High-throughput Chemical Screening Identifies Focal Adhesion Kinase and Aurora Kinase B Inhibition as a Synergistic Treatment Combination in Ewing Sarcoma

Ewing sarcoma is the second most common pediatric bone malignancy of childhood. Therapy for newly diagnosed patients consists of systemic treatment with repeated cycles of chemotherapy combined with either surgical resection, radiotherapy, or both. Recent efforts to improve outcomes for patients who present with nonmetastatic disease by intensifying therapy have resulted in only a modest increase in 5-year event-free survival (EFS) suggesting that additional attempts to intensify therapy may have only limited efficacy (1). For patients with metastatic disease at diagnosis, chemotherapy intensification has done little to improve outcomes with an expected 5-year EFS of approximately 20% (2). Thus, new classes of treatment are needed for patients with Ewing sarcoma.

The hallmark molecular aberration identified in Ewing sarcoma tumors is the presence of genomic rearrangements of TET-family genes with ETS-family transcription factor genes. The resulting fusion protein, most commonly EWS/FLI, is believed to be the oncogenic driving event in this disease (3). However, pharmacologic modulation of transcription factors has been notoriously challenging, and direct inhibitors of EWS/FLI have yet to be successfully applied in the clinic. While there was great hope that genomic profiling of Ewing sarcoma tumors would reveal recurrent somatic dependencies more readily tractable with targeted inhibitors, recent studies demonstrated that these tumors are among the most genomically stable human cancers (4). These studies also demonstrated a stark paucity of recurrently mutated genes currently considered targetable by available anticancer therapies (5, 6).

To identify new therapeutic approaches for Ewing sarcoma, we previously utilized a proteomic approach to identify molecular dependencies that may be more amenable to direct inhibition. We determined that Ewing sarcoma cells are dependent on focal adhesion kinase (FAK) in vitro and in mouse xenograft models of this disease (7). FAK is a non-receptor tyrosine kinase that promotes cellular growth, survival, and migration. FAK expression and activity in cancer have been associated with poor outcome and inhibitors of FAK are under active clinical investigation (8, 9). Recognizing that single-agent targeted therapy is rarely curative in cancer, we performed a small-molecule library screen of Ewing sarcoma cell lines to identify additional targetable dependencies that, when inhibited, synergistically impair cell viability in combination with FAK inhibition. We demonstrate that Aurora Kinase B inhibitors are a class of compounds that act synergistically with FAK inhibitors to impair cell viability in Ewing sarcoma cell lines and suppress tumor progression and improve survival in xenograft models of Ewing sarcoma.

A high-throughput screen was conducted in 1,536-well white flat bottom plates (Corning) on a Kalypsys robotic system. Using a MultiDrop Combi (Thermo Fisher Scientific), 2 μL of media were added to the plates. Compounds were dissolved in DMSO and then added to plates using the EDC ATS100 acoustic dispenser (EDC) creating 6 by 6 matrix blocks using a method described previously (10). Briefly, a 6-concentration-point, threefold dilution of PF-562271, a selective FAK inhibitor (SynKinase), was transferred to the screen plates by transferring a total volume of 10 nL per well (11). The MIPE 4.0 library (Supplementary Table S1) was also transferred by acoustic dispensing into the same plates with each compound plated in 6-point dilution series at fivefold serial dilutions. A673 cells were seeded into plates at a final density of 500 cells in 5 μL of media per well (MultiDrop Combi). After 48 hours of compound incubation, 3 μL of Cell-TiterGlo reagent (Promega) was added to each well. Following a 10-minute incubation, luminescence was read using the Viewlux microplate reader (PerkinElmer). Primary combination screening data are publicly available at https://tripod.nih.gov/matrix-client/?p=442.

Select combinations of interest from the primary screen were retested against a panel of Ewing sarcoma cell lines. Cells were seeded in 384-well white flat bottom plates (Corning) at a final density of 1,000 cells in 50 μL of media per well using a robotic dispenser (BioTek). Immediately after plating, media with PF-562271 (Selleck) at a 7-point, twofold dilution was added robotically by a Bravo Liquid Handler (Agilent) to the plate in combination with media containing either AZD-1152 (Sigma Aldrich) or MLN-8237 (Selleck) at an 11-point, twofold dilution (12, 13). Cells were treated with four replicates of each concentration of each combination. After 48 hours, 10 μL of Cell-TiterGlo reagent (Promega) were added to each well. Luminescence was read using the FLUOstar Omega microplate reader (BMG Labtech). Dose–response curves for each individual compound were generated from wells in microtiter plates treated with only one compound. Cell culture conditions, cell line verification, and analysis of drug synergy are detailed in the Supplementary Material.

