Targeting Hippo-Dependent and Hippo-Independent YAP1 Signaling for the Treatment of Childhood Rhabdomyosarcoma

Rhabdomyosarcoma is the most common soft-tissue sarcoma of childhood. While outcomes for patients with localized disease have improved, patients with metastatic or recurrent disease continue to do poorly, therefore new treatments are needed (1, 2). The two most common subtypes of rhabdomyosarcoma, embryonal and alveolar, are phenotypically and molecularly distinct but often treated with the same modalities. Standard-of-care treatment includes a combination of surgery, radiation, and a backbone of cytotoxic chemotherapies made up of vincristine, actinomycin D, and cyclophosphamide but does not include any targeted therapies. Embryonal rhabdomyosarcoma is characterized by genomic instability with high rates of copy-number alternations, single-nucleotide variations and up to one-third have alterations in the RAS signaling pathway (3, 4). On the other hand, alveolar rhabdomyosarcoma is characterized by a distinct PAX3/7-FOXO1 (PF) chromosomal translation creating a fusion transcription factor. PF expression results in widespread transcriptional disruption and upregulation of oncogenic pathways including FGFR4 and insulin-like growth factor signaling (5). The presence of the PF fusion gene portends a worse prognosis with a 5-year survival rate of 52% for localized disease and 19% when metastatic (6).

The YAP1 protein was initially discovered as the YES-associated protein. YES1, a SRC family tyrosine kinase (SFK), has been shown to be upregulated in rhabdomyosarcoma and drive tumor cell proliferation (7). Although the YES1 kinase is targetable using dasatinib and other kinase inhibitors, to date it has not proven to be effective in clinical trials in rhabdomyosarcoma (8). However, the relationship between YES1 and YAP1 in rhabdomyosarcoma has not been studied.

YAP1 was discovered to be the terminal oncoprotein in the Hippo tumor suppressor pathway, a developmental pathway that regulates organ size and tissue regeneration. MST1/2, the mammalian ortholog of Hippo, is a tumor suppressor that initiates a phosphorylation cascade through LATS1/2 that results in the phosphorylation of YAP1 and its homolog TAZ at conserved serine residues. Once phosphorylated, YAP1 is sequestered in the cytoplasm and can be degraded. MST1/2 is regulated upstream by many inputs including cell-to-cell contact, actin cytoskeleton, and the RASSF family of scaffold proteins. When MST1/2 is inactive, YAP1 is not phosphorylated and translocates to the nucleus where it functions as a transcriptional coactivator, binds the TEAD family of transcription factors, and initiates transcription of pro-growth and antiapoptotic genes (9). Hippo signaling is dysregulated in many cancer types through epigenetic silencing of Hippo activator RASSF1, MST1/2, or LATS1/2, or through general upregulation of YAP1 and/or TAZ expression. YAP1 is highly expressed in human rhabdomyosarcoma tumors where it has been shown to promote proliferation, survival, and stemness, and inhibit myogenic differentiation (10–12). While YAP1 is an ideal therapeutic target in cancer, a specific YAP1 inhibitor has yet to reach the clinic. We aimed to identify a clinically relevant therapy to inhibit YAP1 in rhabdomyosarcoma.

Here we demonstrate that treatment with a DNA methyltransferase inhibitor (DNMTi) can inhibit YAP1 in a Hippo-dependent manner through alteration of RASSF family expression (upstream MST1/2 regulators). Furthermore, YAP1 is regulated in a Hippo-independent manner through YES1 and this regulation can be inhibited through treatment with a SRC family kinase inhibitor, dasatinib. Finally, a combined treatment results in ablation of cell growth and induction of apoptosis. These data suggest a potential therapeutic strategy using the combination of DNMTis and SFK inhibitors for the treatment of recurrent/refractory rhabdomyosarcoma, particularly the alveolar rhabdomyosarcoma subtype.

