Histone Deacetylase Inhibitors Antagonize Distinct Pathways to Suppress Tumorigenesis of Embryonal Rhabdomyosarcoma

Embryonal rhabdomyosarcoma (ERMS) is the most common soft tissue cancer in children. The prognosis of patients with relapsed or metastatic disease remains poor. ERMS genomes show few recurrent mutations, suggesting that other molecular mechanisms such as epigenetic regulation might play a major role in driving ERMS tumor biology. In this study, we have demonstrated the diverse roles of histone deacetylases (HDACs) in the pathogenesis of ERMS by characterizing effects of HDAC inhibitors, trichostatin A (TSA) and suberoylanilide hydroxamic acid (SAHA; also known as vorinostat) in vitro and in vivo. TSA and SAHA suppress ERMS tumor growth and progression by inducing myogenic differentiation as well as reducing the self-renewal and migratory capacity of ERMS cells. Differential expression profiling and pathway analysis revealed downregulation of key oncogenic pathways upon HDAC inhibitor treatment. By gain-of-function, loss-of-function, and chromatin immunoprecipitation (ChIP) studies, we show that Notch1- and EphrinB1-mediated pathways are regulated by HDACs to inhibit differentiation and enhance migratory capacity of ERMS cells, respectively. Our study demonstrates that aberrant HDAC activity plays a major role in ERMS pathogenesis. Druggable targets in the molecular pathways affected by HDAC inhibitors represent novel therapeutic options for ERMS patients.


Introduction
Rhabdomyosarcoma (RMS) is the most common soft tissue malignancy in the pediatric population and pathologically recapitulates many of the phenotypic and biological features of embryonic skeletal muscle.Based on histologic characterization, RMS falls into two major groups in children: embryonal (ERMS) and alveolar (ARMS).Embryonal rhabdomyosarcoma (ERMS) is the most common subtype, accounting for ~60% of childhood cases.In contrast to ARMS, which is genetically characterized by fusion between PAX3 or PAX7 and FOXO1 genes in the majority of cases, ERMS is characterized by complex genetic changes, involving various chromosomal gains and losses [1][2][3].However, RAS mutations are present in at least 25% of ERMS tumors [4][5][6][7].The prognosis for patients with relapsed or metastatic ERMS is dismal, with at least 50% of patients succumbing to the disease, underscoring the need for more effective treatment in these cases.A recent comprehensive genomic study by Shern et al. demonstrates relatively low mutational frequency in rhabdomyosarcoma, with 33 recurrently mutated genes identified, with a higher number of oncogenic mutations found in ERMS compared to ARMS [7].The findings suggest that other molecular mechanisms, such as epigenetic regulation of driver genes, might contribute to RMS tumorigenesis.Interestingly, the same study shows that about 7.4% of fusion-negative RMS harbor mutations in BCOR, a transcriptional repressor that has been shown to interact with class I and II histone deacetylases, implicating the role of histone deacetylases in the pathogenesis of RMS.
ERMS is pathologically characterized by arrested myogenic differentiation and uncontrolled proliferation.We have previously completed a large-scale chemical genetic screen (testing ~40,000 compounds) to identify drugs that induce terminal myogenic differentiation of human ERMS [8].One lead compound identified by our chemical screen, trichostatin A (TSA), a pan-histone deacetylase (HDAC) inhibitor, suppressed tumor growth as well as induced myogenic differentiation of tumor cells in vitro.Histone deacetylases are one of the many classes of enzyme that participate in the epigenetic or post-translational modifications of the genome.There has been increasing interest in the therapeutic use of HDAC inhibitors in some cancers due to their potent effect on proliferating malignant cells in comparison to nonneoplastic counterparts.A large number of HDAC inhibitors are in various phases of clinical trials [9,10].Hematologic malignancies in particular have responded well to treatment by HDAC inhibitors, such that vorinostat and romidepsin have been approved by the Federal Drug Administration (FDA) for treatment of refractory and advanced cutaneous T-cell lymphoma (CTCL) [11,12].There are only a few reports of tumor suppressive effects secondary to HDAC inhibition on several rhabdomyosarcoma cell lines in vitro [13,14].A study in a mouse model of ARMS demonstrates a differential response to entinostat treatment depending on the myogenic lineage of origin of tumor cells [15].The role of histone deacetylase activity in ERMS tumorigenesis and the fundamental mechanisms by which HDAC inhibitors exert their antitumor effects are largely unknown.
In this study, we have revealed diverse roles of HDACs in ERMS tumorigenesis by characterizing the effects of two pan-HDAC inhibitors, trichostatin A (TSA) and suberoylanilide hydroxamic acid (SAHA; also known as vorinostat) in vitro and in vivo using human RMS cell lines and a zebrafish model of ERMS.In addition to suppressing tumor growth by inducing tumor cell differentiation and reducing self-renewal capacity, TSA and SAHA also inhibit the migratory capacity of ERMS tumor cells.By expression profiling and functional analysis of differentially expressed target genes, we show that HDAC inhibitors exert their anti-tumor effects by antagonizing distinct molecular pathways.Using a chemical genetic approach, our study demonstrates that aberrant HDAC activity is a major driver of ERMS pathogenesis.Key pathways targeted by HDAC inhibitors represent potential options for novel targeted therapies in ERMS.

