Secreted Frizzled-Related Protein 3 (SFRP3) Is Required for Tumorigenesis of PAX3–FOXO1-Positive Alveolar Rhabdomyosarcoma

This article is featured in Highlights of This Issue, p. 4741

Rhabdomyosarcoma (RMS) is the most common soft tissue sarcoma of childhood and adolescence (1). Thought to be of skeletal muscle in origin, RMS tumors expresses myogenic markers but are unable to form myotubes or functional muscle units, suggesting a defect in differentiation (2, 3). Like other sarcomas, RMS is biologically diverse and presents as a variety of histologies, the most common being embryonal (eRMS; ∼60% of all cases) and alveolar (aRMS; ∼20% of all cases). Therapy for RMS is multimodal and includes chemotherapy, surgery, and radiation (2). Although eRMS 5-year survival rates have improved over the last 30 years, aRMS 5-year survival rates have remained stagnant at less than 50% (4). Understanding the unique molecular abnormalities for each subtype will be critical for designing new therapies. In the case of aRMS, understanding the tumorigenic influence of the associated signature PAX3–FOXO1 fusion gene will be paramount (5, 6).

The Wnt pathway is an evolutionarily conserved signaling network important for tissue development and homeostasis, directing cell proliferation, polarity, and fate (7) through noncanonical and canonical Wnt/β-catenin branches (8). Canonical Wnt signaling is active when extracellular Wnt binds to a Frizzled (FZD) receptor, recruiting Dishevelled (DVL) and inhibiting the Axin complex that ordinarily phosphorylates β-catenin (shown in Fig. 1D). Once no longer phosphorylated by glycogen synthase kinase 3 (GSK3) and casein kinase (CK1), β-catenin accumulates in the cytoplasm then translocates to the nucleus where it complexes with T cell factor/lymphoid enhancer factor (TCF/LEF) and activates Wnt target gene transcription (7).

Canonical Wnt signaling is inhibited in the absence of Wnt or in the presence of secreted Wnt pathway inhibitors. This permits β-catenin phosphorylation by CK1 and GSK3 and subsequent degradation by the proteasome. Secreted Wnt inhibitors include the dickkopf-related proteins (DKK) and the secreted frizzled-related proteins (SFRP). The SFRPs constitute the largest family of Wnt pathway inhibitors (9, 10), and are divided into two subgroups, with SFRP1, SFRP2, and SFRP5 forming one subgroup, and SFRP3 (also known as FRZB) and SFRP4 forming a second subgroup (9). Because SFRPs share homology to the FZD receptor, they were first identified as Wnt antagonists that bind and inhibit Wnt in the extracellular space (9). SFRPs may also exert their effects by interacting with each other and FZD receptors (9).

The Wnt pathway plays a critical role in skeletal myogenesis (11, 12). In myoblasts arising from satellite cells, it promotes myogenic commitment and lineage progression (3, 11). Wnt signaling also plays an important role in the regeneration of CD45⁺ stem cells in muscle, promoting hematopoietic (non-myogenic) CD45⁺ cells to enter a myogenic lineage. Thus, Wnt upregulation is critical for myogenic commitment and lineage progression toward fully differentiated muscle. When SFRP3 is added to regenerating muscle in vitro, the total number of myotubes formed is decreased (11). Exogenous SFRP3 also reduces the number of CD45⁺ cells recruited to the myogenic lineage, and the number of mononuclear desmin-expressing myoblasts in regenerating muscle (12). Thus, in normal skeletal muscle, SFRP3 exerts an inhibitory effect on myogenic differentiation.

Recently, the Wnt developmental pathway has been implicated in the pathogenesis of RMS. In a murine model of eRMS, Wnt signaling was found downregulated, and activation of Wnt signaling promoted differentiation (13). Similarly, activation of Wnt signaling in human eRMS and aRMS cell lines in vitro promoted differentiation (13–15). After identifying changes in the Wnt pathway in a microarray comparing the transcriptomes of human skeletal muscle myoblast (HSMM) cells with and without PAX3–FOXO1 expression, SFRP3 was noted to be upregulated in response to PAX3–FOXO1 expression and investigated further. We discovered that SFRP3 is necessary for aRMS cell growth in vitro and tumor xenograft growth in vivo. This suggests that SFRP3 is a promising target for future development of therapies for aRMS.