Human RMS cell lines RD (embryonal rhabdomyosarcoma) and Rh30 alveolar rhabdomyosarcoma were obtained from ATCC. Rh28 and RMS559 were obtained from Dr. Javed Khan (NCI, Bethesda, MD). Rh36 embryonal rhabdomyosarcoma cell line was obtained from Dr. Maria Tsokos (Beth Israel Medical Center). Rh41 alveolar rhabdomyosarcoma cell line was obtained from Dr. Peter Houghton (St Jude's Children's Research Hospital, Memphis, TN). Cell line authentication was performed in November 2018 using short tandem repeat analysis (Promega PowerPlex 16 HS System) conducted by the University of Arizona Genetics Core (Tucson, AZ). The cell lines were cultured in RPMI-1640, 100 units/mL penicillin and 100 μg/mL streptomycin, 2 mmol/L glutamine, and 10% heat-inactivated FBS at 37°C in an atmosphere of 5% CO₂. Lentivirus shYES clones 9 and 11 were obtained from Sigma (TRCN0000001609 and TRCN0000001611, respectively) and used as described previously (7). YAPS127A cell line was constructed by transfecting Rh30 cells with pCDNA4/HismaxB-YAP1-S127A (Addgene plasmid, catalog no. 18988) and vector control cell was transfected with pCDNA4/TO (Thermo Fisher Scientific) using Nucleofector (Lonza) program B032, buffer V, using 2 × 10⁶ cells, and 2 μg of plasmid. Cells were batch selected using zeocin (Thermo Fisher Scientific). YEST348I cell line was constructed by stably expressing YES1 (T348I) in pLX303 (Addgene plasmid, catalog no. 42564) or vector control pLX304 (Addgene plasmid, catalog no. 25890) in Rh30 cells using established lentiviral methods (11). Cells were batch selected using blasticidin (Invitrogen).

For growth curves, 1,000–2,000 rhabdomyosarcoma cells per well were plated in 96-well plates (CytoOne) with 5–10 wells per condition. Twenty-four hours after plating cells were treated with 0.5 μmol/L SGI-110, 0.5 μmol/L dasatinib, or combination of SGI-110 and dasatinib at 0.25 μmol/L or 0.5 μmol/L and loaded into the IncuCyte S3 (Essen BioScience). Images were taken every 4 hours for 7 days and percent confluence was calculated. For cleaved caspase-3/7 activity, cells were treated with SGI-110 for 5 days, then media were aspirated and fresh media containing SGI-110 and the IncuCyte Caspase-3/7 Green Apoptosis Assay Reagent (at 1:1,000 dilution) were added. Images were obtained every 2 hours for 3 days.

RNA and DNA extractions were performed on the KingFisher (Thermo Fisher Scientific) as per manufacturer's recommendations. Reverse transcription was performed with the iScript Select cDNA Synthesis Kit (Bio-Rad). qRT-PCR was performed with the PowerUp SYBR Green Master Mix (Thermo Fisher Scientific) as per the manufacturer's protocol. For methylation-specific PCR (MS-PCR), bisulfite conversion was performed with EZ DNA Methylation-lightning Kit (Zymo) and RASSF1 MS-PCR was done with the CpG WIZ RASSF1A Amplification Kit (Millipore). Primer sets are listed in Supplementary Table S1.

