YAP1-Mediated Suppression of USP31 Enhances NFκB Activity to Promote Sarcomagenesis

Soft tissue sarcomas are a heterogeneous group of mesenchymal malignancies arising in muscle, fat, cartilage, and connective tissues (1, 2). Whereas loss or mutation of tumor suppressors (i.e., p53) occurs in approximately 50% of sarcomas (3, 4), sequencing of common adult sarcoma subtypes including fibrosarcoma, liposarcoma, and undifferentiated pleomorphic sarcoma (UPS) have produced no evidence of consistent oncogenic driver mutations (4). The lack of targetable oncogenes has stalled the development of therapeutic modalities (5, 6). We focused our efforts on skeletal muscle UPS given its particularly aggressive nature, lack of nonsurgical treatment strategies, and its high frequency of diagnoses relative to other adult subtypes.

We previously reported that deactivation of the Hippo pathway, a signaling cascade that negatively regulates cell proliferation, promotes sarcomagenesis in skeletal muscle–derived UPS (7). Furthermore, genome-wide analysis of The Cancer Genome Atlas (TCGA) sarcoma dataset confirmed that deregulated Hippo signaling is a contributing factor in sarcomagenesis (2). Inactivation of the Hippo pathway stabilizes the transcriptional effector, YAP1, allowing it to translocate to the nucleus and promote a pro-proliferation gene expression program. Inhibition of YAP1 reduces proliferation in multiple sarcoma subtypes, including UPS, although no genetically engineered mouse model (GEMM) of Yap1-deficient UPS existed prior to this study. While YAP1 activity has been implicated in multiple sarcoma subtypes (7–10), its exact function, regulation, and targets in this context are unclear.

The precise cell of origin for UPS is not known (11). However, in skeletal muscle, these tumors are thought to arise from muscle progenitor cells/satellite cells (12). Whereas investigation of the Hippo pathway in muscle-derived sarcomas has been limited to rhabdomyosarcoma (13), in normal myoblasts, persistent elevated YAP1 and NFκB signaling facilitates proliferation and inhibits differentiation (13–15).

In addition to Hippo pathway deregulation, several studies have recently shown that alterations in the epigenetic landscape can promote sarcomagenesis. Specifically, certain pediatric sarcomas are linked to chromosomal translocations of transcription factor loci encoding chromatin-remodeling factors (16, 17). Copy number loss of chromatin modulators have also been found in some subtypes (15). Together, these studies suggest that disruption of chromatin architecture may be a common event in sarcomagenesis. Our recent work showed that treatment with the histone deacetylase (HDAC) inhibitor vorinostat, also known as suberoylanilide hydroxamic acid (SAHA), leads to reexpression of HIF2α and a corresponding 50% reduction in UPS sarcomagenesis in vivo (18). These findings support the hypothesis that epigenetic modulation can reduce tumorigenesis by returning key transcription factors to the expression and activity levels found in quiescent cells.

In this study, we utilized human skeletal muscle UPS samples and the autochthonous mouse model of UPS to determine that YAP1 promotes proliferation and dedifferentiation through persistent hyperactivation of NFκB. Analysis of the chromatin state in human UPS revealed the presence of massive enhancers, or super enhancers, notably at NFκB target gene loci. Using the epigenetic modulators vorinostat/SAHA and JQ1 (BET bromodomain protein inhibitor), we found that tumor growth was reduced secondary to Hippo pathway reactivation and loss of YAP1 expression. Furthermore, we found that YAP1 suppresses expression of ubiquitin specific peptidase 31 (USP31), a newly identified negative regulator of NFκB signaling (19), resulting in uncontrolled proliferation and tumorigenesis. YAP1 inhibition restored expression of USP31, indicating that aberrant stabilization of YAP1 promotes NFκB activity. We conclude that epigenetic therapy can reclaim control of YAP1-mediated NFκB signaling and subsequently inhibit tumor cell proliferation, induce differentiation, and decrease sarcomagenesis in vivo.

All experiments were performed in accordance with NIH guidelines and were approved by the University of Pennsylvania Institutional Animal Care and Use Committee. We generated Kras^G12D+; Trp53^fl/fl; Yap1^fl/fl (KPY) mice by crossing KP and YAP1^fl/fl animals. Tumors were generated by injection of a calcium phosphate precipitate of adenovirus expressing Cre recombinase (University of Iowa, Iowa City, IA) into the right gastrocnemius muscle of 3- to 6-month-old mice.

For subcutaneously implanted tumors, 1 × 10⁶ KP230 cells were injected subcutaneously into the flanks of 6-week-old nu/nu mice (Charles River Laboratories). Animals were euthanized after 14 days of JQ1 peritoneal injection. For Rela-mediated knockdown KP230 allografts, 1 × 10⁶ cells were injected into the flanks of mice with control tumor cells (scrambled shRNA) and experimental tumor cells (Rela shRNA). Tumor size was measured every other day, and animals were euthanized after 21 days posttumor cells' injection. Tumor volume was calculated by using the formula (ab²)π/6, where a is the longest measurement and b is the shortest.

For in vivo drug studies, total 44 (n = 11 per group) autochthonous KP mice were randomly divided into 4 groups to receive different treatments once tumors reached 100 mm³, and injected for up to 20 days. The mice were euthanized 24 hours after the tumor volume reached 2,000 mm³): (i) Vehicle group (10% Hydroxypropyl-β-cyclodextrin plus DMSO was diluted daily in sterile 45% PEG/55% H₂O); (ii) JQ1 group (JQ1 was diluted daily in 10% HP-β-CD injected into the peritoneal cavity 25 mg/kg twice daily); (iii) SAHA group (SAHA was diluted in sterile 45% PEG/55% H₂O injected into the peritoneal cavity 50 mg/kg daily); (iv) JQ1 and SAHA combination treatment group (drugs were diluted in its vehicles, respectively). Treatment method for drug combination group: (i) 25 mg/kg SAHA + 50 mg/kg JQ1 for first 5 days; (ii) 25 mg/kg SAHA + 25 mg/kg JQ1 each other day for 10 days; (iii) 25 mg/kg SAHA + 50 mg/kg JQ1 for 2 days; and (iv). then mice with tumors received 25 mg/kg SAHA with 25 mg/kg JQ1 for 3 days. Mice without tumors received 5 mg/kg SAHA and 5 mg/kg JQ1 for 3 days. JQ1 was provided by J. Qi (Dana-Farber Cancer Institute, Boston, MA) and SAHA was purchased from Cayman Chemical. HP-β-CD and PEG400 were obtained from Sigma-Aldrich.

