FGFR4 blockade exerts distinct anti-tumorigenic effects in human embryonal versus alveolar rhabdomyosarcoma

Abstract

Purpose

Rhabdomyosarcoma (RMS) is a malignancy with features of skeletal muscle, and the most common soft-tissue sarcoma of childhood. Survival for high risk groups is ~30% at 5 years and there are no durable therapies tailored to its genetic aberrations. During genetic modeling of the common RMS variants, embryonal (eRMS) and alveolar (aRMS), we noted that the RTK FGFR4 was upregulated as an early event in aRMS. Herein, we evaluated the expression of FGFR4 in eRMS compared to aRMS, and whether FGFR4 had similar or distinct roles in their tumorigenesis.

Experimental Design

Human RMS cell lines and tumor tissue were analyzed for FGFR4 expression by immunoblot and IHC. Genetic and pharmacologic loss-of-function of FGFR4 using virally-transduced shRNAs and the FGFR small molecule inhibitor PD173074, respectively, were used to study the role of FGFR4 in RMS cell lines in vitro and xenografts in vivo. Expression of the anti-apoptotic protein BCL2L1 was also examined.

Results

FGFR4 is expressed in both RMS subtypes, but protein expression is higher in aRMS. The signature aRMS gene fusion product, PAX3-FOXO1, induced FGFR4 expression in primary human myoblasts. In eRMS, FGFR4 loss-of-function reduced cell proliferation in vitro and xenograft formation in vivo. In aRMS, it diminished cell survival in vitro. In myoblasts and aRMS, FGFR4 was necessary and sufficient for expression of BCL2L1, while in eRMS, this induction was not observed, suggesting differential FGFR4 signaling.

Conclusion

These studies define dichotomous roles for FGFR4 in RMS subtypes, and support further study of FGFR4 as a therapeutic target.

Introduction

Rhabdomyosarcoma (RMS) is the most common soft tissue sarcoma of childhood and adolescence, accounting for 40% of soft tissue sarcomas in this age range (1). RMS tumors exhibit features of skeletal muscle and are thought to derive from mesenchymal cell precursors that have failed to differentiate or otherwise developed aberrantly along the skeletal muscle axis. RMS treatment includes a multimodal approach of surgery, radiation, and chemotherapy, but despite the evaluation of experimental agents through cooperative group clinical trials, 5-year survival for high risk groups remains ~30% (2).

In addition to clinical group and stage, risk stratification for RMS includes tumor histology. Originally defined by morphology under light microscopy, RMS histologic variants are known to bear unique cytogenetic and molecular abnormalities (3). The embryonal variant (eRMS), often found in the head, neck and trunk, generally portends a favorable outcome, and is associated with LOH and abnormal imprinting on chromosome 11. The alveolar variant (aRMS), often found in the extremities, portends a worse outcome (survival <50% at 5 years) (1), and is associated with a chromosomal translocation resulting in the PAX3-FOXO1 (also known as PAX3-FKHR) mutant fusion transcription factor. Although PAX3-FOXO1 is found exclusively in aRMS (4, 5) and thus a desirable drug target, as a transcription factor it has remained chemically intractable. Microarray studies suggest that RMS variants have distinct transcriptional profiles (6, 7) and therefore are distinct biological entities. We and others have been seeking to understand the differences between the RMS variants, to ultimately identify the most effective treatments for each subtype.

Our approach has been to generate primary cell-based models of RMS variants to shed light on the cumulative mutations required for RMS tumorigenesis. To this end, we have step-wise converted normal human skeletal muscle myoblasts (HSMM cells) to their transformed counterparts (8, 9) using genetic alterations known to occur in RMS. To generate eRMS versus aRMS requires a different combination of genetic changes, including a role for early expression of PAX3-FOXO1 in the aRMS model. When PAX3-FOXO1 is introduced into HSMM cells, followed by two other necessary genetic changes, they become transformed and mimic aRMS xenografts in vivo. However, if PAX3-FOXO1 is introduced later, the cells are not tumorigenic, implying that changes in gene expression and/or cell signaling induced by PAX3-FOXO1 act early to prime cells for transformation. One of the changes we observed was upregulation of the receptor tyrosine kinase (RTK) fibroblast growth factor receptor 4 (FGFR4).