A total of 0.5 × 10⁶ to 1 × 10⁶ cells were plated in 10-cm plates and treated with SGI-110, dasatinib, or combination for 48 hours to 5 days. Cell were harvested with RIPA cell lysis buffer (Cell Signaling Technology) plus protease and phosphatase inhibitors (Thermo Fisher Scientific). The following antibodies were used for immunoblotting: anti-Cleaved PARP (Cell Signaling Technology, catalog no. 5625, 1:1,000), anti-RASSF1 (Abcam, catalog no. ab23950, 1:1,000), anti-RASSF5 (Sigma, catalog no. N5912, 1:200), anti-RASSF4 (Novus Biologicals, catalog no. NBP1-89249, 0.4 μg/mL), anti-P-YAP1 Ser127 (Cell Signaling Technology, catalog no. 13008, 1:1,000), anti-P-Src Family Tyr416 (Cell Signaling Technology, catalog no. 6943, 1:1,000), anti-YES1 (Cell Signaling Technology, catalog no. 3201, 1:500), anti-YAP (Cell Signaling Technology, catalog no. 14074, 1:1,000), anti-Tubulin (Proteintech, catalog no. 66031-1-Ig, 1:2,000), anti-GAPDH (Cell Signaling Technology, catalog no. 5174, 1:1,000), anti-Vimentin (Cell Signaling Technology, catalog no. 5741, 1:1,000), anti-Histone H3 (Cell Signaling Technology, catalog no. 4499, 1:2,000), anti-MEK1/2 (Cell Signaling Technology, catalog no. 8727, 1:1,000), anti-AKT (Cell Signaling Technology, catalog no. 4685, 1:1,000), anti-P-AKT Ser473 (Cell Signaling Technology, catalog no. 4060, 1:2,000), anti-ERK1/2 (Cell Signaling Technology, catalog no. 4695, 1:1,000). Densitometry was performed using ImageJ (NIH, Bethesda, MD) on three biological replicates and all values were normalized to the loading control.

Proximity ligation assay (PLA) was performed to detect colocalization of YES1 and YAP1 using the Duo-Link In Situ Orange Kit (Sigma). Rhabdomyosarcoma cells were plated at 500 cells per well on a 384-well glass bottom plate in 50 μL volume and incubated for 48 hours. The cells were fixed by adding 50 μL of 4% paraformaldehyde (Santa Cruz Biotechnology) per well and allowed to fix for 20 minutes at 4°C. The fixed cells were washed twice with PBS and blocked/permeabilized with 50 μL per well of blocking solution (5% donkey serum, 0.3% Triton X-100 in PBS) for one hour at 37°C. Blocked/permeabilized cells were incubated in primary antibodies overnight with anti-YAP1 (Abcam, catalog no. ab39361) and/or anti-YES1 (Wako, catalog no. 013-14261) at 1:50 dilution. Labeling, ligation, and amplification reactions were performed according to the Duo-Link kit. Visualization and analysis of the labeled cells were done using an Opera Phenix system (PerkinElmer).

LATS biosensor plasmids pCDNA3.1neo-14-3-3-CLuc (Addgene plasmid, catalog no. 107611) and pCDNA3.1neo-NLucYAP15 (Addgene plasmid, catalog no. 107610) were gifts from Xiaolong Yang (Queen's University, Kingston, Ontario, Canada). pcDNA3 Lats1 (Nigg HS189) was a gift from Erich Nigg (Addgene plasmid, catalog no. 41156). The assay was performed as previously described with some modifications (13, 14). Fugene 6 (Promega) was used to transfect 100 ng of each plasmid for the Rh30 cells and 200 ng for the RD cells in a 96-well plate. For an assay positive control, 400 ng pcDNA3 LATS1 was transfected at the same time. Living cell luciferase was measured on the Varioskan LUX multimode microplate reader (Thermo Fisher Scientific) after a 10-minute incubation with 0.05 mg d-Luciferin (Promega). Luciferase assays were replicated in three individual experiments with each containing five replicates per condition.

DNMTi guadecitabine or SGI-110 was obtained from AdooQ and MedChemExpress and resuspended in DMSO at 10 mmol/L for in vitro studies or in PBS at 5 mmol/L for in vivo studies. Dasatinib was obtained from SelleckChem and resuspended in DMSO at 10 mmol/L for in vitro studies or in 4% DMSO + 30% PEG 300 + 5% Tween 80 in ddH2O for in vivo studies.

