The PAX3-FKHR Fusion Gene of Rhabdomyosarcoma Cooperates with Loss of p16^INK4A to Promote Bypass of Cellular Senescence Free

Rhabdomyosarcoma is the most common soft tissue sarcoma of childhood and adolescence. Despite advances in therapy, patients with a histologic variant of rhabdomyosarcoma known as alveolar rhabdomyosarcoma (ARMS) have a 5-year survival of <30%. ARMS is characterized by a chromosomal translocation generating the PAX3-FKHR fusion gene. However, ectopic expression of PAX3-FKHR often induces inhibition of cell proliferation, or cell death, when expressed in nonmuscle cells. This prompted us to explore the effect of expressing PAX3-FKHR in more relevant cells, specifically primary human skeletal muscle cells because these cells can be converted to a tumorigenic state that mimics rhabdomyosarcoma. PAX3-FKHR expression promoted both fetal and postnatal primary human skeletal muscle cell precursors to bypass the senescence growth arrest checkpoint. This bypass was accompanied by epigenetic DNA methylation of the p16^INK4A promoter and correspondingly a loss of expression of this tumor suppressor. Knockdown of p16^INK4A cooperated with PAX3-FKHR to drive proliferation past senescence, whereas reintroduction of wild-type p16^INK4A in post-senescent cells caused growth arrest. Thus, PAX3-FKHR acts in concert with loss of p16^INK4A to promote inappropriate proliferation of skeletal muscle cells. This association between PAX3-FKHR expression and p16^INK4A loss was seen in human ARMS tumor tissue, as both human rhabdomyosarcoma cell lines and tissue microarrays showed a trend toward down-regulation of p16^INK4A protein in alveolar subsets. We surmise that the generation of the PAX3-FKHR fusion protein may require loss of p16^INK4A to promote malignant proliferation of skeletal muscle cells as an early step in ARMS tumorigenesis. [Cancer Res 2007;67(14):6691–9]

Rhabdomyosarcoma is the most common soft tissue sarcoma of childhood and adolescence (1, 2). Through the efforts of the Intergroup Rhabdomyosarcoma Study Group, and now the Children's Oncology Group, patients are stratified into low-, intermediate-, and high-risk populations based on clinical “grouping,” which takes into account extent of disease and surgical resection, and clinical “staging,” which takes into account anatomic site of disease (1). Tumor histology has emerged as a significant predictor of outcome, as patients with the embryonal variant of rhabdomyosarcoma [termed embryonal rhabdomyosarcoma (ERMS)] generally have a better outcome than those with the alveolar type [termed alveolar rhabdomyosarcoma (ARMS); ref. 1]. Histology has thus been incorporated into the risk stratification (e.g., low-stage, low-group ARMS cases are shifted to intermediate risk because of this histology). A cytogenetic hallmark of ARMS is the reciprocal translocation of chromosomes 2 and 13, which fuses in-frame the DNA binding domain of the transcription factor PAX3 with the transactivation domain of the transcription factor FKHR, resulting in the PAX3-FKHR fusion gene (3, 4). This chimeric transcription factor retains the DNA-binding specificity of PAX3 but is much more potent due to the altered transactivation function of the FKHR moiety (reviewed in ref. 5). The expression of PAX3-FKHR in metastatic ARMS portends a particularly poor outcome, with a 4-year survival of only 8% (6). Because of its unique expression in ARMS, PAX3-FKHR continues to be a desirable target for therapeutic purposes (7).

Elucidating the specific roles of PAX3-FKHR in rhabdomyosarcoma tumorigenesis has been challenging. Early studies of PAX3-FKHR were done in avian and rodent cell lines and suggested that PAX3-FKHR was a dominant-acting oncogene that induced transformation (reviewed in ref. 5). The mechanism of transformation is not clear, because although PAX3-FKHR was found through microarray analysis to reactivate myogenic transcription programs (8) in a manner similar, but not identical to the neural/skeletal muscle lineage-control gene PAX3, this reactivation alone is not adequate to transform cells.⁶

Later studies, based on the ectopic expression of PAX3-FKHR in various murine and human ERMS cell lines, suggested that it could have either stimulatory or inhibitory effects on cell proliferation and apoptosis (5), and in some cases was frankly cytotoxic (9). These conflicting data may reflect the differential effects of PAX3-FKHR depending on the context of its expression, including level of expression and cellular milieu. Indeed, expression of other oncogenes, such as c-MYC and RAS, can have very different phenotypic consequences depending on the cell type and cellular milieu in which they are expressed (10). As such, we reasoned that cell type might be a very important consideration when investigating PAX3-FKHR function.