FGFR4 is a member of the FGFR RTK family, which regulates cell processes such as shape change, adhesion, and migration (10, 11), and cell fate decisions such as differentiation, proliferation, and survival. During embryonic and post-natal life, FGFR4 is essential to skeletal muscle myogenesis (12) and repair (13). Dysregulated FGFR4 activity has been found in adult epithelial cancers including head and neck, thyroid, breast, hepatocellular, and prostate tumors ((14, 15) and references therein). A role for FGFR4 in chemoresistance has recently been described for breast cancer, suggesting that aberrant FGFR4 activity can be acquired later in tumorigenesis (16). The earliest evidence for aberrant FGFR4 expression in RMS came from microarray and transcriptional studies (17–19), but the biological roles for FGFR4 in RMS have only recently begun to be elucidated. For example, FGFR4 contains activating mutations in up to 7% of RMS tumors and contributes to metastasis (20).

Given the distinct clinical behavior and molecular profiles of RMS subtypes, the increasing evidence for FGFR4 activation in RMS, our observation of increased FGFR4 expression in human myoblasts bearing the PAX3-FOXO1 fusion gene, and our goal of identifying histologic-specific targets, we hypothesized that FGFR4 would be differentially expressed in RMS subtypes and drive different tumorigenic functions. We therefore examined the expression of FGFR4 in human eRMS and aRMS tumor tissue and the consequence of FGFR4 loss-of-function in cell growth in vitro and tumorigenesis in vivo. We found that while FGFR4 is expressed in both eRMS and aRMS, it is expressed at lower levels in eRMS and appears to stimulate cell proliferation, while it is expressed at higher levels in aRMS and appears to support cell survival. These distinct roles for FGFR4 provide further insight into the molecular origins of RMS subtypes and may be salient for design of RMS histology-specific treatment algorithms.

Results

FGFR4 is differentially expressed in human eRMS versus aRMS cell lines and tumor tissue

While transcriptional profiling has demonstrated upregulation of FGFR4 mRNA in human RMS cell lines and tumors (17, 19), we sought to examine FGFR4 expression at the protein level in aRMS versus eRMS tissue. Using immunoblot for FGFR4 protein expression, we examined a panel of 6 widely-used human RMS cell lines (3 eRMS and 3 aRMS). Since the biological behavior of fusion-negative aRMS is not clear, we studied only aRMS cell lines known to express the PAX3-FOXO1 fusion. Compared to non-transformed primary HSMM cells, FGFR4 protein was increased in all RMS cell lines examined (Fig.1A), with highest expression in those of aRMS histology. Depending upon the exposure, 2 or 3 discrete FGFR4 bands were evident, likely related to differential phosphorylation or glycosylation, which has been described for FGFR4 (21, 22). Because we later found by STR analysis that Rh3 and Rh28 cell lines likely derive from the same tumor (Supp.Table I and as noted in (23)), it was important to confirm FGFR4 expression patterns in a larger cohort of human clinical RMS tumor samples. Using tissue microarrays bearing cores of human RMS tumors and IHC staining for FGFR4, we found that FGFR4 was more highly expressed in aRMS compared to eRMS tissue (Fig.1B). Taken together, these data suggest that FGFR4 protein expression is overall increased in RMS tumor tissue, with differences in expression levels noted between the eRMS and aRMS subtypes.

Figure 1. FGFR4 protein expression is higher in human cell lines and tumors of alveolar (aRMS) histology.

(A) Immunoblot of endogenous FGFR4, FOXO1, and PAX3-FOXO1 protein in human RMS cell lines (eRMS cell lines RD, SMS-CTR, Rh36; aRMS cell lines Rh3, Rh28 and Rh30). Through STR cell line analysis (Materials and Methods and Supp.Table I), Rh3 was found to be identical to Rh28, indicating that it originated from the same tumor. FOXO1 and PAX3-FOXO1 were co-detected by immunoblot for FOXO1. Actin used as a loading control. HSMM, human skeletal muscle myoblasts. (B) Standard IHC was used to measure expression of endogenous FGFR4 in a human RMS tumor microarray. In total, 19 independent eRMS and 39 independent aRMS tumor were analyzed. Staining intensity was scored as described in Materials and Methods. *p=0.0092.

