A High-Content Screen for C/EBPα Expression Identifies Novel Therapeutic Agents in Dedifferentiated Liposarcoma

Among soft tissue sarcomas, which account for about 1% of all cancers, liposarcoma is the most common type, representing 20% (1). The most common biological type of liposarcoma consists of well-differentiated and dedifferentiated liposarcomas (WDLS and DDLS). Primary WDLS, a low-grade disease, can evolve to DDLS, a high-grade form that is associated with higher rates of local recurrence and distant metastasis (2). WDLS and DDLS are most effectively treated by surgery, yet even after complete resection of primary retroperitoneal DDLS, 60% of patients develop local recurrence and 25% develop distant metastasis within 5 years of diagnosis (1). Repeat recurrences eventually become impossible to resect, and the response rate to conventional chemotherapy is low, so that 60% of patients with retroperitoneal DDLS eventually die from disease (1). As few effective systemic treatments are yet available, doxorubicin or doxorubicin-based combination regimens remain the most widely used, despite response rates in the range of 11% to 12% (3–5). Other FDA-approved agents for refractory DDLS include eribulin and trabectedin, but these treatments offer low response rates (<10%), substantial toxicity, and only modest improvement in progression-free survival (6, 7). Thus, the development of new agents for DDLS represents a pressing unmet medical need.

WDLS and DDLS are characterized by amplifications of chromosome 12q13–15, encompassing two major oncogenes, CDK4 and MDM2. Multiple additional copy-number alterations have been identified as specific to DDLS (8), but the molecular mechanisms underlying progression to DDLS remain to be elucidated. In a transcriptomic analysis of tumor tissues from patients with liposarcoma, we noted significant downregulation of CCAAT/enhancer binding protein alpha (C/EBPα) relative to normal fat and WDLS (9), among other genes specifically expressed by adipocytes. We found that CEBPA underexpression is significantly associated with greater risk of distant recurrence-free survival (10). C/EBPα, encoded by CEBPA, is a basic leucine-zipper transcription factor that plays an essential role in the development and homeostasis of hematopoietic, liver, lung, and adipose tissues. In adipocytes, C/EBPα functions as a master regulator of differentiation and induces growth arrest and terminal differentiation of pre-adipocytes (11–13). We observed that adipogenic differentiation culture conditions do not induce C/EBPα expression in DDLS cells, unlike normal mesenchymal stem cells (9), which is concordant with the differentiation-based classification of liposarcomas, in which DDLS is defined to be arrested at an early stage of adipogenic differentiation (14). Moreover, forced expression of C/EBPα in DDLS cells causes growth arrest, and in adipogenic differentiation conditions induces only modest expression of early adipocyte-specific genes, resulting in apoptosis instead of full differentiation (9).

The tumor suppressor function of C/EBPα is well studied in acute myeloid leukemia (AML), in which loss-of-function mutations have been identified in ∼10% of cases, and frequent allelic loss in others (15). Underexpression of C/EBPα has also been observed in a number of solid tumors such as liver, breast, pancreatic, and lung cancer and is associated with tumor progression and poor clinical outcome (16); in these cancers, the CEBPA gene is inactivated by allelic loss, epigenetic silencing through promoter methylation, suppressive microRNAs, or transcriptional repression by oncogenes such as Ras, HIF1α, and TGFβ. In liposarcoma, we detected allelic loss of 19q13, which encompasses CEBPA, in 32% of DDLS patients, which was associated with worse disease-specific survival and higher rates of local recurrence following surgical resection (8). We also identified hypermethylation of the CEBPA promoter in 24% of DDLS, which was associated with a corresponding loss of CEBPA mRNA expression (17). Moreover, treatment of DDLS cells with inhibitors of DNA methylation and histone deacetylase 1 (HDAC1) significantly increase CEBPA mRNA levels and adipocyte-specific differentiation markers, inhibit cell proliferation, induce apoptosis, and suppress tumor growth in xenograft models (17). These findings suggest that multiple epigenetic silencing mechanisms underlie aberrant CEBPA suppression in DDLS and that drugs restoring C/EBPα expression may treat DDLS.

To search for potential therapies that may drive differentiation and inhibit proliferation of DDLS cells, we developed a high-content screen using a library of clinical grade chemical compounds for the ability to induce C/EBPα re-expression and suppress cell growth. The screen was designed to identify agents with effects in DDLS cells but not in normal adipose-derived stem cells (ASC) because such agents may target the specific abnormalities of the C/EPBα-driven tumor suppressor pathway in DDLS.