We have recently developed a human cell model for rhabdomyosarcoma, in which primary human skeletal muscle cell precursors are converted to their tumorigenic counterpart by the serial introduction of a defined set of genetic changes that corrupt the p53, RB, MYC, telomerase, and RAS pathways (11). In this system, conversion of primary human prenatal skeletal muscle cell precursors causes tumors with some features of rhabdomyosarcoma, whereas conversion of primary human postnatal skeletal myoblasts causes tumors resembling ERMS, strongly suggesting that the starting cell type affects rhabdomyosarcoma histology. As these primary human skeletal muscle precursors could be converted to a tumorigenic state identical to rhabdomyosarcoma, we reasoned that these cells might be the very ones in which the PAX3-FKHR oncogenic functions would be manifested. We therefore evaluated the consequences of expressing PAX3-FKHR in a cell type most relevant to rhabdomyosarcoma, human cells of skeletal muscle origin.

Generation of cell lines and evaluation of proliferative capacity. Early passage normal human fetal skeletal muscle (SKMC) cells (Cambrex Corp.) grown in defined media (Clonetics SkGM Bullet kit) were stably infected and selected for drug resistance with amphotrophic retroviruses derived from pK1-PAX3-FKHR, p16INK4AshRNA in pSUPER-retro-GFP-neo, wild-type (WT) p16INK4A, or vector control. Cells were selected in 0.25 μg/mL puromycin (Sigma) or 50 μg/mL hygromycin B for 7 days, or 250 μg/mL G418 (Life Technologies Invitrogen) for 10 days. The first confluent plate after drug selection for the last transgene was designated pd0. SKMCs were defined to be of skeletal muscle origin by the vendor's routine characterization using immunofluorescent staining, which was positive for sarcomeric myosin. Human skeletal muscle myoblast (HSMM) cells were characterized by the vendor as >60% positive for desmin stain at first passage out of cryopreservation, although the lot used in these experiments was >90% desmin positive. IMR90 cells were grown in DMEM/F12 (1:1) with 15% fetal bovine serum (FBS). Human rhabdomyosarcoma cell lines were grown in RPMI 1640 with 10% FBS.

Immunoblotting. Cell lysates were prepared by homogenization in Tris/radioimmunoprecipitation assay buffer with standard protease inhibitors and passage through a 21-gauge needle to shear DNA. Protein concentration was measured by the Bio-Rad detergent-compatible protein assay (Bio-Rad). Each lysate (60–100 μg) was resolved by SDS-PAGE, transferred to polyvinylidene difluoride membrane, and immunoblotted with primary monoclonal antibodies anti-FOXO1A F6928 (Sigma), anti-p16^INK4A (BD PharMingen), anti-desmin M0760 (DakoCytomation), anti–muscle-specific actin M0635 (DakoCytomation), anti-actin SC-8342 (Santa Cruz Biotechnology), or anti-tubulin T-4026 (Sigma), reacted with a secondary horseradish peroxidase (HRP)–labeled goat anti-mouse antibody (Amersham Biosciences) and developed using chemiluminescence (Amersham Biosciences). The anti-FOXO1A antibody was generated against the COOH terminus of FOXO1A (also known as FKHR) protein and therefore recognizes not only FKHR (∼70–80 kDa) but also the PAX3-FKHR fusion, which shows a mobility shift (∼100–110 kDa) in our system. This article does not address the biological significance of changes in FKHR; FKHR bands were only included in the figures to provide molecular weight reference.

Analysis of p16^INK4A promoter methylation and allelic loss. Analysis of INK4A promoter CpG island methylation and INK4A allelic loss was done using methods described (12).

Differentiation assays. Growth media of ∼60% confluence HSMM cell cultures were replaced with fusion media (DMEM/F12 with 2% horse serum) and refreshed every other day for ∼5 days. Myotubes were rinsed with PBS, fixed with cold 70% ethanol/formalin/glacial acetic acid (20:2:1), rinsed again with PBS, and permeabilized for 10 min with 0.10% NP40 in 10 mmol/L Tris (pH 7.4) and 150 mmol/L NaCl. For primary antibody binding, myotubes were first blocked for 30 min with 10% horse serum in PBS, incubated for 1 h with anti–sarcomere-myosin hybridoma MF20 (13), and then rinsed with PBS × 3 for 10 min. Primary antibody binding was detected by incubation for 30 min with biotinylated anti-mouse IgG [diluted to 5 μg/mL in PBS-0.10% bovine serum albumin (BSA)], rinsing with PBS × 3 for 10 min, incubation for 30 min with HRP-streptavidin (diluted to 4 μg/mL in PBS-0.10% BSA), and then rinsing with PBS × 3 for 10 min. A final 5-min incubation was done with 3,3′-diaminobenzidine reagent followed by a water rinse, rendering the antibody binding as brown-black.