HDAC inhibitors suppress ERMS growth by inducing myogenic differentiation and reduce self-renewal and migratory capacity in vitro
The pan-HDAC inhibitor, trichostatin A (TSA), was previously identified in a high-throughput screen as a lead compound with the capacity to induce myogenic differentiation of human ERMS cells [8].To test the effects of HDAC inhibitors on tumor growth and differentiation in vitro, two pan-HDAC inhibitors, TSA and suberoylanilide hydroxamic acid (SAHA), were tested in a panel of ERMS cell lines (RD, 381T and SMS-CTR) and ARMS cell lines (Rh3, Rh5 and Rh30).TSA or SAHA treatment resulted in hyperacetylation of histones including acetylhistone H3 (Lys9), acetyl-histone H3 (Lys27), acetyl-histone H4 (Lys5), and acetyl-histone H4 (Lys8) (results for RD cells shown in To assess the effect of TSA and SAHA on differentiation of RMS cells, immunofluorescence against myosin heavy chain (MF20), a late myogenic differentiation marker, was performed on the same panel of RMS cell lines (RD, 381T, SMS-CTR, Rh3, Rh5 and Rh30).In contrast to the ARMS cell lines (Rh3, Rh5 and Rh30), ERMS cell lines (RD, 381T and SMS-CTR) treated with TSA or SAHA showed at least 2-fold increase in the percentage of MF20-positive cells (Fig 1C -1E), suggesting that the effect of TSA and SAHA on myogenic differentiation is specific to the ERMS subtype.Chromatin immunoprecipitation (ChIP) analysis revealed 2 to 4-fold differential enrichment of acetyl-histone H3 (Lys9) binding on promoters of a number of myogenic genes such as MYOD1, MYOG and MYH4 in ERMS cells treated with TSA compared to the cells treated with vehicle DMSO (Fig 1F), suggesting that these genes are epigenetically regulated by HDAC(s) to drive myogenic differentiation in ERMS.We then performed quantitative RT-PCR to assess the mRNA expression of myogenic genes in ERMS cells treated with TSA or SAHA.At 24 hrs post-TSA treatment, there was variable upregulation of MYOD1 and MYOG and significant upregulation of myosin heavy chain (MHC) in ERMS cells (S2A-S2C Fig) .By contrast, expression of MYOD1 and MYOG was already downregulated in ERMS cells 24 hours post-SAHA treatment, and there was significant upregulation of MHC in ERMS cells (S2D-S2F Fig) .As muscle cells undergo differentiation, MYOD1 and MYOG expression levels are downregulated while late-differentiation myogenic genes, such as myosin heavy chains, are upregulated.Overall, TSA and SAHA induced distinct kinetic changes in the expression levels of myogenic factors in ERMS cells, and the expression profiles reflect a late-differentiation myogenic program by 24 hours post-treatment.
The sphere assay has previously been demonstrated as a powerful surrogate assay to assess self-renewal potential in a variety of tumor types including ERMS [8,16].Here we showed that SAHA or TSA treatment resulted in at least 2-fold reduction in sphere formation of human ERMS cells ( Pan-HDAC inhibitors have been shown to elicit a variety of anti-tumor effects in other cancer types, including a reduction in migratory and invasive behaviors [9,10].Wound closure scratch and transwell assays were used to assess the effect of TSA and SAHA on the migratory capacity of human ERMS cells.In contrast to DMSO treatment, ERMS cells treated with TSA or SAHA showed a significant reduction in the percentage of gap closure in scratch assays (Fig 2A -2G) as well as decreased numbers of migrating cells in the transwell assays (Fig 2H -2K).The reduction in the amount of migratory ERMS cells upon TSA or SAHA treatment was not due to increased apoptosis (S1E and S1F Fig) .Taken together, pan-HDAC inhibitor treatment reduced the migratory capacity of ERMS cells, implicating the role of HDACs in modulating the invasive and metastatic behavior of ERMS cells.