All DDLS tissue samples were collected under IRB-approved protocols after obtaining written informed consent from patients in accordance with the Declaration of Helsinki. The DDLS8817 cell line and PDX-DD0010 xenograft were established from tumor samples from patients at Memorial Sloan Kettering Cancer Center (MSK), and the LPS141 cell line from a patient at Brigham and Women's Hospital. The ASC primary culture was previously established from human subcutaneous fat (18). The identity of DDLS8817, LPS141, and PDX-DD0010 with the primary tumors of origin was confirmed via molecular cytogenetic analysis, array comparative genomic hybridization (aCGH), and whole-genome sequencing as described previously (19), showing amplification of MDM2, CDK4, and HMGA2 on chromosome 12q and allelic loss of CEBPA on 19q in both cell lines, and amplification of JUN on 1p (20) and DDIT3 on 12q (21) in LPS141 only (Supplementary Fig. S1A). Sequencing of TP53 revealed that DDLS8817 is wildtype and LPS141 carries a P72R polymorphism and a W282R mutation; these do not appear to affect regulation of p53 expression, as nutlin3 treatment increases p53 and p21 in both cell lines (22). All cell lines were confirmed as negative for Mycoplasma prior to use in assays. Cell lines were grown in a 1:1 mixture of DMEM high glucose and F12 media with 10% FBS, 100 units/mL penicillin plus 100 μg/mL streptomycin (complete medium) and maintained at 37°C in 5% CO₂. DDLS8817 cells were used after 20 to 40 passages and LPS141 after 40 to 50. Emetine was purchased from Sigma, deguelin from Microsource, and SN-38 from Abatra Technology. 5-Azacytidine (5Aza) and suberoylanilide hydroxamic acid (SAHA) were from Selleckchem. Methylation of the promoter of CEBPA was detected by pyrosequencing as described previously (17).

ASC, DDLS8817, and LPS141 cells were seeded in 40 μL of complete medium into a 384-well microplate (Corning) using a Multidrop 384 dispenser (Thermo Fisher Scientific) and grown overnight in a Steri-Cult incubator. Pilot compounds were pre-plated in a 384-well polypropylene plate (ABgene), and 10 μL was transferred to the assay plates using a PP-384-M Pipettor (Apricot Designs) for a final testing concentration of 10 μmol/L in 1% DMSO (v/v), based on prior screen optimization (23). Negative control wells contained DMSO and positive control wells contained differentiation medium (complete medium with 100 nmol/L GI262570, 0.25 mmol/L 3-isobutyl-1-methylxanthine, 5 μg/mL insulin, 0.2 mmol/L indomethacin, 1 μmol/L dexamethasone, and 17 μmol/L pantothenic acid), which significantly increases C/EBPα expression in ASCs (9). After 96 hours incubation, assay plates were fixed with 4% paraformaldehyde. For C/EBPα immunostaining, assay plates were washed once with PBS using an Elx405 washer (Biotek), then cells were permeabilized with 0.1% Triton X-100 for 5 minutes. After blocking for 1 hour with 10% bovine serum albumin in PBS, plates were washed, then incubated with anti-C/EBPα antibody (Cell Signaling Technology, 2295, Lot 3, 1:500 dilution) with 2% BSA in PBS for 24 hours at 4°C. Assay plates were washed three times with PBS, then incubated with anti-rabbit detection antibody Alexa Fluor 488 (Invitrogen 1:1,000 dilution) for 1 hour. Finally, nuclei were stained with Hoechst at 10 μmol/L. Images were acquired using an IN Cell Analyzer 3000 (INCA3000, GE Healthcare). Cell numbers (Hoechst-stained nuclei) and C/EBPα expression were analyzed using Raven 1.0 software (GE Healthcare). Proliferation was quantified as the total number of cells following compound treatment divided by the total number of cells in the negative control. C/EBPα was considered to be induced if Alexa Fluor 488 intensity exceeded the mean plus three times the SD of that in negative controls.

The library used for the pilot screen includes 3,119 chemicals: the MicroSource library (MicroSource Discovery Systems) contains 2,000 biologically active and structurally diverse compounds including known drugs, experimental bioactives, and pure natural products (23), and the Prestwick Chemical library contains 1,119 high-purity chemical compounds selected for structural diversity, known safety, bioavailability in humans, and a broad spectrum of activities covering several therapeutic areas (24). A list of all screened chemicals is provided in Supplementary Table S1.

