenic differentiation potential of hMSCs. After 14 days of culture in adipogenic-inductive conditions, MSC-0H-GFP cultures displayed abundant lipid droplets filled adipocyte-like cells (Fig. 3A). In contrast, MSC-3H and MSC-4H cells displayed an impaired pattern of differentiation in which most cells of the culture presented a small amount (one to five) of lipid droplets in their cytoplasm, whereas MSC-5H cultures barely showed even these partially differentiated cells (Fig. 3A). Quantification of the oil red O staining provided evidence that adipogenic differentiation ability decreased gradually with the introduction of oncogenic lesions (Fig. 3B). Importantly, the expression of FC caused a 30% reduction in the adipogenic differentiation potential of MSC-0H cells (Fig. 3A, 3B). In addition, T-4H and T-5H tumoral cell lines showed negligible adipogenic differentiation potential in keeping with the parental MSC-4H and MSC-5H (Fig. 3B, 3C). Notably, MSC-4H and MSC-5H cells, regardless of FC expression, retained full ability to differentiate toward the osteoblastic lineage, suggesting that the introduction of the oncogenic events specifically interfere with adipogenic differentiation (Supporting Information Fig. S7).

Figure 3.

Adipogenic differentiation of control and FC-expressing human MSCs (hMSCs) and tumoral cell lines. Representative images of the indicated GFP- and FC-expressing hMSCs (A), and tumoral cell lines (C) following 14 days culture in adipogenic medium (or control medium in insets) and oil red O staining. (B): Quantification of the adipogenic fold-induction relative to cells cultured in control medium (n = 3; *, p-value <.05 vs. MSC-0H-GFP cells). Abbreviations: GFP, green fluorescent protein; MSC, mesenchymal stem cells.

Previous reports have shown that FC represses master transcription factors controlling adipogenesis such as peroxisome proliferator-activated receptor γ (PPARγ) and CCAAT/enhancer binding protein α (C/EBPα) [19]. Thus, to further investigate the molecular basis of the adipogenic inhibition observed in our hMSC-based MLS model, the expression of master adipogenic factors was analyzed in MSC-0H, MSC-5H, and T-5H cells after 0, 7, and 14 days in adipogenic-inductive culture conditions. At the transcriptional level, the activation of the master adipogenic transcription factors PPARγ-2 and C/EBPα was heavily repressed by the expression of FC in the MSC-0H cells (Fig. 4A). This strong inhibition was also observed after the accumulation of mutations (i.e., MSC-5H and T-5H cells [Fig. 4A]), indicating that oncogenic transformation of hMSCs disrupts the differentiation program. Interestingly, the expression of the early adipogenic transcription factor C/EBPδ was barely affected by the expression of FC and/or further cooperating oncogenic mutations (Fig. 4A). These findings were confirmed at the protein level for both PPARγ-2 and C/EBPδ. PPARγ-2 showed less activation throughout adipogenic differentiation in MSC-0H-FC cells as compared to MSC-0H-GFP cells, and was not activated throughout the adipogenic differentiation of MSC-5H or T-5H cells regardless of the expression of FC (Fig. 4B). However, C/EBPδ displayed similar upregulation during the adipogenic differentiation process in all the hMSC types (Fig. 4B). As a result of the FC-mediated inhibition of PPARγ-2 and C/EBPα, the expression of factors specific to terminal adipogenic differentiation, such as LPL, was also repressed in MSC-0H upon expression of FC (Fig. 4A), confirming the functional blockage of adipogenic differentiation potential observed in MSC-0H-FC (Fig. 3B). This repression of LPL was almost complete in MSC-5H and T-5H cells (Fig. 4A). Together, these data indicate that either FC expression or the accumulation of oncogenic hits disrupts the terminal adipogenic differentiation of hMSCs, likely contributing to the development of liposarcoma.

Figure 4.

FC and the accumulation of oncogenic hits cooperate to disrupt the expression of master adipogenic transcription factors. (A): RT-qPCR and (B) Western blotting analysis showing the induction of master adipogenic-related genes in the indicated human MSCs and tumoral cell lines after 0, 7, and 14 days in adipogenic-inductive culture conditions. The expression of β-actin was used as loading control in the Western blotting experiments. Abbreviations: GFP, green fluorescent protein; MSC, mesenchymal stem cells.