Pan-HDAC inhibitors induce tumor cell differentiation and reduce selfrenewal capacity of ERMS in vivo
As treatment of human ERMS with TSA and SAHA resulted in reduced tumor cell growth, increased tumor cell differentiation and reduced self-renewal capacity in vitro, we then assessed whether these effects could be recapitulated in vivo using a zebrafish transgenic ERMS model [17,18].In this transgenic model, primary ERMS was induced in the skeletal muscle by overexpression of the constitutively active human KRASG12D driven by the rag2 promoter in wildtype zebrafish [18].Zebrafish ERMS tumor cell subpopulations can be labeled with fluorescent reporters based on their differentiation status to allow for cellular and molecular characterization of tumor cell subpopulations [17].Specifically, myf5:GFP labels tumor-propagating cells and early myoblasts, and mylz2:mCherry labels late-differentiating myoblasts.The myf5:GFP + / mylz2:mCherry − cell population enriches for the tumor propagating cell subpopulation and is the only cell population in this ERMS model to have self-renewing capacity.We first showed by Western blot analysis that TSA or SAHA treatment resulted in increased acetylation of histones in zebrafish ERMS tumors, indicating an altered histone acetylation status in ERMS in vivo (Fig 3A).ERMS tumor-bearing fish treated with TSA or SAHA showed significantly reduced tumor growth compared to a DMSO-treated cohort (n = 7 for DMSO, n = 10 for SAHA, n = 11 for TSA; Fig 3B and 3C).By quantitative Fluorescent Activated Cell Sorting (FACS) analysis, tumor cells treated with TSA or SAHA also showed alterations in differentiation, with depletion of myf5:GFP + /mylz2:mCherry − tumor cell subpopulation and expansion of the late-differentiating mylz2:mCherry + /myf5:GFP − tumor cell subpopulation (results for SAHA treatment shown in Fig 3D).ERMS cells isolated from tumor-bearing fish treated with TSA or SAHA also showed concomitant upregulation of the myogenic regulatory factors As HDAC inhibitor treatment resulted in depletion of the myf5:GFP + /mylz2:mCherry − tumor cell subpopulation, which contains the tumor propagating cells with the capacity for selfrenewal [17,18], we next assessed the effect of HDAC inhibitors on the self-renewal capacity of ERMS in vivo by limiting dilution experiments.Either unsorted bulk tumor cells or myf5: GFP + /mylz2:mCherry − sorted cells were transplanted at limiting dilutions into syngeneic juvenile fish hosts and subsequently treated with DMSO (vehicle) or HDAC inhibitor (TSA or SAHA) for 5 days prior to the monitoring period for tumor engraftment.Compared to the DMSO-treated cohort, TSA or SAHA-treated tumor-bearing fish showed approximately 3 to 10-fold reduction in the self-renewal frequency (p < 0.05, limiting dilution experiments using four independent primary tumors are shown in Tables 1-3).Together, our results from characterizing the effects of HDAC inhibitors in human ERMS cell lines and in the zebrafish ERMS model suggest an important role of HDACs in ERMS tumorigenesis by regulation of the balance between differentiation and self-renewal of tumor cells.

NOTCH1 loss-of-function phenocopies HDAC inhibitor-induced effects on differentiation of ERMS
To identify candidate genes that are differentially regulated in HDAC inhibitor-treated ERMS, a gene expression profiling study using the Affymetrix Human Gene 2.0 microarray platform was performed on RD and 381T cells treated with TSA or DMSO, and Ingenuity Pathway Analysis was performed on differentially expressed genes.Instead of global change in gene expression, TSA-treatment resulted in a restricted set of differentially-expressed genes.We focused on validating candidate genes with potential signaling or oncogenic function (Table 4).Among the top upregulated gene candidates, we have validated MPP4 (annotated function of extracellular matrix remodeling function), CPA4 (extracellular peptide processing), SEMA3C (migration and morphogenesis) as well as DNER, PTPRN and TNFRSF12A, genes involved in cytokine-mediated signaling activity.Among top genes/pathways downregulated upon TSA treatment, FGFR1, SMO (receptor for the Hedgehog pathway) and Notch1 have been previously been implicated in promoting tumorigenesis of ERMS [3,7,[19][20][21].
To determine which of the downregulated pathways upon TSA treatment functions downstream of HDAC activity to exert oncogenic or tumor suppressive effects in ERMS, we utilized pathway-specific small molecule inhibitors.Treatment of human ERMS cells with inhibitors specific for the Notch, Hedgehog and FGF-mediated pathways resulted in significant suppression of tumor growth in vitro (S3A-S3C Fig) .However, only treatment with the inhibitor of  the Notch pathway, GSI-IX (DAPT), resulted in significant induction of myogenic differentiation of RD and 381T cells, with at least 5-fold increase compared to treatment with vehicle control (results for treatment in RD cells shown in Fig 4A -4C).To validate the role of the Notch pathway in inhibiting myogenic differentiation, we showed that knockdown of NOTCH1 by gene-specific shRNA resulted in significant myogenic differentiation of ERMS cells (S3D and S3E Fig) .Upon TSA and SAHA treatment, NOTCH1 mRNA and protein expression levels were reduced (Table 4 and Fig

EFNB1 is regulated by HDACs to modulate migratory behavior of ERMS cells
Expression of Ephrin-A3 (EFNA3) and Ephrin-B1 (EFNB1) were downregulated in ERMS cells upon TSA treatment (Table 4).The Eph family of receptor tyrosine kinases and their membrane-associated ephrin ligands are known for their importance in a variety of developmental processes and in tissue morphogenesis [22][23][24].In many cancer types, the Eph receptors and ephrins have also been shown to promote tumor cell invasion and metastasis [25][26][27].To determine if EFNA3 and EFNB1 are essential for regulating the migratory capacity of ERMS cells, we performed loss-of-function studies using two independent siRNAs targeted against EFNA3 and EFNB1.We first showed that .Taken together, these data suggest that HDACs regulate EFNB1 to modulate the migratory capacity of ERMS cells.Results from our functional assays in vitro also suggest that Ephrin-B1 does not play a role in other HDAC inhibitorinduced effects such as abnormal cell cycle progression or differentiation.

EFNB1 is differentially expressed in ERMS in comparison to ARMS patients
In our study, EFNB1 regulated the migratory behavior of ERMS cells, suggesting its potential role in affecting the invasive potential and therefore possibly the clinical prognosis of ERMS.EFNB1 expression is differentially increased in ERMS patients in comparison to ARMS patients, EFNB1 likely serves a unique role in the pathogenesis of ERMS.