Human HT-1080 (fibrosarcoma), RD (rhabdomyosarcoma), LPS224 and LPS863 (liposarcoma), SKLMS1 and SKUT1 (leiomyosarcoma), and HEK-293T cell lines were purchased from ATCC. STS-109 and STS-48 cell lines were derived from human UPS patients, then validated by Dr. Rebecca Gladdy (Sinai Health System, Toronto, Ontario, Canada). KP230 and KIA cell lines were derived from UPS mouse tumors as described in Eisinger-Mathason and colleagues (4). Short tandem repeat analysis was performed at the time of derivation and confirmed in April 2015. Cells were purchased, thawed, and then expanded in the laboratory. Multiple aliquots were frozen within 10 days of initial resuscitation. For experimental use, aliquots were resuscitated and cultured for up to 20 passages (4–6 weeks) before being discarded. Cells were cultured in DMEM with 10% (v/v) FBS and 1% penicillin/streptomycin. All cell lines were confirmed to be negative for mycoplasma contamination.

shRNA-mediated knockdown of AMOT TRCN: 0000166812, 0000165373, 0000162009, 0000159177, 0000162010; Amot TRCN: 00001266880, 0000126883, 0000349515, 0000126881, 0000317410. Yap1 TRCN: 0000095864, 0000095867, 0000095868; YAP1 TRCN: 0000107266, 0000107267, 0000015547; Usp31 TRCN: 0000092218, 0000092219, 0000092220, 0000092221, 0000092222, Rela (p65 NFκB) TRCN: 0000055343, 0000055344, 0000055346, 0000055347, and Scramble shRNA were obtained from Addgene. shRNA plasmids were packaged by using the third-generation lenti-vector system (VSV-G, p-MDLG, and pRSV-REV) and expressed in HEK-293T cells. Supernatant was collected at 24 and 48 hours after transfection and subsequently concentrated by using 10-kDa Amicon Ultra-15 centrifugal filter units (Millipore). Lenti GFP-AMOT p130 (plasmid 32828), Lenti GFP-AMOT p130 Y242/287A (plasmid 32829), YAP1-V5 (plasmid 42555), and YAP1 (S6A)-V5 (plasmid 42562) were purchased from Addgene. pLX304 vector was obtained from Addgene. Lenti-pcdh-EF1 and Lenti-pcdh-GFP were used as the empty vector control in our laboratory.

Twenty human UPS paraffin-embedded tissues obtained from the Surgical Pathology group at University of Pennsylvania (Philadelphia, PA) were stained for both p-p65 and Yap1. IHC was performed on 5-μm tissue sections according to standard protocols. Slides were digitally scanned by the Pathology Core Laboratory at the Research Institute at the Children's Hospital of Philadelphia (Philadelphia, PA). The software of Aperio ImageScope (Leica Biosystems) was used for the quantification of slides. Modified macro (Nuclear v9 parameter) was used to distinguish the intensity of staining. The following antibody concentrations were used: rabbit anti-YAP1 (4912; 1:100; Cell Signaling Technology), anti-phospho p65 (86299; 1:250; Abcam) rabbit anti-Ki-67 (15580; 1:100; Abcam), rabbit anti-MYOD (18943-1-AP; 1:100; Proteintech). IHC of human soft tissue sarcoma and smooth/skeletal muscle samples was performed by using core biopsy arrays (US Biomax, #SO801a). Images were taken by Leica 500 microscope and analyzed by using Photoshop CS3 (Adobe Systems). For quantification, 5 areas of tumor and 5 areas of adjacent muscle (identified by hematoxylin and eosin staining) were captured per section and averaged to determine the mean % positive nuclei/section.

Protein lysate was prepared in SDS/Tris (pH 7.6) lysis buffer, separated by electrophoresis in 8%–10% SDS/PAGE gels, transferred to nitrocellulose membrane, and probed with the following antibodies: rabbit anti-YAP1 (4912; 1:1,000), rabbit anti-p-Yap (Ser397; 13619; 1:1,000), rabbit anti-GAPDH (2118; 1:1,000), mouse anti-AMOT (60156-1-lg 1:500), rabbit anti-V5-tag (13202; 1:1,000), rabbit anti-p65 (8242; 1:500), rabbit anti-p-p65 (Ser536; 3033; 1:500), rabbit anti-caspase-3 (9662; 1:1,000), rabbit anti-HSP90 (4875; 1:1,000; Cell Signaling Technology), rabbit USP31 (12076-1-AP; 1:1,000), rabbit anti-MYOD1 (18943-1-AP; 1:500; Proteintech), rabbit FOXM1 (sc-502; 1:500; Santa Cruz Biotechnology), c-Myc (32072; 1:2,000), p57 Kip2 (75974; 1:500; Abcam), mouse anti-Lamin B2 (D18; University of Iowa Developmental Studies Hybridoma Bank). SuperSep Phos-tag gel was purchased from Wako. Rabbit anti-LATS1 (3477; 1:1,000), rabbit anti-MST1 (3682; 1:1,000), and rabbit anti-MST2 (3952; 1:1,000; Cell Signaling Technology) were used to detect phosphorylated proteins.

Plasmid pHAGE NFκB-TA-LUC-UBC-GFP-W (49343; Addgene) was transfected into 293T cells (ATCC) to generate lentiviral particles in the supernatant. Viral supernatant was harvested and then concentrated by centrifugal filter units (Amicon Ultra-15, Millipore). NFκB reporter virus was transduced into KP230 cells. GFP-positive NFκB reporter cells were sorted generating an approximately 85% pure GFP⁺ cell line. For shRNA assays, the NFκB the reporter cell line was transduced with lentivirus expressing control, Usp31, or Yap1 shRNA. For drug studies, the NFκB reporter cells were treated with SAHA (2 μmol/L)/JQ1 (0.5 μmol/L) and BAY 11-7085 (1.5 μmol/L, Selleckchem) for 48 hours. Twelve hours prior to detecting luciferase activity, TNFα (10 ng/mL, R&D Systems) was added. Luciferase activity was assayed using the Dual Luciferase Assay System (E2920, Promega) according to the manufacturer's protocol on a Luminometer (GLOMAX, Promega). Results were calculated as fold induction.