Gene expression data for AURKA and AURKB were extracted from previously published transcriptome sequencing of 23 primary Ewing sarcoma tumors and nine Ewing sarcoma cell lines (5). AURKA and AURKB gene expression data were also extracted from The Cancer Genome Atlas (TCGA; http://cancergenome.nih.gov; ref. 14). AURKA and AURKB expression were also downloaded from the Cancer Cell Line Encyclopedia (CCLE; http://portals.broadinstitute.org/ccle; ref. 15). One-way ANOVA with ranks test was used to compare expression of genes in tumor subtypes. A two-tailed unpaired t test was used to compare gene expression of genes in Ewing sarcoma cell lines to all others. Analyses were performed in Prism 6 (GraphPad).

Publicly available cell line treatment data and analyses were accessed at the Genomics of Drug Sensitivity in Cancer Project website (http://www.cancerrxgene.org) on July 2016, Release 6 (16). Data and analyses were downloaded in three ways (i) EWS/FLI as a predictor of sensitivity to all drugs, (ii) cancer features predicting sensitivity to GSK-1070916, and (iii) IC₅₀ of all cell lines treated with GSK-1070916 (17). The half-maximal inhibitory concentration (IC₅₀) of the Ewing sarcoma cell lines treated with GSK-1070916 was compared with all other cell lines by a two-tailed Mann–Whitney test. Downloaded data were plotted and analyzed in Prism 6 (GraphPad).

To determine the dependency of Ewing sarcoma cell lines on AURKB, data from the Broad Institute's Achilles Project v2.4.3 were analyzed (www.broadinstitute.org/achilles; ref. 18). The Achilles Project utilized an RNAi library of 56,903 barcoded shRNAs targeting 14,222 genes. A collection of 216 cancer cell lines, including five Ewing sarcoma lines, were transduced with this library and after 16 cell doublings, cells were assessed for relative enrichment or depletion of shRNAs. The dataset contained three hairpins targeting AURKB (Supplementary Table S2). Ewing sarcoma dependencies were scored with the RIGER method using Gene-E v3.0.204 (www.broadinstitute.org/cancer/software/GENE-E; ref. 19). Individual hairpins were ranked by their average dependency across the Ewing sarcoma cell lines based on a z-score. Gene-level depletion scores were computed from the weighted sum of the z-scores for the first (weight of 0.25) and the second (weight of 0.75) ranked hairpins for that gene and the significance was assessed by permutated P value.

A list of 449 potential putative chromatin-regulatory genes was compiled using (i) an NCBI Gene Ontology Annotation Database search with terms including: epigenetic regulation of gene expression, chromatin modification, histone binding, histone kinase, and histone modification and (ii) a search of the UniProt database for genes with functional domains associated with epigenetic regulation. A collection of shRNA pLKO.1 constructs was assembled from the Broad Institute's Genetic Perturbation Platform (http://portals.broadinstitute.org/gpp/public/) such that seven to eight shRNAs were included per gene plus 12 control shRNAs, totaling approximately 3,000 shRNAs. The collection included seven hairpins targeting AURKB (Supplementary Table S3).

For lentivirus production, HEK-293T cells were transfected using the X-tremeGene HP DNA Transfection Reagent protocol (Roche). Lentiviral vector and packaging plasmids (pCMV8.9 and pCMV-VSVG) were transfected along with shRNA-expressing plasmids. Cell lines were transduced with the pooled lentiviral shRNA library and selected with puromycin 12 hours posttransduction. After cells were cultured for 7–10 days postselection, 10 million cells were collected as an “input” sample. Simultaneously, 1 million cells suspended in 30% Matrigel were injected subcutaneously in five NOD SCID gamma (NSG) mice. Once tumors reached 500 mm³, mice were sacrificed and DNA was extracted from each tumor and the input samples using a DNeasy Blood and Tissue Kit (Qiagen). DNA was amplified using nested multiplex PCR with barcoded primers specific to the target region of each shRNA. PCR products were purified and sequenced. Hairpin abundance was estimated by determining the counts-per-million (CPM) of each hairpin barcode identified. We used Wilcoxon rank order tests corrected for multiple hypothesis testing by Benjamini–Hochberg method to identify significant changes in CPM relative to input for individual hairpins or at the gene level by combining all hairpin counts for a given gene.