Analysis of tissue microarrays by immunohistochemistry. Human ERMS and ARMS tumor tissue microarrays (TMA) were obtained from the Cooperative Human Tissue Network, which is funded by the National Cancer Institute, with prior Institutional Review Board approval. The ERMS TMA contained 60 cores derived from 21 unique cases; the ARMS TMA contained 31 cores derived from 13 unique cases. Human p16^INK4A expression was evaluated using the CINtec p16^INK4A Histology kit as described by the manufacturer (DakoCytomation). Normal human tonsil tissue was used as an external positive control, and an isotype-specific antibody was used as a negative control. As internal positive controls, companion slides were analyzed (and showed 100% staining) for expression of MyoD1 and myogenin to prove intact antigenicity (14).

Statistical analysis. Tissue cores were evaluated by C.K. and S.Q. and scored for p16^INK4A staining using a system reported previously (15). Only nuclear staining was considered positive and was graded as follows: 0, negative stain; 1+, rare isolated nuclei positive; 2+, ≤10% nuclei positive; 3+, >10% but <50% nuclei positive; and 4+, ≥50% nuclei positive. Because some cores were replicates from the same tumor, and showed variable staining, these replicate scores were averaged to generate a final score for each unique tumor. A two-sample Wilcoxon test was used to compare the scores from the ERMS and ARMS groups.

PAX3-FKHR drives normal primary human skeletal muscle cells past the senescence checkpoint. To evaluate the phenotypic consequences of PAX3-FKHR expression in primary human cells of skeletal muscle origin, a cDNA encoding its full-length sequence was stably expressed in primary human prenatal SKMCs using amphotrophic retrovirus. Because primary SKMCs senesce in culture after 15 population doublings, experiments were initiated in low passage (pd2) cells. Before genetic manipulation, SKMCs were verified to express desmin, one of the protein markers supporting skeletal muscle origin (Fig. 1A,, top). Following selection for retroviral insertion, the resulting cell lines were examined for appropriate transgene protein expression (Fig. 1A,, bottom); the level of expression of PAX3-FKHR protein was similar to that expressed by two human ARMS cell lines, JR and Rh28. SKMCs expressing PAX3-FKHR (SKMC^PF) or vector control (SKMC^V) were then monitored in culture for changes in their proliferation or morphology. Within 1 month, SKMC^V cells slowed their proliferative rate and acquired morphologic signs of senescence, including cell spreading, flattening, and granularity, with complete cessation of proliferation of two population doublings following the transduction of the last retroviral vector (Fig. 1B,, ○). Although the SKMC^PF cells generally showed a slowing of proliferation, the culture was eventually overtaken by microcolonies of small, refractile cells that eventually dominated the culture and continued proliferating past the predicted senescence checkpoint (Fig. 1B , •). These cells were termed “post-senescent ” because they proliferated past the senescence checkpoint that classically leads to the growth arrest of primary cells in culture. This type of bypass suggests that a primary cell population has overcome the growth-inhibitory stress signals of culture in an artificial environment (16).