We used the publicly available database Detwiller and colleagues through Oncomine Research Premium edition software (version 4.5, Life Technologies) to query AMOT expression in sarcomas and normal tissues. YAP1, MYC, and PHLDA1 gene expression were correlated with overall survival in patients with MFH/UPS and DDLS, using the TCGA sarcoma dataset. Kaplan–Meier analyses were performed for overall survival of patients.

Differential gene expression was tested using significance analysis of microarrays (SAM, samr v2.0), yielding fold change, q-value (false discovery rate) and d-score for each gene. We observed a small number of genes meeting our cutoffs for differential expression and so proceeded to GSEA. Log₂-transformed RMA-sst expression values were used as input to GSEA (20) where enrichment was tested against the hallmark gene sets from the Molecular Signatures Database (MSigDB, v5.1, http://software.broadinstitute.org/gsea/msigdb/index.jsp).

C2C12 murine myoblast cells were obtained from ATCC. The cells grow as undifferentiated myoblasts in growth medium (20% FBS with 1% penicillin/streptomycin), and were passaged every 2–3 days at 50% subconfluence. To induce differentiation, cells were grown overnight to approximately 80% confluence in growth medium, and then switched to DMEM supplemented with 2% horse serum. Differentiation media were refreshed every 2 days.

For tumor samples resected from UPS patients at the Hospital of the University of Pennsylvania, approximately 100 mg of tissue was minced into 1–2 mm pieces and incubated in 1% formaldehyde for 15 minutes. Formaldehyde was quenched with glycine at 0.125 mol/L. Fixed tissue was homogenized for 60 seconds with a Tissue Tearor Homogenizer (Biospec) at 30,000 rpm. Homogenized tissue was washed with ice-cold PBS with 1× HALT protease inhibitor. For cell line ChIP-RX, samples were fixed for 10 minutes in 1% formaldehyde quenched with glycine and washed with PBS as above. 5e6 S2 cells (Drosophila melanogaster) were added to each sample of 2.5e7 for ChIP-RX normalization in downstream analysis.

Tissue preprocessed via mortar and pestle homogenization or 2.5e5 cells were homogenized via Qiashredder column (Qiagen). Samples were then processed via mRNeasy Mini Kit (Qiagen). RNA sequencing (RNA-seq) datasets in fastq format were fed into RSEM for mRNA sequence alignment and quantification (21). Gene counts generated by RSEM were fed into edgeR (22) to compute log fold change (logFC) and P value. False discovery rate is used for multiple comparison correction. The GSEA software and Hallmark gene sets (20) were used for GSEA. Genes were ranked by rank score −log(P) × sign(logFC).

Statistical analysis was performed using Prism (GraphPad software). Data are shown as mean ± SEM or SD. Data are reported as biological replicates, with technical replicates indicated in figure legends. Student t tests (unpaired two-tailed) were performed to determine whether a difference between two values is statistically significant different, with P < 0.05 considered significant. In vitro assays were performed in triplicate unless otherwise stated.

Sequencing data reported in this article have been deposited in NCBI's Gene Expression Omnibus and are accessible through GEO Series accession number GSE97295, GSE97296, GSE97297, GSE109920, and GSE109923.

Deregulation of the Hippo pathway is associated with muscle-derived sarcoma subtypes but the downstream mechanisms are unclear (7, 8, 23). Although we have access to only a small number of TCGA patient samples, which prevents our analysis from reaching statistical significance, expression of YAP1 strikingly correlates with long-term survival in human UPS patients (Fig. 1A). To determine the relevance of these findings, we evaluated YAP1 expression in a variety of sarcomas by IHC of a human biopsy tissue microarray. YAP1 expression is particularly high in muscle UPS (Fig. 1B, group 2, boxed), compared with normal mesenchymal tissues including skeletal and smooth muscle (Fig. 1B, group 1). The clinical details of each sample can be found in Supplementary Fig. S1A. To test the role of YAP1 in UPS, we introduced Yap1^fl/fl alleles into our LSL-Kras^G12D/+; Trp53^fl/fl (KP) GEMM of skeletal muscle UPS to generate LSL-Kras^G12D/+; Trp53^fl/fl;Yap1^fl/fl (KPY) animals. Tumors are generated by injection of adenovirus expressing Cre recombinase into the right gastrocnemius muscle. The Cre recombinase activates expression of oncogenic Kras and deletes p53 expression in infected muscle progenitor cells (24, 25). Although Kras mutation is rare in sarcomas, Trp53 mutation and deletion are very common (2). Furthermore, hyperactivation of the MAPK pathway, downstream of KRAS activation, is common in UPS and is an excellent prognostic indicator for recurrence (26). This model and a similar one, Kras^G12D/+; Ink4a/arf^fl/fl (KIA), generate sarcomas that are histologically, transcriptionally, and morphologically identical to UPS and are thus the standard GEMMs in UPS studies (24, 25). Yap1 protein expression is stabilized in KP tumors providing further rationale for the study of Yap1 function in sarcoma using this model (Fig. 1C and D). Loss of Yap1 in KPY tumors was confirmed by Western blot and IHC analyses (Fig. 1C and D; Supplementary Fig. S1B). KP tumors were initially palpated at day 45 post-Adeno-cre injections and were collected when they reached 2 cm³ (maximum tumor volume). KPY tumors were harvested either when they reached 2 cm³ or 60 days post Adeno-cre injection, the time-point at which all KP mice had reached maximum tumor volume. Deletion of Yap1 delayed tumor initiation (median latency KP: 57 days; KPY: 59 days, P < 0.0001) and growth (Fig. 1E–H). Several KPY mice never developed visible tumors and those animals are represented in Fig. 1E with a data point at day 65. Proliferation was reduced by approximately 50% in KPY tumors as defined by Ki67 positivity (Fig. 1D; Supplementary Fig. S1C). To determine the functional role of Yap1 in UPS, we performed microarray analysis of 5 individual KP and KPY tumors. Gene set enrichment analysis (GSEA) revealed that “TNFα-induced NFκB” signaling is significantly reduced in KPY tumors (Fig. 1I and J). Furthermore, using the TCGA sarcoma dataset, we determined that the YAP1-dependent NFκB target PHLDA1 is associated with poor survival in UPS (Fig. 1K). The specific effects of YAP1 on Tnfα-mediated signaling suggest high levels of Tnfα cytokine production in the UPS tumor microenvironment, which activates downstream NFκB signaling. To validate Tnfα expression in UPS tumor tissue, and to determine the source of Tnfα production, we fractionated KP tumors and isolated tumor-associated macrophages (TAM), which are known to produce significant amounts of Tnfα. We compared Tnfα expression in isolated KP tumor cells and TAMs. Tnfα mRNA expression was approximately 45-fold higher in TAMs than in tumor cells (Fig. 1L, left). To determine whether the relative expression of Tnfα in KP TAMs was physiologically significant, we compared this in vivo induction with that of the classical in vitro model of LPS-stimulated murine bone marrow–derived macrophages (BMDM). LPS-stimulated BMDMs express approximately 100 fold more Tnfα than untreated BMDMs (Fig. 1L, right). Together, these data suggest that KP TAMs produce a physiologically relevant amount of Tnfα, which could easily activate NFκB signaling. However, NFκB signaling may be activated by additional mechanisms in this context.

