miR-193b–Regulated Signaling Networks Serve as Tumor Suppressors in Liposarcoma and Promote Adipogenesis in Adipose-Derived Stem Cells

Abstract
Well-differentiated and dedifferentiated liposarcomas (WDLS/ signaling through CRKL and FAK. miR-193b also stimulated DDLS) account for approximately 13% of all soft tissue sarcoma reactive oxygen species signaling by targeting the antioxidant in adults and cause substantial morbidity or mortality in the methionine sulfoxide reductase A to modulate liposarcoma cell majority of patients. In this study, we evaluated the functions survival and ASC differentiation state. Expression of miR-193b of miRNA (miR-193b) in liposarcoma in vitro and in vivo. Deep in liposarcoma cells was downregulated by promoter methyl- RNA sequencing on 93 WDLS, 145 DDLS, and 12 normal fat ation, resulting at least in part from increased expression of the samples demonstrated that miR-193b was significantly under- DNA methyltransferase DNMT1 in WDLS/DDLS. In vivo, miR- expressed in DDLS compared with normal fat. Reintroduction 193b mimetics and FAK inhibitor (PF-562271) each inhibited of miR-193b induced apoptosis in liposarcoma cells and pro- liposarcoma xenograft growth. In summary, miR-193b not moted adipogenesis in human adipose-derived stem cells only functions as a tumor suppressor in liposarcoma but also (ASC). Integrative transcriptomic and proteomic analysis of promotes adipogenesis in ASC. Furthermore, this study reveals miR-193b–target networks identified novel direct targets, key tyrosine kinase and DNA methylation pathways in lipo- including CRK-like proto-oncogene (CRKL) and focal adhesion sarcoma, some with immediate implications for therapeutic kinase (FAK). miR-193b was found to regulate FAK–SRC–CRKL exploration.

Introduction
Liposarcomas account for about 20% of adult sarcomas (1, 2). Well-differentiated liposarcomas (WDLS) and dedifferentiated liposarcomas (DDLS), characterized by chromosome 12q13– 15 amplification, represent the most common biologic group of liposarcoma (3). Primary WDLS, a low-grade disease, can evolve to DDLS, the high-grade form. Despite complete surgical resec- tion, over 60% of patients with retroperitoneal WDLS and DDLS eventually have recurrence and die of disease (4). Once advanced disease develops and surgery is no longer possible, conventional chemotherapy is largely palliative with few long-term responses(5). Therefore, there is pressing need to understand the molecularalterations that drive the development and progression of lipo- sarcoma so as to identify promising therapeutic targets.Small noncoding RNAs (miRNA) play major roles in almost all biological pathways in mammals, and they are predicted to reg- ulate more than 60% of all mRNAs (6). The importance of miRNAs in cancer has been demonstrated by the identification of dysregu- lated miRNA expression and through functional characterization of miRNA target networks (7). miRNAs not only have potential to serve as biomarkers of clinical outcome, but also represent an entry point for the discovery of dysregulated pathways in cancer and for identification of promising therapeutic targets.Although miRNAs have been extensively investigated in many cancers, there is still limited insight into miRNA functions in liposarcoma. Several studies report dysregulation of miRNAs in liposarcoma (8–10), but only a few miRNAs have been demon- strated to be functional (8–13). In this study, we describe the results of miRNA sequencing analysis of tumors from a large cohort of patients with WDLS and DDLS. We discovered that miR- 193b functions as a novel tumor suppressor in liposarcoma.