To assess the reproducibility of this phenomenon, a second type of primary human skeletal muscle cell, known as postnatal human skeletal muscle myoblasts (HSMM) cells, was similarly studied. Low passage cells were verified to be of skeletal muscle origin, as shown by expression of desmin and skeletal-muscle–specific actin (Fig. 1A,, top), stably infected with retrovirus encoding PAX3-FKHR or an empty vector, and validated to express PAX3-FKHR at a protein level similar to other human ARMS cell lines (Fig. 1A,, bottom). Population doublings were monitored over time to assess the proliferative life span of the cells. In agreement with others (9), ectopic expression of PAX3-FKHR initially caused some inhibition of proliferation compared with vector control, but expression was overall tolerated. The proliferation of HSMM^V cells plateaued at about pd15-19 (Fig. 1C,, ○), whereas HSMM^PF cells exhibited a slight delay at this point, but nevertheless proliferated past the senescence checkpoint (Fig. 1C,, •) and continued log-phase growth. As with the SKMCs, HSMM^PF cells bypassing senescence acquired a proliferative appearance with refractile cells in mitosis (Fig. 1D,, left), whereas HSMM^V cells slowed their growth rate and adopted a senescent morphology (Fig. 1D , right). After months of senescence, this population seemed to resume proliferating, albeit at a very slow rate. To determine whether the expression of PAX3-FKHR similarly enabled a nonskeletal muscle primary human mesenchymal cell type to bypass senescence, IMR90 human fetal lung fibroblast cells, which usually senesce at about 50 population doublings, were infected with retrovirus encoding PAX3-FKHR or empty vector. Both cell lines were able to be generated, indicating successful integration of the retrovirus. However, both cell lines senesced, the IMR90^Vector at about pd52 and the IMR90^PAX3-FKHR at about pd36 (data not shown). Last, in agreement with others' observations that PAX3-FKHR can be toxic, we found that although we could express an empty control vector in HT1080 fibrosarcoma cells, PAX3-FKHR expression was not tolerated.

In summary, we found that stable expression of PAX3-FKHR in two types of primary human skeletal muscle cell precursors, SKMC and HSMM cells, enabled them to proliferate past the senescence checkpoint. Stable expression of PAX3-FKHR in a human nonskeletal muscle primary mesenchymal cell type was tolerated but did not enable bypass. Stable expression in a human nonskeletal muscle mesenchymal tumor cell line was toxic. Thus, the phenotypic consequences of PAX3-FKHR expression is cell type specific, and a novel oncogenic function for PAX3-FKHR, bypass of the senescence checkpoint, was revealed when this protein was expressed in the most relevant cells for this disease, human skeletal muscle cells.

PAX3-FKHR–mediated bypass of cellular senescence is accompanied by loss of the tumor suppressor p16^INK4A. Having established that the expression of PAX3-FKHR could coax primary human skeletal muscle cells past the senescence checkpoint, we next sought to identify the molecular changes underlying this phenomenon. We hypothesized that PAX3-FKHR provided a strong proliferative signal that selected for a population of cells able to continue dividing beyond senescence. Previous studies in primary human mammary epithelial cells showed that epigenetic down-regulation of the p16^INK4A tumor suppressor protein was critical in their overcoming a senescence-like state (17) and that this bypass was a step toward tumorigenesis (18).

To investigate the possibility that an epigenetic change such as p16^INK4A silencing might be occurring in the observed PAX3-FKHR–mediated bypass of senescence, p16^INK4A protein expression was evaluated in post-senescent SKMCs expressing PAX3-FKHR. As a positive control, p16^INK4A expression was also assayed in SKMCs expressing the SV40 DNA tumor virus early region, which enables bypass of senescence via inactivation of the p53 and RB pathways, independent of p16^INK4A loss. Whereas post-senescent SKMCs expressing the SV40 early region (SKMC^T/t+V) still expressed p16^INK4A, post-senescent SKMCs expressing PAX3-FKHR (SKMC^V+PF) showed a significant down-regulation in p16^INK4A protein expression (Fig. 2A,, top), although on very long exposures, a faint band could be detected (data not shown). To explore the kinetics of p16^INK4A loss, HSMM^V and HSMM^PF cells were monitored in culture over time; with cumulative population doublings, HSMM^V cells accumulated p16^INK4A (Fig. 2,A, bottom, V_pd), consistent with the onset of senescence, whereas HSMM^PF cells lost p16^INK4A expression (Fig. 2 A, bottom, PF_pd), consistent with the bypass of senescence. Loss of the p16^INK4A tumor suppressor occurs frequently in mammalian tumorigenesis, and mutations that lead to this loss have been catalogued in both cell culture systems and in human clinical tumor specimens (19). In the majority of cases, p16^INK4A expression is lost by one of three genetic or epigenetic mechanisms: transcriptional silencing via DNA methylation-independent trans-acting proteins, allelic deletion at the INK4A locus on chromosome 9p (10), or transcriptional silencing of its gene INK4A via promoter CpG island methylation. Because it is not known whether p16^INK4A is frequently lost in human rhabdomyosarcoma nor through what molecular basis, all three mechanisms were investigated.