Discussion
By characterizing the effects of two pan-HDAC inhibitors (trichostatin A and SAHA) in vitro and in vivo, this study has revealed diverse and essential roles of HDACs in driving the tumorigenesis of ERMS.Through differential expression profiling studies and pathway analysis, we have demonstrated downregulation of key oncogenic pathways including Hedgehog, FGF and Notch pathways in ERMS cells upon treatment of HDAC inhibitors.Using gain-of-function and loss-of-function studies, we further showed that HDACs regulate the differentiation and self-renewal status of ERMS through the Notch1-mediated pathway and migratory capacity through EFNB1-mediated pathway.Our findings suggest that HDAC function represents a major mechanism in driving ERMS tumor growth and progression.Cancer cells show uncontrolled proliferation and fail to undergo tissue-specific differentiation.In differentiation therapy, tumor cells, including those that give rise to treatment resistance and disease recurrence, are induced to undergo terminal differentiation and thereby cease to proliferate.The success of differentiation therapy was first demonstrated in acute promyelocytic leukemia (APL), which was nearly universally lethal before the introduction of alltrans-retinoic acid-induced differentiation therapy.The cure rates now approach approximately 80% [30].Differentiation therapy in solid tumors has shown clinical promise in the past decade [31][32][33].Several pre-clinical in vitro and in vivo studies have demonstrated the efficacy of various agents in inducing tumor cell differentiation in sarcomas such as liposarcoma and osteosarcoma [34][35][36], suggesting that there is a therapeutic potential for differentiation therapy in sarcomas.
ERMS is pathologically characterized by an arrest in myogenic differentiation.In contrast to normal myogenesis, ERMS cells express myogenic regulatory factors such as MYOD1 and MYF5, but they fail to undergo terminal differentiation [37].MYOD1 has been shown in ERMS cells to retain the ability to bind DNA, but is defective in activating myogenic target genes [38].In this study, TSA and SAHA treatment induced myogenic differentiation of ERMS cells and dynamically regulated the expression levels of myogenic regulatory factors.The promoters of myogenic regulatory factors such as MYOD1 and MYOG also showed differential histone acetylation patterns after HDAC inhibitor treatment, suggesting that aberrant epigenetic regulation by HDACs likely contributes to myogenic differentiation arrest in ERMS.In addition, HDAC inhibitor treatment reduced the self-renewal capacity of ERMS cells, suggesting that HDACs modulate the balance between differentiation and self-renewal of ERMS tumor cells.By contrast, TSA and SAHA treatment did not significantly alter the myogenic differentiation status of ARMS cells, suggesting that the myogenic arrest in ARMS is likely regulated by a different molecular mechanism.However, based on our in vitro (MF20 IF) and in vivo (quantitative FACS) results, TSA or SAHA treatment induced myogenic differentiation in a subset of ERMS cells, suggesting that other suppressive events need to be abrogated for complete induction of terminal myogenic differentiation in ERMS cells.
We subsequently showed by expression profiling that multiple oncogenic pathways such as Notch1, Hedgehog, and FGF-mediated pathways were downregulated in ERMS cells upon TSA treatment.The Hedgehog pathway has previously been shown to regulate self-renewal and differentiation of RMS.Work by Satheesa et al. has shown that inhibition of the Hedgehog signaling pathway either by overexpressing SUFU (Suppressor of Fused) or knocking down Smoothened (SMO) decreases self-renewal capacity and increases myogenic differentiation of ERMS cells [39].In a mouse model of RMS with Patched1 (Ptch1) haploinsufficency, Gli1 and Gli2, the Hedgehog-regulated transcriptional factors, have been shown to suppress myogenesis of RMS by repressing the capacity of MyoD to activate transcription [40].In addition, calcitriol, the active form of vitamin D3, inhibits Hedgehog signaling to induce differentiation of RMS derived from mice with Ptch1 mutations [41].In our study, inhibition of Hedgehog signaling by cyclopamine (up to 5 μM) did not induce significant myogenic differentiation of ERMS cells, suggesting variable treatment response among human ERMS cell lines.By contrast, inhibition of NOTCH1 in our study phenocopied the differentiation effect of TSA and SAHA, and overexpression of NICD abrogated the myogenic effect of TSA and SAHA.Overall, our results indicate that Notch represents a major pathway regulated by HDACs to modulate the differentiation status of ERMS cells.The studies by Raimondi et al. and De Salvo et al., have demonstrated that the inhibition of NOTCH3 signaling induces myogenic differentiation of both ARMS and ERMS cells by increasing p38 phosphorylation and p21 Cip1 level [42], and the activated form of Notch3 increases RMS cell proliferation [43].The findings indicate that the Notch gene family plays an important role in driving myogenic differentiation arrest in ERMS.Inhibition of HDAC function and downstream signaling pathways, including Notch1 and Notch3, promises an alternative therapeutic option (i.e.differentiation therapy) in the treatment of ERMS patients.