A combination of computational, gene ontology, and functional analyses revealed pivotal roles of miR-193b–regulated FAK–SRC– CRKL signaling and reactive oxygen species (ROS) signaling in both liposarcomagenesis and adipogenic differentiation.This study was approved by the Memorial Sloan Kettering Cancer Center (MSKCC) Institutional Review Board; allparticipants gave informed consent. Patient studies were con- ducted in accordance with the International Ethical Guidelines for Biomedical Research Involving Human Subjects and in accor- dance with MSKCC Institutional Review Board-approved proto- col #02-060A. Patient characteristics are shown in Supplementary Table S1. Tumor and normal adipose tissue samples obtained during surgical resection were snap-frozen in liquid nitrogen and embedded in cryomolds.Liposarcoma cell lines were established from patient tissue samples: DDLS8817, DD6960-1, and RDD8107 from DDLS samples; WD4847-2, RWD5700, and WD7785-1 from WDLS samples. The cell lines were validated by array comparative genomic hybridization confirming 12q amplification. Primary human adipose tissue-derived stromal/stem cells (ASC) were isolated from subcutaneous fat samples from consenting patients as described (14). Cell lines were maintained as described (10). The sources of antibodies and other reagents are listed in Sup- plementary Table S2.MiRNAs and siRNAs were transfected with Oligofectamine (Invitrogen) and RNAiMAX (Invitrogen). Cells were harvested 72 hours after transfection for protein and mRNA analysis. Vectors for protein overexpression were transfected with Lipofectamine 3000 (Invitrogen) 1 day after miRNA transfection. Cells were collected for assays of protein, cell proliferation, and apoptosis on day 3 after miRNA transfection.

For luciferase reporter assays, 50 nmol/L miRNA and 200 ng of 3′UTR reporter plasmids were cotransfected with Lipofectamine 2000 (Invitrogen). At 48 hours posttransfection, cells were col- lected for luciferase assays according to Promega’s instructions.Total RNA was isolated from cultured cells and analyzed as previously described (15). TaqMan gene expression assays (Life Technology) were used for relative gene expression (Supplemen- tary Table S3). miRNA expression levels were detected by using SYBR Green miRNA-specific primers (Supplementary Table S3). Transcript levels were normalized to levels of GAPDH transcript (for mRNAs) or U6 snRNA (for miRNAs).Cells were lysed in RIPA buffer (with 100× protease inhibitor cocktail and 25 mmol/L MG132) for 30 minutes on ice andsonicated to shear DNA. Protein concentration was determined by the Lowry method. Equal amounts of total protein were loaded and resolved by SDS-PAGE, then transferred to polyvinylidene difluoride membranes for immunoblotting. Secondary antibo- dies (horseradish peroxidase–conjugated) were detected using Western Lightning chemiluminescence reagent (PerkinElmer).Cell viability, cell cycle, apoptosis analysis, and soft agar assays Cells were treated with miRNAs, siRNAs, or FAK inhibitors. Cell viability was assessed using CellTiter-Glo luminescent cell viabil- ity assay (Promega). At 72 hours posttransfection, cell-cycle analysis was performed as previously reported (16).

Apoptosis was quantified using a Muse Annexin V and Dead Cell Kit (EMDMillipore).Soft agar assays were performed in 6-well tissue culture plates by placing miRNA- or siRNA-treated cells (2 × 104/well) in 2 mL of 0.3% soft agar above a 2-mL layer of 0.5% agar. The wells were covered with 1 mL fresh medium. Cell growth medium was exchanged every 2 days. After 2 weeks’ incubation, cellswere stained with 1 mg/mL MTT in medium for 1 hour. Colonies were detected and counted using GelCount technology (Oxford Optronix Inc.).Adipogenic differentiation was induced with initiating medi- um (regular growth medium plus 100 nmol/L insulin, 1 mmol/L dexamethasone, 250 mmol/L 3-isobutyl-1-methylxanthine, 33 mmol/L biotin, 17 mmol/L pantothenic acid, and 5 mmol/L of the PPARG agonist rosiglitazone). After 4 days, media were changed to maintenance medium (initiating medium without 3-isobutyl-1-methylxanthine or rosiglitazone). Cells were fed with maintenance medium every 4 days thereafter. After 12 days of adipogenic induction, cytoplasmic lipid droplets were stained with Oil Red O as described (17) and with DAPI. Stained cells were quantified by manual counting under a microscope. Three biological replicates were analyzed for each treatment group. At least 400 cells that were positive for lipid droplets were counted in each group.ROS measurement was performed 48 hours after transfection by using H2DCFDA (Molecular Probes, Invitrogen) according to the manufacturer’s instructions. At 48 hours after transfection, the cells were treated with 10 mmol/L H2DCFDA (Molecular Probes,Invitrogen) and incubated for 30 minutes at 37◦C. Cells were washed, then plated at 104 cells/well in 96-well plates. Thefluorescence intensity (excitation = 485 nm; emission = 530 nm) was measured by a SpectraMax L microplate reader (Molecular Devices).For assessment of protein oxidation, lysates were prepared as described for immunoblotting.