Microarray analysis of PAX3–FOXO1-expressing HSMMs was previously described (16); the associated dataset has been deposited in the Gene Expression Omnibus database (accession GSE40543). Microarray analysis of SFRP3 shRNA-expressing Rh28 cells was performed as previously described (16). The robust multiarray average (RMA)–normalized data were selected by significance analysis of microarrays (SAM) to identify genes that were consistently induced or repressed by both SFRP3 shRNAs. The associated dataset has been deposited in the Gene Expression Omnibus database (accession GSE67999).

Human RMS cell lines RD (17), Rh28 (18), and Rh30 (19) were gifts from Tim Triche (Children's Hospital of Los Angeles, Los Angeles, CA) in 2005 and Rh36 (20), SMS-CTR (21), and Rh3 (22) were gifts from Brett Hall (Columbus Children's Hospital, OH) in 2006 and cultured as previously described (23). Cell line authentication was performed in 2011 using STR analysis (Promega PowePlex 1.2) conducted by the Fragment Analysis Facility at the Johns Hopkins Genetic Resources Core Facility (Baltimore, MD) and in 2014 (Promega Powerplex 18D) conducted by the Duke University DNA analysis facility (Durham, NC). HSMM (Lonza) stably expressing an empty vector or PAX3–FOXO1 were previously described (16). SFRP3 shRNA sequences were obtained from the RNAi consortium (Broad Institute, Cambridge, MA) and annealed oligos ligated into pLKO.1 puro (Addgene 8453), Tet-pLKO-puro (Addgene 21915), and Tet-pLKO-neo (Addgene 21916) plasmids. HA*-SFRP3 was a gift from Ugo Borello (INSERM/Stem Cell and Brain Research Institute, Bron Cedex, Lyon, France), subcloned into the pBABE puro retroviral backbone, and stably expressed in Rh28 cells (24). Nucleotide mismatches between the shRNA sequences (two in SFRP3 sh3, three in SFRP3 sh5) and HA*-SFRP3, render it shRNA-resistant. All cell lines expressing an shRNA or cDNA are polyclonal. Using STR analysis, we retrospectively discovered that the Rh3 and Rh28 cell lines derive from the same patient, and therefore should be considered related (25).

Reverse transcription PCR was performed as previously described (23). Quantitative real-time PCR was performed as previously described (16). Primer sets for this work can be found in Supplementary Table S1.

Immunoblotting was performed as previously described, except 50 μg of lysate was used per sample (26). The following antibodies were used for immunoblotting: anti-p21 (Santa Cruz Biotechnology; #sc-6246), anti-HA (Roche; #11583816001), anti-cleaved caspase-3 (Cell Signaling Technology; #9661), and anti-actin (Sigma; #A5441).

The MTT assay, used as a surrogate for measuring cell growth, and BrdUrd assay, used to measure cell proliferation, were previously described (26). The cell-cycle analysis protocol was previously described and cells were analyzed by the Duke University Flow Cytometry core (27).

Rh28 cells stably expressing doxycyline-inducible SFRP3 shRNAs (or empty vector) were implanted subcutaneously into the flanks of SCID/beige mice. Mice were monitored twice weekly, and upon observing palpable tumors (Figs. 4 and 6) or a 150-mm³ tumor (Fig. 5), the drinking water was supplemented with 1 mg/mL doxycycline (Sigma-Aldrich) in 5% w/v sucrose or 5% w/v sucrose (control). Tumors were measured using calipers and tumor volume calculated as [((width × length)/2)³]/2. Mice were sacrificed at 23 to 24 days (Fig. 4), at 14 days (Fig. 5), or upon reaching an IACUC-defined maximum tumor burden or decline in health (Figs. 4, 5, and 6). This duration of therapy (23–24 days) was chosen as some of the control mice were reaching the maximum tumor burden. For Fig. 5, the later initiation and shorter duration of treatment were chosen to allow for the greatest observation of effects due to SFRP3 suppression before the emergence of a resistant cell population. Portions of tumors were preserved in RNAlater (Qiagen) for PCR or formalin-fixed for IHC. All animal studies were conducted in accordance with policies set forth by the Duke University IACUC.