Aurora kinase, FAK, and EWS/FLI expression were downregulated by transducing cells with either CRISPR Cas9 gene editing plasmids or inducible shRNA plasmids packaged in lentivirus as described above. For CRISPR studies, target guides were designed using the Broad Institute Genetic Perturbation Platform's sgRNA design tool (www.broadinstitute.org/rnai/public/analysis-tools/sgrna-design). Selected guides (Supplementary Table S4) were cloned into the lentiCRISPR V2 plasmid (Addgene) as described previously (20). Inducible shRNA plasmids were obtained from Dharmacon (GE; Supplementary Table S5). For transduction, Ewing cells were incubated for 2 hours with 2 mL of virus and 8 μg/mL polybrene. Cells were selected with puromycin 2 days after transduction. Cells transduced with inducible shRNA plasmids were transferred 2 days later to medium containing puromycin and 1 μg/mL of doxycycline. Experiments were initiated within 3–5 days of puromycin selection and, for cells with inducible plasmids, two days after the addition of doxycycline. Downregulation of target gene expression was confirmed by Western immunoblotting as described in the Supplementary Material.

To determine the effects of perturbations on Ewing sarcoma cell viability, cells were plated in 384-well plates at a concentration of 1,000 cells per well in 50 μL of medium. Cell viability was measured by adding 10 μL of Cell-Titer Glo ATP-based assay (Promega). Luminescence was read using the FLUOstar Omega microplate reader (BMG LabTech). VS-4718 and GSK- 2461070916 were obtained from Selleck for in vitro treatment of cells. Cells undergoing apoptosis were stained with Annexin V using the Apoptosis Detection Kit-APC (eBioscience) and cellular DNA content was measured by propidium iodide staining (Invitrogen). For intracellular phospho-protein staining, cells were fixed and permeabilized using the BD Cytofix/Cytoperm Kit (BD Biosciences) and stained with phycoerythrin (PE) anti-phospho-S6 (S240, BD Biosciences) and analyzed by flow cytometry. A minimum of 10,000 stained cells were analyzed in all flow cytometry experiments. All experiments testing viability, apoptosis, cell cycle, and measurements of phosphorylation of S6 were performed with two or more experimental replicates and each experiment repeated a minimum of two times. Experiments shown are representative of experimental and biologic replicates.

A673 and TC32 cell lines were treated for 36 hours with FAK inhibitors (PF-562271 and VS-4718 at 2.5 μmol/L) and Aurora kinase B inhibitors (AZD-1152 and GSK 1070916 at 20 nmol/L). Cell pellets were harvested, lysed and quantified, and applied to the chip-based RPPA array and the RPPA assay was performed and analyzed by the MD Anderson Cancer Center core facility as described previously (21). Heatmaps were generated with Morpheus software (https://software.broadinstitute.org/morpheus/) with normalized protein or phospho-protein levels.

Methods for zebrafish studies and mouse pharmacokinetic and tolerability studies are described in the Supplementary Material. For all mouse xenograft studies, AZD-1152, PF-562271, and VS-4718 were purchased from MedChem Express. For all studies, AZD-1152 was dissolved in 0.3 mol/L Tris base pH 9.0 and dosed at 25 mg/kg daily by intraperitoneal injection. PF-562271 was dissolved in 0.5% HPMC, 20% Tween80, and administered by oral gavage twice daily at a dose of 100 mg/kg. VS-4718 was dissolved in 0.5% carboxymethyl cellulose and 0.1% Tween 80 in sterile water and administered by oral gavage twice daily at a dose of 50 mg/kg. Mouse xenograft tumor progression and survival studies were approved by the DFCI Animal Care and Use Committee. For cell line xenograft studies, 3 million A673 cells were collected and injected subcutaneously into the right flank of 7- to 8-week-old female NCr nude mice (Charles River Laboratories). For patient-derived xenograft (PDX) studies, fragments of tumor were implanted subcutaneously into the right flank of 7-week-old female NCr nude mice (Charles River Laboratories). Tumor volume was monitored using calipers (volume = 0.5 × length × width²). For each study, mice were divided into four groups: vehicle control, AZD-1152 alone, FAK inhibitor alone (PF-562271 in the A673 xenograft study and VS-4718 in the PDX study), and AZD-1152 in combination with a FAK inhibitor. In both studies, treatment began when tumors reached at least 100 mm³, with AZD-1152 on days 1–4 and 8–11 and FAK inhibitor on days 1–14. Tumor volumes were measured twice a week and animals were sacrificed when tumors reached the institutional limit of 2,000 mm³.

A table summarizing the reagents and chemicals used in this study is available in the Supplementary Material.