RNA was isolated from LPS141, DDLS8817, and ASC cell lines using the RNeasy Mini Kit (Qiagen). Reverse transcription was performed using random hexamer priming and TaqMan reverse transcription reagents (Applied Biosystems). Quantitative real-time PCR (qPCR) using TaqMan Gene Expression Assays (Applied Biosystems) was done on the ABI Prism 7900HT Sequence Detection System and analyzed using SDS version 2.1 software (Applied Biosystems). Expression assays were used according to the manufacturer's protocol to detect CEBPA RNA (Hs00269972_s1) and 18s rRNA (Hs99999901_s1). The relative expression of CEBPA was calculated by normalizing its expression to 18s rRNA.

Cells were plated in 96-well plates at 1,000 cells per well for proliferation assays or in 6-well plates at 30,000 cells per well for apoptosis assays. Cells were incubated overnight, then drugs were added to triplicate wells at the indicated concentrations. Media containing each drug were replaced every other day and cells were collected on days 4 and 6 of treatment. For proliferation assays, DNA content was estimated using the CyQuant Cell Proliferation Kit (Molecular Probes) and the Spectramax M2 fluorescence microplate reader (Molecular Devices) at 480/520 nm excitation/emission, as per the manufacturer's instructions. For apoptosis assays, cells were stained with Annexin V: PE Apoptosis Kit (BD Biosciences) per manufacturer's instructions, then analyzed using a Guava PCA (Guava Technologies).

Cells were seeded and grown overnight on chamber slides (Nunc Lab-Tek) prior to addition of drugs at the indicated concentrations. Media containing each drug were replaced every other day during the 96-hour growth period, after which cells were fixed in 4% paraformaldehyde. Cells were permeabilized in 0.1% Triton X-100/1X PBS, blocked in 10% BSA, and incubated for 24 hours with anti-C/EPBα antibody (Cell Signaling Technology, 2295; and Santa Cruz Biotechnology, SC-61 [14AA]) in 2% BSA/1X PBS in a humidified chamber at 4°C. Slides were developed with goat anti-rabbit Alexa Fluor 488 (Invitrogen) in 2% BSA/1X PBS for 1 hour in a humidified chamber and stained with Hoechst nuclear stain (Invitrogen). Fluorescent images were captured using an Olympus IX71 microscope.

Sets of pLKO.1 lentiviral short-hairpin RNAs (shRNA; Thermo Scientific Open Biosystems) targeting CEBPA were individually tested for knockdown in DDLS8817 and LPS141 cells. Five distinct sequences were assessed, and the two that yielded greatest knockdown (G4 and G6) were used in subsequent analyses. A scramble (SCR) sequence not known to target any human genes served as negative control. Viruses harboring the shRNA sequences were produced in HEK 293T cells (Invitrogen) as previously described (10). Cells were infected with lentivirus using polybrene (Sigma-Aldrich) to increase infection efficiency and selected with 1 μg/mL puromycin (Sigma-Aldrich). To knock down p53 and HIF1α, cells were transfected with siRNA constructs carrying scrambled control sequences or ON-TARGET Smartpool for TP53 (Dharmacon, L-00332900), or HIF1A (Dharmacon, L-00401800), at 20 nmol/L using Oligofectamine (Invitrogen).

Proteins were isolated from cells by lysis in high-SDS lysis buffer (2.5% SDS, 62.5 mmol/L Tris-HCl, pH 6.8, 10% glycerol); lysates were immediately boiled to denature proteins. For protein isolation from tumor tissues (snap-frozen immediately after dissection), cut samples were first ground using Sample Grinding Kit (GE Health) in SDS lysis buffer and then sonicated to reduce viscosity. Protein concentration was determined using the DC Protein Assay (BioRad Laboratories). Cell lysate was resolved on a NuPAGE Novex 4% to 12% Bis-Tris gel (Invitrogen) in an XCell SureLock MiniCell (Invitrogen), then transferred to nitrocellulose membrane (Millipore). Antibodies against the following proteins were used for western blotting: C/EBPα (Cell Signaling Technology, 2295, 8178), cleaved PARP (Cell Signaling Technology, 9541), PARP (Cell Signaling Technology, 9542), cleaved caspase 3 (Cell Signaling Technology, 9661), anti-HIF1α (Cell Signaling Technology, 14179), vinculin (Sigma, V4505), p53 (Santa Cruz Biotechnology, sc-263), p21 (Santa Cruz Biotechnology, sc-6246), and histone H3 (Cell Signaling Technology, 9751).