Total protein was incubated with 2,4-dinitrophenylhydrazine according to the instructions for the OxyBlot oxidized protein kit (Chemicon). b-Actin was used as a loading control.Small RNA cDNA libraries were prepared from 145 DDLS, 93 WDLS, and 12 normal fat tissue samples, as previously described(10). Data were normalized using the edgeR calcNormFactors function. The normalized counts were rescaled back to the geo- metric mean (mean of logs) of the raw data. The mean clone counts were also the geometric mean.Illumina arrays were used to generate triplicate gene expression profiles in DD8817 and WD4847-2 cell lines 3 days after trans- fection with miR-193b or scramble miRNA. The expression pro- files were processed using the LUMI package (Bioconductor) with the default options: background subtraction, normalization with the quantile method, and log2 transformation of expression values. Data were normalized using the variance-stabilization method (lumiExpresso). LIMMA (Bioconductor) was used to compute differential expression, and the false discovery rate (FDR) method was used for multiple testing corrections.

Themicroarray data are available from GEO (accession number GSE83690).DD8817 and WD4847-2 cells were grown in media supple- mented with either unlabeled L-arginine and L-lysine at 50 mg/L (light medium) or equimolar amounts of the isotopic variants [U-13C6]-L-arginine (Arg6) and [U-13C6]-L–lysine HCl (Lys6; heavy medium; Cambridge Isotope Laboratories). After five cell doublings, cells were >99% labeled with the isotopes. Cells grown in the light medium were transfected with 50 nmol/L miR-193b mimic, while those grown in heavy medium were transfected with scramble miRNA control. Total protein from light and heavy labeled cells (10 mg each) was mixed. Cysteine residues were reduced and alkylated, then protein was digested with trypsin. Digested protein was centrifuged to pellet precipitated lipids and cleared supernatant desalted on a SepPak C18 cartridge 500 mg (Waters). Desalted peptides were separated through strong cation exchange chromatography and analyzed by the Q-Exactive massspectrometer (Thermo Fisher Scientific). All MS/MS samples were(50 mg/kg in vehicle) by oral gavage twice per day for 3 weeks. Tumors were measured twice per week using calipers. All animal care was in accordance with the Institutional Animal Care and Use Committee at MSKCC (protocol # 02-09-024).Data for all experimental groups are presented as means SE. Statistical significance was assessed using the Student t test, and P values < 0.05 were taken as statistically significant.

Results
We applied deep RNA sequencing to 93 WDLS, 145 DDLS, and12 normal fat samples. The 50 miRNAs most differentially expressed between DDLS and normal fat (Supplementary Table S4) were subjected to unsupervised hierarchical clustering (Fig. 1A). Three miR-193 family members were underexpressed in DDLS compared with normal fat. miR193b was underexpressedanalyzed by the Proteomics and Microchemistry Core Facility at MSKCC using MaxQuant (technical details are given in Supple- mentary Methods).Genomic DNA was extracted from L090310 and liposarcoma cell lines using the DNeasy Blood and Tissue Kit (Qiagen). DNA methylation status was established by methylation-specific PCR as described (18). Briefly, bisulfite conversion of DNA was per- formed by using the EpiTect Bisulfite Kit (Qiagen). As positive control, we used CpGenome Universal Methylated DNA (Che- micon/Millipore). PCR was performed in a thermal cycler usingthe EpiTect MSP Kit (Qiagen) with the following cycling condi- tions: 95◦C for 10 minutes; then 35 cycles of 94◦C for 15 seconds, 52◦C for 30 seconds, 72◦C for 30 seconds; then final extension of 10 minutes at 72◦C.