To identify new candidate combination treatment approaches with FAK inhibitors in Ewing sarcoma, we performed a high-throughput screen of 1,912 compounds in combination with the FAK inhibitor, PF-562271. We screened the Mechanism Interrogation PlatE (MIPE) compound library 4.0, which includes compounds selected for anticancer activity, established mechanistic annotation, and clinical relevance (Supplementary Table S1). A673 Ewing sarcoma cells were treated with PF-562271 at six concentrations in combination with each compound from the MIPE 4.0 library at six concentrations, resulting in a treatment matrix of all possible combinations for each compound pair (Fig. 1A). Cells were treated for 48 hours and then cell viability was measured. Each treatment matrix was scored by four metrics for evidence of synergistic inhibition of cell viability (Supplementary Table S6). We found that seven Aurora kinase inhibitors ranked in the top 5% of compounds when treatment matrices were analyzed for synergy by the sum of excess over highest single agent (HSA; Fig. 1B). We also noted that the top 10 scoring Aurora kinase inhibitors in the screen are classified as either pan-Aurora kinase or Aurora kinase B–selective inhibitors (Fig. 1B; Supplementary Table S6). Tozasertib, a pan-Aurora inhibitor, is an illustrative example of the synergistic anti-Ewing activity observed in the screen by combining Aurora kinase inhibition with FAK inhibition (Fig. 1C). Indeed, we found that multiple dose combinations of tozasertib and PF-562271 in this screen induced synergistic inhibition of A673 cell viability as determined by combination index (Fig. 1D).

We next evaluated the expression pattern of Aurora kinases in Ewing sarcoma cell lines and primary tumors. Previous studies have demonstrated that Aurora kinases are highly expressed in many cancer types (22), with Aurora kinase A and B expressed in Ewing sarcoma cell lines (23, 24). We examined RNA-Seq data from a recently published collection of 23 primary Ewing sarcoma tumors and nine Ewing sarcoma cell lines and found that Aurora kinases A and B were highly expressed but Aurora kinase C was not (Fig. 1E; ref. 5). We found that Aurora kinase A and B expression were similar to expression levels observed in eight other cancer subtypes available through the TCGA (Supplementary Fig. S1A and S1B; ref. 14). We also found that Aurora kinase B is expressed in a collection of nine Ewing sarcoma cell lines (Supplementary Fig. S1C) and expression was in the upper third of all cancer cell lines in the CCLE, while Aurora kinase A expression was significantly lower than the average expression of cancer cell lines (Supplementary Fig. S1D and S1E; ref. 15).

Having confirmed that both Aurora kinase A and B are expressed in Ewing sarcoma, we next determined whether inhibition of Aurora kinase A, B, or both were synergistic with FAK inhibitors in Ewing sarcoma. First, we confirmed that Ewing sarcoma cells are sensitive to both Aurora kinase A- and B-specific inhibitors alone. For these experiments, we utilized MLN-8237, an Aurora kinase A–specific inhibitor, and AZD-1152, an Aurora kinase B–specific inhibitor. We chose these compounds for our confirmation experiments because they were the top ranking inhibitors in our initial screen with specificity for each Aurora kinase (Supplementary Tables S1 and S6), both have been tolerated in clinical trials and both continue to be actively developed for clinical use (25–31). We treated nine Ewing sarcoma cell lines with MLN-8237 and AZD-1152 across a broad range of concentrations for 2 days. We found that both inhibitors impaired cell viability at nanomolar concentrations (Fig. 1F and G). Slower-growing Ewing sarcoma cell lines did not achieve 90% inhibition of viability after 2 days of treatment (Fig. 1F and G; Supplementary Fig. S2A), an anticipated finding because Aurora kinase inhibitors impair cell division. When the slower growing TC32 cell line is treated for 5 days with AZD-1152, 90% inhibition is achieved at a similar concentration as in the faster growing cells at the 2-day time point (Supplementary Fig. S2B).

Next, we confirmed that Aurora kinase inhibitors are synergistic with FAK inhibitors in a collection of nine Ewing sarcoma cell lines. Cells were treated with the drug combinations AZD-1152 and PF-562271, AZD-1152 and VS-4718, or MLN-8237 and PF-562271 across a range of concentrations that specifically inhibited the target of each drug (Supplementary Fig. S2C–S2F). Each combination of specific concentrations was tested in quadruplicate. Cells were incubated with each drug for 2 days and viability was measured. To identify individual dose combinations that were synergistic, we calculated the combination index and excess over Bliss Independence based on the average inhibition of viability of the four replicates. In all Ewing sarcoma cell lines treated with AZD-1152 in combination with PF-562271 or VS-4718, we found that a larger range of drug concentrations had a combination index < 0.7 compared with cells treated with MLN-8237 in combination with PF-562271 (Fig. 2A–C; Supplementary Figs. S3 and S4).