Experiments were carried out under a protocol approved by the Institutional Animal Care and Use Committee, and institutional guidelines for proper and humane use of animals were followed. The identity of ex-mouse tumors with the original primary tumors and cell lines was verified by histology and either array comparative genomic hybridization (aCGH) for DDLS8817 or shallow whole-genome sequencing for PDX-DD-0010 (Supplementary Fig. S1A). LPS141 tumors were grown subcutaneously in ICR-scid mice (Taconic) and DDLS8817 and patient-derived xenograft (PDX)-DD-0010 tumors were grown subcutaneously in female NSG mice (stock #005557; The Jackson Laboratory). Tumors were excised when they reached a volume of approximately 500 mm³. Tumors were minced and passed through a cytosieve to yield a single-cell suspension, mixed 1:1 with Matrigel (BD Biosciences), and implanted subcutaneously in the right flank of 6- to 8-week-old female mice. Once tumors reached 100 to 150 mm³, mice were randomized to receive 3 weeks' treatment with irinotecan (100 mg/kg for DDLS8817 or 10–100 mg/kg for LPS141) once a week, doxorubicin (0.9 mg/kg twice per week for DDLS8817 or 1 mg/kg three times per week for LPS141 or PDX-DD-0010), or vehicle control (PBS) twice a week; all treatments were delivered intraperitoneally to 5 to 10 mice per cohort. Mice were sacrificed when the control tumors reached the maximum allowable size. In the experiment in which LPS141 xenograft-bearing mice were treated with varying doses of irinotecan, 2 mice per group (n = 10 total) were sacrificed at day 2 of treatment for tumor protein extraction. Tumor size was measured twice weekly by caliper. Tumor volume was calculated using the formula (π/6) × (small diameter)² × (large diameter), and the mean tumor volume for each group was calculated. At the end of the study mice were euthanized by CO₂ asphyxiation.

Tumor tissue samples were fixed overnight in 4% paraformaldehyde, embedded in paraffin, then sectioned. IHC for C/EBPα was performed using Bond Polymer Refine detection kit (Leica) according to the manufacturer's instructions, and that for Ki-67, cleaved caspase 3, and HIF1α using Discovery XT processor (Ventana Medical Systems). Slides were heated for 30 minutes for antigen retrieval and incubated with either anti-C/EBPα (Cell Signaling Technology, 8178) for 30 minutes, anti–Ki-67 (Abcam, ab15580, 0.1 μg/mL) for 4 hours, or anti–cleaved caspase 3 (Cell Signaling Technology, 9661, 0.1 μg/mL) for 3 hours, and anti-HIF1α (Novus Bio, NB100–479) for 5 hours (all antibodies are rabbit polyclonal). After washing, slides were incubated for 20 minutes with biotinylated goat anti-rabbit IgG (Vector Labs PK6101, 5.75 μg/mL), blocked with Blocker D, and staining detected using a 3,3′-diaminobenzidine (DAB) Detection Kit (Ventana Medical Systems) according to the manufacturer's instructions. Slides were counterstained with hematoxylin and cover-slipped with Permount (Thermo Fisher Scientific). Whole stained slides were scanned, and 4 to 11 randomly selected representative fields containing ≥100 nuclei were counted to determine the ratio of cells with C/EBPα or Ki67 nuclear staining relative to all nuclei by manual counting or using Image J (NIH).

Data were analyzed using GraphPad Prism version 7.0 and the significance of differences was tested by one-way ANOVA. Error bars indicate SD.

Genome sequencing data are available via GEO (accession #GSE184702). Cell lines and the PDX model used in this study are available via materials transfer agreement upon request.

DDLS cell lines were first characterized for copy-number alterations in CEBPA and other genes relevant to liposarcoma. aCGH and whole-genome sequencing showed that a single copy of the CEBPA locus is preserved in DDLS8817 and LPS141 cells (Supplementary Fig. S1A). Both cell lines show significantly reduced CEBPA mRNA expression in response to differentiation medium compared with ASCs (Supplementary Fig. S1B; ref. 9). This lower expression results in part from hypermethylation of the CEBPA promoter ∼−1.5 Kb from the transcription start site (Supplementary Fig. S1C), reported as a repressive methylation site in other cancers (9, 25–27). DDLS cells treated with inhibitors of DNA methylation (5Aza) or HDAC activity (SAHA) increased CEBPA mRNA expression when given alone and displayed synergistic effects when combined (Supplementary Fig. S1D; ref. 17). These results indicate that despite single copy loss of CEBPA in these DDLS cell lines, epigenetic silencing of CEBPA can be overcome (or reversed).

