Recent studies have shown that germ-line determination occurs early in development and that extracellular signaling can alter this fate. This denial of a cell's fate by counteracting its intrinsic signaling pathways through extrinsic stimulation is believed to be associated with oncogenesis. Using specific populations of multipotent skeletal muscle-derived stem cells (MDSCs), we have been able to generate tumors by subjecting cells with specific lineage predilections to concomitant differentiation signals. More specifically, when a stem cell that had a predilection toward osteogenesis was implanted into a skeletal muscle, tumors formed in 25% of implanted mice. When cells predilected to undergo myogenesis were pretreated with bone morphogenetic protein 4 (BMP4) for 4 days prior to implantation, they formed tumors in 25% of mice. These same myogenic predilected cells, when transduced to express BMP4 and implanted into either a long-bone or cranial defect, formed bone, but they formed tumors in 100% of mice when implanted into the skeletal muscle. The tumors generated in this latter study were serially transplantable as long as they retained BMP4 expression. Furthermore, when we impeded the ability of the cells to undergo myogenic differentiation using small interfering RNA to the myogenic regulator MyoD1, we stopped transformation. Based on our findings, we postulate that specific MDSC populations can undergo concomitant signal-induced transformation and that the initial stages of transformation may be due to changes in the balance between the inherent nature of the cell and extrinsic signaling pathways. This theory represents a potential link between somatic stem cells and cancer and suggests an involvement of the niche/environment in transformation.
Disclosure of potential conflicts of interest is found at the end of this article.
Increasing evidence demonstrates that tumors are populated by a fraction of cells with stem cell-like properties . This has been shown for certain leukemias , breast , and brain tumors . Furthermore, it has also been shown that leukemic stem cells express antigens similar to those expressed by hematopoietic stem cells . Although there is evidence of a relationship between stem cells and cancer, the nature of this relationship remains unclear.
The ectopic expression of the telomerase catalytic subunit human telomerase reverse transcriptase (hTERT) in combination with two oncogenes, the simian virus 40 large-T oncoprotein (T/t-Ag) and an oncogenic allele of H-ras, can transform normal human epithelial and fibroblast cells . Using these defined genetic events, researchers have been able to generate a malleable model of rhabdomyosarcoma by converting undifferentiated human skeletal muscle cell precursors and committed human skeletal muscle myoblasts into their malignant counterparts by targeting pathways altered in rhabdomyosarcoma . Furthermore, the expression of mammalian proteins that inactivate the tumor suppressors Rb and p53, in conjunction with the oncoproteins Ras and Myc and the telomerase subunit hTERT, is sufficient to drive normal human somatic cells to a tumorigenic fate . These data potentially provide a blueprint of events that can lead to cancer and allow different cancers to be genetically modeled from normal cells.
The aforementioned studies detail a series of genetic events that are necessary for the transformation of somatic cells, and they investigate anaplasia, the change in cells that allows cancerous growth by dedifferentiation. Researchers have believed for many years that external agents, such as chemicals or viruses, cause cancer by inducing dedifferentiation of mature (differentiated) postnatal cells. Mounting evidence supporting the role of stem cells in cancer suggests an alternative hypothesis based on a modified “embryonal rest” model . According to this model, cells in the bloodstream renew all tissues; hence, both normal tissues and cancer can arise from stem cells in the blood. Furthermore, the embryonal rest hypothesis posits that tumors arise from displaced placental tissue or activated germinal cells in adult tissues. The current version of this theory states that cancer in adults develops from embryonal rudiments (stem cells) that remain in tissues of the fully matured organ. This hypothesis was first introduced in 1874  and revisited again in 2004, where it was postulated that tissue stem cells are the modern-day equivalent of embryonal rests and that most tumors arise from the maturational arrest of a cellular lineage derived from a tissue stem cell . It is our hypothesis that although a defined series of events is required for transformation, these events need not be necessarily genetic in nature. These occurrences may be due to the inherent “nature” of the cell and to a balance of intrinsic and extrinsic signaling pathways.
Herein, we report the transformation of specific populations of postnatal skeletal muscle-derived stem cells (MDSCs) following exposure to concomitant differentiation signals. Specifically, when muscle-derived progenitor cells were exposed to simultaneous osteogenic and myogenic signals in vivo, osteorhabdomyosarcomas were generated. When stimulated by multiple differentiation signals, these stem cells appear to proliferate uncontrollably and undergo malignant transformation. Furthermore, this transformation appears to be environment-specific and dependent on the ability of the exposed progenitor cells to respond simultaneously to multiple differentiation/growth signals. These findings suggest that stem cells stimulated by conflicting differentiation signals can become tumorigenic and therefore potentially represent a new paradigm for both postnatal stem and cancer cell biology. Furthermore, this work suggests a novel means of transformation of somatic stem cells that we believe is applicable to several human cancers and provides a model for the exploration of the initiating stages of cancer.