One would expect that the most successful drug combinations in the clinic involve drugs that are also active as single agents against that disease. Our own prior work focused on validation of FAK as a therapeutic target in Ewing sarcoma (7). Published studies have supported a dependency of Ewing sarcoma cells on Aurora kinase B activity for viability in vitro (24). However, in light of Aurora kinase B emerging as a top target for synergistic inhibition with FAK inhibitors in Ewing sarcoma, we sought to more thoroughly validate this target in vitro and in vivo. We first searched the Genomics of Drug Sensitivity in Cancer Project database for compounds for which sensitivity was predicted by the expression of EWS/FLI, the genomic feature defining Ewing sarcoma cell lines in this dataset. Indeed, EWS/FLI was the top-scoring feature predicting response to GSK1070916, an Aurora kinase B–selective inhibitor, in this dataset (Fig. 3A). Moreover, GSK1070916 ranked ninth ranked among 265 compounds ranked by confidence (P value) in the correlation between EWS/FLI expression and sensitivity to treatment (Fig. 3B). Accordingly, the average IC₅₀ of Ewing sarcoma cell lines treated with GSK1070916 was significantly lower than the average IC₅₀ of all other cancer cell lines screened (Fig. 3C).

Next, we utilized publicly available data generated by the Broad Institute's Project Achilles to determine the relative dependency of Ewing sarcoma cell lines on Aurora kinase B expression (18). We found that Aurora kinase B ranked in the top 10% of genes sorted by the significance of hairpin depletion across Ewing sarcoma cell lines (Fig. 3D). We were unable to assess the effect of Aurora kinase A in this dataset as variability in the activity of the hairpins precluded analysis. Hairpins targeting Aurora kinase C were not significantly depleted which was expected as Ewing sarcoma cell lines do not express Aurora kinase C. We also utilized a previously developed focused shRNA library targeting 449 genes, including AURKB, to perform an in vivo dependency screen in Ewing sarcoma xenografts. A673 and TC32 cells transduced with this library were injected subcutaneously in NSG mice. The tumors that developed were examined for the changes in shRNA abundance to determine which genes are necessary for tumor establishment and progression in these models. Genes were ranked on the basis of the significance (by P value) of fold-change depletion of shRNA constructs targeting that gene. AURKB ranked seventh (out of 449 targeted genes) in the A673 in vivo model and scored in the top 10% (44 of 449) of genes in the TC32 in vivo model (Fig. 3E and F).

To further confirm the dependency of Ewing sarcoma on Aurora kinase B, we next utilized a CRISPR-Cas9 gene editing technique to target AURKB expression. Aurora kinase B expression was downregulated with two Aurora kinase B–specific Cas9-targeting guides (Fig. 4A and B). We found that knockout of Aurora kinase B decreased phosphorylation of histone H3, a substrate of Aurora kinase B (Supplementary Fig. S5A), and significantly impaired cell viability in the A673 and TC32 cell lines compared with cells treated with a nontargeting control guide (P < 0.0001 for all comparisons; Fig. 4C). We also examined the effect of Aurora kinase B knockout on cell cycling and apoptosis. Previous studies have reported that Aurora kinase inhibition primarily induces apoptosis in TP53-wild-type cancer cell lines while inducing cell-cycle arrest and a more delayed induction of apoptosis in TP53-mutated cancer cell lines (32, 33). Indeed, we found that Aurora kinase downregulation in the TP53-mutated cell line, A673, induced cell-cycle arrest in nearly 100% of the cells with evidence of endoreduplication, while 25% or less of cells exhibited induction of apoptosis. Apoptosis was induced in 40% of the TP53-wild-type Ewing sarcoma cells, TC32, with more modest changes in cell-cycle arrest and an absence of endoreduplication (Fig. 4D and E; Supplementary Fig. S5B–S5E).