A library of 3,119 compounds (full list in Supplementary Table S1) was screened in DDLS8817 and LPS141 cells and normal ASCs. In the screen, 85 compounds (2.7%) induced C/EBPα expression in both liposarcoma cell lines (Fig. 1). Of these, 57 (1.8% of all compounds) did not induce C/EBPα in ASCs and thus were selective. Seventy-three compounds (2.3%) inhibited proliferation as assessed by nuclei count in both liposarcoma cell lines, and 32 (1%) did so selectively, that is, in the DDLS cell lines but not in ASCs. Nineteen compounds were both selective C/EBPα inducers and selective proliferation inhibitors. Among the hits, deguelin, emetine, and SN-38 (7-ethyl-10-hydroxy-camptothecin) were chosen for screen validation based on their commercial availability and safety and antitumor efficacy in humans (28–31).

We first sought to confirm the screen results with deguelin, emetine, and SN-38 using a cell proliferation assay. Deguelin (4-day treatment) decreased proliferation of LPS141 cells by 40% at 2 μmol/L and 60% at 4 μmol/L (P < 0.001; Fig. 2A). DDLS8817 cells were less sensitive; 2 μmol/L deguelin had an insignificant effect on proliferation, yet 4 μmol/L deguelin inhibited proliferation by 28% (P < 0.05). Compared with deguelin, emetine had more pronounced effects, even at nanomolar concentrations. At 4 days, emetine (50 or 100 nmol/L) decreased proliferation of LPS141 cells by 57% to 59% (P < 0.001) and of DDLS8817 cells by 47% to 55% (P < 0.001; Fig. 2B). The most potent of the three compounds analyzed was SN-38, with an IC₅₀ of 5 nmol/L in DDLS8817 cells and 9.18 nmol/L in LPS141 as determined by assessment of cell viability (Supplementary Fig. S2A). At 4 days, treatment with 10 nmol/L SN-38 decreased proliferation of LPS141 by 43% and DDLS8817 by 58% (P < 0.001; Fig. 2C). None of the drugs affected proliferation of ASCs. These results validate the screen, as all drugs affected proliferation at concentrations much lower than the 10 μmol/L concentration used in the screen.

We examined whether inhibition of proliferation was mediated by apoptosis by staining cells with annexin V and 7-aminoactinomycin-D (7-AAD). All three drugs induced apoptosis in both cell lines at the above concentrations by 6 days (Fig. 2D–F), to a greater extent in LPS141 compared with DDLS8817 cells, with SN-38 having the greatest effects (Fig. 2D–F). Thus, these drugs appear to inhibit proliferation at least in part by inducing apoptosis.

We further validated the effects of the three hit drugs on expression of C/EBPα in DDLS cells using immunofluorescence. After 2 days of treatment, and more so after 4, both cell lines had increased C/EBPα expression in the nucleus and cytoplasm (Fig. 3A–C; Supplementary Fig. S2B). To confirm the specificity of C/EBPα staining with the antibody used in the screen, we repeated the experiment with another C/EBPα mAb (14AA), which detected C/EBPα exclusively in the nucleus upon treatment with SN-38 (Fig. 3D).

To test whether C/EBPα induction is necessary for drug-induced apoptosis, we used shRNA against CEBPA in DDLS8817 and LPS141 cells, which knocked down expression by >80% relative to scramble control (Fig. 4A). As expected, deguelin and SN-38 induced apoptosis as measured by annexin V and 7-AAD staining in cells expressing scramble shRNA, whereas these drugs failed to induce apoptosis in C/EBPα-deficient cells (Fig. 4B). Emetine increased apoptosis somewhat in C/EBPα-deficient cells, although these increases were smaller than those in scramble controls (Fig. 4B). We further validated induction of apoptosis by SN-38, the drug with the lowest IC₅₀ and thus best potential for translation and the focus of the remainder of this investigation, by measuring protein expression of the markers cleaved caspase 3 and cleaved PARP. Knockdown of CEBPA reduced the increase in expression of these markers in response to SN-38 treatment (Fig. 4C). Together, these results indicate that these drugs' cytotoxicity to DDLS cells depends on induction of C/EBPα in DDLS cells.