Materials and Methods
Isolation of MDSCs
The viruses [11, 12] and MDSCs [11, 13] used in these studies were described previously, and the cells were generated by a modified preplate technique [14, 15]. Skeletal muscle from hind limbs of C57BL/10J mice was minced and processed through a series of enzymatic dissociations: 0.2% collagenase type XI (Sigma-Aldrich, St. Louis, http://www.sigmaaldrich.com) at 37°C for 1 hour, 2.4 units/ml dispase (Gibco-BRL, Gaithersburg, MD, http://www.gibcobrl.com) for 45 minutes, and 0.1% trypsin-EDTA (Gibco-BRL) for 30 minutes. After the enzymatic dissociation, the muscle cells were centrifuged, resuspended in proliferation medium (Dulbecco's modified Eagle's medium [DMEM] supplemented with 10% fetal bovine serum [FBS], 10% horse serum [HS], 0.5% chicken embryo extract, and 1% penicillin-streptomycin [all from Gibco-BRL]), and serially replated. Using this modified preplate technique, different populations of muscle-derived cells have been isolated based on their adhesion characteristics [14, –16]. Small, round cells appear in the late preplate cultures (PP6). These cells remain quiescent, but following approximately 2 weeks in culture they begin to proliferate to become long-term proliferating (LTP) cultures. MDSCs are derived from these LTP cultures. To generate the MDSC populations used in these studies, the LTP cells were subplated at 100 cells per well in a 12-well tissue culture plate, and the resulting cells were screened for multipotentiality [13, –15].
The predilected MDSC populations used in these studies are highly purified populations of cells and were determined to be osteo- or myo-MDSCs by in vitro and in vivo analyses (described below). MDSC populations were considered to be osteo-MDSCs when their in vivo myogenic capacity was lower than the average MDSC and their alkaline phosphatase (ALP) expression was higher than other MDSC populations following bone morphogenetic protein 4 (BMP4) stimulation, whereas myo-MDSCs were populations that had a higher regenerative capacity in skeletal muscle and low responsiveness to BMP4 exposure, compared with other MDSC populations. Osteo-MDSCs and myo-MDSCs were transduced with a BMP4 expression virus that was described previously [11, 12]. BMP4 expression and activity was confirmed by a C2C12 myotube inhibition bioassay [11, 12] as well as a human BMP4 ELISA (R&D Systems Inc., Minneapolis, http://www.rndsystems.com). Noggin expression was confirmed using a modified C2C12 bioassay, as described previously [11, 12].
Generation of Small Interfering RNA Virus
A small interfering (si)RNA retrovirus to murine MyoD1 was generated using the RNAi-Ready pSIREN-RetroQ vector (Clontech, Mountain View, CA, http://www.clontech.com) and procedures described previously . The sequences inserted into the viral backbone were as follows: sense, GAT CCG CTG CGC GAC CAG GAC GCC TTC AAG AGA GGC GTC CTG GTC GCG CAG CAG TTT TTT CGA AAG; antisense, CTT AAT TTC GAA AAA ACT GCT GCG CGA CCA GGA CGC CTC TCT TGA AGG CGT CCT GGT CGC GCA GCG.
The myogenic capacity of the cells used in these studies was assayed using established techniques . To examine in vitro myogenic potential, cells were plated at high confluence (1,500 cells per cm2) in low-serum fusion medium (2% HS in DMEM) for 4 days and examined for myotube formation by immunocytochemistry staining for the differentiated myogenic marker myosin heavy chain (MyHC) (1:250, clone MY32; Sigma-Aldrich) using biotinylated IgG (1:250; Vector Laboratories, Burlington, ON, Canada, http://www.vectorlabs.com) and streptavidin-Cy3 (1:500; Sigma-Aldrich) . The in vivo myogenic potential of the cells was determined by the transplantation of 1 × 105 cells into the gastrocnemius muscles of 5- to 8-week-old mdx/SCID mice (Jackson Laboratory, Bar Harbor, ME, http://www.jax.org) as described previously . Two weeks after transplantation, the mice were sacrificed, and the muscles were sectioned by cryostat (10 μm). Immunohistochemistry was performed to identify dystrophin-positive myofibers. Tissue sections were fixed with cold methanol, and immunostaining was performed using the mouse-on-mouse kit (Vector Laboratories) with the dystrophin antibody DYS2 (1:50; Novocastra Ltd., Newcastle upon Tyne, U.K., http://www.novocastra.co.uk) using previously described protocols .