First, to investigate the possibility that PAX3-FKHR was serving as a DNA methylation-independent transcriptional repressor of INK4A, we stably expressed PAX3-FKHR in SKMC^T/t+V cells, predicting that p16^INK4A levels should decrease. However, SKMC^T/t+PF cells showed persistent p16^INK4A expression (Fig. 2A , top), indicating no effect of PAX3-FKHR. We also considered the possibility that PAX3-FKHR was affecting p16^INK4A expression by transcriptionally activating BMI-1, a polycomb group repressor of INK4A transcription that is implicated in stem cell self-renewal but down-regulated in primary cells as they senesce (20). As expected, BMI-1 levels decreased in both SKMC and HSMM vector populations as they senesced and remained low. However, a similar loss of BMI-1 expression was seen in those cells expressing PAX3-FKHR as they approached the senescence checkpoint, indicating no up-regulation of BMI-1 in the presence of PAX3-FKHR overexpression. Interestingly, after senescence bypass, both SKMCs and HSMM cells expressing PAX3-FKHR showed dramatically increased BMI-1 levels that paralleled their p16^INK4A loss, indicating that other factors must be influencing BMI-1 expression. We also found that SV40-transformed HSMM cells expressing high levels of p16^INK4A protein simultaneously expressed high levels of BMI-1 (data not shown), implicating other inputs to p16^INK4A modulation (19). Thus, the control of the INK4A locus is complex, and although not exhaustive, these analyses provide no evidence for a trans-acting effect of PAX3-FKHR on the expression of p16^INK4A.

To evaluate for homozygous allelic INK4A loss, genomic DNA was isolated from HSMM cells and assayed for the presence or absence of exon 2 of INK4A, using a standard nested PCR approach (12). In both pre-senescent HSMM^V and post-senescent HSMM^PF cells, PCR product was visible in both the INK4A and D9S196 reactions, indicating no homozygous loss (Fig. 2B,, top). To evaluate for heterozygous INK4A allelic loss, genomic DNA isolated from HSMM cells was assayed for the presence of three different microsatellite markers, D9S126, D9S741, and D9S1748, again using a standard PCR approach (12). PCR product patterns were identical in pre- and post-senescent cells, indicating that at these three loci, there was no loss of heterozygosity (Fig. 2B , bottom).

Last, to determine whether the INK4A gene was transcriptionally silenced by promoter methylation of CpG islands, genomic DNA collected from both pre-senescent HSMM^V and post-senescent HSMM^PF cells was subject to methylation-dependent PCR; in pre-senescent HSMM^V cells, the INK4A promoter was unmethylated, whereas in post-senescent HSMM^PF cells, the INK4A promoter became methylated (Fig. 2C,, top). This assay was repeated in SKMCs, with similar findings (Fig. 2C,, bottom). Because the INK4A promoter was methylated in cells expressing PAX3-FKHR that had bypassed senescence, INK4A mRNA should be decreased; this was verified using reverse transcription-PCR (RT-PCR), which showed that INK4A transcription was down-regulated (Fig. 2D).

In summary, we found that in both SKMCs and HSMM cells that had bypassed senescence vis-à-vis PAX3-FKHR expression, p16^INK4A protein expression was greatly diminished, seeming to occur by transcriptional down-regulation via DNA methylation of the INK4A promoter. We hypothesize that PAX3-FKHR expression drives cells to divide sufficiently for the selection of cells with INK4A promoter methylation, thereby overcoming the senescence checkpoint.