Loss-of-function of FGFR4 in eRMS cells inhibits cell proliferation in vitro and tumorigenesis in vivo

Because eRMS and aRMS are biologically distinct (2), we examined the role of FGFR4 in each separately. Using 2 independently targeting shRNAs (designated sh1 and sh2), we knocked down FGFR4 protein expression in RD eRMS cells via stable viral transduction (Fig.2A, left). Knockdown by either shRNA resulted in a decrease in intensity of FGFR4 bands, and in RD cells expressing shRNA1, the appearance of a faint intermediate-sized band (also seen in Fig.3D) and a shift in the migration of the lower band. The significance of these intermediate bands is not known. Following selection, polyclonal FGFR4-deficient RD cell populations emerged, but displayed an obvious decreased growth rate in culture. This decrease in growth was quantified by MTT assay (Fig.2B), and reproduced with manual counting (Supp.Fig.1). Morphologically, there was no increase in dead cells as assessed by trypan positivity compared to vector control (data not shown), suggesting that while cell viability was not affected, cell proliferation was slowed. To this end, we performed BrdU incorporation assays, which showed defective RD cell proliferation in the setting of FGFR4 suppression (Fig.2C, left). A second eRMS cell line, SMS-CTR, was also rendered FGFR4-deficient (Fig.2A, right), and similarly showed defects in BrdU incorporation (Fig.2C, right).

Figure 2. Genetic suppression of FGFR4 in human eRMS cell lines inhibits cell proliferation in vitro and in vivo.

(A) Immunoblot of RD or SMS-CTR eRMS cells stably expressing control or FGFR4 shRNA vectors. The antibody used for FGFR4 detects multiple bands that likely represent post-translational modifications (see Materials and Methods). (B) MTT (OD 540nm) assay of RD cells stably expressing control for FGFR4 shRNA vectors. (C) BrdU (OD 370nm) assays of RD or SMS-CTR cells stably expressing control or FGFR4 shRNA vectors. *p<0.001. (D) Murine xenograft assays of RD or SMS-CTR cells stably expressing control (○) or FGFR4 shRNAs (●). In the SMS-CTR experiment, one FGFR4 shRNA xenograft was mis-injected into the thoracic cavity, so excluded from analysis. Insets; immunoblots showing FGFR4 protein knockdown. Erk and tubulin used as loading controls.

Figure 3. Genetic suppression of FGFR4 in aRMS cell lines inhibits cell survival in vitro.

Immunoblot was used to measure FGFR4 expression in (A) HSMM cells expressing PAX3-FOXO1 (HSMM^PAX3-FOXO1) and (B) HSMM^PAX3-FOXO1 cells expressing empty vector or FGFR4 shRNA. (C) MTT (OD 540nm) assays of HSMM^PAX3-FOXO1 cells stably expressing control or FGFR4 shRNA vectors. *p<0.0001. (D) Immunoblot of RD, Rh30, and Rh28 (E) cells stably expressing vector or FGFR4 shRNAs and assayed for cleaved caspase 3. A bracket indicates the mobility of the cleaved caspase 3 bands. In (D) the entire cell population was analyzed, while in (E) an effort was made to focus on aRMS cells with highest knockdown of FGFR4, which was the non-adherent population, as described in Results. RT-PCR was used to assess expression of BCL2L1 in (F) HSMM cells stably expressing FGFR4 and (G) Rh30 or RD cells stably expressing vector or FGFR4 shRNAs. For FGFR4 RT-PCR, Rh30, 24 cycles, RD, 25 cycles. β-tubulin, Actin, Erk, and GAPDH used as loading controls. (H) Rh28 cells with no treatment (control), empty vector (pLKO.1), 10uM PD173074 for 72h, or FGFR4 shRNAs for 48h, were subject to flow cytometric measurement of propidium iodide (PI) and Annexin V staining. Percentages of cells in early apoptosis (Annexin V positive/PI negative, right lower quadrant) and late apoptosis (Annexin V positive/PI positive, right upper quadrant) are indicated. Vincristine treatment (1uM for 72h) was used as a positive control (57).

To determine whether FGFR4 suppression would block tumorigenesis in vivo, FGFR4-knockdown eRMS cells were evaluated as subcutaneous xenografts in immunodeficient mice. Compared to vector control, mice inoculated with RD or SMS-CTR cells expressing FGFR4 shRNA still developed tumors (Fig.2D), but at a longer latency and with smaller tumors at necropsy (not shown). Xenograft sections were probed for expression of Ki67, TUNEL and CD31 to quantitate changes in proliferation, apoptosis and vessel density, respectively (data not shown). In RD xenografts, while TUNEL and CD31 staining were unchanged, expression of FGFR4 shRNA was associated with a significant decrease in expression of Ki67, from 56.9% to 36.6% (p<0.0003). These data suggest that similar to the in vitro studies, in eRMS cells FGFR4 was stimulating cell proliferation, although we cannot rule out diminished clonogenicity or poor survival of cells injected into mice as additional reasons for delayed xenograft growth.