Paraffin-embedded formalin-fixed xenograft tumor samples were sectioned and stained with H&E (Sigma) to assess morphology. Tumor sections were also labeled with Ki67 (Dako; #M7240) and TUNEL (Trevigen; #4810-30-K) to assess proliferation and apoptosis, respectively. Slides were evaluated by a pathologist (R.C. Bentley) with experience in the evaluation of pediatric sarcomas. Ki67 and TUNEL slides were photographed, and positive and negative cells were counted manually with the aid of cell counting software (ImageJ; NIH, Bethesda, Maryland). Three tumor sections per treatment condition were labeled with β-catenin (1:200; BD Biosciences; #610154). β-Catenin slides were photographed (Microscope: Leica Microsystems #DMLB, Camera: Leica Microsystems #DFC425, Software: Leica Microsystems #LAS V 3.7), and the staining intensity of individual cells was counted manually with the aid of cell counting software (ImageJ; NIH, Bethesda, Maryland).

For in vitro work, vincristine sulfate (Sigma) was dissolved in methanol, diluted in growth media, and added to culture media for final concentrations as indicated. For in vivo work, mice were treated with PBS or vincristine sulfate (1 mg/mL, Hospira) via intraperitoneal injection at a dose of 1 mg/kg weekly for 10 weeks.

Unless otherwise noted, data are presented as the mean and SE. Statistical analysis was performed using GraphPad Prism (GraphPad). One-way ANOVA, two-way ANOVA, log-rank (Mantel–Cox) test, and unpaired t test were used as appropriate. P values were considered significant under 0.05.

Prior work from our laboratory identified a role for the PAX3–FOXO1 fusion gene in permitting bypass of HSMM cells past the senescence checkpoint (28), thus priming cells for additional genetic changes that generate the aRMS phenotype (23). To identify genes that are downstream from or cooperate with PAX3–FOXO1 in this event, we performed gene expression analysis of HSMM cells ectopically expressing the PAX3–FOXO1 fusion cDNA as they transited through presenescent and postsenescent stages (16). These cell populations were compared with presenescent HSMM cells expressing an empty vector. Although approximately 1,000 genes were found to be upregulated or downregulated over 3-fold, one of the pathways that emerged from the analysis was the Wnt pathway.

Review of the gene list revealed changes in both Wnt receptors and ligands, including secreted Wnt activators and inhibitors. In total, seven Wnt pathway members were identified: the secreted inhibitors SFRP1, SFRP3, SFPR4, and DKK2; the canonical and noncanonical Wnt WNT5B (29, 30); and two noncanonical members WNT5A and FDZ5 (Fig. 1A). Other genes known to be altered in PAX3–FOXO1-positive aRMS, including CXCR4, FGFR4, MYOD, MEF2A, served as internal controls for the array (Fig. 1A). Expression of the Wnt family genes was verified by RT-PCR of cDNA generated from the original cell set analyzed in the microarray (Fig. 1B). Because these genes were found upregulated in the setting of ectopic expression of PAX3–FOXO1 in HSMM cells, their expression was next evaluated in three human aRMS cell lines (Rh3, Rh28, and Rh30) and compared with a panel of three human eRMS cell lines (RD, Rh36, and SMS-CTR). Although we found SFRP1, SFRP3, and SFRP4 to all be upregulated in aRMS cell lines, SFRP3 was most interesting because its pattern of expression in aRMS cell lines mirrored that in the HSMM cells expressing PAX3–FOXO1, and it was preferentially upregulated in aRMS compared with eRMS cells (Fig. 1C).

To probe the role of SFRP3 in aRMS, we used a loss-of-function approach and generated five lentivirally delivered shRNA sequences against human SFRP3. Upon stable infection into Rh28 and Rh30 cells, two of the shRNAs reproducibly suppressed SFRP3 expression as measured by RT-PCR (Fig. 2A and B). Monitoring of endogenous SFRP3 at the protein level was not feasible because commercially available antibodies did not detect SFRP3 protein in our system. As measured by the MTT assay, stable expression of the shRNAs in both Rh28 and Rh30 aRMS cells inhibited cell population growth (Fig. 2C and D).