PAX3-FKHR cooperates with p16^INK4A loss to promote bypass of senescence. Because the expression of PAX3-FKHR promoted a bypass of cellular senescence, but only in the setting of decreased p16^INK4A expression, we next explored whether PAX3-FKHR gain-of-function and p16^INK4A loss-of-function cooperated to promote bypass of cellular senescence. Specifically, a p16^INK4A short hairpin RNA (shRNA; 21) to knock down p16^INK4A expression was coexpressed with PAX3-FKHR in primary cultures of HSMM cells. As controls, HSMM cells were engineered to express p16^INK4A shRNA alone, PAX3-FKHR alone, or infected with two empty control vectors. All four cell lines were assessed for proliferative capacity, expression of p16^INK4A and PAX3-FKHR protein, and morphology (Fig. 3). As observed previously, the vector cells (Fig. 3A,, V+V, ⋄) adopted a flattened, granular appearance without mitotic figures (Fig. 3C, V+V) and entered senescence, although in this experiment, a clonal population later escaped senescence and resumed proliferating after months of quiescence, in which p16^INK4A expression was again down-regulated as assessed by immunoblot (data not shown). Also as expected, after an extended period of selection, expression of PAX3-FKHR drove cells beyond senescence (Fig. 3A,, PF+V, □). These cells were rounded, refractile, and contained many mitotic figures (Fig. 3C, PF+V). Consistent with the known role of p16^INK4A in the maintenance of cellular senescence, cells stably expressing the p16^INK4A shRNA also eventually bypassed senescence (Fig. 3A,, V+p16^sh, ♦). These cells were flattened and granular but with occasional mitotic figures (Fig. 3C, V+p16^sh). However, concomitant loss of p16^INK4A with expression of PAX3-FKHR drove cells to proliferate faster and overcome senescence with little or no lag in proliferation (Fig. 3A,, PF+p16^sh, ▪). These cells were rounded, refractile, and contained many mitotic figures (Fig. 3C, PF+p16^sh). Analysis of p16^INK4A protein expression by immunoblot (Fig. 3B) verified the expected expression level of p16^INK4A for each cell line, with the senescent cells showing an increase in p16^INK4A expression compared with pre-senescence cells, the p16^sh cells showing some knockdown of p16^INK4A, whereas those cells expressing PAX3-FKHR, or PAX3-FKHR with the p16^INK4A shRNA construct, showing dramatic loss of p16^INK4A expression. PAX3-FKHR expression remained similar whether the p16 short hairpin was present (Fig. 3B).

We next evaluated the importance of p16^INK4A down-regulation in the phases of cell culture just before and just after the bypass of senescence. First, we tested the effect of forced expression of WT p16^INK4A in HSMM^PF cells just before their entering the time frame when we expected them to encounter the senescence checkpoint. Pre-senescent HSMM^V cells at pd8 or pre-senescent HSMM^PF at pd9 were stably transduced with a second empty vector or vector encoding WT p16^INK4A. Both pre-senescent populations could, as expected, tolerate the expression of a second control empty vector, with the HSMM^V+V cells ultimately senescing and the HSMM^PF+V cells ultimately bypassing senescence. On the other hand, both the pre-senescent populations acutely growth arrested and died when forced to express WT p16^INK4A (data not shown), suggesting that down-regulation of p16^INK4A in the phase of cell culture just before bypass was critical.

Second, to examine the requirement for persistent loss of p16^INK4A after senescence, we tested the effect of forced expression of WT p16^INK4A cDNA in post-senescent HSMM^PF cells. Whereas post-senescent HSMM^PF control cells continued to proliferate well after transduction with empty vector, as reflected by their refractile morphology and numerous mitotic figures (Fig. 3D,, top), those transduced with p16^INK4A cells showed a growth arrest (Fig. 3D , bottom), indicating that continued down-regulation of p16^INK4A is required for persistent proliferation of these cells in the post-senescent phase. Three colonies later emerged from this cell population and were found to lack detectable p16^INK4A expression, further validating our observation that p16^INK4A expression must continue to be suppressed in post-senescent HSMM^PF cells.

In summary, we found that whereas ectopic expression of PAX3-FKHR, or loss of p16^INK4A, each enabled slow bypass of the senescence checkpoint, expression of PAX3-FKHR with concomitant knockdown of p16^INK4A stimulated the most rapid bypass of senescence. Expression of WT p16^INK4A at levels tolerated in other cells either immediately prior to, or after the bypass of senescence, caused acute cell growth arrest and cell death, indicating the absolute requirement for p16^INK4A loss. Thus, loss of p16^INK4A cooperates with gain of PAX3-FKHR to drive inappropriate proliferation. These data support the role for the cooperation of PAX3-FKHR gain-of-function with p16^INK4A loss-of-function in the ability of human cells of skeletal muscle origin to bypass the senescence checkpoint.

PAX3-FKHR does not promote senescence bypass by interfering with myoblast differentiation. PAX3-FKHR could theoretically cause bypass of senescence either by promoting proliferation or by interfering with the ability of cells to respond to differentiation-inducing signals. Although we showed that PAX3-FKHR does stimulate proliferation, it nevertheless remained possible that inhibition of differentiation was a major contributor to overcoming senescence. It has been suggested that PAX3-FKHR contributes to rhabdomyosarcoma by thwarting the ability of skeletal muscle cells to terminally differentiate to myotubes (5). Furthermore, experiments in human ERMS cell lines showed that loss of p16^INK4A specifically contributes to their resistance to differentiate to myotubes (22).

