Identification of Mithramycin Analogues with Improved Targeting of the EWS-FLI1 Transcription Factor

This article is featured in Highlights of This Issue, p. 3985

Ewing sarcoma is a bone and soft-tissue sarcoma with an overall survival of only 55% (1, 2). Survival is 70% for patients with localized disease treated on a compressed schedule (3), but patients with high-risk relapsed or metastatic disease have a survival rate of less than 30% (4). In addition, patients receive chemotherapy that has significant short- and long-term side effects (5). Therefore, there is a need to develop new, less-toxic, and more effective therapies for this tumor type.

Ewing sarcoma has a unique dependence on the EWS-FLI1 transcription factor for cell survival (6). Since the identification of EWS-FLI1 in the early 1990s, independent studies have established that the tumor absolutely depends on the activity of EWS-FLI1 for continued proliferation (6, 7), yet the clinical realization of an EWS-FLI1–directed therapy has not been achieved.

A number of compounds have been identified as EWS-FLI1 inhibitors, including cytarabine, YK-4-279, trabectedin, mithramycin (MMA), midostaurin, low-dose actinomycin, shikonin, and HCI2509 (8–14). Most of these compounds show an effect on the EWS-FLI1 transcriptional program and reverse the expression of well-established EWS-FLI1 targets such as NR0B1 and PHLDA1 (midostaurin, shikonin) and/or reverse the gene signature of EWS-FLI1 on a genome-wide scale (cytarabine, trabectedin, MMA, low-dose actinomycin, and HCI2509). Unfortunately, the compounds that have made it to the clinic have failed in phase II trials (15, 16). Furthermore, we do not know whether these compounds achieved the exposure necessary to suppress EWS-FLI1 in these trials.

We screened more than 50,000 compounds to identify MMA as an inhibitor of EWS-FLI1 (11). We showed that the drug blocks the expression of critical EWS-FLI1 downstream targets at the mRNA and protein levels both in vitro and in vivo, and it reverses the expression of the EWS-FLI1 gene signature on a genome-wide scale (11). In addition, MMA shows an excellent (low nmol/L) IC₅₀in vitro and good suppression of Ewing sarcoma xenograft growth. These results reflected clinical reports from the 1960s of the activity of the drug in Ewing sarcoma patients. Therefore, we translated the compound to the clinic in a phase I/II trial (17, 18). The compound was well tolerated, but liver toxicity limited serum concentrations of the drug to values that our preclinical models predicted would not be high enough to inhibit EWS-FLI1 (17 nmol/L vs. 50 nmol/L; manuscript in preparation). Therefore, the trial was closed to accrual.

The goal of this study was to identify a second-generation MMA that can achieve serum levels high enough to block EWS-FLI1 activity in patients. The approach was to characterize either a compound with a similar toxicity profile but more potent inhibition of EWS-FLI1 or a compound that maintained suppression of EWS-FLI1 at similar concentration but was less toxic, thus allowing larger doses to be administered. In order to accomplish this, we generated a panel of more than 20 MMA analogues for their ability to reverse EWS-FLI1 activity. MMA chemical space was expanded by genetic engineering of the MMA biosynthesis pathway and enzymatic biocatalysis to generate mithralogs showing both lower toxicity and higher biologic activity (19–21). We found several mithralogs that suppressed EWS-FLI1 to a comparable or greater extent than MMA. In this report, we show that EC-8105 was a more potent EWS-FLI1 inhibitor than MMA and yet maintained a comparable toxicity profile. We also show that a different analogue, EC-8042, maintained comparable suppression of EWS-FLI1 but was one order of magnitude less toxic than the parent compound. Both compounds suppressed EWS-FLI1 at the mRNA and protein levels in vitro and in vivo and showed excellent activity in Ewing sarcoma xenografts. Together, the results provide a basis for the further development of these compounds as targeted therapies for Ewing sarcoma.

TC32 and TC71 Ewing sarcoma cells were the gift of Dr. T. Triche (The Saban Research Hospital, Children's Hospital of Los Angeles, CA). HepG2 cells were obtained from the American Type Culture Collection (ATCC). RH30, RD, and U2OS cells were the gift of Lee Helman. The identity of all cells was independently authenticated by short tandem repeat genotyping. All cells were maintained in culture in RPMI 1640 (Invitrogen) with the exception of HepG2 which was cultured in EMEM (ATCC). Medium was supplemented with 10% FBS (Gemini Bio-Products), 2 mmol/L L-Gln, 100 U/mL and 100 μg/mL penicillin and streptomycin, respectively (ThermoFisher).