FGFR4 promotes cell survival in human aRMS cells in vitro

In our survey of FGFR4 protein expression in human RMS tissue, we observed higher expression in PAX3-FOXO1-positive aRMS cells and histologically-defined aRMS tumors compared to eRMS tissue (Fig.1), suggesting a correlation between PAX3-FOXO1 and FGFR4 expression. To investigate the relationship between PAX3-FOXO1 and FGFR4, we stably expressed PAX3-FOXO1 cDNA in HSMM cells, and found that endogenous FGFR4 protein expression was induced (Fig.3A), suggesting that FGFR4 is downstream from PAX3-FOXO1. To determine the functional significance of this increased expression, we stably knocked down FGFR4 in HSMM cells expressing PAX3-FOXO1 (Fig.3B), and found that cell viability was inhibited as measured by MTT (Fig.3C). Additional experiments verified a correlation between FGFR4 expression and cell viability (Supp.Fig.2).

We next investigated FGFR4 loss-of-function in human fusion-positive aRMS cell lines. As opposed to eRMS cells (described above), which survived selection into polyclonal, passageable populations after viral transduction with FGFR4 shRNAs, human aRMS cells (Rh28, and Rh30) selected for stable FGFR4 knockdown became non-adherent and trypan blue-positive (data not shown), suggesting that FGFR4 loss was not compatible with cell survival. Despite repeated attempts with the identical reagents used in the eRMS studies, passageable aRMS cells exhibiting stable FGFR4 knockdown could not be generated. This phenotype also precluded in vivo analysis in our murine xenograft system. Suspecting that FGFR4 was required for aRMS cell survival, we examined Rh30 and Rh28 cell populations (completing selection for stable shRNA expression but beginning to detach from tissue culture plate) for biochemical evidence of cell death. As expected, FGFR4 was appropriately suppressed by the shRNAs (Fig.3D, lanes 5–6), but caspase 3 cleavage was also evident (Fig.3D, lanes 5–6 and Fig.3E, lanes 2–3), suggesting induction of apoptosis. As a second measure of apoptosis, we subjected Rh28 aRMS cells expressing FGFR4 shRNAs to flow cytometric measurement of propidium iodide and Annexin V. We found that compared to untreated (control) or empty vector (pLKO.1)-expressing cells, which had a baseline of 10–13% of cells in early or late apoptosis, the percentage of FGFR4 shRNA-expressing cells in late apoptosis increased to 32–44% (Fig.3H). There was not a significant increase in early apoptosis, suggesting we had sampled the aRMS cells late in the apoptotic process.

FGFR4-associated cell survival has previously been noted in human breast cancer cells, thought to be mediated by the anti-apoptotic protein BCL2L1 (BCL-X(L)) (16, 24). Since BCL2L1 has also been linked to PAX3-FOXO1 expression (25, 26), we predicted that in our system FGFR4 was promoting aRMS survival through BCL2L1. To this end, we examined expression of BCL2L1 mRNA in HSMM cells ectopically expressing FGFR4, and observed an increase in BCL2L1 (Fig.3F). Ectopic PAX3-FOXO1 did not increase BCL2L1 levels (data not shown). Next, we suppressed FGFR4 in Rh30 aRMS cells, and found that correspondingly, BCL2L1 decreased (Fig.3G). Interestingly, in RD eRMS cells, no effect on BCL2L1 was observed with FGFR4 loss. Last, we generated aRMS cell lines stably expressing epitope-tagged-BCL2L1, followed by vector, FGFR4 shRNA1 or shRNA2. Although BCL2L1 was able to block caspase 3 cleavage in these cells, they could not re-adhere to the tissue culture plate and ultimately lost viability as determined by trypan blue staining (data not shown), suggesting that additional cell survival mechanisms were disrupted by FGFR4 suppression.