To increase our understanding of the impaired adipogenic differentiation program in our MLS model, we performed whole genome gene expression profiles (GEP) to identify genes altered (differentially expressed) in MSC-0H-FC, MSC-5H-FC, and T-5H-FC cell types compared to the control MSC-0H-GFP cells. Using IPA, we searched for adipogenic regulators that were significantly altered among the differentially expressed genes. In the MSC-0H-FC cells, several regulators, including the most important early (C/EBPβ and C/EBPδ) and late (PPARγ and C/EBPα) adipogenic factors, were altered (Fig. 5A). However, many more regulators were upregulated/downregulated in MSC-5H-FC and T-5H-FC cells including PPARγ and C/EBPα (Fig. 5A, 5B). RT-qPCR analysis confirmed the altered expression of these master adipogenic transcription factors (Fig. 5B). The expression of the PPARγ isoform that most efficiently induces terminal adipogenic differentiation (PPARγ-2) was mainly unaffected by the expression of FC or by oncogenic cooperating mutations (Fig. 5B). Nevertheless, the level of the total PPARγ mRNA (isoforms 1 and 2) increased in MSC-5H and T-5H cells, suggesting that introduction of the oncogenic mutations has an impact on the PPARγ-1 isoform, which seems to play a role earlier in adipogenesis [35] (Fig. 5A, 5C). Conversely, the expression of C/EBPα was downregulated in MSC-5H and T-5H (Fig. 5B). Thus, accumulation of the oncogenic hits appears to initially predispose the cells to a process of adipogenic differentiation, as suggested by the upregulation of the adipogenic positive regulators PPARγ-total/ PPARγ-1, MYB, or BMP2/4. However, adipogenic differentiation seems blocked at a later step, as suggested by the downregulation of late adipogenic positive regulators such as C/EBPα, USF1, KLF2, and STAT5, and the upregulation of TFAP2A and CHOP which function as repressors of C/EBPα (Fig. 5C). In order to validate this, we compared the GEP of the MSC-0H-FC, MSC-5H-FC, and T-5H-FC with the GEP produced by Sekiya et al. [36] from different adipogenic stages in hMSCs. We found that most of the genes reported by Sekiya et al. [36] to be upregulated during the earlier phases of hMSC adipogenesis, and significantly modulated in the FC-expressing cells, were also upregulated in our MSC-0H-FC, MSC-5H-FC, and T-5H-FC cells. Conversely, those genes reported by Sekiya et al. [36] to be upregulated later in adipogenesis were downregulated in our FC-expressing cells (Fig. 5D), suggesting that FC/oncogenic hits impede late/terminal rather than early adipogenic differentiation.

Figure 5.

Gene expression analysis confirms a late/terminal blockage in adipogenic differentiation of MSC-5H-FC and T-5H-FC MLS cellular models. Genes differentially expressed (p value < .005; expression more than twofold or down) in MSC-0H-FC, MSC-5H-FC, and T-5H-FC cells as compared to MSC-0H-GFP (wild-type/control cells) were analyzed using the IPA software. (A): List of positive (top) and negative (bottom) regulators of adipogenesis differentially expressed between MSC-0H-FC, MSC-5H-FC, and T-5H-FC versus MSC-0H-GFP. Gray color indicates alteration of the upstream regulators signaling without significant upregulation or downregulation of their own expression. Green and red colors indicate significant downregulation and upregulation, respectively, as compared to MSC-0H-GFP. The fold change values are indicated. (B): RT-qPCR analysis confirming the regulation of key adipogenic-related genes in the indicated human MSCs (hMSCs) and tumoral cell lines relative to MSC-0H-GFP cells. (C): Cartoon depicting the activation/repressive role of altered regulators during adipogenesis. (D): Comparison between the genes upregulated during the adipogenic differentiation process of hMSCs (data from Sekiya et al. [36]; left panel) and the genes significantly altered in MSC-0H-FC, MSC-5H-FC, and T-5H-FC cells (relative to MSC-0H-GFP). Green and red colors indicate significant downregulation and upregulation, respectively. The fold change values are indicated. The genes reported by Sekiya et al. to reach maximum activation early during adipogenesis are mainly found upregulated in FC-expressing cells whereas those genes found by Sekiya et al. upregulated later in adipogenesis are mainly downregulated in FC-expressing cells. Abbreviations: GFP, green fluorescent protein; MSC, mesenchymal stem cells.