MMA and all analogues were obtained from EntreChem SL Biotechnology. All compounds were aliquoted, stored frozen, and thawed immediately before use.

TC32 cells stably expressing the NR0B1 luciferase reporter were incubated in triplicate with each of the analogues of MMA over concentrations from 500 to 0.1 nmol/L for 12 hours. Cell were lysed and the bioluminescence was quantified using Steady-Glo luciferase (Promega) as previously described (11).

TC32 cells (0.3 × 10⁶) were exposed to compound, and RNA was collected using the RNEasy Kit with QIAshredder (Qiagen), immediately reverse-transcribed using a High Capacity cDNA Synthesis Kit (Life Technologies) on a Veriti thermocycler (Life Technologies), and PCR-amplified using SYBR green master mix (BioRad) and the CFX 384 Real Time System (BioRad) with the following program: 95°C for 10 minutes, 95°C for 30 seconds, 55°C for 30 seconds, and 72°C for 30 seconds for 40 cycles. The expression of target genes was determined using standard ΔΔCT methods and normalized to GAPDH control. See Supplementary Table S1 for the list of target genes and corresponding primers. Heat maps were created using R v 3.2.2 (R Foundation for Statistical Computing) and comprise ΔΔCT scores truncated between −3 and 3 to prevent very large scores from oversaturating the color gradient.

TC32 and TC71 cells (1.5 × 10⁶) were incubated with drug, collected, washed with PBS, lysed and boiled in 4% LDS buffer (0.125 mol/L Trizma hydrochloride buffer solution, pH 7.5) and 4% lithium dodecyl sulfate (Sigma-Aldrich). Protein concentrations were determined after diluting the detergent using the bicinchoninic acid assay Kit (Pierce Protein Biology Products). Thirty micrograms of protein was resolved on a 4% to 12% NuPAGE Bis-Tris Mini gels (Invitrogen) in 1×4-morpholinepropanesulfonic acid (MOPS) sodium dodecyl sulfate buffer (Invitrogen), transferred to nitrocellulose (GE Healthcare Life Sciences), and probed with the following antibodies: rabbit monoclonal anti-EZH2 (1:1,000; Cell Signaling Technology), mouse monoclonal anti-FLI1 (1:1,000; Abcam), mouse polyclonal anti-ACTB (1:1,000; Cell Signaling Technology), rabbit polyclonal anti-NR0B1 (1:500; Abcam), mouse monoclonal anti–phospho-histone H2A.X (Ser139; 1:10,000; Millipore), and rabbit monoclonal anti-ID2 (1:1,000; Cell Signaling Technology). The protein was visualized by using horseradish peroxidase (HRP)–conjugated secondary antibody and ECL (Amersham).

IC₅₀s were determined by nonlinear regression from at least three independent experiments at 48 hours using Prism GraphPad. Cytotoxicity relative to a panel of other pediatric tumors (including leukemia, lymphoma, and other solid tumor cell lines) was determined by the Pediatric Preclinical Testing Program at 96 hours as previously described (22).

Healthy CD-1 mice (n = 3) provided by the University of Oviedo SPF Vivarium were treated with single or repeated intravenous injections of mithralogs, using saline solution as vehicle. For repeat dose treatment, drugs were administered by intravenous injections every 3 days for 8 doses (q3d × 8). Body weight, deaths, changes in behavior, motility, eating and drinking habits, and any other sign of local or systemic toxicity were recorded daily. All experiments were performed in accordance with the guidelines and regulation of and approved by the Animal Care and Use Committee at University of Oviedo, Spain.

TC32 and HepG2 cells were incubated with drug as above and imaged every 2 hours on the IncuCyte Zoom (Essen Bioscience). Confluence of cells in each well was measured using IncuCyte Zoom software by a proprietary algorithm that determines the percent confluence of each well continuously in real time. End point confirmation was performed by standard MTS assay and the manufacturer's protocol (Promega).

TC32 cells and HepG2 cells were incubated with drug in Lab-Tek II 4 chamber wells (Nunc), fixed in 4% paraformaldehyde in PBS, washed, and permeabilized in 1% Triton X-100. Cells were blocked with 10% goat serum, and expression was determined with anti–phospho-histone H2A.X (ser139) antibody (1:200; Millipore) and Alexa647–labeled anti-mouse immunoglobulin G (1:200; Life technologies) on a Zeiss 510 confocal microscope in the presence of DAPI in VectaShield mounting medium (Vector Labs), with standard settings that were not changed among treatment groups as previously described (23).