For RNA-seq, RNA samples were collected from RD or Rh30 cells treated with 0.5 μmol/L SGI-110, dasatinib, or combination for 5 days in biological triplicates. The library prep and sequencing was performed by the Single Cell, Sequencing, and CyTOF (SC2) Core Laboratory at the Saban Research Institute of Children's Hospital Los Angeles (Los Angeles, California). Transcriptome libraries were prepared from total RNA using the Illumina TruSeq Stranded mRNA Library Prep kit following manufacturer's protocol, which allows for standardized mRNA isolation from total RNA, first- and second-strand cDNA synthesis, end-repair, A-base addition, and adapter ligation. Libraries were PCR-amplified and paired-end cluster generation was performed using the Illumina Nextseq 500 on a High-Output Sequencing Reagent Kit v2 (2 × 75 bp). Quality control and adapter trimming were performed using trim_galore (v0.4.2) with default parameters (https://github.com/FelixKrueger/TrimGalore). Reads were aligned to the GRCh38 reference genome and transcriptome using HISAT2 (v2.1.0; ref. 15), and transcript quantification was performed using feature Counts (v1.5.1; ref. 16). Differential expression analysis was performed using the “DESeq2” R package (v1.16.1; ref. 17), and a rank score calculated as −log₁₀(q_val)*sign(log₂FoldChange) was used as input to the GSEAPreranked tool for pathway analysis (18). Data were analyzed through the use of Ingenuity Pathway Analysis (IPA; QIAGEN Inc., https://www.qiagenbioinformatics.com/products/ingenuitypathway-analysis; ref. 19).

For the Illumina Infinium HumanMethylationEPIC (EPIC) DNA methylation assay, genomic DNA (1,000 ng) was bisulfite converted using the Zymo EZ DNA methylation Kit (Zymo Research) according to the manufacturer's recommendations. The amount of bisulfite-converted DNA as well as the completeness of bisulfite conversion for each sample was assessed using a panel of MethyLight-based real-time PCR quality control assays as described previously (20). Bisulfite-converted DNAs were then repaired using the Illumina Restoration Kit as recommended by the manufacturer. The repaired DNA was used as a substrate for the Illumina EPIC BeadArrays, as recommended by the manufacturer and first described in Moran and colleagues (21). After the chemistry steps, BeadArrays were scanned and the raw signal intensities are extracted from the *.IDAT files using the “noob” function in the minfi R package. The “noob” function corrects for background fluorescence intensities and red-green dye-bias as described by Triche and colleagues (22). The beta value was calculated as [M/(M+U)], in which M and U refer to the (preprocessed) mean methylated and unmethylated probe signal intensities, respectively. Measurements in which the fluorescent intensity is not statistically significantly above background signal (detection P value > 0.05) were removed from the dataset. All the RNA-seq and EPIC array data were submitted to Gene Expression Omnibus and assigned repository #GSE147240 and #GSE147246, respectively.

YAP1 immunofluorescence (IF; Cell Signaling Technology, catalog no. 14074, 1:100) was performed according to the manufacturer's protocol except for the addition of a permeabilization step with 0.3% Triton X in PBS pH 7.4 before blocking step.

A total of 2 × 10⁶ luciferase-labeled (pGL4.18 CMV-Luc, Addgene plasmid, catalog no. 100984) Rh30 cells were resuspended in sterile Hank's Balanced Salt Solution and injected intramuscularly into the right gastrocnemius of female 4–6 week old SCID/beige mice. Treatment began at day 10 after injection when tumors were detectable on bioluminescence imaging and mice were randomized into treatment groups. Dasatinib was given at 100 mg/kg three times a week by oral gavage and DNMTi guadecitabine was given at 2 mg/kg 5 days in a row on 14-day cycles by subcutaneous injections. Mice were observed twice weekly for evidence of malaise, weight loss or inability to ambulate normally, and the limbs were measured with calipers. Tumor volume was calculated by the following formula: V (mm³) = (D x d²)/6 × 3.14, where D is the longest tumor axis and d is the shortest tumor axis as in previous studies (23). Mice were sacrificed at day 33 post tumor cell injection due to weight loss in the combination treated group. Tumor weight was measured by subtracting the weight of the nontumor bearing leg from the weight of the tumor-bearing leg. Tumors were preserved in RNAlater (Qiagen) for PCR analysis or formalin-fixed for IHC. Studies were approved by CHLA's Institutional Animal Care and Use Committee.