We tested the ability of Yap1 to control NFκB activity in vitro using an established luciferase reporter assay (27). Sarcoma cells derived from a KP tumor were infected with a GFP-labeled NFκB reporter construct, sorted to approximately 86% purity, and then infected with lentivirus expressing control or Yap1 shRNA. NFκB activity is reduced by approximately 50% in Yap1 shRNA–expressing cells, treated with Tnfα, compared with treated control shRNA-expressing cells (Fig. 1M). From these findings we conclude that YAP1 promotes NFκB activity, which may enhance sarcomagenesis.

To determine the extent of NFκB signaling in human UPS, we performed multiple genome-wide analyses of these tumor samples directly from patients. We performed acetylated histone 3 lysine 27 (H3K27Ac) chromatin immunoprecipitation with massively parallel DNA sequencing [chromatin immunoprecipitation sequencing (ChIP-seq)] using three independent UPS patient samples originating in skeletal muscle tissue (Supplementary Fig. S2A). Acetylation of lysine residues on histone tails is linked to active euchromatin transcription, especially at H3K27Ac, which is associated with active cis-regulatory elements (28). Therefore, high levels of H3K27Ac at the enhancer regions of NFκB target genes are directly associated with elevated NFκB activity. The observed genome-wide binding of H3K27Ac antibody was consistent in all three UPS samples. We then conducted focused epigenomic analysis on regions of dramatic H3K27Ac enrichment, frequently referred to as super enhancer (SE) analysis, across all three samples to map the regions of highly active enhancers across the whole genome, and identify regions of differential hyperacetylation (SE; Fig. 2A). We identified activation of multiple SE regions associated with NFκB from the set of overall SEs. The SEs associated with NFκB pathway genes correlated well with average H3K27Ac signal density in representative human UPS at SEs (Fig. 2B and C). More convincingly, NFκB pathway components, such as RELA, BCL3, and RELB showed enhanced H3K27Ac signal in all three UPS samples (Fig. 2D–F), the input data are listed in Supplementary Fig. S2B and S2D, and the identified SEs were most significantly associated with the hallmark M5890 gene set referred to as “TNFA signaling via NFκB” (Fig. 2G–I) in all three human samples. As a control assay for H3K27Ac ChIP-Seq, we performed RNA-seq on human UPS tumor sample #1 compared with normal adult skeletal muscle. Using differential gene expression analysis (29), we observed that TNFα and NFκB pathway genes, including RELA itself, were significantly upregulated in UPS, while the muscle differentiation markers, MEF2C and MYOD1, were contrastingly downregulated (Fig. 2J, enlarged in Supplementary Fig. S2E). Likewise, we noted high expression of NFκB targets in UPS, including many of those identified as YAP1-regulated in the KP/KPY microarray [i.e., CCL2 (30), BCL3 (31), PHLDA1 (32); Fig. 1I]. Furthermore, GSEA confirmed that genes of the “TNFA signaling via NFκB” pathway are significantly and increasingly differentially expressed in UPS compared with skeletal muscle (Fig. 2K), whereas “myogenesis” associated gene expression is inhibited in these tumors (Supplementary Fig. S2F). These RNA-Seq observations validate the utility of the H3K27Ac ChIP-Seq predicting the transcriptional profile of UPS. Thus, we have identified a “pro-proliferation” gene signature in human UPS that correlates well with our observations in the KP GEMM.

The ChIP-seq and RNA-seq experiments described above were performed on a limited number of fresh frozen patient samples (n = 3) due to the paucity of available UPS tissues. Therefore, we sought a larger scale approach to evaluate NFκB activity in human UPS tissues. We performed IHC of 20 independent human skeletal muscle UPS tumor sections provided by Dr. Kumarasen Cooper (Department of Pathology & Lab Medicine, University of Pennsylvania, Philadelphia, PA). We evaluated active NF-κB (phospho-p65/p-p65) and YAP1 protein levels and found that both proteins are highly expressed in approximately 80% of tumors (Fig. 3A) relative to normal adjacent muscle tissue (Fig. 3B–D) and the expression is specifically nuclear. To confirm the functional importance of NFκB signaling in UPS, we inhibited Rela (the gene encoding p65) with two independent and specific shRNAs in KP cells. Rela shRNA #1 reduced expression by approximately 50%. shRNA #2 was less effective, reducing expression by only approximately 25% (Fig. 3E and F). Control and Rela shRNA–expressing KP cells were implanted subcutaneously into nude mice. Importantly, Loss of Rela significantly reduced final tumor volume (Fig. 3G and H) and weight (Fig. 3I) and these reductions directly correlated with the degree of knockdown provided by each shRNA. On the basis of these findings, we postulated that sarcoma cells are sensitive to NFκB inhibitors. We compared proliferation of KP cells treated with two independent NFκB inhibitors caffeic acid phenethyl ester (CAPE; refs. 33, 34) and BAY 11-7085 (35) to similarly treated HCT-116 colorectal cancer cells. HCT-116 cells are thought to be particularly sensitive to NFκB inhibition (36). KP cells are two times as sensitive to CAPE inhibition and three times as sensitive to BAY 11-7085 as HCT-116 cells (Fig. 3J and K). Together, these findings clearly show that NFκB activity promotes UPS growth and that YAP1 expression correlates with NFκB upregulation in human UPS.