To verify that the growth inhibitory actions of the shRNAs were not due to off-target effects, we sought to concurrently rescue the decreased cell growth with a gain-of-function approach using an ectopically expressed SFRP3. To this end, we took advantage of an shRNA-resistant murine SFRP3 construct with an HA (epitope) tag (24). Because SFRP3 is proteolytically cleaved at both the N and C-termini (31), the HA tag is positioned between the CRD and NTR domains and the construct termed HA*-SFRP3 (Fig. 2E). Using immunoblot to the HA epitope, we verified the ability of the cDNA to be expressed (Fig. 2F), then stably coexpressed it in Rh28 cells containing an empty vector or a doxycycline-inducible shRNA against SFRP3 (inducible system shown in Supplementary Fig. S1). As predicted, the decreased cell growth observed in the cells expressing SFRP3 shRNAs was largely rescued in the presence of HA*-SFRP3 (Fig. 2G). Taken together, these data suggest that SFRP3 has a progrowth role in human PAX3–FOXO1-positive aRMS cell lines.

To investigate the etiology of the decreased cell growth, we first investigated the possibility that SFRP3 suppression was interfering with cell proliferation. We found that in aRMS cells stably expressing SFRP3 shRNAs, bromodeoxyuridine (BrdUrd) incorporation declined (Fig. 3A and B), suggesting that loss of function of SFRP3 did indeed inhibit cell proliferation. Because a decrease in cell proliferation can reflect unrecognized differentiation, senescence, or cell-cycle arrest, all three processes were further investigated. No consistent or robust differences in the number of differentiated (as assessed by ability to form myotubes and express the myogenic marker MF20) or senescent cells were observed in response to the stable expression of SFRP3 shRNAs in Rh28 cells (data not shown). Similarly, we saw no evidence of Wnt pathway activation (which is required for terminal myogenic differentiation) in aRMS cells stably expressing SFRP3 shRNAs, although in response to Wnt3a-conditioned media and LiCl (a GSK3β inhibitor) they retain the ability to activate β-catenin signaling, similarly to seen previously (15). However, in Rh28 cells both shRNAs caused a 16% to 22% increase in the percentage of cells accumulated in the G₁-phase of the cell cycle (Fig. 3C). In Rh30 cells, the effect of the shRNAs was more modest, with a 3% to 5% increase in the percentage of cells at G₁ (Fig. 3D). To investigate mechanism, levels of the cell-cycle inhibitor p21 (32) were measured by semiquantitative RT-PCR and immunoblot. We found p21 upregulated at both the mRNA (Fig. 3E) and protein (Fig. 3F) levels. Together, these data suggest that inhibition of SFRP3 in aRMS cells inhibits cell proliferation via a G₁ cell-cycle arrest that is partially mediated through p21.

The decrease in cell proliferation observed in aRMS cells expressing SFRP3 shRNAs was much less than the profound decrease observed in overall cell growth. This suggested that another mechanism such as cell death might be contributing. Although the previous cell-cycle analysis (Fig. 3C and D) showed no change in the sub-G₁ peak (reflective of apoptosis), that analysis had only included adherent cells. And, because cells with SFRP3 shRNAs morphologically demonstrated an increase in the number of floating (nonadherent) cells, to capture possible missed sub-G₁ signal, we next assayed the entire (adherent and nonadherent) cell population by flow cytometry. Rh28 cells expressing SFRP3 shRNAs showed a large increase in the percentage of cells in sub-G₁ (Fig. 3G), while Rh30 cells again showed a more modest increase (Fig. 3H). For confirmation of an apoptotic effect of SFRP3 suppression, we measured cleaved caspase-3 by immunoblot. Although no increase in apoptosis as measured by cleaved caspase-3 was observed in the adherent cell population, a significant increase in apoptosis was observed in nonadherent Rh28 cells expressing either SFRP3 shRNAs (Fig. 3I). SFRP3 suppression also increased cleaved caspase-3 in Rh30 cells (Supplementary Fig. S2). In summary, suppression of SFRP3 causes decreased cell proliferation, a cell-cycle arrest at G₁, increased p21 expression, and an increase in apoptosis, suggesting that a combination of these mechanisms is responsible for the observed decreased aRMS cell growth.