The EWS/ETS translocations are the only known oncogenic driver in Ewing sarcoma. To determine whether these translocations contribute to the sensitivity of Ewing sarcoma cells to Aurora kinase B inhibitors, we downregulate EWS/FLI expression with inducible shRNA. We found that cells with depleted EWS/FLI expression were resistant to concentrations of AZD-1152 that cause loss of viability in Ewing cells expressing EWS/FLI (Fig. 4F). In short-term cultures, we saw that downregulation of EWS/FLI expression had a mild effect on cell growth in the absence of AZD-1152 treatment, as expected (Fig. 4G). One possible explanation for our results is that loss of EWS/FLI expression reduces cell growth which reduces the rate of cell cycling causing cells to be resistant to Aurora kinase B inhibition. However, Ewing sarcoma cell lines remained sensitive to AZD-1152 when cell growth was reduced by restricting FBS levels in the culture medium (Fig. 4H and I). The only exception was in TC32 cells grown in 0% FBS, but in this case, cells had completely arrested. These studies demonstrate that the rate of cell growth cannot be the primary reason that the expression of EWS/FLI in Ewing sarcoma sensitizes cells to inhibition of Aurora kinase B.

These findings strongly support a therapeutic opportunity for targeting Aurora kinase B in Ewing sarcoma. We next examined whether the combination of FAK inhibition enhances the effects of Aurora kinase B inhibition on cell-cycle arrest and apoptosis. We treated the A673 and TC32 cell lines with either control, PF-562271 alone, AZD-1152 alone, or both drugs together at concentrations corresponding to the IC₅₀ (PF-562271 = 2.5 μmol/L and AZD-1152 = 20 nmol/L), half the IC₅₀, and one quarter of the IC₅₀. We also found that the combination of these compounds induced a significant increase in G₂ arrest using a concentration of PF-562271 (1.25 μmol/L or half the IC₅₀) where FAK inhibition alone had no effect on cell cycle (Fig. 5A and B). Similarly, apoptosis was significantly increased in both cell lines across multiple drug concentrations when comparing the combination with each compound alone (Fig. 5C and D).

We next focused on validating the synergistic activity of combining Aurora kinase B inhibition with FAK inhibition. To confirm that loss of Aurora kinase B activity sensitizes cells to FAK inhibition, we examined the effect of genetic downregulation of Aurora kinase B on sensitivity of A673 and TC32 cells to treatment with PF-562271. For these experiments, we chose an inducible shRNA system because CRISPR-Cas9 knockout of Aurora kinase B led to such a profound loss of viability that there were insufficient numbers of cells to treat with PF-562271. Downregulation of Aurora kinase B with shRNA, in contrast, significantly impaired cell viability but less profoundly than with CRISPR knockout. We found that downregulation of Aurora kinase B in A673 and TC32 cells resulted in sensitivity to lower concentrations of PF-562271 compared with cells treated with nontargeting control shRNA (Fig. 5E and F). To demonstrate that FAK downregulation sensitizes Ewing sarcoma cells to Aurora kinase B inhibition, we next generated CRISPR Cas9 FAK-targeting guides that were specific for the kinase region of the gene and had no effect on expression AURKB or PYK2, a FAK homolog that is poorly expressed in A673 and TC32 (Fig. 5G; Supplementary Fig. S6). We found that loss of FAK expression sensitized A673 and TC32 cells to lower concentrations of AZD-1152 (Fig. 5H).

To identify mechanisms contributing to the synergistic activity of FAK and Aurora kinase B inhibition in Ewing sarcoma cells, we analyzed the effect of these inhibitors on the phosphoproteome using a RPPA. Ewing sarcoma cells were treated for 36 hours with two FAK inhibitors (PF-562271 and VS-4718) and two Aurora kinase B inhibitors (AZD-1152 and GSK-1070916) at the IC₅₀ for each drug. Cells were treated with each drug alone and with each combination of FAK plus Aurora kinase B inhibitors and protein from cell lysates were analyzed (Supplementary Table S7). As we previously demonstrated, Ewing sarcoma cells treated with FAK inhibitors have a decrease in phosphorylated levels of S6 compared with vehicle-treated control cells (7). Interestingly, we found that phosphorylation levels of S6 and several other members of the mTOR pathway, were downregulated more when cells were treated with FAK and Aurora kinase B inhibitors together than cells treated with FAK inhibitors alone (Fig. 6A).

To confirm the finding that FAK and Aurora kinase inhibitors suppress mTOR pathway activity more than FAK inhibitors alone, we then treated Ewing sarcoma cells for 24 hours with VS-4718 and AZD-1152, alone and in combination, at one quarter the IC₅₀ (Fig. 6B) and one half the IC₅₀ (Fig. 6C). We chose low concentrations and a short treatment duration to make sure we were capturing on-target activity while cell viability was still minimally impacted by treatment. Indeed, we found that the combination resulted in lower levels of phosphorylated S6 than treatment of cells with vehicle or either compound alone. The dependence of Ewing sarcoma cells on mTOR pathway activity has been well described (34, 35). These results support a model where Aurora kinase B inhibition enhances the anti-Ewing effect of downregulating mTOR activity in Ewing sarcoma.