On the basis of our identification of YAP1 as a mediator of tumorigenesis, we sought inhibitory small-molecule approaches in vivo. The existing YAP1 inhibitor (i.e., verteporfin) has nonspecific effects as well as cell permeability challenges. Recent findings from our group and others suggest a strong epigenetic basis for soft tissue sarcoma development and efficacy of HDAC inhibitors in preclinical UPS models (18, 37–39). Similarly, Mazur and colleagues published an RNA-seq dataset of KP-driven PDAC that suggested epigenetic modulation increased expression of the key YAP1 inhibitor AMOT (40). Therefore, we queried whether inhibitors of chromatin modifiers/readers would be effective against muscle-derived UPS in a YAP1-dependent manner. We hypothesized that the chromatin landscape in UPS is permissive with regard to NFκB signaling, which promotes constitutive activation of NFκB SEs. Therefore, we focused on inhibitors that modify the SE landscape, such as the bromodomain inhibitor, JQ1 (41). To test this hypothesis, we initially treated three independent sarcoma cell lines with increasing doses of JQ1, including human fibrosarcoma cells, KP cells, and a cell line derived from a second sarcoma muscle UPS GEMM, the KIA model. Although not specifically a muscle-derived tumor, some fibrosarcomas are now thought to be genetically indistinguishable from UPS (2). Consistent with this observation, HT-1080 fibrosarcoma cells require YAP1 for proliferation and share critical proliferation mechanisms with UPS cells (7). Therefore, we included HT-1080 cells in our current studies. All three cell lines were sensitive to JQ1 (Supplementary Fig. S3A). Therefore, we tested the effect of 50 mg/kg daily JQ1 on our KP-derived subcutaneous allograft sarcoma model in nude mice. JQ1 dramatically reduced tumor progression (Fig. 4A) with a 5-fold reduction in final tumor volume (Fig. 4B) and a 6-fold reduction in final tumor weight (Fig. 4C) compared with DMSO-treated controls. Interestingly, we also found that JQ1 inhibits expression of Yap1 in vivo (Fig. 4D; Supplementary Fig. S3B). Given JQ1's many potential molecular effects, a rescue assay was performed to determine whether JQ1-mediated inhibition of proliferation is YAP1-dependent. We used HT-1080 cells, which are sensitive to 250 nmol/L JQ1, easily transduced, and proliferate under control conditions at a less rapid pace than the other sarcoma cell lines. This last parameter is important because it allowed us to ectopically express a constitutively nuclear/active YAP1 mutant, YAP1(S6A), and observe the expected significant increase in proliferation (Fig. 4E), a necessary positive control for the rest of the experiment. YAP1(S6A) rescued approximately 40% of the JQ1-induced proliferation deficit by day 5 (Fig. 4E). To confirm that the role of YAP1 in proliferation is dependent on its nuclear functions, we performed a similar rescue assay using WT-YAP1, which is shuttled back and forth from the nucleus to the cytoplasm. WT-YAP1 also rescued proliferation, but the effects were less substantial as predicted (Fig. 4F). Similar, but less dramatic results were observed in the more proliferative KP cells (Supplementary Fig. S3C). Western blot analysis of mutant and WT YAP1–expressing cells shows the level of expression of V5-tagged constructs in each condition (Supplementary Fig. S3D). Consistent with the above findings, we observed that 250–500 nmol/L JQ1 treatment lowers YAP1 mRNA levels by 50% in HT-1080, KIA, and KP cells (Supplementary Fig. S3E and S3F).

In some cancer contexts, the efficacy of JQ1 is dependent on its ability to inhibit the MYC oncogene. However, that is not the case in the sarcoma subtypes we studied (Supplementary Fig. S4A–S4D). Our previously published work showed that the pan-HDAC inhibitor SAHA suppressed sarcoma proliferation in vitro and tumor growth in vivo by 50% (18) and we hypothesized that combination SAHA/JQ1 treatment would be more effective than either drug individually. MTT proliferation studies comparing dose escalation of SAHA and JQ1 separately and in combination showed that the combination of both drugs at lower doses is more effective than either drug alone in vitro (Fig. 4G and H; Supplementary Fig. S4E). Consistent with our proliferation findings, combination therapy [SAHA (2 μmol/L)/JQ1 (0.5 μmol/L)] most effectively inhibited expression of both Yap1 and its transcriptional target Foxm1 in KP cells (Fig. 4I). We saw similar effects on YAP1 and FOXM1 in six additional human sarcoma cell lines representing multiple subtypes including rhabdomyosarcoma, liposarcoma, leiomyosarcoma, and UPS (Supplementary Fig. S4F). These findings suggest potential broader applicability of SAHA/JQ1 against human sarcomas.

We sought to determine how SAHA/JQ1 inhibits YAP1 activity so effectively. In addition to AMOT-mediated inhibition of YAP1, its activity is regulated by Hippo kinase LATS1/2-dependent phosphorylation (42). SAHA/JQ1 treatment does not change expression of the upstream hippo kinases MST1/2 or LATS1 (Supplementary Fig. S4G and S4H). However, we did observe a modest increase in LATS1 phosphorylation, suggesting a possible but likely minor increase in activity (Supplementary Fig. S4G*). Importantly, YAP1 phosphorylation and cytoplasmic localization increased substantially in response to SAHA/JQ1 (Supplementary Fig. S5A and S5B). Therefore, we focused on AMOT-mediated YAP1 inhibition.