Here, we show that the ability of PAX3-FKHR to drive cells past senescence is not due to resistance of these cells to properly respond to differentiation signals. Specifically, HSMM cells expressing PAX3-FKHR, PAX3-FKHR with a p16^INK4A shRNA targeting vector, or p16^INK4A shRNA targeting vector alone were cultured in myoblast fusion media to induce myotube formation. We found that in a manner similar to control native HSMM cells (Fig. 4A,, a and b), HSMM cells expressing PAX3-FKHR could be stimulated to differentiate (Fig. 4A,, c and d), whereas cells expressing the p16^INK4A shRNA construct could not (Fig. 4A,, g and h), similar to ERMS cell lines lacking p16^INK4A (22). However, those cells expressing both PAX3-FKHR and the p16^INK4A shRNA construct were able to be differentiated to myotubes (Fig. 4A,, e and f), indicating that the block to differentiation induced by p16^INK4A loss could be overcome by expression of PAX3-FKHR and, therefore, that the combination of PAX3-FKHR expression and p16^INK4A loss was not stimulating proliferation past senescence by inhibiting differentiation. Interestingly, differentiated cells exhibited reduced expression of PAX3-FKHR (Fig. 4B), but nevertheless the transgene was still expressed.

In summary, the expression of PAX3-FKHR in HSMM cells that had bypassed senescence does not inhibit their ability to terminally differentiate to skeletal myotubes. Rather, we conclude that the cooperation of PAX3-FKHR with p16^INK4A loss instead provides a permissive environment for PAX3-FKHR to drive proliferation.

Down-regulation of p16^INK4A protein expression in human rhabdomyosarcoma cell lines and tumors. The above in vitro data suggested that expression of PAX3-FKHR cooperated with the loss of p16^INK4A, resulting in a robust proliferative phenotype. To determine if in vivo during human rhabdomyosarcomagenesis a similar cooperative event might also occur, the expression of p16^INK4A protein was assayed in human ARMS and ERMS cell lines and in human rhabdomyosarcoma tumors. In three of three tested ERMS cell lines, which do not harbor the PAX3-FKHR fusion gene, p16^INK4A protein was readily detected, whereas in three of four ARMS cell lines that all expressed the PAX3-FKHR fusion gene, p16^INK4A was expressed at a lower level or not at all (Fig. 5A). A fifth ARMS cell line, Rh18, which, although it is morphologically alveolar, does not have the t(2;13) translocation and therefore does not express PAX3-FKHR, also showed down-regulation of p16^INK4A protein (Fig. 5A). It is interesting to note that in the ARMS cell lines harboring a PAX3-FKHR fusion, p16^INK4A and PAX3-FKHR protein levels seem to have an inverse relationship: the higher the level of PAX3-FKHR, the lower the level of p16^INK4A.

We next assayed p16^INK4A expression in human tumor samples by subjecting human rhabdomyosarcoma TMAs to standard immunohistochemical techniques to detect and quantify nuclear p16^INK4A protein expression. Cores were scored using a method described previously (23), and the scores were statistically analyzed. p16^INK4A expression was found to be decreased in the majority of the tissue cores, but because some cores were replicates from the same tumor and the level of p16^INK4A expression varied between replicates, a final average core value was used to represent each unique tumor. Variable expression of p16^INK4A within single tumor samples has been reported by others, perhaps related to microenvironment (20). p16^INK4A expression was completely negative in 33% of the ERMS tumors and 54% of the ARMS tumors (example of these divergent histologies in Fig. 5B), and although there was a trend toward overall decreased expression in the ARMS histology, this was not statistically significant. PAX3-FKHR fusion status was not known for these cores, as fusion testing is not yet done routinely. Thus, although these data are intriguing, a true correlation between the expression of PAX3-FKHR and p16^INK4A in ARMS tumors will require repeat studies with a larger sample size and known PAX3-FKHR fusion status.

In summary, we found that in human ARMS cell lines, p16^INK4A is decreased at the protein level, whereas in primary rhabdomyosarcoma tumor tissue, there is a trend, although not statistically significant, toward preferential lost of p16^INK4A in the alveolar subtype. These results support our in vitro data, suggesting that the cooperation of PAX3-FKHR expression with p16^INK4A loss may be an important biological phenomenon in rhabdomyosarcoma tumorigenesis.