Our study also uncovered a novel role of Ephrin-B1 in regulating the migratory behavior of ERMS cells.The ephrin ligands and the Eph receptors are a cell surface-bound family of tyrosine kinases that have been shown to play an essential role in a variety of developmental and tissue morphogenetic events [24].In particular, their complex roles in axon guidance are well characterized [23].A subset of ephrins and Eph receptors has been shown to either inhibit or promote tumor growth and progression in several cancer types [44,45].In ERMS, high expression of PAX7 upregulates EPHA3 and EFNA1 to promote migration and invasiveness of tumor cells [46].By contrast, suppression of EPHA3 expression increases motility and migration in a subset of ERMS cell lines (TE671 and RD), indicating a context-dependent role of EPHA3 in either promoting or suppressing migratory behavior of ERMS cells.[47].In a murine model of ARMS, Ephrin-B2 has dual signaling roles with its cognate ligand Ephrin-B2 to promote apoptosis and with PDGFRβ to promote proliferation [48].In our study, we have demonstrated the important role of EFNB1 in regulating the migratory behavior of ERMS cells.EFNB1 expression is downregulated in ERMS cells upon treatment with HDAC inhibitors, and EFNB1 knockdown significantly reduces the migratory capacity of ERMS cells.Our findings suggest that aberrant regulation of EFNB1 activity by HDACs likely contributes to the invasive and metastatic potential of ERMS.Analysis of the RMS patients in the study by Davicioni et al. (2009) and Williamson et al. (2010) has also shown that EFNB1 is differentially upregulated in ERMS in comparison to ARMS.Our study as well as previous studies have demonstrated EFNB1 overexpression in a subset of ERMS cell lines and primary samples [49].Taken together, these results indicate that Ephrin-B1 expression likely plays an important role in promoting ERMS tumor progression by modulating the invasive behavior of tumor cells.
The current study identified distinct sets of upregulated and downregulated pathways in ERMS cells upon TSA treatment.There is inconsistent correlation between histone H3 (Lys9) hyperacetylation and gene transcription, as evidenced by the downregulation of NOTCH1 and EFNB1 genes after TSA treatment.Our findings suggest that HDAC inhibitors can alter the epigenetic state of the chromatin by hyperacetylation, but do not necessarily increase gene transcription.Similar observations have been made in other biological systems.For example, the study by Lopez-Atalaya et al. in adult mouse hippocampus demonstrated that TSA-induced histone hyperacetylation has a modest impact on global gene transcription, and histone acetylation changes upon TSA treatment do not always translate into significant changes in gene expression levels [50].In human T cells, HDACs are frequently recruited to transcriptionally active regions of the genome to reset chromatin structure so as to prevent continuous gene activation, indicating that histone acetylation does not necessarily indicate active gene transcription [51].The lack of correlation between histone acetylation status and gene transcription can also reflect molecular events independent of HDAC activity.For example, the HDAC inhibitors, SAHA and panobinostat, have previously been shown by Hedrick et al. to repress expression of oncogenic Sp transcription factors through a reactive oxygen species (ROS) dependent mechanism in RMS cells [52].These findings are consistent with the previous finding of increased oxidative stress in RMS cell lines and xenografts treated with panobinostat [4].Taken together, while our findings suggest that while HDAC activity plays an important role in regulating the tumor biology of ERMS, other histone acetylation-independent or repressive events induced by HDACs may also contribute to the tumorigenesis of ERMS.
In summary, this study has demonstrated the important roles of HDACs in ERMS pathogenesis by characterizing the anti-tumor effects of pan-HDAC inhibitors, trichostatin A and suberoylanilide hydroxamic acid (SAHA), in ERMS in vitro and in vivo.We have delineated Notch and Ephrin-B1 mediated pathways as essential downstream events of HDAC activity in the inhibition of myogenic differentiation and enhanced migratory capacity of ERMS cells, respectively.Aberrant epigenetic regulation by HDACs represents a major molecular mechanism driving tumorigenesis of ERMS.Potential druggable targets within the molecular network regulated by HDAC activity represent promising therapeutic options for the treatment of ERMS.