To determine the effects of combining Aurora kinase B and FAK inhibition on Ewing sarcoma tumor progression in vivo, we first utilized an established zebrafish xenograft model of Ewing sarcoma (36). We determined that zebrafish tolerated the combination of 5 μmol/L PF-562271 with 6 μmol/L AZD-1152 and treatment with this combination significantly impaired tumor progression in an A673 zebrafish xenograft model of Ewing sarcoma compared with vehicle or single-agent treatment. (Supplementary Fig. S7A–S7D). We then tested the effects of the combination of PF-562271 or VS-4718 and AZD-1152 on tumor progression and survival in two mouse xenograft models of Ewing sarcoma. First, we assessed the potential for drug–drug interactions and resultant effect on the drugs' metabolism and exposure profiles. Pharmacokinetic experiments were performed with both single-agent and combination dosing in NCr nude mice. No significant change in plasma exposure was observed for any of the drugs in either combination compared with their single-agent arms (Supplementary Fig. S7E–S7G). Next, we established the tolerability of the combination of 100 mg/kg orally twice a day of PF-562271 and 25 mg/kg intraperitoneally daily of AZD-1152 for 5 days or 50 mg/kg orally twice a day of VS-4718 and 25 mg/kg intraperitoneally daily of AZD-1152 in CD1 nude mice (Supplementary Fig. S7H and S7I). We then treated Ncr nude mice, after establishment of palpable A673 xenograft tumors, with PF-562271 for 14 days and AZD-1152 on days 1–4 and 8–11. In this experiment, we chose a dose of PF-562271 that was substantially lower than we had previously used to maximize our ability to detect the synergistic activity of the combination (7). We found that the FAK inhibitor alone did not have a significant effect on tumor progression at this reduced dose, but treatment with both AZD-1152 alone and the combination treatment significantly impaired tumor progression by day 14 of treatment (Fig. 6D). Furthermore, treatment with the combination (P = 0.004) significantly improved survival over treatment with vehicle. Treatment with the combination also improved survival compared with PF-562271 treatment alone (P = 0.01) and AZD-1152 treatment alone (P = 0.02; Fig. 6E). Finally, mice with established palpable Ewing sarcoma patient-derived xenografts (PDX) were treated with VS-4718 for 14 days and AZD-1152 on days 1–4 and 8–11. Again, we found that the FAK inhibitor alone did not have a significant effect on tumor progression, but treatment with both AZD-1152 alone and the combination treatment significantly impaired tumor progression by day 21 (P < 0.0001; Fig. 6F). However, only the mice treated with combination therapy had a significantly prolonged survival compared with vehicle-treated mice (P = 0.01; Fig. 6G).

Despite a growing understanding of the oncogenic mechanisms that cause Ewing sarcoma, there remains a relative paucity of new candidate drug targets for the treatment of this disease. We recently identified FAK as a targetable dependency in Ewing sarcoma (7). To identify potential drug combinations for Ewing sarcoma that incorporate FAK inhibitors, we performed an unbiased screen of a larger collection of anticancer agents that includes traditional chemotherapeutic drugs as well as targeted inhibitors. We found that multiple Aurora kinase inhibitors demonstrated synergistic anti-Ewing activity when combined with FAK inhibition.

Aurora kinases are serine/threonine kinases that regulate chromosomal alignment and segregation during cell division (37). Aurora kinases are frequently amplified and highly expressed in cancer (22). Thus, significant effort has been focused on developing Aurora kinase inhibitors for cancer therapy, and several inhibitors have been studied in early-phase clinical trials in patients with leukemia and solid tumors (22). Aurora kinase inhibitors, including pan-Aurora inhibitors and those that selectively target Aurora kinase A or B, have been used in patients, with bone marrow suppression being the most common dose-limiting toxicity (25, 26, 30). Studies have also demonstrated responses and disease stabilization in patients with cancer treated with single-agent Aurora kinase inhibitors (28, 30, 31). Aurora kinase inhibitors are also tolerated in children, and two recent studies demonstrated efficacy for some children with aggressive solid tumors treated with an Aurora kinase A inhibitor, MLN-8237 (27, 31). However, chronic exposure to these agents have induced some undesirable toxicities. To further improve the therapeutic window, one group has developed a nanoparticle formulation of an Aurora kinase B inhibitor (38). In vivo models demonstrated that this delivery approach reduces bone marrow toxicity while increasing drug exposure to the tumors. This formulation is now being tested for safety and efficacy in a clinical trial for patients with advanced solid tumors (https://clinicaltrials.gov/ct2/show/NCT02579226).