To test DNMT1 binding of the miR-193b promoter, chromatin immunoprecipitation (ChIP) analysis was performed in WD4847-2 cells. Briefly, after cells were exposed to crosslinking with 1% formaldehyde for 10 minutes at room temperature, chromatin was purified and sonicated until the DNAs were around 500 bp long. ChIP was performed using the ChIP-IT High Sensitivity Kit (53040, Active Motif) with ChIP-grade DNMT1 antibody (39204) and native IgG, according to the manufacturer's instructions. DNA was analyzed via quantitative PCR. All ChIP experiments were completed with two biological replicates.ICR SCID mice (Taconics) were implanted subcutaneously with serially transplanted DDLS8817 tumors, together with Matrigel. After palpable tumors developed (typically 100–150 mm3), mice were ready for miRNA or drug treatment. For theThe miR-193b/365 cluster is underexpressed in liposarcoma. A, Heat map of the50 most differentially expressed miRNA profiles in DDLS tissue compared with normal fat (NF). Columns, profiled DDLS and NF tissue samples; rows, miRNAs ordered by unsupervised clustering. The details of heatmap constructions are in Supplementary Methods. B, miRNA expression in cell lines. The values (normalized to miR-365 in L090310 cells) represent the mean SE of three independent experiments.by 9-fold (FDR = 0.006; Supplementary Fig. S1A). Among patients with primary WDLS or DDLS, miR-193b expression was associated with distant recurrence-free survival (HR = 0.58, P = 0.015; Fig. Supplementary S1B).miR-193b derives from the same primary transcript as another miRNA, miR-365.

Both these miRNAs were downregulated approximately 2-to 3-fold in multiple liposarcoma cell lines derived from liposarcoma tissues compared with normal ASCs (Fig. 1B).We investigated miRNA function by introducing miRNAs into liposarcoma cells and ASCs. miR-193b mimic significantly inhib- ited proliferation of liposarcoma cells, but not ASCs (Fig. 2A; Supplementary Fig. S2). miR-193b mimic repressed multiple cell- cycle regulators (Fig. 2B) and induced increases in the G0–G1 population (Fig. 2C), implying G0–G1 arrest. miR-193b also induced significant apoptosis in liposarcoma cells (Fig. 2D and E). Furthermore, miR-193b strongly inhibited anchorage-inde- pendent growth of liposarcoma cells (Fig. 2F). miR-365 mimic had weaker effects.Next, we assessed whether the miR-193b/365 cluster regulates adipogenic differentiation. During induction of adipogenesis in ASCs (Fig. 2G), miR-193b and miR-365 levels increased about 16 fold (Fig. 2H). miR-193b mimic markedly upregulated adipo- genic markers (Fig. 2I). Consistent with this, miR-193b mimic promoted the formation of lipid droplets, while miR-193b inhib- itor (anti-miRNA) completely blocked this effect (Fig. 2J). miR- 365 showed no effects on differentiation. These results indicate that miR-193b functions as a tumor suppressor in liposarcoma and an activator of adipogenesis in ASCs.

To identify miR-193b targets, we carried out mRNA microarray and SILAC analysis in miR-193b–treated cells. Similar to a prior study (15), the frequency of downregulated genes or proteins(fold change ≤ —1.5, FDR < 0.05) was no more than 10% (Supplementary Table S5A and S5B). In contrast to non-predictedgenes, the majority of predicted miR-193b target genes exhibited reductions in both protein and mRNA levels (Fig. 3A; Supple- mentary Fig. S3A and S3B). Eight known targets of miR-193b (20, 21) showed downregulation by miR-193b in either our micro- array or SILAC analysis (Table S5C).We developed a strategy to identify miR-193b targets and the processes affected by them in liposarcoma (Fig. 3B). Eight target prediction programs (DIANA-mT, miRDB, miRanda, miRWalk, RNAhybrid, PICTAR5, RNA22, Targetscan; refs. 20–28) were applied to reduce bias. There were 2,056 genes predicted to be miR-193b targets by at least 3 of the 8 prediction programs.Among these 2,056 were 50 genes were downregulated in our SILAC analysis in DD8817 or WD4847-2 cells (fold change ≤—1.5, FDR < 0.05; Supplementary Table S6). These 50 genes,subjected to gene ontology enrichment analysis, were significantlyassociated with multiple biological processes and molecular func- tions (P ≤ 0.05;Supplementary Fig. S3C, S3D, and Supplementary Table S7). The KEGG pathway annotations of these target genesrevealed enrichment of several pathways, including focal adhe- sion, regulation of actin cytoskeleton, and pyrimidine metabolism (P < 0.05, Fig. 3C). Four of these pathways involved the putativemiR-193b targets CRKL and FAK (focal adhesion kinase; HUGO name PTK2), so we investigated these genes further.CRKL plays important roles in signal transduction in cancer(29).