Gene Expression Analysis Identifies Specific Signaling Pathways Affected by FC Expression and the Oncogenic Mutations (5H) in the MLS Model

Many (91 out of 135, 67.4%) of the genes that were differentially expressed in MSC-0H-FC cells were also differentially regulated in both MSC-5H-FC and T-5H-FC cells (Fig. 6A). These 91 genes commonly regulated in MSC-0H-FC, MSC-5H-FC, and T-5H-FC cells (Supporting Information Table S4) represent the contribution of FC expression to the MLS model. Similarly, the GEP of MSC-5H-FC and the corresponding ex vivo-derived tumoral cell line T-5H-FC were very similar, sharing 78%–81% differentially expressed genes (Fig. 6A). The list of genes commonly regulated between MSC-5H-FC and T-5H-FC (2,087 genes; excluding the 91 genes also regulated in MSC-0H-FC cells) represent the contribution of the oncogenic mutations (5H) to the MLS model. The combination of the genes commonly regulated by FC (91 genes) and 5H (2,087 genes) explain, at least in part, the molecular basis underlying the development of MLS in vivo. Pathway analysis was used to group the differentially regulated genes into canonical signaling pathways that were significantly altered by FC (Fig. 6B) and by the oncogenic mutations (5H) (Fig. 6C). As expected, the canonical signaling pathways that were altered in the MLS model (MSC-5H-FC/T-5H-FC) were a combination of those signaling pathways independently regulated by FC and by 5H. Importantly, three signaling pathways (AHR-, IL-17-, and VDR/RXR-mediated pathways) were commonly altered by both FC and by the oncogenic mutations (5H) (Fig. 6D).

Figure 6.

Signaling pathways significantly regulated by FC and/or by the oncogenic mutations (5H) in the MSC-5H-FC and T-5H-FC MLS model. (A): Venn diagram representing the number and overlay of genes differentially expressed in MSC-0H-FC, MSC-5H-FC, or T-5H-FC relative to MSC-0H-GFP cells. As many as 91 out of 135 genes (67.4%) differentially expressed in MSC-0H-FC cells were shared by both MSC-5H-FC cells and T-5H-FC tumoral cell lines. Furthermore, between 77.6% and 81.5% of the genes differentially expressed in MSC-5H-FC and T-5H-FC cells are common. (B–D): Lists of significantly (p < .05) modulated canonical signaling pathways generated using IPA software with the genes annotated in (A). (B): Signaling pathways commonly altered in MSC-0H-FC, MSC-5H-FC, and T-5H-FC cells (91 genes), identifying the contribution of the FC expression to the development of the MLS model (red dotted line). (C): Signaling pathways specifically altered in both MSC-5H-FC and T-5H-FC cells (2,087 genes) identifying the contribution of the oncogenic mutations (5H) to the MLS model (blue dotted line). (D): Signaling pathways commonly regulated by the synergic contribution of FC and oncogenic mutations 5H (91 + 2,087 genes), identifying the molecular basis underlying the development of MLS in vivo (purple dotted line). Abbreviations: MLS, mixoid liposarcoma; MSC, mesenchymal stem cells.

Discussion

Sarcomas are generally studied using primary patient samples in which full transformation events have already occurred and therefore, the mechanisms of transformation are not amenable to analysis. Hence, there exists the need to establish bona fide models to recapitulate sarcomagenesis in vitro and in vivo. Mounting evidence indicates that MSCs may represent the putative target cell of origin for a variety of human sarcomas, linking MSCs and cancer and encouraging the development of hMSC-based sarcoma models by targeting hMSC with the appropriate oncogenic events. Such models, like the one presented here, would improve our understanding of the mechanisms governing sarcomagenesis, eventually facilitating the testing of specific therapies directed against the sarcoma-initiating cell. In this regard, the analysis of the transcriptional profile in the collection of gradually transformed hMSCs used in this study identified several upregulated enzymes that could represent putative novel antisarcoma targets [37].