Given the loss of viability in FGFR4-deficient aRMS cells, we re-evaluated the eRMS cell lines in vitro and in vivo to assess for biochemical evidence of apoptosis. In vitro, stable expression of FGFR4 shRNAs in RD cells did not cause caspase 3 cleavage (Fig.3D, lanes 2–3). In vivo, early harvesting of RD eRMS xenografts (day 11 after inoculation) did not reveal a “missed” increase in apoptosis (Supp.Fig.3). That is, while the baseline level of TUNEL staining was mildly elevated relative to that seen in the long-term tumor assays (5–7% in Supp.Fig.3 versus 2–4%), there was no difference in TUNEL staining between vector and FGFR4 shRNA-expressing cells. To test the hypothesis that high levels of FGFR4, such as seen in aRMS cells, are associated with dependence on FGFR4 for cell survival, we ectopically over-expressed wild type FGFR4 in eRMS cells (data not shown). This over-expression did not render them sensitive to apoptosis under conditions of FGFR4 suppression, as assayed by caspase 3 cleavage, suggesting that level of expression alone is not a determinant of phenotypic response to FGFR4 suppression. In summary, while FGFR4 suppression diminished cell survival in aRMS cells, in eRMS cells FGFR4 suppression had no such effect, suggesting a persistent dichotomy in the function of FGFR4 in RMS subtypes.

Small molecule inhibition of FGFR4 in RMS tumorigenesis in vitro and in vivo

Since knockdown of FGFR4 in eRMS and aRMS cells using a genetic approach revealed different cellular responses, we next evaluated the effect of pharmacologic blockade of FGFR4 using a commercially available small molecule inhibitor of FGF receptors, PD173074. PD173074 was originally observed to inhibit FGFR1 (27), but also shown to antagonize FGFR4 autophosphorylation and activity (28). In MTT assays, PD173074 inhibited both RD eRMS and Rh28 aRMS cell growth in a dose-dependent fashion (Fig.4A,B). Similar results were observed in the SMS-CTR eRMS and Rh30 aRMS cell lines (Supp.Fig.4). Doses as low as 5µM and 3µM (not shown) inhibited FGFR4 autophosphorylation (Fig.4C) and signaling through the canonical ERK pathway, while there was no effect on AKT activation (Fig.4D). To determine whether FGFR blockade by a pharmacologic inhibitor would still yield dichotomous phenotypes in RMS subtypes, we evaluated aRMS and eRMS cells for caspase 3 cleavage in response to PD173074. In both Rh28 and Rh30 aRMS cells, PD173074 caused caspase 3 cleavage, while in RD eRMS cells, there was no evidence of such (Fig.5A). Apoptosis was validated by PI/Annexin V staining, showing that aRMS cells treated with PD173074 displayed both increased early and late apoptosis (Fig.3H). Interestingly, treatment with PD173074 caused downregulation of BCL2L1 in Rh30, but not RD cells (Fig.5B), again suggesting that FGFR4 is preferentially sustaining survival signals in aRMS cells.

Figure 4. Pharmacologic inhibition of FGFRs inhibits both eRMS and aRMS cell growth in vitro.

MTT (OD 540nm) assays of (A) RD and Rh28 cells treated with increasing doses of PD173074, and (B) RD and Rh28 cells treated with 10µM PD173074 or DMSO vehicle for 48h. *p≤0.0001. (C) Phosphotyrosine immunoblot of FGFR4 immunopurified from Rh28 cells treated with 5µM PD173074 or vehicle for 24h. (D) Immunoblot of Rh28 cells that were serum starved, then pulsed for 10 min with 100ng/mL FGF19, with or without pretreatment with 5µM PD173074 or vehicle. Blots were probed for total and phospho-Erk and Akt.

Figure 5. Pharmacologic inhibition of FGFRs exerts distinct biochemical effects in eRMS versus aRMS cells in vitro.

(A) Immunoblot of Rh28, Rh30, or RD cells treated with vehicle, 5, 10, or 20µM PD173074 for 48h. Blots were probed for caspase 3 cleavage. (B) RT-PCR was used to assess downregulation of BCL2L1 in Rh30 or RD cells treated with 10µM PD173074. Erk and GAPDH used as loading controls.