Two million TC71 cells were injected intramuscularly in the left gastrocnemius of 6-week-old female homozygous nude mice (Crl; Nu-Foxn1^Nu; Charles River Laboratories) and established to a minimum diameter of 0.5 cm. Four cohorts of 12 mice were treated with vehicle; 1 or 1.5 mg/kg of EC-8105; or 24 mg/kg of EC-8042, administered either intraperitoneally and intravenously starting on day zero and on a Monday/Wednesday/Friday (M/W/F; IP) schedule or Q3D schedule (IV) for eight doses. Tumor volume was measured 3 times per week and determined using the equation (D × d²)/6 × 3.12 (where D is the maximum diameter and d is the minimum diameter). Tissue was collected and fixed in 10% formalin from 2 mice in each cohort on days 2 and 4 for immunohistochemical analysis. The remaining mice were sacrificed when the tumor diameter reached 2 cm in any dimension. All experiments were performed in accordance with the guidelines and regulation of, and approved by, the Animal Care and Use Committee at Vanderbilt University, Nashville, TN, or in accordance with Animal Care and Use Committee of Southern Research Institute. Investigators were not blinded to the treatment groups.

Paraffin-embedded tissue was sectioned into 5 micrometer sections and mounted on colormark plus charged slides. Antigen retrieval was performed in Ventana CC1, and automated staining was performed using the Ventana Discovery, NR0B1 primary (1:50), Ventana Ultramap Rb (HRP; 16 minutes), and Ventana Discovery Cy5 amplification.

ANOVA with Tukey-HSD post-hoc tests was used to determine if the means of multiple groups were significantly different. Normality was assumed for all tests and was verified visually. Homoscedasticity was assessed via the Bartlett test; if this test was significant, then Welch t tests with Bonferroni multiple testing corrections were used instead of ANOVA. Linear mixed-effects models with random slopes were used to determine if tumor growth rates were significantly different between treatment groups, while mice were on treatment. Last, log-rank tests were used to determine if survival times were significantly different between multiple treatment groups. Analyses were performed in R V 3.2.2 (https://www.r-project.org/) and Graphpad V 6.0F (http://www.graphpad.com/).

To identify MMA analogues that more potently target the EWS-FLI1 transcription factor, we performed a luciferase screen of 22 MMA analogues and evaluated the effect of treatment on the activity of EWS-FLI1 in cells expressing a stable EWS-FLI1–driven luciferase construct (Fig. 1A; refs. 24, 25). These cells utilize the NR0B1 promoter to drive expression of luciferase. This promoter contains the GGAA microsatellite that EWS-FLI1 utilizes to drive gene expression (26). Most of the compounds suppressed EWS-FLI1 to a comparable extent as MMA itself. However, EC-8105 (gray arrow) improved the suppression of EWS-FLI1 by almost 10-fold, with an IC₅₀ of 2 nmol/L [95% confidence interval (CI), 2–3] as opposed to 17 nmol/L (95% CI, 15–18) for MMA (Fig. 1B). Furthermore, the analogue that does not bind DNA, EC-8041, showed a virtual loss of activity (IC₅₀ 348 nmol/L; 95% CI, 316–381), more than 150 times lower than EC-8105 activity (Fig. 1B).

To confirm these results, we evaluated the effect of drug treatment on the mRNA expression of NR0B1 using qPCR. Treatment of TC32 Ewing sarcoma cells with 50 nmol/L EC-8105 improved the suppression of NR0B1 expression by a factor of 5, from the MMA fold change of 0.48 (95% CI, 0.37–0.61, P < 0.0001) to the EC-8105 fold change of 0.10 (95% CI, 0.06–0.1, P < 0.0001; Fig. 1C). Again, the non–DNA-binding MMA analogue EC-8041 showed no suppression of NR0B1, with a fold change of 0.99 (95% CI, 0.94–1.0). In addition, at concentrations that should be achievable in patients (see below), there was marked suppression of EWS-FLI1 activity as measured by NR0B1 mRNA expression, with EC-8105 at 15 nmol/L showing a fold change of 0.19 (95% CI, 0.17–0.20) and at 5 nmol/L showing statistically significant suppression at a fold change of 0.68 (95% CI, 0.58–0.78; P = 0.0002; Supplementary Fig. S3A). Finally, the suppression of EWS-FLI1 observed in these studies translates into a marked suppression in cell viability and an IC₅₀ of 3.29 nmol/L (95% CI, 3.1–3.5) for EC-8105, which is again substantially lower than the IC₅₀ of MMA of 15.5 nmol/L (95% CI, 14.7–17.1; Fig. 1D).