To determine their ability to undergo osteogenic differentiation, cells were plated at a density of 1,500 cells per cm2 and stimulated with BMP4 (200 ng/ml) for 4 days, with medium being changed every 48 hours. Following stimulation, the cells were either fixed and stained for ALP activity (Leukocyte kit; Sigma-Aldrich) using manufacturers' instructions  or lysed with 0.1% Triton-X in water, and assayed using SIGMA FAST p-nitrophenyl phosphate tablets (N-2770; Sigma-Aldrich) for ALP activity quantification. For ALP enzymatic activity, the relevant protocols recommended by the manufacturer were followed. Total protein concentration was determined using a bicinchoninic acid protein assay kit (Pierce, Rockford, IL, http://www.piercenet.com) with bovine serum albumin as the standard. The enzyme activity is expressed as nanomoles of p-nitrophenyl phosphate produced per minute per mg of protein.
Soft Agar Analysis
To monitor adhesion independent cell growth, cells were plated at various densities on a layer of 1% agar in proliferation medium. Cells were plated at 100, 500, and 1,000 cells per 3.8 cm2 in triplicate, and the number of colonies was counted at 2 and 3 weeks postseeding.
All of the animal experiments were approved by the Animal Research and Care Committee of the Children's Hospital of Pittsburgh and performed a minimum of two times to ensure reproducibility. A synopsis of key in vivo experiments can be found in Table 1.
Table Table 1.. A synopsis of the information pertaining to the key in vivo experiments/results with the osteo- and myo-MDSCs
Implantation into a Muscle Pocket.
To generate tumors or ectopic bone, cells were seeded in gelatin-based scaffolds (Gelfoam; Pfizer, New York, http://www.pfizer.com) (1 × 104 to 1 × 105 cells per 0.5 cm2 Gelfoam) and implanted intramuscularly into C57BL/6J (normal) or SCID (immunodeficient) mice (Jackson Laboratory). Tumor formation was monitored radiographically (Faxitron x-ray cabinet; Faxitron X-Ray, Wheeling, IL, http://www.faxitron.com) and histologically with von Kossa staining, followed by hematoxylin and eosin staining (described below). Dr. Aaron Pollett at Mt. Sinai Hospital, Toronto, performed pathological analysis of biopsies.
Long-bone and skull defects were created as described previously [12, 18, 19]. Cells were seeded at 1 × 105 to 5 × 105 cells per 0.5 cm2 Gelfoam and implanted either into C57BL/6J mice (Jackson Laboratory) (skull defects) or nude rats (Charles River, Wilmington, MA) (long bone defects). Bone healing was monitored radiographically (Faxitron x-ray cabinet).
Samples were surgically excised, fixed in 10% neutral buffered formalin (Richard-Allan Scientific, Kalamazoo, MI, http://www.rallansci.com) overnight, and then stored in 70% ethanol (EtOH) until being embedded in paraffin. The samples were embedded using a Hypercenter XP enclosed tissue processor (Thermo Electron Corporation, Waltham, MA, http://www.thermo.com). Following the second paraffin treatment, the samples were blocked in paraffin with a Tissue-Tek instrument (Sakura, Torrance, CA). Samples were cut at a thickness of 5 μm using a Reichert-Jung Biocut 2030 (Leica Microsystems Inc., Bannockburn, IL, http://www.leica.com) and placed on microscope slides (SP Laboratory Products, War minster, PA, http://www.spindustries.com), baked overnight at 43.4°C (in an economy digital incubator by Boekel Industries Inc, Feasterville, PA, http://www.boekelsci.com), and stored at room temperature until staining.
Slides underwent hematoxylin and eosin staining for pathological analysis. The slides were deparaffinized. Following deparaffinization, the tissue was incubated in Harris hematoxylin (Surgipath Medical Industries, Richmond, IL, http://www.surgipath.com) for 2 minutes, washed in tap water for 5 minutes, dipped five times in 70% EtOH, stained for 30 seconds in eosin (Sigma-Aldrich), dehydrated in 95% EtOH for two dips (two 1-minute dips in 100% EtOH), and then finally two 2-minute dips in xylene. The slides were coverslipped using cytoseal-XYL, xylene-based (Richard-Allan Scientific). Von Kossa staining was performed as described previously . Deparaffinized slides were rinsed in distilled water three times. The slides were stained in 2% silver nitrate solution in the dark for 15 minutes, rinsed three times in distilled water, and exposed to light for 20 minutes until appropriate stain development and then counterstained with hematoxylin and eosin.
Green fluorescent protein (GFP) was detected by immunohistochemistry. Frozen sections were fixed in 10% formalin for 10 minutes and were rinsed three times in phosphate-buffered saline. The sections were processed for 3,3′-diaminobenzidine tetrahydrochloride (DAB) immunostaining as suggested in the manufacturer's protocol (Vectastain Elite ABC Kit, DAB Substrate Kit for Peroxidase; Vector Laboratories). The monoclonal GFP antibody (1:100 dilution, AF757; R&D Systems) was used for the immunostaining. These sections were counterstained with hematoxylin.
