Rhabdomyosarcoma (RMS) is the most common soft tissue sarcoma of childhood and adolescence. Despite advances in therapy, patients with a histologic variant of RMS known as alveolar (aRMS) have a 5-year survival rate of <30%. aRMS tissues exhibit a number of genetic changes, including loss-of-function of the p53 and Rb tumor suppressor pathways, amplification of MYCN, stabilization of telomeres, and most characteristically, reciprocal translocation of loci involving the PAX and FKHR genes, generating the PAX7-FKHR or PAX3-FKHR fusion proteins. We previously showed that PAX3-FKHR expression in primary human myoblasts, cells that can give rise to RMS, cooperated with loss of p16INK4A to promote extended proliferation. To better understand the genetic events required for aRMS formation, we then stepwise converted these cells to their transformed counterpart. PAX3-FKHR, the catalytic unit of telomerase hTERT, and MycN, in cooperation with down-regulation of p16INK4A/p14ARF expression, were necessary and sufficient to convert normal human myoblasts into tumorigenic cells that gave rise to aRMS tumors. However, the order of expression of these transgenes was critical, as only those cells expressing PAX3-FKHR early could form tumors. We therefore suggest that the translocation of PAX3 to FKHR drives proliferation of myoblasts, and a selection for loss of p16INK4A/p14ARF. These early steps, coupled with MycN amplification and telomere stabilization, then drive the cells to a fully tumorigenic state. [Cancer Res 2008;68(23):9583–8]
Rhabdomyosarcoma (RMS) is the most common soft tissue sarcoma of childhood and adolescence. Most cases can be classified as embryonal (eRMS) or alveolar (aRMS), depending upon appearance under light microscopy. Although ongoing clinical trials have led to improved survival for patients with eRMS, children with aRMS still face a 5-year survival of <30% (1). Even more discouraging is the outcome for children with aRMS whose tumors harbor the PAX3-FKHR fusion gene; when metastatic, their 5-year survival is <10% (2). Although it is accepted that PAX3-FKHR is a characteristic and most likely founding mutation of aRMS (3–5), functioning in part by illegitimately activating myogenic transcription programs (6), it is not clear how this protein cooperates with other changes to promote aRMS.
To elucidate the molecular changes giving rise to RMS, we first showed that human myoblasts could be driven to a tumorigenic state by expression of the SV40 DNA tumor virus large T and small t antigens, which disable the p53 and Rb pathways and activate Myc pathways, respectively; oncogenic H-Ras, which provides a proliferative signal; and hTERT, which reactivates telomerase and immortalizes cells (7–9). This work identified myoblasts as a putative cell of origin for RMS, and validated the roles of pathways commonly dysregulated in human cancer in the development of RMS (10). Based on these studies, we recently showed that such myoblasts can be induced to proliferate inappropriately by expression of PAX3-FKHR, and that this was accompanied by epigenetic silencing of p16INK4A via methylation of its promoter (11). Similarly, loss of p16INK4A and tissue-specific expression of Pax3:Fkhr in murine mouse models led to aRMS-like tumors (12). Taken together, we speculate that PAX3-FKHR and the accompanying loss of p16INK4A, which disables the Rb pathway, may be initiating events in aRMS. Because p16INK4A silencing is often accompanied by loss of p14ARF (13), thereby also disabling the p53 pathway, and both MYCN amplification (14) and telomere stabilization (summarized in ref. 9) are observed in clinical samples of RMS, we tested whether these additional changes cooperated with PAX3-FKHR loss to drive myoblasts to become aRMS tumors.