Statistical analysis was performed using GraphPad Prism (GraphPad). Unless otherwise noted, data are presented as the mean and SE. One-way ANOVA, two-way ANOVA, and unpaired t test were used as appropriate. P values were considered significant at *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

We analyzed the effect of treatment with a DNMTi, SGI-110 also known as guadecitabine, on the growth of rhabdomyosarcoma cell lines using live cell imaging. DNMTi treatment markedly decreased rhabdomyosarcoma cell growth in both alveolar rhabdomyosarcoma and embryonal rhabdomyosarcoma cells (Fig. 1A; Supplementary Fig. S1A). To determine whether DNMTi treatment causes rhabdomyosarcoma cell death, we analyzed apoptosis by a cleaved caspase-3/7 fluorescence reporter and by immunoblots of cleaved PARP. DNMTi treatment markedly increased apoptosis in alveolar rhabdomyosarcoma cells and modestly increased apoptosis in embryonal rhabdomyosarcoma cells (Fig. 1B and C; Supplementary Fig. S1B and S1C). The alveolar rhabdomyosarcoma cells were more sensitive to DNMTi treatment exhibiting a 5× increase in the percentage of cleaved caspase-3/7 positive cells compared with control while the embryonal rhabdomyosarcoma cells only had a 2× increase in apoptosis.

After treatment with a DNMTi, we also observed a striking phenotypic change in the percentage of the rhabdomyosarcoma cells. In addition to the obvious apoptotic cells, we observed large flat elongated cells that resembled skeletal muscle myoblasts (Fig. 1D). qRT-PCR for MYF6, a myogenic factor involved in the later stages of skeletal muscle differentiation, showed a large increase in MYF6 expression (Fig. 1E). These data suggest that DNMTi treatment promotes changes toward myogenic differentiation, which may contribute to the growth inhibitory effects of drug treatment. The decrease in growth may also be due to cellular senescence, which was not explored in this study but has been observed after Hippo pathway activation in rhabdomyosarcoma (11, 24).

To validate that DNMTi treatment could successfully demethylate the rhabdomyosarcoma genome, we analyzed changes in whole genome DNA methylation by the Illumina EPIC methylation array, which analyzes the methylation status of over 850,000 sites in the genome. The EPIC array revealed that DNMTi treatment for 5 days was sufficient to cause widespread DNA demethylation in both the Rh30 and RD cells (Fig. 2A). To determine the effect of genome demethylation on gene expression, we performed RNA-seq on Rh30 and RD cells after DNMTi treatment. RNA-seq revealed changes in expression of the RASSF family, regulators of Hippo (MST1/2). RASSF1, 5, 6, which have roles in activating MST1/2, were elevated at the mRNA and protein level (Fig. 2C and D; Supplementary Fig. S2A). On the other hand, RASSF4, which has been shown to inhibit MST1/2 in alveolar rhabdomyosarcoma (11), decreased in expression (Fig. 2C and D; Supplementary Fig. S2A). YAP1 itself also decreases at the RNA level and several myogenic differentiation genes including MYF5, MYF6 increase in accordance with our observations in Fig. 1D and E (Fig. 2B). FGFR4, another oncogenic signaling pathway in rhabdomyosarcoma (25, 26), was also significantly decreased (Fig. 2B). These data demonstrate that genes relevant to rhabdomyosarcoma tumorigenesis are altered in expression after DNMTi treatment. Furthermore, we confirmed by MS-PCR that the promoters of RASSF1 and RASSF5 are ordinarily methylated in alveolar rhabdomyosarcoma cells and embryonal rhabdomyosarcoma cells but become partially demethylated after DNMTi treatment (Fig. 2E; Supplementary Fig. S2B). Interestingly, RASSF4 was also partially demethylated as analyzed on the EPIC methylation array (Fig. 2F). These data demonstrate that DNMTi treatment in rhabdomyosarcoma cells results in upregulation of tumor suppressor RASSF family members 1, 5, 6 and downregulation of pro-growth RASSF4.