Last, we evaluated the effect of PD173074 on the growth of Rh28 aRMS subcutaneous tumor xenografts in vivo. At a dose of 25mg/kg, PD173074 caused tumor regression of established xenografts after 5–7 day (Supp.Fig.5A), and prevention of tumor growth after 8 days (Supp.Fig.5B). Xenografts from these studies displayed increased TUNEL staining (12.4% in drug-treated compared to 4.6% in vehicle-treated, p<0.0001), suggesting an apoptotic effect. However, this PD173074 dose caused weight loss leading to early study termination, raising the possibility that its anti-tumor effects were due to non-specific toxicity. Repeat trials were performed at 1, 5, 10, and 20mg/kg. 1–10mg/kg dosing was non-toxic and without anti-tumorigenic effect. While 20mg/kg had minimal toxicity, and anti-tumor effects remained (Supp.Fig.5C), the effects on apoptosis (2.20% TUNEL-positive in drug-treated compared to 1.17% in vehicle-treated, p=0.176, data not shown) lessened. We conclude that the FGFR inhibitor PD173074 has a narrow therapeutic window with relatively high toxicity, and that future studies on the impact of pharmacologic blockade of FGFR4 will require alternative approaches.

Discussion

Despite advances in supportive care and the evaluation of new agents, survival for children with high risk RMS remains ~30% at 5 years (2). Expression of the PAX3-FOXO1 gene fusion in patients with metastatic aRMS portends a particularly poor outcome (29). While it would be ideal to therapeutically interfere with tumor-specific proteins such as PAX3-FOXO1, transcription factors are currently chemically intractable. Identification of alternate RMS-specific proteins will be critical, and understanding their differential expression and regulation in RMS variants may permit refining of therapy.

The current work focuses on FGFR4, a member of the FGFR RTK family. In our primary cell-based modeling of aRMS, we noted that stable expression of PAX3-FOXO1 in human myoblasts induced elevated FGFR4 protein. Analysis of human aRMS cell lines and tumor tissue, described here and by others (19, 20, 30, 31), continues to support a relationship between acquisition of the PAX3-FOXO1 fusion and expression of FGFR4. One limitation of the current study is our retrospective discovery that the Rh3 and Rh28 cell lines likely derive from the same tumor, so that we in fact evaluated FGFR4 expression in two rather than three independent aRMS cell lines. However, the tissue microarray studies of human clinical tissue do validate the observed pattern of greater FGFR4 protein expression in aRMS compared to eRMS tumor tissue. Moreover, mouse development studies show that Pax3 modulates Fgfr expression via a 3’ cis regulatory element (32), so it is reasonable to suspect that PAX3-FOXO1 is similarly co-opting PAX3 targets (33, 34) such as FGFR4. Recent ChIP-Seq analysis confirmed the presence of a PAX3-FOXO1 binding site in a distal regulatory element of human FGFR4 (35).

On the other hand, the significance and mechanism of FGFR4 expression in eRMS is not clear. While we observe “upregulation” of FGFR4 in eRMS compared to primary human myoblasts, it is unknown what the baseline level of FGFR4 expression should be, since the precise cell of origin for eRMS is not identified. However, since wild type PAX3 is expressed in eRMS, and genetically upstream of FGFR4 (36–38), it could have some role in controlling FGFR4 expression in eRMS. Post-transcriptional and post-translational mechanisms (as suggested by the multiple bands seen on FGFR4 immunoblot and the shift in these bands in response to shRNA knockdown) likely further modulate FGFR4 levels and activity. Future studies are required to determine the significance and roles of these FGFR4 protein species.

While the mechanisms of expression of FGFR4 in the RMS subtypes are intriguing, even more so are the consequences of FGFR4 axis blockade. In our work, FGFR4 blockade in eRMS inhibited proliferation in vitro and tumorigenesis in vivo. On the other hand, FGFR4 blockade in aRMS induced cell death. BCL2L1-mediated reversal of caspase 3 induction did not completely rescue FGFR4 loss, raising the likelihood that additional survival pathways and processes such as anoikis might be involved. However, the phenotypic difference we note in response to FGFR4 suppression does suggest a dichotomy between the function of FGFR4 in eRMS compared to aRMS cells. Based on our aRMS modeling, we speculate that acquisition of the PAX3-FOXO1 fusion is an early event in tumorigenesis, providing survival signals such as increased FGFR4 and BCL2L1 expression, which permit accumulation of later mutations resulting in aRMS. BCL2L1 does not appear to be a direct transcriptional target of PAX3-FOXO1, which is consistent with recent microarray studies in which BCL2L1 expression is not changed in the presence of PAX3-FOXO1 (30). Since our results are based on a limited number of human cell lines, we cannot exclude the possibility of cell-line specific effects. For example, RD cells harbor an activating Ras mutation (39), which may render them resistant to apoptosis. The profound response to FGFR4 suppression in aRMS cells does suggest an “addiction” to FGFR4 (40), consistent with the recent recognition of FGFR4 as an oncogene (41).