Materials and Methods
Generation of cell lines. Early passage normal human skeletal muscle myoblasts (HSMMs; Lonza), grown in defined medium (Clonetics SkGM-2 Bullet kit) were stably infected with amphotrophic retroviruses derived from pK1-PAX3-FKHR-puro, pBABE-hTERT-hygro (9), pWZL-FLAG-murine-MycN-blasticidin, or vector. Cells were selected in 0.25 μg/mL puromycin (Sigma) or 50 μg/mL hygromycin B for 7 d, or 250 μg/mL G418 (Life Technologies Invitrogen) for 10 d. HSMMs were characterized by the vendor as >90% desmin positive. Unless indicated, for presenescent expression, retroviral infections were initiated at population doubling (pd) 2 to 4. Senescence bypass in HSMMs generally starts at ∼pd 15, although can be influenced by culture conditions. HSMMPF+H+M, M+H+PF, and PF+M+H cells contained all transgenes by pd 21, 31, and 13, respectively. Human RMS cell lines were grown in RPMI 1640 (Life Technologies) with 10% fetal bovine serum. HSMMs previously engineered to generate eRMS morphology, via expression of SV40 T/t-oncoproteins, hTERT, and oncogenic H-Ras (9), provided control cell lysates for transgene expression.
Immunoblotting. Cells were lysed in Tris/radioimmunoprecipitation assay buffer with standard protease inhibitors and passaged through a 21g needle to shear DNA. Protein concentration was measured by the DC assay (Bio-Rad). Sixty to 100 μg of lysate were resolved by SDS-PAGE, transferred to polyvinylidene difluoride membrane, and immunoblotted with primary monoclonal antibodies anti-FOXO1A (FKHR) F6928, FLAG F3165, tubulin T4026 (Sigma), p16INK4A 554079 (BDPharmingen), actin SC-8462 (Santa Cruz), or primary polyclonal antibody anti-PAX3 1607802 (Geneka). Membranes were reacted with a secondary horseradish peroxidase–labeled goat anti-mouse or anti-rabbit antibody (Invitrogen-Zymed), and developed using chemiluminescence (Amersham).
Reverse transcription-PCR. Total cellular RNA was isolated using the RNA-Bee kit (TEL-TEST). After spectrophotometric quantitation, 2 μg were subject to reverse transcription using the Omniscript RT kit (QIAGEN) with Oligo-dT primers (Life Technologies Invitrogen). Standard PCR using primer sets for MYOD1, MYOGENIN (15), PAX3-FKHR, FLAG-hTERT (16), murine-FLAG-MycN, p14ARF, and p16INK4A (11) was performed, with product separated on 2% agarose (Supplementary Table S1). Glyceraldehyde-3-phosphate dehydrogenase and water controls were included to verify equal RNA and specificity of cDNA input, respectively.
Tumor xenografts assays to measure in vivo tumorigenic ability. Although cell transformation may be measured in vitro as colony growth in soft agar, because neither the human aRMS JR nor HSMMPF+M+H cell lines formed colonies, we proceeded directly to tumorigenesis assays in vivo. Under Institutional Animal Care and Use Committee–approved protocols, and as performed (9), cell lines were proven to be free of replicating retrovirus or Mycoplasma, expanded in culture, then 10 million cells per cell line were injected s.c. into the flanks of severe combined immunodeficient (SCID)/beige mice in triplicate or quadruplicate. Mice were monitored biweekly. At maximal tumor burden, or if demonstrating a decline in health, mice were sacrificed and underwent necropsy, with portions of tumor fixed in formalin or snap-frozen in liquid N2 for later analysis.
Immunohistochemistry. Paraffin-embedded samples were sectioned and stained with H&E to assess tumor morphology or for immunohistochemical analysis using monoclonal antibodies from DakoCytomation (9). First-tier immunohistochemical analysis using anti-desmin, myoglobin, and skeletal muscle-specific actin determined resemblance to RMS. Second-tier analysis using MyoD1 and myogenin determined RMS subtype. Pathologists with experience in the evaluation of pediatric sarcomas (S.Q., R.C.B.) evaluated slides.