Fluorescent In Situ Hybridization
Slides were deparaffinized using serial washes of xylene (three 10-minute washes), rehydrated with a series of EtOH washes (100%, 95%, 80%, and 70%) for 3 minutes at each successive concentration, air-dried, and then denatured in 70% formamide, 2× saline-sodium citrate (SSC; 0.3 M NaCl, 0.03 M sodium citrate, pH 7.0) at 75°C for 5 minutes. Slides were immediately quenched with ice-cold 70% EtOH followed by a series of EtOH washes (80%, 95%, and 100%) for 3 minutes at each successive concentration. Y chromosome paint probe directly conjugated to digoxigenin was made by degenerative oligonucleotide primer-polymerase chain reaction (DNA courtesy of Dr. R. Stanyon, National Cancer Institute, Frederick, MD) mixed with hybridization buffer and mouse Cot-1 DNA (Invitrogen, Carlsbad, CA, http://www.invitrogen.com). The probe mix was denatured at 70°C for 10 minutes and allowed to reanneal for 60 minutes at 37°C. The probe was placed on denatured sections, covered with plastic coverslips, sealed with rubber cement, and hybridized overnight at 37°C. The probe was detected on the following day by rinsing with 2× SSC solution, pH 7.0, at 45°C for 5 minutes; 50% formamide, 2× SSC for 12 minutes; and 4× SSC-Tween 20 for 2 minutes. The hybridized probe was detected with anti-dig-Rhodamine Fab fragments (Roche Diagnostics, Indianapolis, IN, http://www.roche-applied-science.com) diluted 1:200 in 4× SSC-Tween 20–10% FBS for 45 minutes at 37°C. Subsequently, excess antibody was washed four times with 4× SSC-Tween-20 at 45°C for 5 minutes each. Nuclei were counterstained with 4′,6′-diamidino-2-phenylindole in Vectashield mounting medium (Vector Laboratories).
Means and SDs for ALP activity were calculated and compared using a one-way analysis of variance. Means and SDs for colony formation were calculated and compared using either a 1 × 6 (cell type × treatment) one-way analysis of variance, or using a 2 × 6 (cell type × treatment) two-way analysis of variance. Significant intergroup differences were determined using the least significant differences multiple comparisons test. Differences were considered statistically significant when a p value of less than 0.05 was achieved.
Coexpression of BMP4 and Noggin at Specific Ratios in MDSCs Induces Tumor Formation In Vivo
Members of our laboratory have previously reported that retroviral delivery of the BMP antagonist Noggin inhibits heterotopic ossification induced by BMP4 [11, 12]. Furthermore, ratios of BMP4- to Noggin-expressing postnatal MDSCs higher than 1:1 (protein ratios higher than 1:2) resulted in a significant decrease in radiographically visible bone formation at 4 weeks [11, 12]. However, when the experiment was continued for an additional 2 weeks (8 weeks after transplantation), reproducible tumor formation occurred in mice treated with BMP4- and Noggin-expressing cells at ratios of 1:2 and 1:3 (protein ratios of 1:4.8 and 1:7) (Fig. 1A) but not in mice treated with cells at a ratio of 1:1. This phenomenon occurred in a reproducible and dose-dependent manner; tumors formed in mice that received cells expressing BMP4 and Noggin at protein ratios ranging from 1:3.7 to 1:7.5, but not at protein ratios above or below this range (supplemental online Table 1). Histologic analysis of the tumors revealed a neoplastic growth comprising two distinct and unrelated sarcomas (Fig. 1B), with areas of osteosarcoma (Fig. 1C) and rhabdomyosarcoma (Fig. 1D) mixed with undifferentiated neoplastic cells (Fig. 1E). Because female cells were implanted into male mice, fluorescence in situ hybridization with Y chromosome probes was used to confirm that donor cells formed the majority of the tumors (Fig. 1F). This type of tumor is referred to as an osteorhabdomyosarcoma, a type of malignant mesenchymoma. We posit that the oncogenic transformation described above occurred because of aberrant signaling, resulting from the simultaneous exposure of MDSCs to BMP4 and Noggin, and that the simultaneous presence of high levels of Noggin and BMP4 tempered the BMP4 signal and caused the intramuscularly transplanted cells to receive both an osteogenic signal (from BMP4) and a myogenic signal (from the skeletal muscle environment). Because of this cross-signaling, most of the cells likely remained undifferentiated and became trapped in a state of transient amplification. We believe that such stimulation induced the cells to grow and divide, but never to fully differentiate or terminally commit to a single lineage.