Previous attempts to reproduce human sarcomas by expressing fusion genes in hMSCs have been able to partially mimic the transcriptional expression pattern of several types of sarcomas [13, 14] but failed to transform hMSCs. We previously showed that the expression of FC in transformed leiomyosarcoma-forming p53-deficient mASCs [20] was able to redirect the tumor phenotype toward the formation of liposarcoma [9]. However, in the human setting, the expression of FC failed to transform either wt or p53-deficient hASCs [9]. We thus hypothesized that cooperating mutations are needed to transform hMSCs, so that the expression of FC together with cooperating oncogenic hits could result in liposarcoma formation similar to that reported in mASCs [9]. Here, we harnessed an available collection of sequentially mutated hMSCs ranging from MSC-0H (no oncogenic hits) to fully transformed MSC-5H (targeted with five oncogenic hits: p53 deficiency, Rb deficiency, hTERT expression, C-MYC stabilization, and H-RASV12 mutation) [21] to address whether expression of FC in hMSCs harboring different oncogenic insults cooperates to develop a bona fide hMSC-based model reproducing the MLS phenotype. This hypothesis is further supported by previous studies demonstrating: (a) the acquisition of a liposarcoma phenotype upon expression of FC in a fibrosarcoma cell line [38]; (b) the progressive accumulation of these oncogenic hits induces a gradual global hypomethylation (a hallmark of cancer) of hMSCs [39]; (c) immortalized/transformed hMSCs harbor oncogenic mutations in genes related to relevant signaling pathways altered in liposarcoma-forming FC-expressing p53−/− mMSCs but not in nontransformed equivalent hMSCs [9], indicating that the introduction of these oncogenic mutations may cooperate with FC to trigger liposarcoma development from hMSCs (these pathways include p53; Wnt and platelet-derived growth factor signaling, associated with c-myc stabilization; and PTEN, PI3K/AKT, G-coupled receptors, and fibroblast growth factor signaling pathways, linked to H-RAS activation); and (d) although there are not many recurrent alterations found in human MLS, apart from the translocation encoding FC, the oncogenic targets mutated in these hMSCs or related signaling pathways are among the most relevant alterations previously described in human MLS. Thus, frequent alterations have been reported in p53, Rb, and p16INK4a/p14ARF (involved in Rb regulation) in MLS tumors [30, 40]. Moreover, abnormal expression of G1 phase cell cycle proteins [32] including CDK2, which bind to FC and CHOP [41], has been reported in MLS tumors. Additionally, a p53 null background was needed for the development of a recently reported mouse model of MLS where FC was expressed by the mesoderm-specific Prx1 promoter [6], suggesting that FC synergizes with p53 in the formation of MLS. Furthermore, genome analysis of STS patients has identified a gain in chromosome 8, where the MYC gene is located, as the most common copy-number variation in MLS [33]. Point mutations in PIK3CA (associated with AKT activation) and KIT, which constitute important downstream signaling pathways for RAS oncogenic signals [42], have also been found in 18% and 4.8% of MLS cases, respectively [33].

The differences of FC expression observed between the different MSCs are probably due to the lower lentiviral transduction efficiency achieved in primary hMSCs as compared to transformed cells [43]. Nevertheless, experimentally induced FC expression is within the expression range observed in human MLS samples. The fact that MSC-3H-FC did not form tumors indicates that the expression of appropriate levels of FC in BM-hMSCs (lacking functional p53 and Rb and expressing hTERT) is not enough to achieve tumoral transformation, in line with previous reports where FC or other sarcoma-related fusion genes are expressed in nontransformed hASCs or BM-hMSCs [9, 13, 14]. Nevertheless, we cannot discard that higher levels of FC expression or alternative oncogenic lesions could facilitate the MLS development.

The expression of FC fully transforms immortalized—nontransformed—MSC-4H cells (p53 and Rb deficiency, hTERT overexpresion, and c-myc stabilization) giving rise to in vivo tumors resembling MLS. Furthermore, FC expression was able to redirect the tumor phenotype of the MSC-5H cells (MSC-4H plus H-RASV12) from UPCS toward MLS. This MLS phenotype could be reproduced upon serial transplantation, suggesting the existence of MLS-propagating cells. The ability of FC to redirect tumor genesis/phenotype from a UPCS to MLS indicates that this fusion oncogene has an instructive role in lineage/tissue commitment during transformation. The development of MLS tumors from MSC-4H-FC cells indicates that H-RASV12 is not necessary for the development of the MLS phenotype, but that it shortens tumor latency. To the best of our knowledge, this is the first human model of sarcomagenesis resulting from the expression of a fusion oncogene in hMSCs, and constitutes a proof-of-concept to use the same strategy to model other types of fusion gene-associated human sarcomas.