To confirm that the effect of EC-8105 treatment extends to other EWS-FLI1 targets, we evaluated the expression of a gene signature of EWS-FLI1. EWS-FLI1 is known to both induce and suppress expression of its target genes by binding DNA to either establish enhancers or displace the binding of other ETS family members (26). This translates into a change in expression of more than 500 genes (27), but no definitive list of EWS-FLI1 targets exists. Therefore, we randomly selected a panel of EWS-FLI1 target genes to reflect both direct and indirect targets established by both genome-wide techniques and specific dedicated studies (see Supplementary Table S2 for evidence, Supplementary Table S1 for primers; refs. 11, 12, 24, 27–45). Next, we verified that siRNA silencing of EWS-FLI1 did in fact lead to the suppression of the EWS-FLI1–induced targets and induction of the suppressed targets (Supplementary Fig. S3B and S3C). We used both induced and suppressed targets to account for both mechanisms of EWS-FLI1 activity and to rule out a general effect on transcription, because a general inhibitor of RNAPII would not be expected to induce gene expression.

We utilized qPCR to show that 15 nmol/L EC-8105 reversed the gene signature of EWS-FLI1 for both induced and suppressed targets as shown in the heat map (Fig. 1E). All five EWS-FLI1–repressed genes were induced with a 3-hour treatment at 15 nmol/L EC-8105. In addition, all 14 EWS-FLI1–induced target genes were suppressed by treating TC32 Ewing sarcoma cells with 15 nmol/L EC-8105 for 18 hours (Fig. 1E).

Next, we showed that this suppression of EWS-FLI1 targets at both the promoter and mRNA levels extends to the protein level in Ewing sarcoma cells. Treatment of TC32 and TC71 cells over a range of concentrations of EC-8105 or MMA for 18 hours suppressed expression of the EWS-FLI1 target genes EZH2, NR0B1, and ID2 without suppressing EWS-FLI1 expression or the housekeeping gene ACTB (Fig. 1F). It is notable that EC-8105 achieved similar suppression of EWS-FLI1 targets at 5 to 15 nmol/L as was seen at 50 to 100 nmol/L of MMA (Fig. 1F).

In order to determine whether the improved suppression of EWS-FLI1 comes at the expense of increased toxicity, we performed toxicity studies on the majority of the analogues to determine the MTD (Table 1; structures of lead compounds shown). Mice tolerated EC-8105 and MMA to a similar extent: EC-8105 had an MTD less than 4 mg/kg intravenously, versus 2 mg/kg for MMA. The majority of the analogues were at least equally as well-tolerated as MMA; inactive compounds were not evaluated (see Supplementary Fig. S4A and S4B for structures)

One MMA analogue, EC-8042, was substantially less toxic than any of the other compounds. EC-8042 had an MTD in mice of 200 mg/kg intraperitoneally and 64 mg/kg intravenously that is 130 or 32 times higher than the MTD of MMA (EC7071). In order to confirm that the drug is in fact less toxic, we compared the toxicity of MMA to EC-8042 in another species, the rat. We found that treatment of the rat with 0.8 mg/kg of MMA intravenously recapitulated the human toxicity profile causing almost no myelosuppression and instead lead to an immediate increase in circulating ALT and AST, to 46.99 U/L (SD ± 4.30) and 149.20 U/L (SD ± 6.64) following drug administration (Fig. 2A and C). In contrast, there was no elevation in ALT or AST with a 5 times higher dose of EC-8042. Indeed, the animals required almost a 10 times higher dose to show the same increase in circulating liver enzymes (Fig. 2A). Furthermore, EC-8042 showed no evidence of other toxicities such as myelosuppression or more progressive liver damage leading to a compromise of liver synthetic function even at 8 mg/kg dose (Fig. 2B and C). This improved toxicity profile translated into substantially higher serum levels of drug in mice from 385 nmol/L for MMA (at the MTD) to 4,295 nmol/L with EC-8042 (at <20% of the MTD; Fig. 2D). In addition, based on PK data from rats and dogs dosed intravenously every 3 days, we predicted by allometric scaling an increase in the MTD for patients from 0.07 mg/kg for MMA to 0.59 mg/kg for EC-8042, an order of magnitude higher than MMA (ref. 46; Supplementary Fig. S4C).