Postnatal MDSCs are not intrinsically oncogenic and have significant therapeutic value [21, , , –25]. They have been used to regenerate dystrophin-positive myofibers in dystrophin-deficient (mdx) mice, for gene delivery to promote bone healing, and for cardiac tissue repair and the treatment of urinary dysfunction [22, 26, 27]. Furthermore, MDSCs implanted in lethally irradiated mice have no deleterious effects and have been shown to reconstitute bone marrow for 5 months with no signs of oncogenesis . Implantation of nontransduced MDSCs and MDSCs transduced with Noggin had no oncogenic effects in vivo (>12 months post-transplantation), and MDSCs transduced to express BMP4 remained disease-free for more than 6 months postimplantation (data not shown). We observed oncogenesis only after implantation of a mixture of BMP4- and Noggin-transduced cells at specific ratios.
Mixtures of BMP4- and Noggin-Expressing Cells Form Soft Agar Colonies
To study this perceived transformation event in vitro, MDSCs transduced with BMP4 were cocultured with Noggin-expressing MDSCs. Following 7 weeks of culture (the time required for observation of tumorigenesis in vivo), soft agar analysis was performed to examine anchorage-independent growth, a common hallmark of transformed cells. Significant differences in colony formation were observed (F = 5.08, p < .05), with cells cultured at a ratio of 1:3 (BMP4-:Noggin-expressing cells) forming more colonies than cells cultured at a ratio of 1:1 or unmixed parental cells (Fig. 2A, p < .01). Furthermore, these cocultured cells (1:3) had obvious cytogenetic structural and numerical abnormalities. They exhibited translocations and chromosomal modality between 37 and 39 (normal modality is 40) (Fig. 2B). We repeated this experiment using conditioned media from mixed transduced cell populations to treat two other MDSC isolations. MDSC populations treated for only 1 week with conditioned media from 1:3 mixed cell populations formed more colonies than did those treated with media from the other test populations (Fig. 2C). However, subcutaneous or i.v. implantation of cells that were treated with conditioned media for 1 week produced no tumors during the 1-year duration of the study. This result suggests the following: (a) MDSCs themselves are not oncogenic, and (b) the induction of transformation requires continuous exposure to differentiation signals.
Characterization of Muscle-Derived Progenitor Cells with a Predilection for One Lineage over Another
As the above-described transformation appeared to occur through the activity of multiple extrinsic signals, we next explored whether a similar transformation could occur through conflicting intrinsic and extrinsic signals. We have previously reported the isolation and characterization of MDSCs from murine skeletal muscle according to their adhesion characteristics and proliferation behaviors using a modified preplate technique [14, 15]. These cells have been shown to undergo self-renewal, display long-term proliferation, and differentiate into multiple lineages, and they have been used in preclinical studies for regenerative medicine [22, 26, 27]. In the process of characterizing MDSCs, it was observed that these cells were heterogeneous in nature and displayed various degrees of differentiation to specific lineages [13, 28]. While we were trying to further understand the heterogeneous nature of MDSCs, we noted that two populations of cells, which were capable of undergoing multilineage differentiation in vitro, had various affinities toward two mesodermal lineages: one population was more predilected to the myogenic lineage (myo-MDSCs), whereas the other had a predilection toward osteogenesis (osteo-MDSCs). Both cell types can undergo in vitro myogenic and osteogenic differentiation, as determined through MyHC (a differentiated myogenic marker) staining (Fig. 3A, 3B) and ALP (an early marker for osteogenesis) activity (Fig. 3C, 3D), respectively. Although myo- and osteo-MDSCs can undergo both types of differentiation, the regeneration index of the myo-MDSCs in skeletal muscle is approximately seven times higher than that of osteo-MDSCs (Fig. 3E–3G), and significant differences were found in the ALP activity of MDSCs (F = 768.674, p < .01), with osteo-MDSCs stimulated with BMP4 having more than 53 times the ALP expression than that of similarly stimulated myo-MDSCs (Fig. 3H, p < .01).