We next assessed the role of SFRP3 in supporting tumorigenesis in vivo, using a conditional subcutaneous xenograft model of aRMS. Because SFRP3 suppression causes aRMS cells to grow poorly, we used a doxycyline-inducible shRNA system to control for effects on tumor implantation. This allowed for expression of the shRNAs after a tumor had formed. SFRP3 sh3 and sh5 were cloned into an inducible vector and tested for their ability to suppress SFRP3 mRNA expression following three days of 4 μg/mL doxycycline (Supplementary Fig. S1A and S1B). The ability of the inducible shRNA to reduce Rh28 and Rh30 cell growth was also verified (Supplementary Fig. S1C and S1D).

Xenograft tumors arising from Rh28 cells with a doxycycline-inducible shRNA against SFRP3 or an empty vector were generated, and mice received either control treatment (sucrose water) or doxycycline. Mice receiving the control treatment showed no difference in tumor growth, regardless of SFRP3 shRNA status, showing that in the absence of doxycycline the shRNAs were not leaky (Fig. 4A and C). Mice receiving doxycycline and bearing xenograft tumors from cells expressing empty vector similarly showed robust tumor growth (Fig. 4B and D). However, mice receiving doxycycline and bearing xenografts expressing SFRP3 shRNA3 or shRNA5 showed a profound inhibition of xenograft growth (Fig. 4B and D), with an associated decrease in tumor weight (Fig. 4E) To prove target knockdown, after necropsy levels of human SFRP3 mRNA were measured in the tumors by semiquantitative RT-PCR and analyzed using densitometry (Fig. 4F). Although there was some variation, tumors in which SFRP3 was suppressed using the shRNAs showed lower levels of SFRP3 than the controls.

To understand the mechanism of decreased tumor growth and identify changes caused by SFRP3 suppression, tumors were embedded, sectioned, and stained for H&E, Ki67, and TUNEL to assess tumor cell morphology, proliferation, and apoptosis, respectively. Although H&E showed no obvious changes (data not shown), there was a trend toward a reduction in Ki67 and an increase in apoptosis in tumors bearing SFRP3 shRNAs (Fig. 4G and H), mirroring the observed decrease in proliferation and increase in apoptosis observed in vitro.

Although we had not observed a consistent or robust increase in differentiation in RMS cells under conditions of SFRP3 suppression in vitro, we nevertheless examined the SQ tumor xenografts for evidence of such, and found that tumors expressing the doxycycline-inducible SFRP3 shRNA constructs under conditions of doxycycline treatment had evidence of myogenic differentiation. Specifically, the early differentiation marker MyoD (Fig. 4I) was upregulated in response to SFRP3 sh5, while the intermediate and late markers Myogenin (Fig. 4J) and Myf6 (Fig. 4K) were upregulated in response to both SFRP3 sh3 and sh5. Given the requirement for Wnt signaling during myogenic differentiation, we next examined the same tumors for evidence of Wnt pathway activation using nuclear β-catenin immunohistochemical staining and Axin2 expression (15, 33–35). Although we could not detect an increase in nuclear β-catenin staining in those tumors bearing the SFRP3 shRNAs under conditions of doxycycline treatment (data not shown), we did find Axin2 modestly upregulated (Fig. 4L).

Because these xenograft tumors were harvested over 20 days after initiation of doxycycline, potentially permitting tolerance to shRNA expression, we repeated the xenograft experiment but harvested the tumors after only 14 days doxycycline exposure in order to capture early events. As before, those tumors expressing the SFRP3 shRNAs under conditions of doxycycline showed decreased tumor volume and weight (Supplementary Fig. S3A–S3C), decreased Ki67 (Fig. 5A) and increased TUNEL staining (Fig. 5B). However, they also demonstrated an altered morphology [statistically significant fewer cells per field (5C)] and an increase in nuclear β-catenin staining (Fig. 5D). Taken together, these results suggest that in the in vivo setting, suppression of SFRP3 in aRMS does promote myogenic differentiation and Wnt pathway activation.