To test whether it is a direct target of miR-193b, we con- structed CRKL 3'UTR reporters. miR-193b mimic inhibited the luciferase activity of wild-type reporter, but not of reporter bearing seed site mutations (Fig. 3D). Furthermore, miR-193b mimic repressed mRNA and protein levels of CRKL, while addition of anti-miRNA completely blocked the effect on the protein level(Fig. 3E). These results demonstrated that miR-193b directly targets CRKL. Overexpression of CRKL partially blocked the miR-193b–induced cell growth inhibition and apoptosis (Fig. S4A–C), providing evidence that derepression of CRKL is responsible in part for the effects of low miR-193b expression in liposarcoma cells.siRNA knockdown of CRKL significantly repressed cell prolif- eration in liposarcoma cells (Fig. 3F). Similar to miR-193b trans- fection, CRKL knockdown induced G0–G1 phase arrest, inhibited multiple cell-cycle regulators (Fig. 3G; Supplementary Fig. S4D), and induced substantial apoptosis (Fig. 3H). Furthermore, CRKL knockdown reduced colony formation of DD8817 cells by 90% (Fig. 3I). CRKL has been reported to mediate signaling of FAK through SRC family kinases (30). Consistent with this, CRKL knockdown repressed the phosphorylation of FAK and SRC (Fig. 3J).

Taken together, these results indicate that inhibition of CRKL may suppress liposarcomagenesis. In a test for direct targeting of FAK, miR-193b repressed more than 80% of the activity of a wild-type FAK 3'UTR reporter, but did not affect reporter bearing seed-site mutations (Fig. 4A). miR- 193b mimic inhibited FAK expression, while miR-193b inhibitor blocked the effect on protein (Fig. 4B). Thus, miR-193b directly regulates FAK. Knockdown of FAK induced significant cell growthinhibition, G0–G1 phase arrest (40%–80%), and apoptosis (Fig. 4C–E). Restoring FAK expression through exogenous over- expression significantly reduced miR-193b–induced growth inhi- bition and apoptosis (Fig. S5A–C). These observations demon- strate the oncogenic function of FAK in liposarcoma cells.To define the molecular mechanisms of FAK function in lipo- sarcoma, we assessed the regulation of effectors of FAK signaling. miR193b mimic strongly repressed phosphorylation of both SRC and BCAR1, which are phosphorylated by FAK (31), while miR- 193b inhibitor blocked this inhibition (Fig. 4F). Similar regula- tion was observed with FAK knockdown (Fig. 4G). These results suggest that miR-193b regulates oncogenic FAK–SRC–BCAR1 signaling in liposarcoma through directly targeting FAK.Next, we tested two FAK inhibitors, PF-573228 (PF-228) and PF-562271 (PF-271). Both inhibitors strongly repressed FAK activation as measured by phosphorylation of FAK, SRC, and BCAR1 (Fig. 4H). These compounds inhibited cell growth (Fig. 4I) and induced apoptosis in liposarcoma cells (Fig. 4J).