Osteogenic Priming Coupled with a Myogenic Niche Induces Transformation of Multilineage Skeletal Muscle-Derived Progenitor Cells
It has been shown that germ-line determination occurs early in differentiation, and it has been hypothesized that the denial of a cell's fate by counteracting its intrinsic signaling pathways through extrinsic stimulation may be linked to oncogenesis [29, 30]. We wanted to determine whether altering a cell's fate through environmental/external stimulation could cause concomitant signal-induced oncogenesis of cells that had a predilection for one lineage over another. To explore this, we implanted our predilected osteo-MDSC line into a myogenic niche. After implantation into a muscle pocket, these osteo-MDSCs reproducibly formed a tumor in 25% of mice (Table 1). The mice that did not form tumors displayed neither bone formation nor abnormal histology. Conversely, intramuscularly implanted myo-MDSCs remained untransformed. The tumor formed by the osteo-MDSCs contained rhabdoid and undifferentiated cells with sporadic bone nodules throughout (Fig. 4A, 4B). Yet, when osteo- and myo-MDSCs were implanted into a skull defect, no significant bone formation was detected over controls, and neither osteo- or myo-MDSCs induced tumor formation (data not shown). Since tumor formation was observed only when osteo-MDSCs were implanted into a myogenic environment, we posit that the local environment altered the cells' fate and promoted tumorigenesis. However, the myo-MDSCs did not undergo transformation when implanted into an osteogenic environment, and neither tumor formation or bone growth was observed. From this, we posit that the osteogenic signals were not sufficient to induce transformation. Hence, we consequently wanted to determine whether increased BMP stimulation could induce a myogenic cell to undergo active denial-induced oncogenesis. To explore this, myo-MDSCs and osteo-MDSCs were treated with BMP4 for 4 days before implantation into a myogenic environment. The BMP4-stimulated osteo-MDSCs did not undergo transformation when implanted intramuscularly. Thus, it appears that pretreatment of the stem cells with BMP4 resulted in unalterable commitment of the osteo-MDSCs to the osteogenic lineage, which inhibited their ability to respond to environmental cues. The BMP4-stimulated myo-MDSCs, which were primed to undergo osteogenesis but implanted intramuscularly, formed tumors in 25% of host mice, and these tumors resembled the ones generated by untreated osteo-MDSCs implanted in muscle (Fig. 4C), containing rhabdoid and undifferentiated cells but no bone nodules (Fig. 4D). Similar to the previous experiment, the remaining 75% of the mice displayed no abnormal histology or physiology.
Transduction of Predilected MDSCs with BMP4 Leads to Site-Specific Tumorigenesis
To further investigate the potential relationship between altering cell fate and oncogenesis, we transduced myo- and osteo-MDSCs to constitutively express BMP4 (myo-MDSC-B4s and osteo-MDSC-B4s, respectively). We predicted that only myo-MDSC-B4s would become oncogenic after receiving a myogenic signal. Indeed, although BMP4-transduced myo-MDSCs implanted in a long-bone (Fig. 5A, 5B) or cranial (Fig. 5C, 5D) defect produced bone, BMP4-transduced myo-MDSCs implanted in skeletal muscle formed malignant mesenchymomas in 100% of the mice between weeks 4 and 8 (Fig. 5E), whereas osteo-MDSC-B4s generated bone when implanted in skeletal muscle (Fig. 5F). The tumors generated were osteorhabdomyosarcomas (Fig. 5G) (similar to those generated with BMP4- and Noggin-expressing cells) and were mostly GFP-positive, indicating that the neoplastic mass originated almost entirely from our implanted cells (Fig. 5H, 5I). All myo-MDSC-B4-implanted mice showed significant weight loss and large tumor growths, as well as rapid spontaneous death occurring in several mice. Hence, the apparent environmentally induced oncogenesis of the BMP4-expressing myo-MDSCs suggests that simultaneous exposure to different differentiation signals likely caused these cells to undergo tumorigenesis.
BMP4 Expression Is Necessary for Site-Specific Oncogenesis
Cells were isolated from three myo-MDSC-B4-generated tumors. All cells retained GFP expression, whereas only two populations had detectable BMP4 expression. To assess whether the cells were terminally transformed, we reimplanted all three populations into mouse muscle pockets. The two BMP4-expressing cell populations generated tumors (Fig. 5J), whereas the third (BMP4-deficient) cell line did not produce bone or tumor (Fig. 5K). This result suggests that for tumor formation to occur, these MDSCs required three key factors: (a) the ability to respond to both myogenic and osteogenic signals, (b) a source of myogenic signals (i.e., the environment), and (c) an osteogenic stimulus (i.e., BMP4 expression). Thus, if the myogenic potential of myo-MDSC-B4 cells is inhibited, they should no longer be able to undergo concomitant signal-induced transformation.
Inhibition of MyoD1 Decreases the Myogenic Potential and Increases the Osteogenic Potential of Myo-MDSCs In Vitro
To investigate whether inhibition of the myogenic pathway could indeed abrogate transformation, myo-MDSCs were transduced with a retrovirus expressing siRNA for the master myogenic regulator MyoD1  (siMyoD1). These cells had impaired myotube formation when cultured in differentiation medium (Fig. 6A, 6B), whereas control cells underwent myotube formation (Fig. 6C, 6D). The inhibition of MyoD1 also increased the in vitro osteogenic potential of myo-MDSCs, as these cells displayed an increase in the percentage of ALP staining compared with controls following BMP4 stimulation (from 8% to 40% 4 days after BMP4 stimulation) (Fig. 6E, 6F). Hence, downmodulation of MyoD1 levels decreased the myogenic potential and increased the osteogenic potential of myo-MDSCs.