Results and Discussion
Genetic creation of alveolar RMS driven by PAX3-FKHR. We previously showed that ectopic expression of PAX3-FKHR, in cooperation with silencing of the p16INK4A locus, drives cultured HSMMs to bypass the growth-arrest state of senescence. Moreover, PAX3-FKHR–positive aRMS tumor cell lines, and some aRMS tumor specimens, show profound down-regulation of p16INK4A (11). Taken together, these data suggest that gain-of-function of PAX3-FKHR combined with loss-of-function of p16INK4A, or more broadly, the Rb pathway, are genetic changes common to aRMS. Because overcoming senescence is presumably an early event in tumorigenesis, we hypothesized that the PAX3-FKHR translocation, coupled with loss of p16INK4A, is an initiating event driving the expansion of a susceptible population of skeletal muscle cell progenitors. However, aRMS tumors are also characterized by other genetic changes, including inappropriate up-regulation of the MYCN oncogene (14) and telomere stabilization (summarized in ref. 9). Prior modeling data suggests that these latter genetic changes are critical to human tumorigenesis. Thus, we tested whether addition of MycN, and the catalytic subunit of telomerase known as hTERT, cooperated with the expression of PAX3-FKHR and the silencing of the p16INK4A locus in HSMMs to promote tumor growth.
The HSMM cells used in this and prior studies were validated to be of skeletal muscle lineage by their expression of desmin and skeletal muscle–specific actin, and by their ability to differentiate to myotubes (9, 11). However, because we anticipated examining the expression of MyoD1 and myogenin in tumors derived from these cells, we first assessed the levels of expression of these muscle-specific transcription factors. As shown in Fig. 1A, native HSMMs appropriately express low levels of MyoD1 and myogenin, associated with myogenic determination and differentiation, respectively (17).
Next, to further characterize postsenescent HSMMs expressing PAX3-FKHR (HSMMPF), we verified that p16INK4A was down-regulated as seen previously (11), and also found that the p14ARF tumor suppressor was down-regulated in these cells and its derivatives (Fig. 1B). This was somewhat expected, as p16INK4A and p14ARF are transcribed from the same INK/ARF locus. In this regard, the HSMMPF cells were expressing oncogenic PAX3-FKHR in a cellular background in which both Rb and p53 tumor suppressor pathways were inactivated by loss of p16INK4A and p14ARF, respectively. hTERT and MycN were then stably introduced into HSMMPF cells using retroviral transduction and appropriate transgene expression confirmed (Fig. 1B). To evaluate tumorigenicity, these genetically modified cells were injected as s.c. xenografts in SCID/beige mice and monitored for tumor appearance and volume (Fig. 1C). The human aRMS cell line JR (naturally expressing PAX3-FKHR) and the HSMMPF cell line expressing only hTERT (HSMMPF+H) were included as positive and negative controls, respectively. As anticipated, JR cells rapidly formed tumors, whereas HSMMPF+H cells failed to generate tumors during the 90 days that mice were monitored (Fig. 1C). Importantly, HSMMPF cells additionally expressing hTERT and MycN, and lacking p14ARF/p16INK4A expression (HSMMPF+H+M), also formed tumors (Fig. 1C), albeit at a longer latency than the positive controls. This latency is consistent with the variation observed for other human RMS cell lines, such as the eRMS cell line RD, which requires over 3 weeks to induce s.c. tumor growth (Fig. 1C).
To assess the morphologic characteristics of the resulting xenograft tumors, they were subject to standard H&E staining (Fig. 1D). Whether derived from the human RMS cell lines or from HSMMPF+H+M cells, all tumors consisted of monomorphic small round blue cells, a morphology characteristic of pediatric sarcomas. The tumors deriving from HSMMPF+H+M cells also contained interspersed myoblastic cells, reflecting their recent derivation from primary skeletal muscle (Fig. 1D). Tumors were next subject to a tripartite of immunohistochemical stains that are the clinical standard for identifying RMS: desmin, skeletal muscle–specific actin, and myoglobin (18). As expected, control RD eRMS and JR aRMS tumors stained positive for desmin and skeletal muscle-specific actin (Fig. 2A,, a–b and d–e). Similarly, the HSMMPF+H+M-derived tumors stained positively for these markers (Fig. 2A,, c and f). Myoglobin staining was strongest in the HSMMPF+H+M xenograft (Fig. 2A , i). Thus, ectopic expression of hTERT and MycN in postsenescent HSMMPF myoblasts promotes tumor growth resembling RMS as judged by clinical immunohistochemical standards.