Since its discovery in 1993 as a genetic lesion specific to rhabdomyosarcoma (3, 4), the PAX3-FKHR fusion gene has remained a desirable, but elusive, target for the treatment of rhabdomyosarcoma. A variety of cell systems have been used to explore its role in rhabdomyosarcomagenesis, ranging from the earliest studies of ectopically expressed PAX3-FKHR in avian and rodent fibroblasts, which first illuminated its role as a weakly transforming oncogene (5), to the most recent studies of ectopically expressed PAX3-FKHR in murine cell lines, which highlight its growth-inhibitory properties (9). These seemingly contradictory roles suggested to us that effects of PAX3-FKHR might be cell type specific, and given that primary human skeletal muscle cells could be driven to a rhabdomyosarcoma-like state (11), we studied the phenotypic consequences of expressing PAX3-FKHR in this primary cell type.

We found that primary human skeletal muscle cells normally exhibit a prolonged period of senescence before the selection of rare clones that acquire an extended life span. However, this time interval was greatly reduced by ectopic expression of a shRNA to the p16^INK4A tumor suppressor or by ectopic expression of the PAX3-FKHR fusion protein. In the latter case, this was accompanied by p16^INK4A DNA promoter methylation. These observations suggested a cooperative effect between p16^INK4A loss-of-function and PAX3-FKHR gain-of-function, and indeed, synergy was noted when the p16 shRNA was coexpressed with PAX3-FKHR, greatly accelerating the bypass of senescence. Loss of p16^INK4A expression was required not only in the period of culturing just before senescence bypass but also in the post-senescent phase. Post-senescent cells expressing PAX3-FKHR were still able to be terminally differentiated to myotubes, indicating that PAX3-FKHR was not enabling the bypass of senescence by inactivating the signaling pathways that mediate differentiation. We conclude that the expression of PAX3-FKHR cooperates with the silencing of the p16^INK4A to enable primary human cells of skeletal muscle origin to bypass the senescence checkpoint.

We speculate that the relationship between gain-of-function of PAX3-FKHR and loss-of-function of p16^INK4A may reflect cooperating tumorigenic events in rhabdomyosarcomagenesis. In support of this, conditional expression of PAX3-FKHR in differentiating skeletal muscle myofibers of mice with prior INK4A/ARF pathway inactivation generates tumors with ARMS histology. This cooperation may even extend to other cancers. For example, the overexpression of the micropthalmia transcription factor (a transcriptional target of PAX3) in cooperation with p16^INK4A loss results in the development of human malignant melanoma (24). Loss of p16^INK4A in actual human rhabdomyosarcoma tumors has not been studied comprehensively, although three independent groups reported previously that human rhabdomyosarcoma cell lines and tumors showed p16^INK4A loss through allelic deletion or INK4A promoter methylation (25–27). The total number of rhabdomyosarcoma tissues examined in these studies was small, and no consistent correlation with ARMS/ERMS histology was made. Nevertheless, these studies provided evidence that p16^INK4A silencing is occurring in human rhabdomyosarcoma. Our TMA studies suggest that p16^INK4A loss occurs on a larger scale, and it is interesting to consider this in view of the multiple mouse models that show a role for p16^INK4A pathway inactivation in rhabdomyosarcoma formation.

In summary, we have found that expression of the ARMS-specific fusion gene PAX3-FKHR in primary human skeletal muscle cells cooperates with the epigenetic loss of the p16^INK4A tumor suppressor to promote bypass of the senescence checkpoint. This overcoming of senescence may be an early, initiating oncogenic event that eventually leads to rhabdomyosarcoma in a susceptible cell type. Our system provides a platform on which to further model the serial accumulation of genetic insults that result in this sarcoma.

Grant support: Hope Streets Kids and Bear Necessities 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, Philadelphia, PA) for sharing the PAX3-FKHR cDNA; Tim Triche (Children's Hospital of Los Angeles, Los Angeles, CA) for providing human rhabdomyosarcoma cell lines; John Sedivy and Utz Herbig (Brown University, Providence, RI) for providing the p16^INK4A cDNA shRNA; and Natalia Mitin (University of North Carolina, Chapel Hill, NC) for advice on the myoblast differentiation assay. The anti–sarcomere-myosin hybridoma MF20 was obtained from the Developmental Studies Hybridoma Bank, developed under the auspices of the National Institute of Child Health and Human Resources, and maintained by the Department of Biological Sciences, University of Iowa, Iowa City, IA.