Having demonstrated that EC-8042 is less toxic, we next wanted to confirm that the compound maintains suppression of EWS-FLI1 activity. We confirmed that 50 nmol/L EC-8042 suppressed NR0B1 expression to the identical degree as MMA, with a fold change of 0.62 (95% CI, 0.50–0.72, P < 0.0001) versus 0.62 (95% CI, 0.50–0.72, P < 0.0001; Fig. 3A). Next, we showed that the suppression extended to our panel of EWS-FLI1 target genes (Fig. 3B). All the EWS-FLI1–induced targets were suppressed, and all the repressed targets were induced to a comparable extent as MMA (Fig. 3B). Similar to the case with EC-8105, this is not a general suppressive effect on transcription because of the marked induction in expression of the repressed targets (Fig. 3B).

To further confirm that this is not a general repression of transcription and is linked to EWS-FLI1 blockade, we evaluated the effect of treatment with MMA, EC-8105, and EC-8042 on a panel of childhood sarcoma cell lines; U2OS osteosarcoma, RD embryonal rhabdomyosarcoma, and RH30 alveolar rhabdomyosarcoma cell lines. Notably, these cell lines represent a different bone tumor (U2OS) and a cell line driven by an alternate PAX3-FOXO1 transcription factor (RH30). Although some genes showed a minor degree of suppression (WRN, IL1RAP), most did not change or were even induced with drug treatment with the three compounds (ID2 in RD and RH30 cells; Fig. 3C). Importantly, because EWS-FLI1 is not found in these cell lines, there was no consistent alteration in expression of this panel of genes like with EC-8042, EC-8105, or MMA.

To confirm suppression of EWS-FLI1 targets at the protein level, we treated TC32 and TC71 cells with either MMA or EC-8042 for 18 hours. In TC32 cells, 50 nmol/L of MMA or EC-8042 suppressed the EWS-FLI1 targets EZH2, NR0B1, and ID2 (Fig. 3D). In TC71 cells, 100 nmol/L of either compound suppressed the EWS/FL1 targets NR0B1 and EZH2 (Fig. 3D).

Importantly, in contrast with MMA, the suppression of EWS-FLI1 by EC-8042 happened in the absence of DNA damage, as measured by the phosphorylation of γH2AX (Fig. 3D). Furthermore, the concentration that causes EWS-FLI1 target suppression more closely approximates the cell viability IC₅₀ of 37.8 nmol/L (95% CI, 35.6–40.0), for EC-8042 than the 15.5 nmol/L (95% CI, 14.7–17.1) value for MMA (Fig. 3E). These results suggest that EC-8042 suppresses cell viability by blocking EWS-FLI1, whereas MMA gains additional cytotoxicity from nonspecific DNA damage leading to a lower IC₅₀ but a broader toxicity profile.

To model this cleaner cytotoxicity of EC-8042 and EC-8105, we evaluated the effect of drug treatment using time-lapse microscopy. Because the major toxicity of MMA is liver toxicity, we compared the effects of drug treatment in Ewing sarcoma versus HepG2 cells immortalized liver cells. Although not a perfect model, HepG2 cells have been used in several studies to model liver toxicity (47, 48). They appear to be a reasonable model when the toxicity is due to changes in gene expression, but not when they are due to changes in drug-metabolizing enzymes. Therefore, we used these cells as a model system to compare toxicity induced by suppression of EWS-FLI1 that should be specific to Ewing sarcoma cells versus general mechanisms of toxicity such as DNA damage, which should occur in both types of cells.

We first established concentrations of each compound that showed equivalent suppression of EWS-FLI1 targets NR0B1, EZH2, and ID2: 50 nmol/L MMA, 50 nmol/L EC-8042, and 7.5 nmol/L EC-8105 (Fig. 4A). Further, those concentrations also released the EWS-FLI1–mediated repression of LOX1, PHLDA1, and TAF1C to a similar degree (Fig. 4A).