Given SN-38′s translational potential, we confirmed its restoration of C/EBPα expression by qRT-PCR and investigated its mechanism of action. SN-38 increased CEBPA mRNA by approximately two-fold in both DDLS8817 and LP141 cells (Fig. 5A). SN-38, an active metabolite of irinotecan and topoisomerase-1 (Topo1) inhibitor, binds the Topo1-mediated DNA cleavage complex that causes DNA double-strand breaks; subsequent DNA damage increases expression of the pro-apoptotic tumor suppressor p53 as a DNA damage response (32, 33). We therefore examined the contribution of p53 to SN-38–triggered apoptosis. We first assessed SN-38′s effects on expression of p53 and found that it increased protein levels of p53 at days 2 and 4 (Fig. 5B, top row). We also examined expression of the p53 transcriptional target p21, another tumor suppressor that induces cell-cycle arrest, and found that SN-38 increased its protein levels (Fig. 5B, second row). Surprisingly, although C/EBPα is a known target of p53 (34), knockdown of C/EBPα reduced SN-38–induced p53 and p21 upregulation (Fig. 5C), indicating that C/EBPα regulates p53 expression in DDLS cells, at least indirectly. To test whether SN-38–induced apoptosis requires p53, we used small-interfering RNA (siRNA) to knock down p53, which nearly eliminated p53 protein expression in DDLS8817 cells (Fig. 5D, top row). p53 knockdown also significantly reduced SN-38 induction of p21 expression (Fig. 5D, second row). However, p53 knockdown did not interfere with the ability of SN-38 to induce apoptosis as measured by cleavage of PARP and caspase 3 (detected by Western blot analysis; Fig. 5D, rows 3 and 5). We observed similar effects of p53 knockdown in LPS141 cells (data not shown). Instead, nearly eliminating p53 enhanced the degree to which SN-38 increased levels of these apoptosis markers, indicating that SN-38′s effects on apoptosis are independent of p53.

The enzymatic function of Topo1 is to relax DNA supercoils to enable replication and transcription, and camptothecin derivatives such as SN-38 are known to control gene expression through Topo1 inhibition (35, 36). Among the genes regulated by Topo1 and its inhibitors is HIF1α, an oncogenic transcription factor (37). HIF1α is also reported to inhibit CEBPA expression under hypoxic conditions in breast cancer (38). Because HIF1α expression is associated with worse clinical outcomes in sarcoma and has been shown to promote its metastasis (39, 40), we investigated SN-38′s effects on HIF1α. We found that SN-38 efficiently reduces HIF1α protein expression in both DDLS8817 and LPS141 cells at 4 and 6 days of treatment (Supplementary Fig. S3A). Irinotecan-induced suppression of HIF1α was also observed in vivo by IHC at early time points (day 4) when xenograft tumors were small and at the end of the experiment (day 10) when control tumors were quite large, whereas HIF1α expression increases with tumor size in controls (Fig. 6C; Supplementary Fig. S3B). To further investigate whether HIF1α mediates SN-38′s effects on C/EBPα expression, we knocked down HIF1α using Smartpool oligonucleotide transfection, which reduces HIF1α expression in both DDLS8817 and LPS141 cell lines (Supplementary Fig. S3C). HIF1α knockdown increased C/EBPα expression in DDLS8817 but not in LPS141 cells, likely because the latter carry amplification of DDIT3 (9), a known dominant-negative regulator of C/EBP-induced transcription. In addition, combining SN-38 treatment with HIF1α knockdown did not further affect C/EBPα expression. These results suggest that HIF1α may be involved in regulating C/EBPα expression in some DDLS.

To test the effects of SN-38 on DDLS in vivo, we used irinotecan (also called camptothecin-11 or CPT-11), which is activated in vivo to SN-38, in xenograft models of liposarcoma. Immunodeficient (ICR-scid) mice were engrafted subcutaneously with LPS141 or DDLS8817 tumors and treated with irinotecan, vehicle control, or doxorubicin, a standard chemotherapy for liposarcoma. In mice treated with vehicle, DDLS8817 tumors grew approximately 15-fold and LPS141 tumors grew approximately 20-fold (Fig. 6A–C). Irinotecan (100 mg/kg weekly) almost completely prevented growth of both types of DDLS tumors (Fig. 6A–C), exceeding the efficacy of doxorubicin, which reduced growth of DDLS8817 tumors by approximately 50% and of LPS141 tumors by 30% (Fig. 6A and B) at a dose high enough to cause weight loss (Supplementary Figs. S4A and S4B). To determine the minimum dose of irinotecan required for antitumor effects in vivo, we evaluated a range of doses to as low as 10 mg/kg, which revealed dose-dependent effects (Fig. 6C). All doses significantly inhibited tumor growth without causing weight loss (Supplementary Fig. S4B). To further validate the antitumor effects of irinotecan, we used a patient-derived xenograft (PDX) model directly derived from a patient DDLS tumor [DD0010, which carries no loss or deletion of the CEBPA locus (Supplementary Fig. S1A)], implanted into NSG mice. Results in this independent model were similar to those in cell line–derived xenografts; treatment with irinotecan at 50 or 100 mg/kg once weekly inhibited tumor growth in a dose-dependent manner without inducing weight loss, to a greater degree than the somewhat toxic dose of doxorubicin (Fig. 6D; Supplementary Figs. S4C and S4D). Upon withdrawal of irinotecan, tumors resumed growth, indicating that it did not completely eradicate tumors (Supplementary Fig. S4C). Irinotecan treatment reduced Ki67 staining and increased cleaved caspase 3 expression in tumors (Fig. 6E and F), demonstrating that it efficiently inhibits DDLS cell proliferation and induces apoptosis. Furthermore, irinotecan increased expression of C/EBPα in the nucleus detected by IHC of tumor tissue (Fig. 6G; Supplementary Fig. S4E), and by immunoblotting (Fig. 6H; Supplementary Fig. S4F). Together, these data indicate that irinotecan induces C/EBPα expression and inhibits HIF1α expression, leading to induction of DDLS cell apoptosis and inhibition of DDLS tumor growth in vivo.