Given the findings in the xenograft tumors, to gain additional insight into the molecular consequences of SFRP3 suppression we used microarrays to analyze global gene expression after SFRP3 suppression by the two doxycycline-inducible SFRP3 shRNAs (Supplementary Fig. S1A and S1B) in triplicate. The RMA-normalized data were selected by SAM to identify genes that were consistently induced or repressed by both SFRP3 shRNAs. With the false discovery rate (FDR) of 1.13%, we identified 474 induced and 711 repressed probe-sets (Supplementary Table S2). We analyzed the Gene Ontology (GO) enrichment and found the induced genes were enriched in striated muscle development/differentiation (Supplementary Table S3). In contrast, the repressed genes were enriched in response to stimulus and cell-cycle/mitosis genes (Supplementary Table S3). These results indicate that SFRP3 suppression induced muscle differentiation and repressed proliferation, consistent with the in vivo data (Figs. 4G and 5A–D).

Given the role of SFRP3 in Wnt signaling, we also interrogated the microarray data for effects on Wnt-related genes. One hundred and nine genes were chosen for analysis (based on the literature and their inclusion in commercially available Wnt-related signaling arrays.) We were most interested in changes seen across both shRNAs, derived by zero-transformation (36), and as expected found downregulation of SFRP3 (FRZB; Fig. 5E; Supplementary Table S4) but also downregulation of Wnt pathway-repressing genes such as CTBP2 [a transcriptional repressor of TCF, similar to CTBP1 (37)] and NAV2 [which is downstream from APC (38)]. Interestingly, APC (which promotes the degradation of β-catenin) was also downregulated by the shRNAs, but just missed the threshold of inclusion. Conversely, we noted upregulation of genes, including CCND1 (cyclin D1) and SNAI2 (SLUG), both Wnt signaling target genes (39, 40) and WNT6, which is known to inhibit myoblast proliferation but induce myoblast elongation (41). Taken together, these results suggest that suppression of SFRP3, in part, activates β-catenin–Wnt signaling pathway as we had seen in the tumor xenografts.

Finally, as a preclinical assessment of SFRP3 suppression in aRMS, we investigated the role of genetic SFRP3 suppression in combination with chemotherapy, as combination therapy approaches have been successful in treating pediatric malignancies. We hypothesized that combining SFRP3 suppression with vincristine, a chemotherapy agent used for aRMS treatment since the 1970s (42), would be more effective at inhibiting aRMS cell growth than either alone. To this end, we first performed an MTT assay using Rh28 cells expressing an empty vector or SFRP3 sh3 with vincristine (0–10 nmol/L). Although both SFRP3 inhibition and vincristine inhibited cell growth alone, the combination produced the greatest inhibition (Fig. 6A), as evidenced by the shRNA–drug dose curve shifting to the left.

Although Rh28 cells appeared sensitized to vincristine in the presence of SFRP3 suppression in vitro, we next investigated the combination in our in vivo xenograft model of aRMS. Xenograft tumors were generated using Rh28 cells containing a doxycycline-inducible SFRP3 shRNA (SFRP3 sh3) or empty vector as done previously. Once the tumors were palpable, mice were randomized to treatment with doxycycline plus weekly vincristine (or vehicle) injections. As expected, we found that vincristine alone or SFRP3 suppression alone inhibited tumor growth (Fig. 6B and C). However, the combination of vincristine and SFRP3 suppression was more effective than either treatment alone (Fig. 6D and E), and unexpectedly caused tumor regression. [To verify that doxycycline was not interfering with the chemotherapeutic property of vincristine, a shadow control group of mice were treated with vincristine plus the vehicle for doxycycline and compared with those being treated with vincristine. There was no difference between the groups, suggesting that doxycycline was not interfering with vincristine (Supplementary Fig. S4).] In survival analysis, while SFRP3 suppression alone and vincristine alone prolonged survival to an IACUC-approved endpoint, the combination was most effective (Fig. 6F).

At necropsy, the tumors arising from the SFRP3 suppression alone or vincristine alone treatment groups showed no notable changes in the histology when stained with H&E. However, both groups showed decreased cell proliferation as measured by Ki67, with the change caused by SFRP3 suppression being significant (Fig. 6G). Both also demonstrated a trend toward an increase in apoptosis, as measured by TUNEL staining (Fig. 6H). In the SFRP3 suppression plus vincristine group, 5 of 7 mice had no detectable tumors. The remaining two mice had tiny remnant subcutaneous masses, the larger of which we embedded in paraffin and found on H&E to be composed mostly of fat cells. In summary, these data demonstrate the preclinical utility of SFRP3 suppression in combination with chemotherapy in vivo, suggesting that this approach should be explored further for its potential use in clinical care of patients with aRMS.