Next, we used time-lapse microscopy to observe the effect of treating TC32 cells with these concentrations of drugs. We previously showed that EWS-FLI1 target suppression occurs at the protein level in vitro after 12 to 18 hours of treatment with MMA (ref. 11 and Figs. 1H and 3D). Consistent with these kinetics, treatment of TC32 cells with the established concentrations of all three drugs markedly impaired proliferation 18 hours after exposure (black arrow; Fig. 4B). Treatment of HepG2 cells with 50 nmol/L MMA impaired their proliferation to a comparable extent, although with much different kinetics, consistent with an alternative mechanism of cytotoxicity and with the liver toxicity observed in the clinic (Fig. 4D). In contrast, neither EC-8105 or EC-8042 had any effect on the proliferation of HepG2 cells (Fig. 4D). Finally, in order to exclude changes in cell shape as the cause for the change in percent confluence, we confirmed the results using a standard MTS endpoint assay (Fig. 4C and E).

We then evaluated the effect of this treatment on the DNA integrity of the cells. MMA at 50 nmol/L generated significant DNA damage as measured by the phosphorylation of γH2AX by confocal microscopy (Fig. 4F) or Western blot analysis (Fig. 4G). In contrast, EC-8105 showed substantially less DNA damage, and EC-8042 had no DNA-damaging effects (Fig. 4F and G).

Finally, to further explore the activity of these three agents, we screened a panel of pediatric cell lines as part of the Pediatric Preclinical Testing Program. Consistent with a common mechanism of action, the panel responded similarly to all three compounds as measured by a Pearson correlation analysis (Supplementary Fig. S5A). In addition, both compounds had a nmol/L IC₅₀ in all of the Ewing sarcoma cells tested. Interestingly, there was an unexpected sensitivity of rhabdoid tumor and acute lymphoblastic leukemia cell lines to all three agents (Supplementary Fig. S5B). Finally, all cell lines that were p53 mutated were statistically more sensitive to MMA and EC-8105 but not EC-8042 consistent with the known DNA-damaging properties of these agents (Supplementary Fig. S5C).

We previously demonstrated that treatment of TC32 and TC71 Ewing sarcoma xenografts with a dose of 1 mg/kg of MMA IP on a M/W/F schedule markedly suppresses and even regresses tumor growth, particularly for the TC32 xenograft (11). Because the TC71 xenograft was more resistant to MMA than TC32, in the current study, we evaluated EC-8105 and EC-8042 in the more resistant TC71 xenograft model and compared the IP route we previously employed to IV administration on a similar schedule (11).

We treated mice with EC-8105 at 1 mg/kg IP and the M/W/F schedule, and found limited suppression of tumor growth (Supplementary Fig. S6). In contrast, when mice were injected IV with a slightly higher dose of 1.5 mg/kg, every mouse in the cohort showed suppression of tumor growth (Fig. 5A). While on treatment, tumors in this group grew an average of 171.7 mm³ less per day than the control (P < 0.0001, 95% CI, 112.4–231.1). In addition, there was a delayed regression of tumors that was seen in 3 of 8 mice and persisted long after treatment discontinuation, including 1 mouse that was cured (see thin lines in Fig. 5A). Overall, the effect translated into a statistically significant survival advantage as measured by time to a tumor size of 2 cm (P < 0.0001; Fig. 5B). It is notable that, overall, the drug was well tolerated, with some transient weight loss that resolved for every mouse (Supplementary Fig. S7).

EC-8042 also showed excellent activity in the resistant TC71 model, but the activity was independent of route of administration. In parallel to the EC-8105 study, mice were treated with EC-8042 IV at 37% of its IV MTD (24 mg/kg) on a Q3D X * schedule (Fig. 5C). While on treatment, tumors in this cohort grew an average of 151 mm³ less per day than the control (P < 0.001, 95% CI, 92.0–211.1). This marked suppression of tumor growth extended the survival of the EC-8042–treated cohort (P < 0.0001; Supplementary Fig. S8A). Again, regressions following a period of growth were seen, but unfortunately 2 of the mice that were responding to drug died of unknown causes (Fig. 5C, red asterisks).

A separate cohort of mice were treated by the intraperitoneal route in order to compare the activity of EC-8042 to our previously published report for MMA. Mice were again treated at 24 mg/kg to compare with the IV dose, even though this dose was only 12% of the IP MTD. Even at this low dose of EC-8042, every tumor in every mouse showed some level of regression following variable periods of tumor growth, remaining suppressed until therapy was discontinued (Fig. 5D; Supplementary Fig. S9A). Not only did this low dose result in consistent repression, the overall effect was quite impressive, conferring a tumor growth rate that was, on average, 268.2 mm³ less per day than the control (while being treated; 95% CI, 206.0–330.5, P < 0.0001). This regression translated into a difference in tumor volume on day 11 of treatment: the mean tumor size (or final measurement) for control mice was 3,177 mm³ (SEM ± 308.2), versus a mean tumor size in treated mice of 422.5 mm³ (SEM ± 48; P < 0.0001; Supplementary Fig. S9B). This regression extended survival but lacked permanence, and with cessation of treatment, the tumors grew back (Fig. 5D and Supplementary Fig. S8B). The mice were not rechallenged with drug, despite the fact that more drug would likely have been well tolerated. Finally, examination of tumor tissue post-mortem showed clear suppression of EWS-FLI1 activity on day 3 of treatment with EC-8042 as measured by immunofluorescent staining for NR0B1 expression (Fig. 5E). Similar results were obtained with immunofluorescent staining for NR0B1 after treatment with EC-8105, although these tumors were larger at the time of collection and so show more variability (Supplementary Fig. S10).