To determine the effect of changes in RASSF expression on downstream Hippo pathway signaling, we performed a biosensor for LATS1/2 activity. Luciferase is expressed when YAP1 is phosphorylated at S127 by LATS1/2 and is sequestered by 14-3-3 (13). There is a baseline amount of luminescence or LATS1/2 activity in rhabdomyosarcoma, which then doubles after DNMTi treatment, suggesting DNMTi treatment activates canonical Hippo signaling (Fig. 3A). This increase in LATS1/2 activity after DNMTi treatment is comparable with the increase seen after transient expression of LATS1 in these cells (Fig. 3A). Furthermore, by immunoblot, DNMTi treatment results in an increase in inactive P-YAP1 (Fig. 3B) and the DNMTi-induced growth inhibition can be partially rescued by overexpression of a constitutively active YAPS127A (a mutation that prevents YAP1 serine phosphorylation and cytoplasmic localization), suggesting that the growth inhibition after DNMTi treatment is in part due to YAP1 inactivation (Fig. 3C and D). DNMTi treatment also decreased both YES1 expression and Y416 phosphorylation, a marker of YES1 activity (Fig. 3B). These data suggest that DNMTi treatment can inhibit both YAP1 and YES1 signaling. Overall, these data demonstrate that DNMTi treatment can inhibit YAP1 signaling in a Hippo-dependent manner through modulation of RASSF family expression and can decrease YES1 activation.

We decided to focus more closely on the alveolar rhabdomyosarcoma subtype in the rest of the study because the alveolar rhabdomyosarcoma cells were more sensitive to DNMTi treatment. To further investigate the connection between YES1 and YAP1, we performed PLAs and discovered that YES1 and YAP1 interact in the nucleus of the Rh30 alveolar rhabdomyosarcoma cell line as in other cell types (Fig. 4A). Furthermore, shRNA knockdown of YES1 promotes YAP1 cytoplasmic localization and decreases expression of the YAP1 target genes, suggesting a role for YES1 in regulating YAP1 cellular localization in rhabdomyosarcoma (Fig. 4B and C). Similarly, treatment with the SRC family kinase inhibitor dasatinib decreases expression of the YAP1 target genes (Fig. 4D) and results in YAP1 cytoplasmic localization shown by YAP1 IF (Fig. 4E). The localization of YAP1 to the cytoplasm after dasatinib treatment can be rescued by expression of a dasatinib-resistant YES1 T348I (Fig. 4E; Supplementary Fig. S3), suggesting YES1 activity is necessary to control YAP1 localization. Expression of YEST348I was confirmed by immunoblot (Fig. 4F). Together these data propose a dual regulation of YAP1 in rhabdomyosarcoma through Hippo-dependent and Hippo-independent (via YES1) mechanisms.

To target both mechanisms of YAP1 regulation, we evaluated combination treatment of DNMTi and dasatinib in rhabdomyosarcoma cells in vitro. The combination ablated cell growth, showing greater growth inhibition than either drug alone (Fig. 5A; Supplementary Fig. S4A). Combination treatment also resulted in a marked increase in apoptosis, shown by immunoblots for cleaved PARP (Fig. 5B). These data demonstrate that DNMTi and dasatinib combination treatment has greater activity than each individual drug, which results in an ablation of rhabdomyosarcoma cell growth.