This work demonstrates a novel role for the secreted Wnt inhibitor SFRP3 in supporting aRMS tumorigenesis. SFRP3 is necessary for in vitro and in vivo aRMS cell growth. Although SFRP3 levels were increased in PAX3–FOXO1-expressing cells (human myoblasts stably expressing PAX3–FOXO1 and human PAX3–FOXO1-positive aRMS cell lines), SFRP3 most likely cooperates with PAX3–FOXO1 to support aRMS tumorigenesis, rather than being a direct downstream target. Indeed neither SFRP3, nor any other SFRP family members, appear in a genome-wide screen identifying PAX3–FOXO1 binding sites (43). This cooperation is not unusual, as other genetic changes supporting aRMS tumorigenesis, including inactivation of p16^INK4a, work in concert with PAX3–FOXO1 rather than being transcriptional targets (28). Future work will elucidate the connection between PAX3–FOXO1 and SFRP3, which may also provide insight into the regulation of SFRP3.

In both our in vitro and in vivo studies, suppression of SFRP3 in aRMS cells blocked cell growth both by reducing proliferation and inducing apoptosis. These phenotypes are likely connected, as cell-cycle arrest can lead to secondary apoptosis if specific checkpoints are activated (44). In adult renal cancer, SFRP3 supports cell survival, suggesting that it may also have a primary role in suppressing apoptosis (45). Future work examining the timing of the cell-cycle arrest and apoptosis in SFRP3-suppressed aRMS cells may reveal if they are related.

Previous studies have shown that SFRP3 can inhibit the differentiation-inducing effects of Wnt signaling on murine myoblasts, and that the addition of recombinant SFRP3 in regenerating murine skeletal muscle decreases nascent myotube formation, demonstrating that SFRP3 has a clear role in blocking differentiation (11). To this end, we examined both myogenic differentiation and Wnt pathway status following SFRP3 suppression in vitro and in vivo. SFRP3 suppression induced myogenic differentiation as observed using a microarray approach and examination of markers of differentiation in tumors, but did not alter terminal differentiation (MF-20 staining) in vitro. In addition, while both shRNAs promoted myogenic differentiation in the tumors, differences emerged as to which markers of differentiation were induced. Similarly, while we did not see evidence of Wnt pathway activation in the stable suppression system, the microarray analysis and levels of AXIN2 in the tumors did reflect this activation. Although we do not fully understand the mechanisms causing these inconsistencies, most likely a combination of effects are responsible, including differences in the microenvironment, differences between the shRNAs, and differences in the amplitude of SFRP3 suppression. Indeed, Wnt activation and repression has been shown to be finely tuned to SFRP levels (46). Future work will attempt to reconcile these differences.

Increased SFRP3 expression has been previously reported in metastatic renal carcinoma, required for cell growth and invasion (45). However, several other malignancies, including prostate cancer, breast cancer, and fibrosarcoma, demonstrate downregulation or inhibition of SFRPs (47, 48). Although SFRPs were originally identified as Wnt signaling inhibitors, recently SFRPs, such as SFRP1, have been shown to both potentiate and inhibit canonical Wnt signaling depending on a number of factors, including SFRP concentration (mentioned above), cell type, and Frizzled receptor expression (46). This suggests that secreted Wnt inhibitors may have more complex signaling functions than previously thought, and that each SFRP will need to be investigated in a cell and cancer-specific context.

Most interesting for the preclinical aspects of this work were the combined SFRP3 genetic suppression plus vincristine xenograft experiments, which showed frank tumor regression. The mechanism of this treatment effect is not clear. One possibility is that SFRP3 suppression worked in combination with the microtubule-poisoning properties of vincristine (49) to kill the aRMS cells within the xenograft. However, because the combination in vitro was not as profound as that seen in vivo, other mechanisms may be at work, including the possibility that vincristine was exerting effects on the tumor microenvironment, including inhibiting angiogenesis (50). Regardless of mechanism, while genetic inhibition of SFRP3 in combination with vincristine is currently not feasible for treatment of human aRMS patients, a monoclonal antibody against another secreted frizzled–related protein (SFRP2) has been developed, and is effective in a murine model of breast cancer (51). A similar monoclonal antibody against SFRP3 could be developed and tested for clinical efficacy.