Inhibition of MyoD1 Decreases the Oncogenic Potential of Myo-MDSC-B4 Cells In Vivo
Myo-MDSC-B4 cells incubated in fusion medium for 4 days maintained their ability to form myotube-like structures, even while expressing high levels of BMP4 (Fig. 6G). However, myo-MDSC-B4 cells transduced with siMyoD1 failed to undergo myogenesis (Fig. 6H). When myo-MDSC-B4 cells transduced with the siMyoD1 virus were implanted into a mouse muscle pocket, no tumor or bone growths were observed (Fig. 6I), whereas control mice implanted with myo-MDSC-B4s formed tumors (Fig. 6J). This result further supports our hypothesis that transformation of MDSCs is dependent on the presence of, and the responsiveness of these cells to, conflicting differentiation signals. The observation that no bone formation occurred in the myo-MDSC-B4 cells transduced with siMyoD1 is not unexpected. Although we previously found that inhibition of MyoD1 increased the osteogenic potential of our cells, it has also been shown that MyoD is required for the osteogenic effect of BMPs . This osteogenic inhibition appears to occur in vitro at the ALP stage in embryonic cell lines and satellite cells when they are stimulated with BMP7  and in vivo when the cells are myo-MDSCs transduced to produce BMP4. We believe that the differences in cell types and BMP molecules used explain the discrepancies in effects. Regardless, in both systems, MyoD inhibition decreases the osteogenic potential of the cells.
Herein, we have shown that specific postnatal stem cells isolated from the skeletal muscle of mice, although capable of undergoing multilineage differentiation, are also able to undergo malignant transformation when exposed to conflicting differentiation signals. Furthermore, we have found that transformation appears to be dependent on altering the balance of intrinsic and extrinsic signaling pathways and can be abrogated when the ability of a cell to undergo differentiation is removed.
It was once stated that “cancer is a problem of developmental biology” . Here, we have adopted a developmental biology-based approach to the investigation of concomitant signal-induced transformation. We have mentioned earlier that the ectopic expression of hTERT in combination with T/t-Ag and H-ras can transform normal human cells [6, –8]. If a defined series of genetic events is necessary for the transformation of somatic cells, how is our observed transformation occurring? In all of the aforementioned studies, telomerase and other genes are used to escape the cell cycle (i.e., T/t-Ag or inhibitors of Rb and p53). The cells used in the studies detailed here are activated progenitor cells. Hence (similar to the aforementioned experimentally transformed cells), they would contain telomerase activity and would not be trapped in the cell cycle (we have shown that our cells can be expanded for more than 300 population doublings ). To drive a somatic cell to a tumorigenic fate, Rb and p53, in conjunction with the oncoproteins Ras and Myc, and the telomerase subunit hTERT, were necessary [7, 8]. We believe that the coupling of BMP4 expression with endogenous signals could mimic the effects of mutant Ras and/or Myc.
Previously, we have used MDSCs (and BMP4) to generate bone, to regenerate dystrophin-positive myofibers in a muscular dystrophy animal model, to repair cardiac tissue, to treat urinary dysfunction, and to regenerate the hematopoietic system following lethal irradiation of mice, all with no deleterious effects [15, 18, 19, 21, 26, 27]. This previous work was primarily performed with either an MDSC clone or MDSC populations that had no lineage predilections. Of all of the MDSC populations screened to date, very few predilected populations have been identified. For future work, we plan to further characterize the differences between our predilected populations and isolate each of them from primary human sources.
An important question that arises from this research is whether the cells used in this study are already pretransformed. We have performed numerical chromosomal analysis on all of the cell lines used in our studies and observed no abnormalities. Furthermore, we have never observed more than one colony per 100 cells plated in soft-agar colony formation assays. These findings, along with the results showing that our cells form tumors only in a niche-specific fashion and the reimplantation studies demonstrating the need for continued BMP4 expression, strongly suggest that the cells used in our studies are not pretransformed.
The finding that a multilineage stem cell can undergo environment-specific transformation is not new. It has been reported many times that teratomas can be experimentally produced in mice from embryonic stem (ES) cells transplanted into extrauterine sites. Furthermore, cells from these teratomas can be isolated and grown in culture [34, 35]. These reisolated cells lose their malignancy and become benign, and they participate in the normal development when injected into a blastocyst [34, 35]. Hence, ES cells, when implanted extrauterinely, receive multiple and conflicting environmental signals. These signals cause a change in the balance between the intracellular and extracellular signaling pathways. As a result, the cells become trapped in a state of transient amplification, where they grow and divide, with the majority of cells never fully differentiating or committing to a single lineage, thus leading to tumor formation. Fortunately, as ES cells have a remarkable ability to maintain their genome, they do not amass mutations throughout this process and become benign when reimplanted into a blastocyst . The cells used in this study are similar to ES cells in that they are capable of undergoing multilineage differentiation in vitro and, when implanted in an environment where they appear to receive conflicting differentiation signals, undergo transformation. Unfortunately, MDSCs, unlike ES cells, do not possess the ability to perpetually self-renew, although they are able to undergo long-term expansion. This lack of ability to undergo perpetual self-renewal potentially results in the amassment of mutations during tumorigenesis and hence terminal transformation.