Furthermore, both FAK siRNA and FAK inhibitor (PF-271) not only resulted in upregulation of adipogenic makers (Fig. 4K), but also induced 2.5- to 3.5-fold increases in lipid droplet formation (Fig. 4L). These results show that FAK inhibition suppresses tumor activity and promotes adipogenic differentiation in normal ASCs.was induced by 5-aza-dC (Supplementary Fig. S8D). 5-Aza-dC treatment also caused dose-dependent increases of 3- to 15-fold in the miR-193b and miR-365 levels (Fig. 6B). The miR-193b targets (CRKL, FAK, and MSRA) were downregulated with 5-aza-dC treatment (Fig. 6C). Furthermore, Annexin V assays showed that 5-aza-dC induced apoptosis in DD8817 (74%) and WD4847-2 (35%; Fig. 6D). Thus, pharmacologic inhibition of methylation in liposarcoma cells reverses several of the effects of miR-193b underexpression.In data from our earlier microarray analysis of mRNA in WDLS/ DDLS tissue samples versus normal fat (35), DNMT1 expression was elevated on average 3-fold in DDLS and 2-fold in WDLS. In addition, DNMT1 was highly expressed in most liposarcoma cell lines compared with ASCs (Supplementary Fig. S8E). Intriguingly, DNMT1 was strongly downregulated by miR-193b in both micro-array and SILAC analyses. Transfected miR-193b strongly repressed DNMT1 expression (Fig. 6E and F). However, multiple miRNA-target prediction tools found no miR-193b seed sites in the DNMT1 3'UTR region.Similar to 5-aza-dC treatment, DNMT1 knockdown specificallyincreased miR-193b/365 expression by 2- to 4-fold (Fig. 6G). Correspondingly, DNMT1 knockdown resulted in downregula- tion of miR-193b targets (Fig. 6H). ChIP analysis demonstrated that DNMT1 directly bound to the miR-193b promoter, and 5-aza-cD treatment markedly reduced this binding (Supplemen- tary Fig. S8F). DNMT1 knockdown resulted in partial demethyl- ation of the miR-193b promoter, as evidenced by amplification by unmethylation-specific primers (Fig. 6I). However, the demethylation was not complete, suggesting that other genes may also regulate miR-193b promoter methylation. DuringTo assess the antitumor activity of miR-193b in vivo, mice bearing subcutaneous DD8817 tumor xenografts were treated by intratumoral miR-193b injections, which resulted in a 14-fold increase in the miR-193b level (Supplementary Fig. S9A).

After 3 weeks of treatment, miR-193b–injected mice had significantly lower tumor volume (242 mm3) than those receiving the control miRNA injections (1,431 mm3), without significant difference in body weight of the mice (Fig. 7A and Supplementary S9B). The miR-193b–treated xenografts showed reduced Ki-67 staining (Supplementary Fig. S9C). Three miR-193b targets (CRKL, FAK, and MSRA) had reduced protein levels in the miR-193b–injected tumors (Fig. 7B). These results demonstrate that miR-193b has antitumor effects in vivo in a human liposarcoma xenograft model. We also evaluated the FAK inhibitor PF-271 in vivo. After 3 weeks of treatment of DD8817-xenografted mice, tumor growth was inhibited more than 50% in PF-271-treated mice compared with vehicle-treated mice (Fig. 7C), while body weight was not affected (Supplementary Fig. S9D). Phosphorylation of FAK, SRC, and BCAR1 was markedly inhibited, demonstrating that PF-271 inhibited FAK and its downstream effectors (SupplementaryFig. S9E).

Discussion
In this study, we demonstrated the significance of miR-193b in liposarcoma. miR-193b, by directly targeting FAK, CRKL, and MSRA, regulates focal adhesion signaling and ROS signaling, which play pivotal roles in liposarcomagenesis and adipogenic differentiation (Fig. 7D). In addition, we showed that DNMT1 upregulation in WDLS and DDLS promotes miR-193b promoter methylation, leading to miR-193b underexpression in liposar- coma (Fig. 7D). To our knowledge, this is the first demonstration of the tumor suppressive activity of miR-193b in liposarcoma, and of the oncogenic roles of FAK–SRC–CRKL signaling and MSRA- regulated ROS signaling in liposarcoma.A strength of this study is that the miRNA profiles were derived from 145 patient samples, which to our knowledge is a much larger number than in prior studies of miRNAs in liposarcoma. Our study confirmed prior reports (9, 11, 12) of dysregulation in liposarcoma for miR-143, miR-145, miR-451, and miR-486 (Sup- plementry Table S2 and data not shown).Of the 50 miRNAs most differentially expressed in DDLS tissues in our analysis, 45 were underexpressed. This general downregu- lation of miRNAs in liposarcoma could be explained in part by the downregulation of Dicer that has been observed in liposar- coma (8).Two of the direct miR-193b targets (FAK and CRKL) are known to be upregulated in multiple tumor types. FAK regulates cell motility, the cell cycle, survival, and differentiation in many cell types (31). CRKL, along with other CRK family members, is closely associated with FAK–SFK signaling (30).