Our previous work revealed that underexpression of the tumor suppressor C/EBPα, either through epigenetic alterations with or without allelic chromosomal loss but not mutations or deep deletions, is critical for DDLS survival and strongly associated with progression of WDLS towards DDLS and risk of distant recurrence (9). These findings led us to screen for compounds that restore CEBPA expression in DDLS cells. In this study, we conducted a high-content screen of bioactive compounds and FDA-approved drugs for DDLS-selective inhibition of proliferation and restoration of C/EBPα nuclear expression. This screen identified 19 highly selective compounds, of which 3, deguelin, emetine, and SN-38, were validated. We identified irinotecan and its metabolic derivative SN-38 as potent and selective compounds, with a nanomolar IC₅₀ in vitro and remarkable antitumor efficacy in both PDX and xenograft DDLS models in vivo without noticeable toxicity.

In this study, we show that all three validated hit chemicals, deguelin, emetine, and SN-38, induce apoptosis in a C/EBPα-dependent manner in DDLS8817 and LPS141 cells, as knockdown of CEBPA abolishes apoptosis caused by these compounds. C/EBPα regulates differentiation through transcriptional control of lineage-specific gene expression and by inducing growth arrest via cell-cycle regulators including p21, CDK4, CDK2, and E2F1 through protein–protein interactions. Although restoration of C/EBPα exerts antitumor effects mostly by inhibiting cell-cycle progression, ectopic expression of C/EBPα also increases cell death in certain cancers including breast, lung, and DDLS (41, 42). We show here that in DDLS, SN-38 transcriptionally upregulates C/EBPα and consequently increases its nuclear expression, as well as the expression of p53 and p21. Interestingly, however, neither p53 nor p21 is essential for drug-induced apoptosis. These results suggest that irinotecan and related agents have anti-proliferative and pro-apoptotic effects in cancer cells in which C/EBPα acts as a tumor suppressor by transcriptionally restoring C/EBPα expression through an unknown but p53-independent mechanism. The molecular pathway mediating C/EBPα-induced apoptosis should be elucidated as it could identify which tumor types would be sensitive to C/EBPα restoration like DDLS. Interestingly, the antitumor effects of C/EBPα may be enhanced by its promotion of myeloid-derived suppressive cells' (MDSC) proliferation and function, demonstrated in immunocompetent mouse models of liver cancer (43).

Compared with doxorubicin, the current standard chemotherapy for liposarcomas, irinotecan exerted remarkably better antitumor efficacy in xenograft DDLS models with less toxicity, highlighting the potential of translating our results to a clinical trial, and opening the realistic chance of improving systemic treatment for patients with unresectable aggressive DDLS. Moreover, we observed a clear dose dependence of irinotecan's antitumor effects, with 25 and 50 mg/kg being nearly as effective as 100 mg/kg (Fig. 6B), suggesting the potential of a broad treatment window and use in combination with other therapies. Although 25 mg/kg was sufficient to exert maximum C/EBPα restoration, irinotecan's antitumor effects were maximal at 100 mg/kg, suggesting that they involve additional mechanisms. Since the first approval of irinotecan more than two decades ago to treat colorectal cancer, it remains the first-line systemic treatment for the metastatic or advanced stage of a variety of solid tumors including colon, pancreatic, and gastric cancers, largely in combination regimens with 5-fluorouracil with or without leucovorin (36). Recently, newly developed versions of irinotecan have become available; a liposomal nanoparticle formulation has demonstrated much improved efficacy in preclinical xenograft models and is now FDA-approved for pancreatic cancer (44, 45) and an oral version of irinotecan is being evaluated in early-phase clinical trials (46), further enhancing the ability to deliver irinotecan selectively to tumor cells with potentially less toxicity.