The identification of FGFR4 kinase domain activating mutations in 7% of RMS tumors, which promote metastasis in xenotransplanted models, further supports FGFR4 as a novel and important mediator of RMS progression (20). Since none of the human RMS cell lines routinely used in cell culture harbors an FGFR4 activating mutation (20, 31), it is not possible to strictly correlate our observations with the phenotypes described for the FGFR4 mutations. However, in murine cell systems designed to assess the impact of the activating mutations, the FGFR4 activating mutations K535 and E550 do provide additional drive to proliferate (31) and resist apoptosis (20). Since the majority of RMS tumors do not harbor these mutations, it will be important to continue to understand how upregulated, wild-type FGFR4 contributes to RMS.

In a manner similar to that being elucidated for the RTK MET (42), development of FGFR4 inhibitors will be critical. While PD173074 itself is too toxic for therapeutic use, either because of broad FGFR inhibition or off-target effects, other small molecule inhibitors of FGFR4 are under development (43), as are high-affinity or neutralizing monoclonal antibodies to FGFR4 (44) and its cognate ligand FGF19 (45). It is not clear how important it will be for FGFR4 specificity, since although complete FGFR inhibition will likely be toxic, it is possible that FGFRs in addition to FGFR4 will be important in tumorigenesis. For example, FGFR1 has recently been noted upregulated in RMS (46).

In summary, in this work we show that FGFR4 protein expression is upregulated in human RMS cell lines relative to non-transformed human myoblasts, with highest expression in PAX3-FOXO1-positive aRMS cell lines. IHC of RMS tumor microarrays supports this finding in a larger cohort of patient tumor samples. PAX3-FOXO1 expression mediates upregulation of FGFR4, leading to increased myoblast viability. Loss-of-function of FGFR4, whether through genetic or pharmacologic means, appears to cause distinct effects in aRMS versus eRMS cell lines, suggesting dichotomous roles in these histologic subtypes. These results provide insight into the unique molecular origins of eRMS and aRMS, may be useful in the design of RMS histology-specific treatment algorithms, and support further study of FGFR4 as a rational drig target in RMS.

Materials and Methods

Generation of Cell Lines and Constructs

Early passage normal human skeletal muscle myoblasts (HSMM, Lonza) grown in defined media (SkGM-2) were stably infected with amphotrophic retroviruses to express pK1-PAX3-FOXO1-puro (a gift from Dr. Fred Barr, NCI) or empty vector, then selected in 0.25µg/ml puromycin (Sigma). Human RMS cell lines RD (47), SMS-CTR (48), Rh36 (49), Rh3 (50), Rh28 (51), and Rh30 (52) were gifts from Dr. Tim Triche in (Children’s Hospital of Los Angeles, CA) in 2005 and Dr. Brett Hall (Nationwide Childrens, Columbus, OH) in 2006 and cultured as described (9). Cell line identity was investigated in 2011 using STR analysis (Promega PowerPlex 1.2) performed by the Fragment Analysis Facility at the Johns Hopkins Genetic Resources Core Facility (Supp.Table I). FGFR4 shRNA sequences were designed de novo or obtained from the literature (53) and annealed shRNA oligos (Supp.Table II) ligated into the pSUPER-retro-GFP/neo or pLKO.1puro (Addgene 8453) plasmids. Wild-type FGFR4 was cloned from a JR cDNA library using high-fidelity PCR and subcloned into the EcoRI-SalI multicloning site of pBabe-puro and validated by sequencing. Cell lines expressing cDNAs or shRNAs were polyclonal.