In line with previous reports, the expression of FC in MSC-0H cells largely prevented the activation of the master adipogenic factors PPARγ and CEBPα [19], resulting in a partial inhibition of the adipogenic differentiation. Similarly, the oncogenic mutations introduced in MSC-3H, MSC-4H, and MSC-5H cells largely impaired adipogenesis of hMSCs. The oncogenic mutations introduced in MSCs appear to display opposing effects on the adipogenic differentiation. For example, p53 and Rb inhibit adipogenic differentiation through repression of PPARγ [35, 44] suggesting that the deficiency of p53 and/or Rb in MSCs favors adipogenesis [44, 45]. Conversely, c-myc inhibits terminal adipocyte differentiation by suppressing PPARγ and CEBPα [46]. The introduction of oncogenic H-RASV12 in MSC-5H cells almost completely blocks adipogenesis of hMSCs. This observation is supported by the potent effect of RASV12 in the blockage of lipid accumulation and repression of adipocyte gene expression [47]. According to this, GEP analysis indicates that the proadipogenic (p53 and Rb deficiency) and antiadipogenic (c-myc and RASv-12) cooperating mutations strongly alter the adipogenic differentiation process by inducing both positive (upregulation of PPARγ-total, c-myb, BMPs, or KLF5 and downregulation of FOXO1) and negative (upregulation of GATA2/3, REL-A, TFAP2A, or CHOP and downregulation of CEBPα, USF1, or STAT5A) regulators of adipogenesis. Most likely, these stimulatory/inhibitory adipogenic effects of the oncogenic mutations cooperate with FC expression in the development of MLS. This is supported by the Prx1-FC mouse model of MLS in which MSCs were committed to adipocytic differentiation but unable to terminally differentiate [6]. Also, the comparison of our GEP (MSC-5H-FC and T-5H-FC) with the genes reported by Sekiya et al. [36] to be upregulated during hMSCs adipogenic differentiation revealed that there is a positive correlation at earlier time points but a negative correlation at later phases. These findings suggest that the combination of FC and the oncogenic hits impede late/terminal rather than early adipogenic differentiation.

Our GEP studies revealed that IL-6 signaling is the most significantly altered pathway upon FC expression in MSCs. Accordingly, in keeping with this observation, it was previously reported that IL-6 is upregulated in fibrosarcoma cell lines with ectopic expression of FC [48], and that IL-6 expression appears to be necessary and sufficient to enhance MSC proliferation, to inhibit MSC apoptosis, and to prevent adipogenic and chondrogenic differentiation [49]. In addition, we found that other signaling pathways known to be involved in the modulation of adipogenesis were also altered following FC expression in our MSCs including PPAR signaling [35], VDR/RXR activation [50], NFkB [51], IL-17 [52], and aryl hydrocarbon receptor (AHR) activation [53]. Conversely, our GEP analysis revealed that DNA damage response, cell cycle control, and pathways controlling cell fate, proliferation, and differentiation (such as WNT/β-catenin, cAMP, G-coupled receptor, protein kinase A, and PTEN signaling) are the pathways most significantly regulated by the introduction of oncogenic mutations (5H) in MSCs.

Conclusion

In summary, we have developed and characterized the first human stem cell-based model that reproduces a sarcoma phenotype upon the expression of a sarcoma-associated fusion oncogene in hMSCs. In this model of MLS, FC appears to trigger a lineage instructive role, and cooperates with oncogenic hits to block the adipogenic differentiation potential of hMSCs. We envision that this model will facilitate the in vivo testing of novel small-molecules directed against liposarcoma-initiating cells, and paves the way for modeling subsequent fusion gene-driven human sarcomas.

Acknowledgments

We thank the Andalusian Platform of Bioinformatics (PAB; University of Málaga) for providing access to IPA software and Dr. Patrick Aebischer for the pLVUHshp53 plasmid. This work was supported by the Instituto de Salud Carlos III/Fondos FEDER (PI10/00449 to P.M., CP11/00024 to R.Ro, Tercel [RD12/0019/0006] and RTICC [RD12/0036/0015]), the Junta de Andalucía/FEDER (P08-CTS-3678 to P.M.), the Spanish Association Against Cancer (Junta Provincial de Albacete-CI110023 to P.M. and Junta Provincial de Granada to R.Ro), Grupo Español de Investigación en Sarcomas (beca J.M. Buesa-2012 to R.Ro), Health Canada (H4084–112281 to P.M., R.Ro, and M.R-M.), Health Canada's Genomics Research and Development Initiative (to M.R-M, C.Y and A.W), and Obra Social Cajastur-IUOPA. R.Ro is supported by the Miguel Servet program of the ISCIII/FEDER. R.Ru was supported by a fellowship of the ISCIII/FEDER. P.M. is an ICREA Research Professor supported by the Generalitat of Catalunya.

Disclosure of Potential Conflicts of Interest

The authors indicate no potential conflicts of interest.

Ancillary