Importantly, AMOT is highly expressed in human skeletal muscle (43, 44), but is lost in all commonly diagnosed sarcomas according to the Detwiller and colleagues' sarcoma dataset (Supplementary Fig. S6A; ref. 45). Our RNA-seq of human UPS confirmed this finding (Fig. 2J), suggesting that AMOT may function as a tumor suppressor in sarcoma. To test whether Amot is the predominant inhibitor of Yap1 in our cells, we used the Tankyrase inhibitor XAV939, which stabilizes Amot protein levels (46). Increasing doses of XAV939 clearly result in higher levels of YAP1 phosphorylation and loss of total YAP1 (Supplementary Fig. S6B). To confirm that expression of AMOT suppresses proliferation in sarcoma, we transduced human HT-1080 cells with GFP-tagged WT-AMOT. We sorted the GFP-AMOT⁺ population to 85% purity (Supplementary Fig. S6C). By immunofluorescence (IF), we observed that GFP-AMOT appropriately localizes to the perinuclear region as has been previously reported (Supplementary Fig. S6D; ref. 47). Control vector and GFP-AMOT expressing cells were pulsed with fluorescent CellTrace dye. CellTrace levels were appropriately depleted during each control cell division as dye molecules are distributed to daughter cells. GFP-AMOT–expressing cells quantified by flow cytometry contained approximately 50% more CellTrace than control cells, indicating that cell division in experimental cells was reduced by half (Supplementary Fig. S6E). Furthermore, analysis of the control GFP⁺ and GFP-AMOT⁺ cells at day 5 revealed that AMOT-GFP⁺ cells only are reduced from approximately 85% to approximately 45% of the population, having been overtaken in culture by GFP⁻ cells (Supplementary Fig. S6F). Our findings show that expression of AMOT is a strong YAP1 inhibitor in our cells and accordingly suppresses sarcoma cell proliferation.

Together, our studies revealed an AMOT-dependent mechanism of YAP1 degradation in SAHA/JQ1–treated sarcoma cells. Consistent with this observation, loss of YAP1 transcription, together with increased expression of AMOT, resulted in near ablation of YAP1 targets FOXM1 and BIRC5 in SAHA/JQ1–treated cells (Fig. 4I–K). We expanded our observations to include additional murine (KIA) and human (STS-48, STS-109) UPS cell lines (Fig. 4L). Next we asked whether SAHA/JQ1 treatment altered NFκB signaling. Using the NFκB luciferase system in KP cells, we evaluated NFκB activity after 12 or 48 hours of SAHA/JQ1 treatment. The NFκB inhibitor BAY 11-7085 was used as a control and reduced NFκB activity by 50% at both time points. We observed that NFκB activity doubled after 12 hours of SAHA/JQ1 treatment and then plummeted to 50% of the TNFα-treated positive control at 48 hours (Fig. 4M). This finding is consistent with published observations that NFκB transcriptional activity is extremely dynamic and oscillates over time in response to stimuli (48). However, we observed that phosphorylation of p65, a key readout of NFκB activity, remains high after 12 hours of SAHA/JQ1 treatment (Fig. 4N). We hypothesize that elevated NFκB activity after 12 hours of treatment is due to increased nuclear localization of p-p65. By 48 hours of treatment, p-p65 levels were substantially reduced. These observations suggest that SAHA/JQ1 may induce oscillation, but that treatment ultimately reduces NFκB signaling.

Combination SAHA/JQ1 treatment inhibits Yap1 and dramatically reduces proliferation of sarcoma cells (Fig. 4G–I). One possible explanation for this finding is induction of apoptosis or differentiation. Importantly, persistent YAP1 and NFκB signaling are central components of myoblast proliferation and are suppressed during differentiation (14, 16, 49–51). Therefore, we tested the possibility that epigenetic therapy differentiates sarcoma cells into a less malignant muscle-like cell. First, we evaluated NFκB signaling in the C2C12 murine model of normal myoblast differentiation. We treated undifferentiated C2C12 myoblasts with differentiation media (DM) for up to 6 days and evaluated differentiation markers, NFκB targets, and myotube formation. Differentiation markers (Myh1, Myh2, Mef2c, and Myh3) were all upregulated dramatically by day 6 (Fig. 5A), coincident with myotube formation (Fig. 5B). We also observed that expression of the YAP1-dependent NFκB targets Ccl2 and Hbegf oscillated over time (Fig. 5C), consistent with our NFκB luciferase activity observations (Fig. 4M) and published findings (48). Furthermore, YAP1-mediated NFκB targets/regulators (Ccl2, Hbegf, Areg, and Phlda1) did not oscillate in proliferating C2C12 cells in growth media (GM) for 3 days (Fig. 5D). Beyond 3 days, confluence initiates expression of differentiation markers, and we can no longer accurately measure proliferation-associated gene expression. We next evaluated expression of these targets in KP cells treated with SAHA/JQ1 and observed the oscillation associated with myoblast differentiation (Fig. 5E). Interestingly, NFκB targets do not oscillate synchronously either in SAHA/JQ1–treated KP cells or in differentiating C2C12 cells (Fig. 5C), suggesting that timing of target expression may be dependent on production of specific NFκB cofactors (52). The temporal oscillation of individual targets will have to be studied in depth to fully understand the role of these patterns in differentiation and proliferation. Our data suggest that epigenetic therapy of sarcoma cells restores the NFκB oscillation observed in differentiating myoblasts. Consistent with this hypothesis, SAHA/JQ1–treated KP cells have a significantly different morphology than control or individually drug-treated cells, although they do not form myotubes, many cells appear flat, multinucleated, and in some cases seem to be merging (Fig. 5F, arrows). To determine whether a muscle differentiation program is induced transcriptionally, we analyzed differential changes in gene expression by microarray of DMSO and SAHA/JQ1–treated KP cells. Metascape analysis of genes enriched in DMSO-treated cells showed most significant association with cell cycle and cell division (Fig. 5G). However, genes enriched in SAHA/JQ1–treated cells were associated with numerous processes linked to normal muscle function, including endoplasmic reticulum (ER) stress, autophagy, catabolic metabolism, mitochondrial function, and lipid processing (Fig. 5H). (53). Similarly, Ingenuity Pathway Analysis (IPA) of our KP/KPY tumor microarray revealed that Yap1 loss induces ER stress and unfolded protein response (UPR), as well as the expected changes in TNF receptor signaling suggesting that YAP1 contributes specifically to these processes during muscle differentiation (Supplementary Fig. S7A). The lack of induction of pathways associated with terminal skeletal muscle differentiation including hallmark “Myogenesis” suggests that SAHA/JQ1 promotes incomplete differentiation. To validate upregulation of a muscle phenotype in SAHA/JQ1–treated cells, we evaluated expression of the muscle differentiation markers MEF2C, MYOD1, and CDKN1C (p57). Both MEF2C and CDKN1C mRNA levels were substantially increased due to SAHA/JQ1 treatment in KP and HT-1080 cells (Fig. 5I and J). However, MYOD1 mRNA levels were unaffected (Fig. 5I). MYOD is regulated at the protein level by its physical interaction with p57 (54); therefore, we tested expression of Myod and p57 by Western blotting of murine (Fig. 5K) and human (Fig. 5L) UPS cells and saw both proteins were increased in drug-treated cells. Importantly, expression of YAP1 and its transcriptional target FOXM1 (7) were nearly abolished (Fig. 4I–L). We saw similar effects in HT-1080 cells. Together, these findings show that combination SAHA/JQ1 treatment promotes differentiation, potentially due to inhibition of YAP1-mediated transcriptional regulation.