Ewing sarcoma is dependent on the continued expression of the EWS-FLI1 transcription factor for cell survival. We have previously characterized MMA as an EWS-FLI1 inhibitor and translated it to the clinic, but were unable to achieve high enough serum levels to block the target in patients. In this study, we sought to improve the targeting of EWS-FLI1 by identifying MMA analogues that widen the gap between efficacy and toxicity.

We screened 22 MMA analogues and identified EC-8105 as more potent and EC-8042 as less toxic. We validated the improvement in toxicity in multiple species, showed some preference for the Ewing sarcoma histotype, demonstrated suppression of EWS-FLI1 both in vitro and in vivo, and showed good activity in xenograft models of the disease.

It is unclear which of the two analogues should be prioritized for clinical development and/or will ultimately achieve the therapeutic suppression of EWS-FLI1, which highlights the challenge of using preclinical testing to predict the value of agents in the clinic. Although both compounds were superior to MMA in all our assays, neither was better than the other in every assay. For example, EC-8042 shows a more dramatic effect against xenograft tumors when administered IP, and it was active by both IV and IP routes (even though the IP dose was 1/10th of the MTD). Unfortunately, by this dose, schedule, and route, the effect was reversible. In contrast, EC-8105 was more active via IV but completely inactive by IP injection. However, 3 of the 8 mice showed impressive regressions of the xenograft following a period of initial growth, including a complete cure in 1 mouse. Therefore, future studies will focus on optimizing the dose, route, and schedule of both agents in an effort to prioritize one agent for development. However, because these results are not necessarily predictive of activity in the clinic, they will need to be weighed against practical considerations when prioritizing the two analogues, such as activity in other tumor types, ease of synthesis, and stability over time.

This study also highlights the challenge of predicting toxicity in patients based on preclinical models. In our original studies with MMA, we saw little toxicity in the mouse, which allowed us to achieve high serum levels of drug determined in this study to be over 300 nmol/L. These concentrations far exceed the 50 to 100 nmol/L concentration that blocks EWS-FLI1 activity. However, in the clinic, there was more toxicity, and serum levels were a fraction of those achieved in the mouse; as a result, therapeutic activity was limited in patients.

The gold standard is to test toxicity across several species, and because the basis for selecting EC-8042 was an improved toxicity profile, we felt it necessary to carry out such tests in this study. This presented the opportunity to evaluate other traditional surrogates of toxicity such as evaluation of toxicity in control cell lines. We evaluated the cytotoxicity of MMA, EC-8042, and EC-8105 against a panel of cell lines for a histotype preference. We found a limited preference for Ewing sarcoma cell lines. Since we know EC-8042 is less toxic, this approach as a surrogate for toxicity is disfavored. Instead these studies should be reserved to highlight particularly unique sensitivities of specific tumor types to agents as was seen for rhabdoid tumor and perhaps ALL in this study.

In a complementary approach, we also modeled the liver toxicity in patients by evaluating the compounds in immortalized hepatocytes at concentrations that effectively suppress EWS-FLI1 as the mechanism of cytotoxicity in Ewing sarcoma cell lines. Both analogues were superior to MMA: both suppressed EWS-FLI1 and Ewing sarcoma growth without affecting the growth of HepG2 cells. In contrast, concentrations of MMA that suppressed EWS-FLI1 induced marked cytotoxicity in HepG2 cells. These differences are at least partially explained by the differences in the DNA-damaging properties of the three drugs. Both analogues achieve suppression of EWS-FLI1 with limited (EC-8105) or no (EC-8042) associated DNA damage, whereas MMA suppressed EWS-FLI1 at relatively high concentrations that produced marked DNA damage.