To further classify the HSMMPF+H+M tumors as either eRMS or aRMS, they were subject to a second tier of immunohistochemical stains (MyoD1 and myogenin) that distinguishes these variants (18). Using nuclear expression of these transcription factors, the control eRMS RD xenografts were found to appropriately exhibit diffuse staining for both markers, whereas the control aRMS JR xenografts appropriately exhibited diffuse staining for MyoD1 but patchy staining for myogenin (Fig. 2B,, a–b and d–e). Tumors derived from HSMMPF+H+M cells exhibited diffuse staining for MyoD1, but patchy staining for myogenin, consistent with an alveolar pattern (Fig. 2B , c and f). Thus, HSMMs engineered to stably express PAX3-FKHR (with concomitant silencing of the INK/ARF locus), hTERT, and MycN can recreate alveolar RMS.
Effect of altering order of acquisition of genetic changes. Because PAX3-FKHR is found exclusively in aRMS, it is thought to be a genetic lesion acquired early in the stepwise process of tumorigenesis. Although this is not proven, other chromosomal translocation–driven malignancies such as chronic myelogenous leukemia are similarly thought to result from early translocation events, based on the paradigm of BCR-ABL expression in susceptible precursors (19). To determine whether the order of acquisition of genetic changes is important in aRMS, we generated a second set of genetically-defined HSMM cell lines whereby, as before, PAX3-FKHR was introduced first, followed by bypass of senescence, then introduction of hTERT, then MycN (HSMMPF+H+M); or whereby MycN was introduced first, followed by hTERT, then PAX3-FKHR (HSMMM+H+PF). hTERT was not used as an initial genetic change because although it prevents crisis associated with telomere shortening, it does not enable HSMMs to bypass the senescence checkpoint (9). Thus, here hTERT is used as a second or third genetic change. Cell lines were evaluated for stable transgene as well as p14ARF/p16INK4A expression (Fig. 3A), then tested as xenografts (Fig. 3B). As before, we found that only HSMMs expressing all three genetic changes—PAX3-FKHR, hTERT, and MycN, in that order—could form tumors (HSMMPF+H+M). However, HSMMs expressing MycN, hTERT, then PAX3-FKHR (HSMMM+H+PF) were unable to do so. Because PAX3-FKHR–mediated bypass of senescence (11) is an early step in this model, we asked whether this event might be the pivotal difference between these matched cell lines. Therefore, we assessed the effect of early MycN expression by stably expressing it in presenescent HSMMs, and found that although it enabled bypass of senescence (Fig. 3C), likely through epigenetic down-regulation of p14ARF (20), MycN-mediated bypass was not accompanied by the low p16INK4A and high PAX3-FKHR levels seen in PAX3-FKHR–mediated bypass (11). To the contrary, postsenescent HSMMMycN cells exhibited high p16INK4A (Fig. 3C,, inset M postsenescent; Fig. 3A,, M+H, M+H+PF) and low PAX3-FKHR protein expression (Fig. 3A).
Although we suspected that the altered order of genetic changes in the HSMMM+H+PF underlay their lack of tumorigenesis, mechanistically we wondered whether this was due to failure of p16INK4A down-regulation (or PAX3-FKHR up-regulation) to a critical threshold. Because p14ARF was similarly down-regulated in both MycN and PAX3-FKHR–mediated bypass, it was not likely the cause. To this end, we attempted to knock down p16INK4A (using an shRNA; ref. 11) or increase PAX3-FKHR (using overexpression constructs) in these cells. However, p16INK4A levels paradoxically increased in response to the shRNA, and PAX3-FKHR levels remained constant (data not shown), suggesting a tolerated setpoint of PAX3-FKHR expression, a phenomenon noted previously (21). Thus, it seems that early expression of PAX3-FKHR results in a specific level of expression of both itself and the INK/ARF locus, which affect later tumorigenesis.