Next we sought to determine whether Yap1 inhibition via SAHA/JQ1 mediates the differentiation phenotype and identify the associated mechanism. First, we investigated whether SAHA (2 μmol/L)/JQ1 (0.5 μmol/L) treatment induced apoptosis. We performed Annexin V/Propidium Iodide (PI) assays by flow cytometry using 1 μmol/L staurosporin as a positive control. Whereas staurosporin treatment dramatically increased Annexin/PI positivity compared with vehicle control, SAHA/JQ1 did not alter apoptotic levels (Fig. 6A and B). Consistent with these findings, we observed no change in cleaved caspase-3 levels due to SAHA/JQ1 treatment, whereas staurosporin induced the predicted increase in caspase-3 cleavage (Fig. 6C; Supplementary Fig. S7B). Next, we tested the prediction that dampening NFκB signaling recapitulates the effect of SAHA/JQ1 on the cell cycle. We treated sarcoma cells with the NFκB inhibitor CAPE and performed bromodeoxyuridine (BrdUrd) analysis by flow cytometry and compared the results with SAHA/JQ1–treated cells. Both treatments induced cell-cycle arrest in G₂ phase, although SAHA/JQ1 is significantly more effective (Fig. 6D and E). To determine whether NFκB inhibition alone is sufficient to induce Myod expression and differentiation, we investigated Myod levels in sarcoma cells treated with the NFκB inhibitor BAY117085 and found that NFκB inhibition alone could not induce MYOD1 protein expression (Fig. 6F). Additional epigenetic changes due SAHA/JQ1 are necessary to facilitate the muscle differentiation program.

Given the widespread effects of YAP1 signaling on NFκB activity (Fig. 1I–K and M), we hypothesized that a critical upstream regulator of the NFκB pathway is modulated by YAP1. Recently identified as a critical upstream negative regulator of NFκB, the peptidase USP31 controls ubiquitination of the TRAF molecules, which convey signals initiated by TNFα receptor engagement downstream to p65/NFκB (19). Usp31 levels increase in response to SAHA/JQ1 in murine KP cells (Fig. 6G). Interestingly, the time point at which USP31 induction is maximal varies between cell lines according to rate of proliferation. Rapidly growing KP cells require 72 hours of treatment. Consistent with this observation, 72 hours of drug treatment in KP cells decreased expression of NFκB targets we identified as upregulated in human UPS (Fig. 2A, Litaf) and as Yap1-mediated in KP tumors (Fig. 1I and K, Phlda1; Fig. 6H). The importance of this observation will be defined in later studies. Slower growing human fibrosarcoma (HT-1080) and UPS (STS-109) cells only need 12–48 hours to induce maximal USP31 (Fig. 6I and J). Importantly, these findings support the conclusion that SAHA/JQ1 inhibits NFκB activity in multiple cell lines and sarcoma subtypes. Furthermore, genetic inhibition of Yap1 via specific shRNA also induces Usp31 expression, indicating that this phenotype is mediated by YAP1 (Fig. 6K). To ascertain the relevance of Usp31 expression in muscle cell differentiation, we evaluated differentiating C2C12 cells and observed that Usp31 expression increases during the 8-day course of differentiation (Fig. 6L). Amot levels increase during precisely the same time period in C2C12 cells, indicating that loss of Yap1 activity occurs simultaneously with induction of Usp31. Importantly, Usp31 levels remained consistent in proliferating C2C12 cells (Fig. 6M). Together, these findings support the hypothesis that YAP1 suppresses USP31 and in doing so maintains persistently high levels of NFκB activity, signaling, and proliferation.

To determine the functional outcome of Usp31 expression, we silenced it with specific shRNA. Fifty percent loss of Usp31 expression significantly increased in vitro KP cell proliferation (Fig. 7A). These data are consistent with increased NFκB activity. Moreover, in the luciferase reporter system Usp31 shRNA expression rescued SAHA/JQ1–mediated suppression of NFκB activity. This finding clearly shows that Usp31 mediates the effects of SAHA/JQ1 on NFκB activity (Fig. 7B). To confirm that drug treatment inhibits NFκB signaling through upregulation of Usp31 we evaluated the early NFκB regulator and direct target of Usp31 peptidase activity, Traf2. K63 linked ubiquitination of Traf2 is dependent upon Usp31 (19). Interestingly, Traf2 protein expression is lost in SAHA/JQ1–treated cells (Fig. 7C, top), whereas our microarray analysis of KP cells showed that Traf2 mRNA levels were not significantly altered relative to Yap1 and its transcription targets Foxm1 and Birc5 (Fig. 7C, bottom). These data suggest that Usp31-mediated K63-linked ubiquitination affects Traf2 protein stability. However, this hypothesis and the nature of YAP1-mediated suppression of USP31 expression (direct vs. indirect) require additional investigation. To show that Yap1's effect on proliferation is predominantly due to its control of Usp31 expression, we performed an in vivo rescue allograft assay. KP cells were infected with control or Yap1 shRNA–expressing lentivirus, puromycin selected for 48 hours, then infected with control or Usp31 shRNA lentivirus. Western blot analysis of Yap1 and qRT-PCR expression of Usp31 confirms inhibition of expression in the appropriate cells (Fig. 7D). As we have previously reported, expression of Yap1 shRNA significantly reduced tumor growth and weight compared with Scr shRNA control (7). Consistent with our in vitro proliferation findings (Fig. 7A), Usp31-specific shRNA increased tumor growth (Fig. 7E and F). Most importantly, loss of Usp31 expression rescued the Yap1 shRNA-mediated reduction in tumor growth. We therefore conclude that Usp31 suppression is critical for YAP1-dependent tumorigenesis.