On the basis of our observations, it appears that stem/progenitor cells can become transformed when their intrinsic and extrinsic signaling pathways become conflicted because of multiple differentiation signals. Hence, when a stem cell is either stimulated with growth/differentiation factors in vitro or is placed in an environment that exposes it to signals conflicting to its intrinsic pathways, the cell may undergo transformation. We believe that this means of transformation may explain the teratoma formation observed by ES cells and testicular stem cells when they are implanted outside of their native environment [37, , –40]. Furthermore, it has been reported recently that coculturing human ES cells with immortalized human astrocytes from fetal midbrain tissue cells (making them more similar to adult progenitor cells prior to implantation) resulted in substantial and long-lasting restitution of motor function when the ES cells were transplanted into the neostriata of 6-hydroxydopamine-lesioned parkinsonian rats . Unfortunately, the researchers noted a component of the regenerated tissue that was comprised of slowly dividing cells that were potentially tumorigenic. These cells do not appear to be a result of contamination from either the undifferentiated ES cells or immortalized astrocytes, but rather a population of neuroepithelial cells that had failed to differentiate . Although the work presented here was performed with postnatal MDSCs, this work should eventually be repeated with clonal populations. Our findings, taken together, should be considered when using other progenitor cells for cell-based therapies. As well, the balance between intrinsic and extrinsic signaling cascades should be taken into account and potentially screened in vitro prior to use in vivo. By understanding the form of oncogenesis presented in this article, we may be able to generate better tools for regenerative medicine, as well as investigate the initiating events behind several forms of human cancers.
Herein, we have described a series of experiments that has led us to a hypothesis that potentially explains how somatic stem cells can become transformed when they receive conflicting differentiation signals. We propose that the model presented here may be used to examine the initiating stages of several distinctly different forms of human cancers: (a) rhabdomyosarcoma, (b) osteosarcoma, (c) adamantinoma, and (d) (although not normally diagnosed) malignant mesenchymoma. We have observed the transformation of specific postnatal MDSCs through concomitant conflicting signals. Using this model, we have repeatedly generated tumors that are relatively ambiguous in their morphology and have been classified by pathologists as containing components of all four of the aforementioned types of cancers. Although this work was performed using murine cells, we hypothesize that this model represents a means to explore the human analogs of the above-mentioned tumors. This will provide insights into the initiating events underlying these tumors and in the future will help provide a means of better understanding and potentially treating these diseases.
In 1921, Rotter hypothesized that adult cancers arise from the inappropriate reactivation of embryonal cell-like progenitor cells within adult tissues . Although many researchers originally dismissed this hypothesis, they are now re-evaluating its merit. The studies detailed here suggest that postnatal multipotent muscle-derived cells simultaneously exposed to different differentiation signals can undergo transformation. These results indicate that cancer might arise when activated somatic stem cells receive contradictory intrinsic and extrinsic signals. Such cells would become locked in a continuous loop of transient amplification: they would grow and divide but not terminally differentiate. This process results in altered gene expression and genetic instability leading to transformation. As the process described here may underlie the development of several human cancers, our research provides a model system to examine the initiating events leading to oncogenesis. In addition, this work may have a significant impact on research in the fields of stem cell therapy, cancer cell biology, tissue engineering, and cancer treatment.
Disclosure of Potential Conflicts of Interest
The authors indicate no potential conflicts of interest.
We thank Dr. Aaron Pollett for performing the histologic analysis of the tumor samples, Dr. David Parham and Dr. Corrine Linardic for expert advice, and Dr. Arvydas Usas for technical assistance. Furthermore, we are grateful to Dr. Ronald Herberman and Dr. Tao Cheng for their advice and input on our experimental design and manuscript; Dr. Mary Hitt, Dr. Jin Jen, and Uchenna Nwosu for external reviews; Dr. Hairong Peng, Dr. David Hannallah, and Anne Olshanski for providing samples and reagents; Dr. Baohong Cao, Dr. Guangheng Li, and Jim Cummins for reviewing the paper and offering advice and encouragement; and Ryan Sauder and David Humiston for excellent editorial assistance. This work was supported by NIH Grant R01-DE13420-01 (J.H.) and by the William F. and Jean W. Donaldson Chair at the Children's Hospital of Pittsburgh and the Henry J. Mankin Chair for Orthopaedic Research at the University of Pittsburgh.