To analyze possible mechanisms of this combination, we used IPA to compare the RNA-seq transcriptional profiles of Rh30 and RD cells treated with the combination compared with DNMTi alone treated cells. IPA uses a knowledge network to determine the canonical pathways predicted to be activated or inactivated in the dataset. The top canonical pathways enriched in the combination group showed a variety of pathways including several pathways regulating the actin cytoskeleton (Supplementary Table S2). Of interest were the inactivation of NF-kB signaling and the activation of PTEN signaling because of their roles in cell proliferation and survival. When we analyzed signaling downstream of PTEN, we saw a near elimination of AKT signaling (analyzed by P-AKT) in the alveolar rhabdomyosarcoma cell lines while this decrease in P-AKT was not seen in the embryonal rhabdomyosarcoma cell lines (Fig. 5C, top). This is in line with the higher rates of apoptosis seen in the alveolar rhabdomyosarcoma cell lines. While multiple pathways are likely altered with DNMTi and dasatinib treatment, we show at least two possible pathways (NFκB and PTEN) that are altered by the combination treatment and may contribute to the activity observed.

Because of the striking inhibition of cell growth after combination treatment, we pursued the combination treatment in orthotopic xenografts in immunocompromised mice. Here we focused on the alveolar rhabdomyosarcoma subtype because the alveolar rhabdomyosarcoma cells were more sensitive to DNMTi treatment than the embryonal rhabdomyosarcoma cells (Fig. 1). We generated orthotopic alveolar rhabdomyosarcoma xenografts from the Rh30 cell line and began treatment once tumors were detectable on imaging. DNMTi treatment alone significantly decreased tumor growth over time as measured by calipers (Fig. 6A). Meanwhile, dasatinib treatment alone did not significantly inhibit tumor growth due to variability, as has been seen in previous studies (7). Unfortunately, toxicity arose in the combination group after the second round of DNMTi treatment, so the study was terminated early. However, there was a trend toward decreased tumor growth in the combination treated group, but it did not reach significance due to variability (Fig. 6A). In addition, tumor weight was significantly smaller in the combination treated group (Fig. 6B). Like in vitro, DNMTi treatment also results in an upregulation in expression of the myogenic differentiation gene MYF6, albeit a smaller magnitude of change is observed in vivo (Fig. 6C). However, hematoxylin and eosin analysis did not show any obvious differences between the groups (Supplementary Fig. S5). Together these data are suggestive that combination treatment could be effective in vivo but further exploration is needed.

In conclusion, in rhabdomyosarcoma cells, YAP1 is regulated by both Hippo-dependent and Hippo-independent signaling. DNMTi treatment leads to widespread genomic demethylation including at the RASSF1 and RASSF5 tumor suppressor sites. These alterations in RASSF expression lead to activation of canonical Hippo signaling and inactivation of YAP1. In a Hippo-independent manner, YES1 associates with YAP1 and promotes YAP1 nuclear localization. This regulation of YAP1 can be inhibited with dasatinib treatment. Finally, targeting both mechanisms of YAP1 regulation by combination DNMTi and dasatinib treatment ablates rhabdomyosarcoma cell growth laying the foundation for clinical investigations of this combination treatment (Fig. 6D).

Here we describe one mechanism by which DNMTi treatment inhibits rhabdomyosarcoma cell growth through Hippo-dependent YAP1 inhibition. However, DNMTis are not targeted or specific to YAP1. As shown, DNMTi treatment causes widespread genomic demethylation and transcriptional changes and future studies would be needed to fully examine the multitude of possible mechanisms causing rhabdomyosarcoma growth inhibition. However, we hope this study further emphasizes the need for a specific YAP1 inhibitor for cancer treatment, and further demonstrates the utility of such a compound for rhabdomyosarcoma treatment. To date, a specific YAP1 inhibitor has not reached the clinic despite continued research efforts (27). There has been continued effort in identifying alternative approaches to inhibit YAP1 signaling such as the use of statins or G protein-coupled receptor modulators to regulate the signaling upstream of YAP1 in particular cell types (ref. 28 and reviewed in ref. 29). In addition, verteporfin (VP) was identified in a screen as an inhibitor of the interaction between YAP1 and the TEAD family of transcription factors. However, subsequent studies showed VP had off-target effects and significant solubility issues that limit its use clinically (10, 30, 31). Here we suggest use of DNMTis as an alternative approach to YAP1 inhibition until a specific modulator is available.