The role of Wnt in RMS, like other developmental pathways Hedgehog, Hippo, and Notch, is only beginning to be understood (15, 16, 26, 52). Unlike in most adult epithelial cancers, where Wnt/β-catenin activity is upregulated and oncogenic, in a p53^−/−/c-fos^−/− mouse model of eRMS and in human eRMS cell lines, β-catenin activity is downregulated. Boosting β-catenin activity pharmacologically induces MyoD expression and promotes differentiation (13), and increasing Wnt signaling inhibits eRMS cell proliferation and induces differentiation (14). Three studies of human RMS tumor samples (including fusion-positive aRMS) showed low or absent nuclear β-catenin staining, with some cytoplasmic β-catenin staining (15, 53, 54). These data suggest that while β-catenin is present in RMS, Wnt pathway signaling is diminished. Importantly, this decrease is not due to inability to be activated, because treating the cells with Wnt3a ligand results in activation of the pathway (15). These cell biologic findings will need to be reconciled with recent genomic data identifying activating β-catenin mutations in 3% to 7% of human eRMS samples, with 20% of eRMS samples demonstrating nuclear (active) β-catenin staining (55). No β-catenin mutations were found in PAX3–FOXO1-positive aRMS samples, consistent with the observation that aRMS harbors fewer mutations than eRMS (55–57). Taken together, these data indicate a possibility for active β-catenin contributing to a subset of eRMS tumors, but show no evidence of a contribution to aRMS tumors.

In summary, we demonstrate that SFRP3 is upregulated in PAX3–FOXO1-positive human myoblasts and human aRMS cells. Genetic suppression of SFRP3 reduced aRMS growth in vitro and in vivo, suggesting that SFRP3 is required for aRMS tumorigenesis. aRMS cells in which SFRP3 has been suppressed are more sensitive to vincristine and this combination results in tumor regression in a xenograft model of aRMS. These data suggest that SFRPs and suppression of the Wnt pathway play an important role in aRMS, provide insight into the molecular mechanisms underlying aRMS, and may inform new strategies for aRMS treatment.

No potential conflicts of interest were disclosed.

Conception and design: J.J.G. Kephart, L.E.S. Crose, C.M. Linardic

Development of methodology: J.J.G. Kephart, A.R. Hinson, C.M. Linardic

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J.J.G. Kephart, R.G.J. Tiller, K.K. Slemmons, A.R. Hinson, R.C. Bentley, C.M. Linardic

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J.J.G. Kephart, R.G.J. Tiller, K.K. Slemmons, P.-H. Chen, J.-T.A. Chi, C.M. Linardic

Writing, review, and/or revision of the manuscript: J.J.G. Kephart, A.R. Hinson, J.-T.A. Chi, C.M. Linardic

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J.J.G. Kephart, C.M. Linardic

Study supervision: C.M. Linardic

The authors thank the laboratories of Drs. Christopher Counter, David Kirsch, and Dan Wechsler (Duke University School of Medicine) for helpful discussions, and the laboratories of Drs. Oren Becher (Duke University School of Medicine) and Ben Alman (The Hospital for Sick Children, Toronto, Ontario, Canada and Duke University School of Medicine), for reagents and advice on β-catenin IHC. The authors also thank the Duke Microarray Core facility (a Duke National Cancer Institute and a Duke Genomic and Computational Biology shared resource facility) for their technical support, microarray data management, and feedback on the generation of the microarray data reported in this article.

This research was supported by NIH grants R01 CA122706 (to C.M. Linardic), 5T32GM007184-35 (to J.J.G. Kephart), T32 CA059365 (to K.K. Slemmons and L.E.S. Crose), and the V Foundation for Cancer Research, St. Baldrick's Foundation, Sarcoma Foundation of America, and A Collaborative Pediatric Cancer Research Award from Bear Necessities Pediatric Cancer Foundation and Rally Foundation (to C.M. Linardic).

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.