Last, to confirm the requirement for not only early PAX3-FKHR expression, but for PAX3-FKHR–mediated bypass of senescence accompanied by p16INK4A loss in this model of aRMS, we generated a cell line (HSMMPF+M) in which PAX3-FKHR was expressed first and MycN second, but importantly, both expressed before the senescence checkpoint. After bypass, p16INK4A and PAX3-FKHR levels were high and low, respectively (Fig. 3D,, top) reminiscent of bypass mediated by MycN in the nontumorigenic HSMMM+H+PF cells. This predicted that HSMMPF+M cells, even after addition of hTERT, would not form tumors. However, when assayed as xenografts, all HSMMPF+M+H injected sites formed tumors, albeit after a delay (Fig. 3B). HSMMPF+M+H tumor lysates showed down-regulated p16INK4A compared with that in preinjection cultured cells (Fig. 3D,, bottom), suggesting a selective advantage for p16INK4A loss in vivo, and supporting our hypothesis that PAX3-FKHR–driven tumors require p16INK4A down-regulation. PAX3-FKHR transcripts from these tumors were not increased (Fig. 3D,, bottom), suggesting that low expression can be adequate for tumorigenesis. However, this must be interpreted cautiously, as we cannot rule out other mutations acquired in vivo, and at least in HSMMPF+M+H cells, PAX3-FKHR transcript level does not predict protein expression (Fig. 3D), suggesting regulation of PAX3-FKHR at the translational or posttranslational level. Taken together, these data suggest a specific order of acquisition of genetic lesions required to convert HSMMs to tumors mimicking aRMS, with early expression of PAX3-FKHR critical for tumorigenesis.
Summary. We have used a rational modeling approach to identify a set of genetic changes required to generate aRMS in the laboratory. Based on mutations identified in human aRMS tumor specimens, this includes gain-of-function of PAX3-FKHR with concomitant loss-of-function of p16INK4A/p14ARF (RB/p53 pathways), and gain-of-function of MycN and hTERT (Fig. 4). In addition, the order of acquisition of genetic lesions is critical, as only those HSMMs serially transduced to express PAX3-FKHR first, followed by hTERT/MycN, formed tumors in vivo. This supports the prediction that oncogenic translocations mediate important early events underlying later tumorigenesis. In RMS, early expression of PAX3-FKHR in a susceptible cell may provide the initiating step of senescence bypass; acquisition of subsequent critical oncogenes enables full conversion to the malignant phenotype. The biological value of this approach is in demonstrating that the same precursor cells, HSMMs, can be steered toward an eRMS or aRMS phenotype depending upon the genetic lesions introduced. The therapeutic value is in defining a minimum number of genetic lesions (cellular pathways) required to convert normal human skeletal muscle precursors into cells that can form aRMS. This knowledge may be useful in the development of rational drug combinations to treat this cancer.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
S. Naini and K.T. Etheridge contributed equally to this work.
Grant support: Hope Street Kids and Alex's Lemonade Stand pediatric cancer research foundations, and NIH grants 5K12-HD043494 (C.M. Linardic), and R01-CA94184 (C.M. Counter).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
We thank Fred Barr (University of Pennsylvania) and Rob Wechsler-Reya (Duke) for sharing the human PAX3-FKHR and murine FLAG-MycN cDNAs, respectively, and Tim Triche (Children's Hospital of Los Angeles) for providing human RMS cell lines. This manuscript is dedicated to the memory of Dr. Stephen Qualman, a superb clinician, scientist, and mentor.