Animal protocol approval
Zebrafish tumor studies were approved by the Institutional Animal Care and Use Committee at University of Washington under the protocol 4330-01.

Chemicals
Trichostatin A and suberoylanilide hydroxamic acid (SAHA) were obtained from Sigma Aldrich.Each compound was reconstituted in dimethylsulfoxide (DMSO) as 10-100 mM stocks.For chemical treatment of human cell lines, TSA or SAHA was diluted in 0.1% DMSO/ growth medium.

Cell-based assays
ATP-based viability assay (CellTiter-Glo; Promega), EdU flow cytometry-based assay (Life Technologies), differentiation, sphere, scratch, and transwell migrations were performed as described previously [53] except for the following modifications.In the differentiation assay, cells were cultured in 2% horse serum and treated with TSA or SAHA for 3 days prior to fixation and immunostaining for MF20 (1:200; R&D Systems) and DAPI (1:500; Life Technologies).Images were taken using the Evos microscopy imaging system.In the sphere assay, 5000 cells per well were cultured in neurobasal medium supplemented with bFGF (1500 pg/μL), PDGF-AA (733 pg/μL), PDGF-BB (733 pg/μL) and EGF (5.2 pg/μL) in low-binding 24-well plates.Treatment with DMSO (vehicle), TSA (200 nM) and SAHA (1 μM) began on the day of plating and the number of spheres per well was scored 3 days post-treatment.For serial replating, spheres were harvested every 4 days and dissociated using StemPro Accutase (Life Technologies).The cells were resuspended in the neurobasal medium supplemented with growth factors and replated into a new low-binding plate with continued drug treatment.The number of spheres per well was scored 3 days post-replating.In the scratch assay, treatment of human ERMS with TSA or SAHA began 6 hours prior to scratch.Each treatment condition was performed in duplicates.Bright field images of pre-and 18 hours post-scratch were taken using the Evos microscopy imaging system (Life Technologies).In the transwell migration assay, after a 22-hour migration period, three random fields of the membranes containing migrated cells were imaged using the Nikon stereoscope (SMZ18, 3x magnification) and counted using ImageJ software.A Student's t-test was performed to assess differences between the control and experimental groups.

Chromatin Immunoprecipitation
RD and 381T cells cultured in growth medium were cross-linked with 1% formaldehyde 24 hours following chemical (TSA or DMSO) treatment or EFNB1 siRNA transfection.Chromatin immunoprecipitation was performed using the MAGnify Chromatin Immunoprecipitation system (Life Technologies) with the following modifications.Chromatin was initially digested with 2-3 units of micrococcal nuclease per million cells (New England Biolabs) for 15 minutes prior to sonication using the Misonix S-4000 sonicator (Amp 2, 25 cycles).Sheared chromatin fragments were in the range of 300-500 bp as detected on an ethidium bromide-stained 2% agarose gel. 1 μg of anti-acetyl-histone H3 (Lys9) (Active Motif) was used in binding chromatin.Rabbit IgG antibody supplied by the kit was used as a negative control in the ChIP assay.Purified DNA fragments were used in quantitative PCR (95C for 2 min, 95C for 15 sec, 63C for 30 sec, 72C for 30 sec, 40 cycles) using a Rotor-gene 3000 (Qiagen) to assess for enrichment of gene-specific promoter regions.Primers used in the ChIP assay are listed in S1 Table.

Chemical treatment of zebrafish with ERMS for tumor growth and limiting dilution assays
Six-week old CG1 syngeneic fish were transplanted with 3x10 5 unsorted tumor cells arising from Tg(myf5:GFP, mylz2:mCherry) ERMS from CG1 strain fish [54].Engrafted animals were treated at 7 days post-transplantation with 1 μM of TSA or 50 μM of SAHA and vehicle control (DMSO) for 5 days (including one 24-hr drug holiday).Tumor volume was assessed by imaging animals pre-and post-treatment.Tumor volume was calculated by multiplying tumor area by fluorescent intensity using ImageJ.A Student's t-test was performed to assess differences between tumor size in the control and experimental groups.For limiting dilution experiments, the myf5:GFP + /mylz2:mCherry − tumor cell subpopulation was sorted by FACS prior to transplanting at limiting dilutions of 10 3 , 10 2 and 10 cells into six-week old CG1 syngeneic fish.Unsorted bulk tumor cells were transplanted at limiting dilutions of 10 4 , 10 3 and 10 2 cells.Selfrenewal frequency was determined by the extreme limiting dilution analysis (ELDA) algorithm [55].
Tumor-bearing zebrafish were euthanized at or before tumor endpoint criteria, defined as 1) tumor invasion from trunk skeletal muscle into adjacent tissues/organs or 2) tumor grows to 50% of host body mass.No tumor exceeded 50% of host body mass.Zebrafish were euthanized by immersion in an overdose of neutral buffered Tricaine followed by decapitation or icing.

Analysis of human clinical data
Clinical and expression data of RMS patients were obtained from the supplemental materials of the study by Davicioni et al. (2009) and Williamson et al. (2010).A Cox regression model with the assumption of a linear relationship between the expression level and the hazard ratio was used to generate the survival curves for ERMS and ARMS cohorts.

Microarray analysis
RD and 381T cells treated with DMSO and TSA (200 nM) were harvested 24 hours post-treatment.RNA was isolated from the cells using the Qiagen RNeasy Kit.Integrity of RNA samples was assessed with an Agilent 2100 Bioanalyzer (Agilent Technologies Inc., Santa Clara, CA).Processing of the RNA samples that passed quality control was performed according to the standard Affymetrix GeneChip Whole Transcript Sense Target labeling protocol.The arrays were scanned with an Affymetrix GeneChip1 3000 scanner.Image generation and feature extraction were performed using Affymetrix GeneChip Command Console Software.The data at the transcript level were summarized using the RMA algorithm, as implemented in the Bioconductor oligo package [56].We filtered out all control probesets, as well as those probesets that appear to be either not expressed, or at such a low level that noise dominates the signal.After filtering there were 19,211 probesets that remained.We then fit a weighted analysis of variance (ANOVA) model [57] and made comparisons using empirical Bayes adjusted contrasts, using the Bioconductor limma package [55].
Microarray data can be accessed at NCBI Gene Expression Omnibus (GEO) as series GSE74970.