Interestingly, the RASSF1 promoter has been reported to be methylated in pediatric rhabdomyosarcoma but not in adult rhabdomyosarcoma (32, 33). This suggests a potential vulnerability of pediatric rhabdomyosarcoma to epigenetic modifiers that can activate RASSF1 and other epigenetically silenced tumor suppressors and may be one advantage of using a DNMTi. Cytosine analogue DNMTis such as guadecitabine are thought to inhibit all DNMT isoforms through their incorporation into the genome. This is an advantage in rhabdomyosarcoma as several isoforms including DNMT1 and DNMT3B have been shown to promote rhabdomyosarcoma tumorigenesis (34, 35). We see a significant effect of guadecitabine treatment at inhibiting rhabdomyosarcoma tumor growth in vivo as a single agent but the use of guadecitabine in combination with other treatments such as traditional chemotherapies, immunotherapy, or other targeted therapies has not been evaluated in rhabdomyosarcoma and warrants further exploration.

Finally, this work lays the foundation for future clinical investigations of combination guadecitabine and dasatinib treatment for patients with rhabdomyosarcoma. Importantly, both of these drugs have been used in children and have minimal toxicities. Guadecitabine is currently being tested in over 23 active clinical trials and currently being evaluated in children with another type of sarcoma, GIST (NCT03165721). Dasatinib is already FDA approved for use in children with chronic myelogenous leukemia (CML) or Philadelphia chromosome–positive ALL (36), and has been used in patients with rhabdomyosarcoma (NCT00464620, NCT03041701). Dasatinib is the preferred SFK inhibitor for rhabdomyosarcoma because it has higher specificity for YES1, the highest expressed SFK in rhabdomyosarcoma (23). Furthermore, a combination of imatinib (another SFK inhibitor) and decitabine (the earlier generation DNMTi) was evaluated in a phase II clinical trial in adults with CML and found to be safe and well tolerated (37). Another case report showed combination azacitidine (another DNMTi) and dasatinib to be safe for the treatment of dual CML and myelodysplasia syndrome (38). These reports suggest that, while we observed toxicity after combination treatment in our murine studies, combination DNMTi and dasatinib may be plausible in the clinic. Our results suggest that patients with alveolar rhabdomyosarcoma would benefit more from guadecitabine treatment than patients with embryonal rhabdomyosarcoma. However, due to rarity of this cancer, it is likely clinical trials would include both subtypes of rhabdomyosarcoma.

K.K. Slemmons is a current employee of Amgen. L.J. Helman is a consultant for Roche/Genetech, Boehringer-Ingelheim, SpringWorks, and Amgen. No potential conflicts of interest were disclosed by the other authors.

Conception and design: K.K. Slemmons, C. Yeung, L.J. Helman

Development of methodology: K.K. Slemmons, L.J. Helman

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K.K. Slemmons, J.T. Baumgart, J.O. Martinez Juarez

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K.K. Slemmons, C. Yeung, J.T. Baumgart, J.O. Martinez Juarez, L.J. Helman

Writing, review, and/or revision of the manuscript: K.K. Slemmons, C. Yeung, J.T. Baumgart, L.J. Helman

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K.K. Slemmons, C. Yeung, A. McCalla

Study supervision: L.J. Helman

The authors thank the Single Cell, Sequencing, and CyTOF (SC2) Core Laboratory at the Saban Research Institute of Children's Hospital Los Angeles for their expertise and help in performing and analyzing the RNA-seq experiments presented and the USC Molecular Genomics Core for their expertise and help in performing the Illumina EPIC methylation array. This research was supported by a Rally Foundation for Childhood Cancer Research Fellowship (19FN14), The Saban Research Institute Research Career Development Fellowship (8030-RRI011550), and a CHLA Core Pilot Project Program grant (RRI012015 to K.K. Slemmons), and a USC/Norris Comprehensive Cancer Center CORE Support grant (P30CA014089), the Nautica Endowment Fund, and the Sarcoma Endowment Fund (to L.J. Helman).

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.