Immunoprecipitation (IP) and Immunoblotting (IB)

Cells were lysed in Tris/RIPA or 20mM Tris pH7.5, 1% NP-40, 137mM NaCl, 10% glycerol, 0.1mM EDTA, 0.1mM EGTA (for IP) with standard protease and phosphatase inhibitors. Protein concentration was measured by the DC assay (Bio-Rad). FGFR4 IPs were performed with anti-FGFR4 (C-16) and protein G sepharose (Santa Cruz). For IB, the following antibodies were used: anti-FOXO1A (FKHR) F6928, FLAG F3165, tubulin T4026 (Sigma), actin 8462, FGFR4 C-16, ERK1 C-16 (Santa Cruz), phospho-ERK1/2 (Thr202/Tyr204), phospho-Akt (Ser273), Akt, caspase 3 (8G10, Cell Signaling), phospho-tyrosine (4G10, Millipore), FGFR4 (R&D Systems), HRP-labeled goat anti-mouse or anti-rabbit antibody (Invitrogen-Zymed). Several bands were detected when IB for FGFR4 using the sc-124 antibody; this was previously reported and is related to post-translational phosphorylation or glycosylation (21, 22).

MTT, BrdU and Apoptosis Assays

The MTT assay was used as a surrogate measure for cell number (and therefore growth over time) and BrdU incorporation was used as a surrogate measure of proliferation according to the procedures described in (54). The double staining propidium iodide/Annexin V apoptosis flow cytometric assay (eBioscience, #88–8007) was used per manufacturer’s protocol as a secondary measure of apoptosis.

RT-PCR

RT-PCR was performed as previously described (9). Primer sets for this work are shown in Supp.Table II.

Tumor Xenografts

Under institutional IACUC-approved protocols, and as performed (9, 55), 9–10 million cells/cell line were injected subcutaneously into the flanks of SCID/beige mice in triplicate or quadruplicate for genetic knockdown studies, or in replicates of 5 for FGFR drug inhibitor studies. In drug studies, mice were treated for short durations with PD173074 at 25mg/kg/day (or DMSO vehicle) intraperitoneally for 2 or 5 days/week as previously reported (28), or 20mg/kg/day for long durations (14 days, accomplished using an induction period of daily injections for 5 days, then every other day dosing for 9 days following). Mice were monitored biweekly. Tumor volume was estimated by external caliper measurements and calculated as ((width)²×length)/2. In some experiments, measurement by calipers was difficult due to the tumor’s flat, oblong shape; therefore decreases in estimated volume were verified by necropsy. Mice were sacrificed when tumors met IACUC-defined tumor burden or ill-health thresholds, and underwent necropsy with portions of tumor fixed in formalin or snap-frozen in liquid N₂.

Immunohistochemistry

Human RMS tissue microarrays (TMAs) were obtained from the Cooperative Human Tissue Network, which is funded by the National Cancer Institute; other investigators may have received specimens from the same subjects. The TMAs contain eRMS and aRMS cores; 134 individual eRMS or aRMS cores representing 19 eRMS and 39 aRMS tumors were scored. Staining for FGFR4 (FGFR4 C-16, Santa Cruz, (56)) was performed with the assistance of the University of Florida Department of Pathology, Immunology, and Lab Medicine Pathology Core. The TMA was scored by 2 independent observers who were blinded to the annotated histologic type. For each core, FGFR4 positivity was determined on the basis of plasma membrane and cytoplasmic staining, and a semi-quantitative score of the percentage of positive cells assigned as an integer (0 = no staining; 1 = <25% staining; 2 = 25–50% staining; 3 = >50% staining). In the case where there were replicate cores from a single tumor, an integer score was assigned based on the overall impression of the cores combined. A two-sample Wilcoxon test was used to compare the eRMS versus aRMS scores. Tumor xenograft IHC from murine studies was performed as described previously [47].

Statement of Translational Relevance.

Rhabdomyosarcomas (RMS) are the most common soft tissue sarcomas of childhood and adolescence. As with other sarcomas, there are several histologic variants of RMS, but the two predominant subtypes are embryonal RMS (eRMS) and the more aggressive alveolar RMS (aRMS). Accumulating evidence suggests that eRMS and aRMS are molecularly distinct entities that may require specific therapies. Therefore, understanding their common and unique cell signaling pathways will be critical for novel therapeutics development. Activating mutations in fibroblast growth factor receptor 4 (FGFR4) have recently been implicated in RMS metastasis. However, how FGFR4 functions in the RMS variants is not known. Herein, we provide evidence that while FGFR4 is expressed in both eRMS and aRMS, this receptor tyrosine kinase contributes to tumorigenesis by preferentially stimulating cell proliferation in eRMS, but promoting cell survival in aRMS. These data support the exploration of FGFR4 inhibitors as therapeutics for RMS.