Supporting Information
Fig 1A), indicating that TSA and SAHA can alter the histone acetylation landscape in RMS cells.By quantitative analysis of cell numbers and an ATPbased cell viability assay, TSA and SAHA inhibited tumor growth in all ERMS cell lines tested (Fig 1B; S1A and S1B Fig).By contrast, ARMS cell lines exhibited a variable response to TSA and SAHA (Fig 1B).Within the ARMS cell lines, SAHA inhibited tumor growth of all three lines, while TSA inhibited tumor growth only in the Rh3 cell line at doses that showed effects in all ERMS cell lines tested.Overall, TSA and SAHA appeared to exert a more potent effect on tumor growth in ERMS cell lines in comparison to ARMS cell lines.Upon cell cycle analysis, ERMS cells treated with TSA or SAHA showed a reduction in S-phase and prolonged G2/M phase (S1C and S1D Fig).TSA or SAHA treatment did not result in any significant apoptosis at doses (200 nM of TSA and 1 μM of SAHA) that induced abnormal cell cycle progression (S1E and S1F Fig).Together, these results indicate that pan-HDAC inhibitors reduce ERMS tumor growth by altering cell cycle progression.
Fig 1G-1I; S2G and S2H Fig), and this effect was further enhanced upon serial replating (S3I and S3J Fig), suggesting that inhibition of HDAC function can reduce selfrenewal potential of human ERMS cells.Together, the results of our in vitro studies implicate the essential role of HDACs in modulating the differentiation and self-renewal status of ERMS during tumor progression.As the self-renewal capacity of tumor cells is an indicator of their relapse potential, HDAC activity may serve as a potential biomarker for poor prognosis.

Fig 1 .
Fig 1. Trichostatin A (TSA) reduced growth and altered differentiation and self-renewal capacity of ERMS cell lines.(A) Western blots demonstrating hyperacetylation of histones using antibodies against acetyl-histone H3 (Lys9), acetyl-histone H3 (Lys27), acetyl-histone H4 (Lys5), and acetyl-histone H4 (Lys8).D: DMSO; S: vorinostat (SAHA); T: TSA.Each band intensity was normalized to GAPDH loading control.Relative fold increase to DMSO (vehicle) treatment is shown.(B) Analysis of cell viability by cell counts.Cell counts were performed on cells treated with DMSO, 200 nM TSA or 1 μM SAHA at day 0 and day 5. Fold change in cell counts was normalized to day 0. (C-D) Representative images of immunofluorescence (IF) against MF20 performed on RD cells treated with DMSO (C) or 200 nM TSA (D) for 3 days in 2% horse serum in DMEM.Green: MF20-positive cells.Blue: DAPI.Scale bar: 20 μm.(E) Summary of IF against MF20 in ERMS (RD, 381T and SMS-CTR) and ARMS (Rh3, Rh5 and Rh30) cell lines treated with DMSO, 200 nM TSA or 1 μM SAHA.(F) Chromatin immunoprecipitation (ChIP) assays showing differential binding of acetylhistone H3 (Lys9) at myogenic promoters.Fold enrichment binding of MYOD1, MYOG, and MYH4 promoter regions were determined by quantitative PCR, normalizing amplification levels to input DNA of each sample.Rabbit IgG was used as a negative control for chromatin immunoprecipitation. (G-H) Representative brightfield images from a sphere assay on RD cells treated with DMSO (G) and 200 nM TSA (H).(I) Summary of sphere assays in RD and 381T cells.Each error bar in panels (B), (E), (F) and (I) indicates standard deviation of 3 technical replicates.* indicates p < 0.05.** indicates p < 0.01.*** indicates p < 0.001.doi:10.1371/journal.pone.0144320.g001