From a mechanistic standpoint, it is not clear if the DNA damage induced by the drug is favorable or not to the activity of this class of compounds. The fact that tumors markedly regressed under EC-8105 treatment and have a substantially lower IC₅₀ with this agent relative to EC-8042 suggests that some DNA damage assists the mechanism of EWS-FLI1 suppression. In addition, in our panel of cell lines, there was a statistically significant sensitivity in cell lines that are p53 wild type (49). This is reflected in our in vivo results, where the less-sensitive TC71 xenograft was p53 mutant while TC32 was p53 wild type and much more sensitive in our previously published study. Importantly, the majority of Ewing sarcoma cases are p53 wild type. Nevertheless, it has been suggested that Ewing sarcoma cells have a baseline tolerance for low-level DNA damage. This may be related to reports that show the direct modulation of the p53 axis by EWS-FLI1. It is possible that by suppressing EWS-FLI1, restoration of the p53 axis while generating low-level DNA damage (as seen with EC-8105) may contribute to the effectiveness of the drug. Further mechanistic studies are in progress.

Finally and perhaps most importantly, this study provides the basis for the use and optimization of DNA-binding drugs as targeted agents. In general, these compounds are thought to be nonspecific inhibitors of transcription. Here, we show that this is not the case, that there is some preference for particular transcription factors and that this preference can be tuned by optimized analogues even when the mechanism of target suppression is not completely understood. By directly addressing the limitations of MMA while preserving the suppression of EWS-FLI1, this study provides the basis for the further development of these compounds for Ewing sarcoma therapy. In addition, the study serves as a precedent for similar strategies with DNA-binding compounds for other tumor types and other transcription factor targets.

S.L. Lessnick holds ownership interest (including patents) in and is a consultant/advisory board member for Salarius Pharmaceuticals. F. Morís holds ownership interest (including patents) in EntreChem SL. No potential conflicts of interest were disclosed by the other authors.

Conception and design: C.L. Osgood, M. He, M. Smith, L.J. Helman, F. Morís, P.J. Grohar

Development of methodology: C.L. Osgood, M. Gebregiorgis, S. Sankar, L. Turner, F. Morís, P.J. Grohar

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C.L. Osgood, C.G. Kidd, S. Kitchen-Goosen, L. Segars, M. Gebregiorgis, G.M. Woldemichael, S.L. Lessnick, M. Kang, L. Turner, F. Morís, P.J. Grohar

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C.L. Osgood, C.G. Kidd, S. Kitchen-Goosen, M. Gebregiorgis, G.M. Woldemichael, M. He, S.L. Lessnick, M. Kang, M. Smith, Z.B. Madaj, M.E. Winn, P.J. Grohar

Writing, review, and/or revision of the manuscript: C.L. Osgood, C.G. Kidd, G.M. Woldemichael, M. He, S. Lessnick, M. Smith, Z.B. Madaj, M.E. Winn, L.-E. Núñez, L.J. Helman, F. Morís, P.J. Grohar

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C.L. Osgood, N. Maloney, M. Gebregiorgis, P.J. Grohar

Study supervision: P.J. Grohar

Other (generation of compounds by combinatorial biosynthesis. Isolation and purification of compounds): L.-E. Núñez

Other (carried out the chemoenzymatic synthesis of some of the compounds involved in this article; elucidated the chemical structure of all the compounds by NMR spectroscopy): J. González-Sabín

The authors acknowledge the contractor, the Southern Research Company, for intravenous treatment of mice in this article. They also thank David Nadziejka for technical editing of the article.

The majority of the funding for this work was from the Hyundai Hope on Wheels Program (to P.J. Grohar, C.G. Kidd, and N. Maloney). Funding was also provided by the T32 training grant conducting research in pediatric oncology (Principal Investigator, D. Friedman) as well as the Vanderbilt Clinical Oncology Research Career Development K12 both for C.L. Osgood (Principal Investigator, D. Friedman). Additional internal funds were provided by the Vanderbilt University Department of Pediatrics (to P.J. Grohar, C.L. Osgood, and N. Maloney). Immunofluorescent assay development was supported by the Alex's Lemonade Stand Reach Award (to P.J. Grohar). Additional funding was provided by the Lily's Garden Foundation (to P.J. Grohar and L. Segars). EntreChem thanks Mariam Hermosilla and Patricia Oro for superior technical assistance. This project has also been funded in part with federal funds from the Frederick National Laboratory for Cancer Research, National Institutes of Health under contract HHSN261200800001E (to G.M. Woldemichael).

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.