Our results suggest a partial mechanism for induction of C/EBPα expression following treatment with irinotecan. Although irinotecan interferes with DNA replication to cause cell-cycle arrest, the observed IC₅₀ of SN-38 in DDLS cells in this study was in the 5 to 10 nmol/L range, much lower than that required to inhibit Topo1 activity and block DNA replication (30–50 nmol/L; refs. 35, 36, 47). Because HIF1α gene expression was shown to be regulated by irinotecan's target, Topo1 (37, 48), and its analog topotecan was identified as the best hit in a comprehensive screen for HIF1α repressors, we investigated SN-38′s effects on HIF1α. We found that SN-38 significantly reduces HIF1α expression in vitro, and irinotecan blocks HIF1α expression in the nucleus of liposarcoma cells in vivo. Furthermore, HIF1α knockdown increases C/EBPα expression in DDLS8817 cells, suggesting that SN-38 may restore C/EBPα through the inhibition of HIF1α. However, HIF1α knockdown did not change C/EBPα levels in LPS141 cells, which are known to harbor amplification of DDIT3, a known negative regulator of C/EBPα. Therefore, HIF1α may function as a C/EBPα suppressor in some DDLS, depending on the copy-number changes and associated epigenetic alterations they harbor.

The two other validated compounds from the screen, emetine and deguelin, have been reported to exert antitumor effects on multiple cancers. Interestingly both chemicals have also been shown to suppress HIF1α expression and thereby tumor angiogenesis (49, 50), supporting the potential of HIF1α as a target to restore C/EBPα expression in sarcoma. Further studies to elucidate epigenetic and genetic mechanisms of C/EBPα suppression in DDLS will inform optimal drug combinations and selection of biomarkers for treatment response in future clinical trials. Recently, another approach to restore C/EBPα through small activating RNA, named MLT-CEBPA, has shown efficacy in a liver cancer mouse model, and is now in early-phase clinical trials for hepatocellular carcinomas (51).

In summary, this study demonstrated the power of a targeted high-throughput drug screen focused on altered gene expression for a cancer with minimal treatment options. This screening method may be effective for other cancers with aberrant C/EBPα downregulation, especially solid tumors, in which epigenetic silencing mechanisms are more common than inactivating mutations (21). We found that irinotecan has promising therapeutic potential for DDLS, supporting its evaluation in clinical trials. Continued investigation of other identified DDLS-selective compounds may lead to further novel discoveries for liposarcoma treatment.

B.A. Nacev reports grants from NIH during the conduct of the study, as well as grants from NIH and the Connective Tissue Oncology Society outside the submitted work. M.A. Dickson reports grants from NCI during the conduct of the study. T. Okada reports grants from NCI during the conduct of the study. S. Singer reports grants from NCI during the conduct of the study. No disclosures were reported by the other authors.

C.V. Angeles: Formal analysis, investigation, writing–original draft. A. Velez: Formal analysis, investigation, writing–original draft. J. Rios: Investigation. B. Laxa: Formal analysis, writing–original draft. D. Shum: Investigation, methodology. P.D. Ruiz: Formal analysis, investigation. Y. Shen: Investigation. I. Ostrovnaya: Formal analysis. R. Gularte-Mérida: Formal analysis, data interpretation. B.A. Nacev: Writing–review and editing, data interpretation. M.A. Dickson: Writing–review and editing, data interpretation. H. Djaballah: Conceptualization, resources, formal analysis, supervision, methodology, writing–review and editing, data interpretation. T. Okada: Formal analysis, supervision, investigation, writing–original draft, project administration, writing–review and editing, data interpretation. S. Singer: Conceptualization, resources, formal analysis, supervision, funding acquisition, methodology, writing–review and editing, data interpretation.

We thank Jessica Moore for editorial assistance with the manuscript and figures, Elisa De Stanchina (Antitumor Assessment Core, MSK) for assistance with xenograft experiments, Marina Asher (Department of Pathology, MSK) and Afsar Barlas (Molecular Cytology Core, MSK) for assistance with IHC, and all members of the Singer lab, particularly Rachael O'Connor, for technical assistance and Young-Mi Kim for helpful comments on the manuscript. This research was supported by the NIH SPORE in Soft Tissue Sarcoma P50 CA140146, P50 CA217694 and the NIH/NCI Cancer Center Support Grant P30 CA008748.

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.