One of our major goals is to identify novel therapeutic approaches for the treatment of UPS and other sarcomas. To test the potential efficacy of SAHA/JQ1 as a therapeutic strategy, we tested effects of the drug combination in vivo in KP GEMM tumors using 25 mg/kg SAHA and 50 mg/kg JQ1 (Fig. 7G). These doses and the precise schedule were optimized in detail due to the loss of several animals in pilot experiments from drug toxicity. The schedule presented here was safe and no animals died from drug administration. SAHA/JQ1 treatment caused statistically significant tumor regression in this very aggressive autochthonous model (Fig. 7H). Furthermore, the combined therapy delayed time to maximum tumor volume by 2-fold (Fig. 7I). Median survival of animals bearing control-treated tumors was 19 days relative to 25 days (SAHA) and 33 days (JQ1). The combination of SAHA and JQ1 was most effective with a median survival of 39 days (χ²P < 0.0001). Importantly, Amot levels were significantly elevated in SAHA/JQ1–treated KP tumor tissue compared with controls (Fig. 7J), which is consistent with our in vitro findings. Subsequent IHC of SAHA/JQ1–treated KP autochthonous tumors revealed that Myod levels were elevated while Ki67 positivity was lost in vivo (Fig. 7K). Together, these findings show that epigenetic modulation of the Hippo pathway restores normal NFκB activity, leads to decreased sarcomagenesis, and increased muscle differentiation in vivo via Usp31.

Targetable oncogenic driver mutations are rare in sarcomas. As a result, there are no effective targeted therapies. Our goal was to identify mutant oncogene-independent mechanisms of sarcomagenesis in the subtypes common to adults (i.e., UPS and fibrosarcoma) and potential therapeutic interventions. Our study highlights the efficacy of combined epigenetic inhibitors against UPS. Although not specific inhibitors, SAHA and JQ1 may provide a new option for these patients.

In this study, we focused on understanding the pathways regulated by YAP1 in sarcomagenesis. Through genetic and epigenetic studies, we found that YAP1 contributes to hyperactivation of NFκB, a signaling pathway with a crucial role in muscle progenitor cell division and differentiation. In fact, ChIP-seq for H3K27Ac and super enhancer analysis of human UPS samples led to the identification of NFκB as the most transcriptionally active pathway in UPS. YAP1-mediated control of NFκB target expression was confirmed in a GEMM of sarcoma. On the basis of our findings, we postulate that NFκB signaling is consistently upregulated in human UPS and is enhanced by persistent YAP1 stabilization.

NFκB signaling is complex and requires the expression and posttranslational modification of multiple key effectors. Our studies revealed that YAP1 promotes NFκB activity by controlling expression and stability of several key upstream regulators of NFκB signaling (i.e., USP31 and TRAF2). Using our YAP1-deficient UPS GEMM, specific shRNAs, and SAHA/JQ1, we determined that NFκB target expression, transcriptional activity, and phosphorylation/activation of p65 (the key dimer of the NFκB transcription factor) are all decreased when YAP1 is lost. Importantly, our studies reveal two parallel alterations in NFκB signaling found in human UPS (i) upregulation of YAP1 alters expression USP31, a negative regulator of NFκB and (ii) chromatin modification at NFκB target loci is modified to allow transcription of critical genes. Together, these conditions promote NFκB activity and sarcomagenesis.

Here we also report that the mechanism by which SAHA/JQ1 inhibits YAP1 expression is multi-pronged. The combination decreases YAP1 mRNA levels by ∼50% and increases expression of the YAP1 inhibitor, AMOT, which is silenced in UPS compared to normal skeletal. SAHA/JQ1 restores AMOT expression, inhibiting YAP1 activity. SAHA/JQ1 treatment restores control of Hippo pathway signaling, inhibiting proliferation and simultaneously initiating a muscle differentiation transcriptional program.

The most critical YAP1 target we identified was the novel ubiquitin peptidase USP31. Virtually nothing is known about USP31 beyond its structure and connection to NFκB signaling (19, 55). USP31 was induced in response to YAP1 inhibition and SAHA/JQ1. Importantly, we observed that Usp31 was upregulated in differentiating C2C12 myoblasts, suggesting that in sarcoma cells SAHA/JQ1 and Yap1 loss may restore differentiation (Fig. 7L). Together, our findings suggest that muscle-derived UPS cells behave like proliferating myoblasts incapable of undergoing differentiation due to persistent YAP1 and NFκB signaling. SAHA/JQ1 inhibits YAP1 and restores NFκB to patterns observed in differentiating myoblasts, forcing tumor cells to differentiate.

Overall, our work establishes NFκB, a key regulator of normal muscle development, as a pathway that becomes persistently upregulated during sarcomagenesis at least in part through aberrant suppression of the YAP1 target USP31. Collectively, our data suggest that YAP1 stabilization, p65 phosphorylation and AMOT suppression could potentially serve as useful biomarkers for UPS and provide the mechanistic rationale for epigenetic therapy to restore Hippo pathway activity to normal levels in the treatment of this disease.

No potential conflicts of interest were disclosed.

Conception and design: S. Ye, J. Qi, T.S.K. Eisinger-Mathason

Development of methodology: A. Rivera-Reyes, J. Qi

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Ye, M.A. Lawlor, S. Egolf, S. Chor, K. Pak, G.E. Ciotti, A.C. Lee, G.E. Marino, J. Shah, D. Niedzwicki, P.M.C. Park, M.Z. Alam, M. Haldar, J. Qi

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Ye, M.A. Lawlor, A. Rivera-Reyes, P.M.C. Park, M. Xu, J. Qi, T.S.K. Eisinger-Mathason

Writing, review, and/or revision of the manuscript: S. Ye, K. Weber, A. Grazioli, J.A. Perry, J. Qi, T.S.K. Eisinger-Mathason

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Qi, T.S.K. Eisinger-Mathason

Study supervision: J. Qi, T.S.K. Eisinger-Mathason

We would like to thank John Tobias for assistance with bioinformatics, Fernando Camargo for providing the Yap^fl/fl mice, and Kumarasen Cooper M.D., and Paul Zhang M.D. for their assistance with human tumor pathology. This work was funded by The University of Pennsylvania Abramson Cancer Center (to T.S.K. Eisinger-Mathason), The Penn Sarcoma Program Center (to T.S.K. Eisinger-Mathason), Steps to Cure Sarcoma Center (to T.S.K. Eisinger-Mathason), and NIH/NCI P50 CA100707 (to J. Qi).

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.