Thus, in the FAK–SFK signaling network, miR-193b directly inhibits both upstream FAK and downstream CRKL signal transduction. CRKL is upregulated in numerous cancer types (29). However, there islimited knowledge of miRNA regulation of CRKL (29), and only four miRNAs (miR-126 and miR-200b/200c/429) were previous- ly identified as direct regulators of CRKL (36, 37). In liposarcoma cells, inhibition of CRKL and FAK significantly represses activa- tion of FAK–SRC signaling, resulting in cell-cycle arrest and apoptosis. Similar to miR-193b, the FAK inhibitor (PF-271) strongly inhibits DDLS tumor growth in mice. This demonstrates that FAK inhibitors have therapeutic potential in DDLS with reduced miR-193b expression.Besides FAK–SRC–CRKL signaling, we also defined the func- tion of ROS signaling in liposarcoma cells, which miR-193b regulates by targeting MSRA. Cross-talk between ROS signaling and FAK–SRC signaling may be a means by which miR-193b regulates tumorigenesis and differentiation. Although ROS, as external or internal cell stimuli, can activate FAK, prolonged oxidative stress induces FAK cleavage and apoptosis (38, 39).

Normal ASCs have lower levels of FAK and CRKL than liposar- coma cells, and they may not be as dependent on FAK–SRC–CRKL signaling for survival, because inhibition of FAK or CRKL does not induce significant cell death in ASCs (data not shown). The fact that overexpression of FAK, CRKL, or MRSA only partially rescued cells from the effects of exogenous miR-193b implies that all three of these proteins contribute to the miR-193b–related phenotypes in liposarcoma. Both repression of FAK and increased production of ROS have been reported to favor adipogenic differentiation of mesenchymal stromal cells (31, 40). Therefore, miR-193b– induced inhibition of FAK and elevation of ROS determine the different fates of tumor cells and ASCs, inducing cell death in tumor and adipogenic differentiation in ASCs.DNA methylation silences numerous cancer-related genes, including miRNAs. Although methylation of the miR-193b pro- moter has been reported in several tumors (34, 41, 42), the mechanisms were not determined. We found that DNMT1 is upregulated in WDLS and DDLS cells and binds to the miR193b promoter, resulting in promoter methylation and decreased miR- 193b expression. The decreased miR-193b expression in liposar- coma likely results from the action of multiple regulators.

For example, DNMT1 may not be the only DMNT contributing to methylation of the miR-193b promoter, as knockdown of DNMT1 did not fully abolish its methylation. Besides DNMTs, histone modification or other factors could also regulate miR- 193b expression. Interestingly, inhibition of DNMT1 was reported to promote adipogenic differentiation (43), suggesting that inhibition of DNMT1-induced miR-193b promoter methyl- ation increased miR-193b expression, which in turn enhanced adipogenic differentiation. The strong inhibition of DNMT1 by miR-193b could result from direct binding to seed sites outsidethe 3'UTR of DNMT1 or indirect regulation via targeting regula- tors of DNMT1 expression. All these possibilities will be exploredmore in our future studies.

In conclusion, this study elucidates several interconnected mechanisms of molecular dysregulation in liposarcomas, with miR-193b modulating liposarcoma genesis and audiogenic differentiation at multiple levels. Our findings suggest PF-562271 possible therapeutic benefit of tumor-targeted miR-193b in liposarcomas with low miR-193b expression. Furthermore, small-molecule inhibitors of FAK and DNA methylation may be useful therapeutic strategies for combination therapy of liposarcoma.