Fig 3 .
Fig 3. TSA and SAHA reduced tumor growth and induced myogenic differentiation in vivo.(A) Western blots demonstrating hyperacetylation of histone H3 (Lys9) and histone H4 (Lys5) in zebrafish ERMS treated with 1 μM TSA or 50 μM SAHA.GAPDH was used as loading control.The values shown represent fold change in band intensity from TSA or SAHA treatment relative to DMSO after normalizing to loading control.(B) Representative pre-and posttreatment images of zebrafish ERMS treated with DMSO (vehicle) or 1 μM TSA.Dotted line outlines the tumor in each fish.Scale bar = 2 mm.(C) Summary of tumor volume change of zebrafish Tg(myf5:GFP; mylz2:mCherry) ERMS treated with DMSO, 50 μM SAHA or 1 μM TSA.Overlaid images of bright field and red fluorescent channel are shown.Error bar indicates standard error of means.n = number of animals treated in each cohort.(D) Summary of quantitative Fluorescence Activated Cell Sorting analysis on ERMS treated with DMSO or 10 μM SAHA.Each pie chart shows relative proportion of each tumor cell subpopulation in an individual treated tumor.Green: myf5:GFP + /mylz2:mCherry − cells; yellow: myf5:GFP + /mylz2:mCherry + cells; red: mylz2: mCherry + /myf5:GFP − cells.(E) Quantitative RT-PCR analysis of myog and myod mRNA expression in ERMS treated with DMSO, 1 μM TSA or 50 μM SAHA.Each bar demonstrates an individual tumor.Each error bar indicates standard deviation of technical triplicates.(F) Annexin V analysis of ERMS tumors treated with DMSO, 50 μM SAHA, or 1μM TSA.6 animals were analyzed per group.Each error bar indicates standard deviation.* indicates p < 0.05.** indicates p < 0.01.doi:10.1371/journal.pone.0144320.g003 4D) and downstream target genes, HEY1, HEY2, and HES1, were also downregulated in ERMS cells (Fig 4E).Interestingly, there was enriched binding of acetyl-histone H3 (Lys9) on the NOTCH1 promoter in ERMS cells treated with TSA or SAHA (Fig 4F and S3F Fig), suggesting that histone hyperacetylation does not necessarily correlate with active transcription and/or other repressive events are present to suppress transcription of NOTCH1 upon TSA or SAHA treatment.To assess the specificity of the Notch pathway as a downstream effector of HDAC-induced myogenic differentiation arrest, we showed that overexpression of Notch Intracellular Domain (NICD), which constitutively activates downstream signaling of the Notch pathway, reversed the myogenic differentiation phenotype resulting from TSA or SAHA treatment (Fig 4G and S3G Fig).Together, our results indicate that the Notch pathway is directly regulated by HDACs to suppress myogenic differentiation in ERMS.
targeting EFNB1 and EFNA3 by gene-specific siR-NAs resulted in effective gene knockdown by quantitative RT-PCR and/or Western blot analysis (S4A and S4B Fig).As assessed by scratch assay, RD or 381T cells transfected with EFNB1 siRNA showed a significant reduction in the percentage of gap closure in comparison to a mock control (Fig 5A).By contrast, knockdown of EFNA3 did not affect the percent of gap closure in the scratch assay (S4C Fig).We subsequently validated the effect of EFNB1 knockdown on the migratory capacity of ERMS cells using a separate transwell migration assay (Fig 5B).EFNB1 knockdown did not affect proliferation or apoptosis of the ERMS cells (S4D and S4E Fig), indicating that the reduced migratory capacity of the ERMS cells is not secondary to altered cell growth or increased cell death.To assess the effect of EFNB1 gain-of-function, a stable line of RD cells overexpressing EFNB1 was generated.Overexpression of EFNB1 mRNA and protein was confirmed by quantitative RT-PCR and Western blot analysis (S4F Fig).In contrast to GFP-expressing control cells, RD cells overexpressing EFNB1 showed at least 3-fold increase in their migratory capacity in both scratch and transwell assays (Fig 5C and 5D).Finally, by ChIP analysis, TSA or SAHA-treatment resulted in enriched binding of acetylated histones on the EFNB1 promoter region (Fig 5E).EFNB1 protein expression level was reduced upon TSA or SAHA treatment (Fig 5F)

Fig 4 .Fig 5 .
Fig 4. Aberrant NOTCH1-mediated signaling activity is essential for inhibiting myogenic differentiation of ERMS.(A) Summary of MF20 IF of RD cells treated with two different doses of GSI-IX, PD173074, or cyclopamine.(B-C) Representative MF20 IF images of RD cells treated with DMSO (B) and 20 μM GSI-IX (C).Scale bar = 20 μm.(D) Western blot of NOTCH1 expression in RD cells treated with DMSO, 200 nM TSA, or 1 μM SAHA.Each band intensity was normalized to GAPDH loading control.Percent intensity relative to DMSO (vehicle) treatment is shown.(E) Quantitative RT-PCR analysis of NOTCH1 pathway downstream targets, HEY1, HEY2 and HES1 in RD cells treated with DMSO, 200 nM TSA, or 1 μM SAHA.(F) ChIP assay summary showing differential binding of acetyl-histone H3 (Lys9) at the NOTCH1 promoter in RD cells treated with DMSO or 200 nM TSA.Rabbit IgG was used as a negative control for chromatin immunoprecipitation. (G) Summary of MF20 IF of control GFP-overexpressing and NICD-overexpressing RD cells treated with DMSO, 200 nM TSA or 1 μM SAHA.Error bar in each graph indicates standard deviation of technical triplicates.Brackets in each group are used to indicate comparison groups.* indicates p < 0.05.** indicates p < 0.01.doi:10.1371/journal.pone.0144320.g004

S1
Fig. SAHA exerted anti-tumor effects in ERMS by reducing tumor growth and altering cell cycle progression.(A-B) CellTiter-Glo viability assays on RD cells treated with DMSO, 3 doses of TSA (A) or 3 doses of SAHA (B).(C-D) Cell cycle analysis of RD and 381T cells treated with TSA (C) or SAHA (D).(E-F) Annexin V analysis of RD, 381T and SMS-CTR cells treated with TSA (E) or SAHA (F).Each error bar indicates standard deviation of technical triplicates.(TIF)

Table 1 .
Summary of limiting dilution experiment of myf5-GFP+ sorted transplants treated with SAHA.

Table 2 .
Summary of limiting dilution experiments of unsorted bulk tumor treated with SAHA.

Table 3 .
Summary of limiting dilution experiments of unsorted bulk tumor treated with TSA.

Table 4 .
Validation of top downregulated and upregulated genes in TSA-treated ERMS cells.
difference in overall survival when comparing high-EFNB1 expressing vs low-EFNB1 expressing cohorts in both studies (S5 Fig).However, this analysis was limited by the low number of death events (11 deaths in the Davicioni study and 34 deaths in the Williamson study).As