Previous studies have described a potential role for Aurora kinase inhibition in the treatment of Ewing sarcoma. Ewing sarcoma cell lines express Aurora kinase A and B, and transcript levels for both kinases correlate with EWS/FLI expression (23, 24). Ewing sarcoma cell lines treated with pan-Aurora kinase inhibitors also demonstrate impaired cell viability and tumor proliferation (24). Interestingly, one study found that EWS/FLI interferes with the interaction of Aurora kinase B and wild-type EWS, impairing normal localization of Aurora kinase B during mitosis (39). On the basis of this finding, one might expect that Ewing sarcoma cells would be highly sensitive to further perturbations of Aurora kinase B activity. We utilized orthogonal small-molecule and functional genomic screening data combined with Aurora kinase B knockout experiments to highlight the dependency of Ewing sarcoma cells specifically on Aurora kinase B. Downregulation of EWS/FLI expression diminished the dependency of cells on Aurora kinase B activity, independent of cell growth rate, suggesting that the EWS/FLI oncogene may be directly responsible for sensitizing Ewing sarcoma to Aurora kinase B inhibitors. Furthermore, we demonstrated that single-agent Aurora kinase B inhibition impairs tumor progression and prolongs survival in a murine xenograft model of Ewing sarcoma. Our data, combined with that from previously published work, provide support for further studies exploring the molecular mechanisms that render Ewing sarcoma cells dependent on Aurora kinase B activity.

While targeted therapy with tyrosine kinase inhibitors remains an attractive therapeutic approach in cancer, numerous studies demonstrate that single-agent therapies are unlikely to result in durable treatment responses (40, 41). Our study was designed to identify drugs that could be combined with FAK inhibitors for the treatment of patients with Ewing sarcoma. Clinical trials have demonstrated that FAK inhibitors are well tolerated, and preclinical studies in mice have demonstrated tolerability of FAK inhibitors with both chemotherapy and targeted agents (42–44). Our screen nominated Aurora kinase B inhibitors for combination with FAK inhibitors in Ewing sarcoma cell lines. Several studies have also demonstrated the feasibility of combining Aurora kinase inhibitors with other anticancer therapies, including a study that combined an Aurora kinase B–selective inhibitor with chemotherapy (29). We found that the combination of FAK inhibition and Aurora kinase B inhibition was well tolerated in zebrafish and mice. Furthermore, we found that this combination impairs tumor progression in multiple xenograft models of Ewing sarcoma. One limitation of our study is that Aurora kinase B inhibition alone significantly impaired tumor progression in our xenograft models, limiting the ability to fully measure the additional benefit of the combination on tumor progression. However, the combination of FAK and Aurora kinase B inhibition did demonstrate an improvement in survival more than treatment with either drug alone, providing enthusiasm for further investigating this candidate combination for Ewing sarcoma. While short-term treatment with these inhibitors prevented xenograft disease progression until treatment was withdrawn, future studies will be needed to determine whether additional cycles of treatment can induce tumor regressions and remissions.

Finally, we utilized an unbiased proteomic screening approach to explore potential mechanisms of synergistic activity of FAK and AURKB inhibitors in Ewing sarcoma. We found that inhibiting both targets suppressed mTOR activity more than inhibition of either FAK or AURKB alone. The mTOR pathway is an important signaling mechanism within cells that senses the availability of adequate nutrients and reagents for cell growth and cell division (45). Many cancers are dependent on the mTOR pathway, including Ewing sarcoma, where both activated FAK and IGF1R promote mTOR activity (7, 34, 35, 46). Interestingly, several studies now demonstrate that mTOR colocalizes with the chromosomal passenger complex (CPC) and AURKB, one of the core members of the CPC, during mitosis (47, 48). AURKB and mTOR appear to cross regulate each other and loss of AURKB activity sensitizes cells to direct inhibition of mTOR activity by rapamycin (48, 49). Therefore, one would expect that downregulation of both AURKB and tyrosine kinases that promote mTOR activity would have a synergistic anti-Ewing sarcoma effect. Indeed, our phenotypic assays demonstrate that the loss of both AURKB and FAK activity profoundly increased cell-cycle arrest, supporting the theory that both AURKB and mTOR play a central role in promoting mitosis in Ewing sarcoma. Furthermore, in another recent study, IGF1R and CDK4/6 inhibitors were found to have synergistic activity in Ewing sarcoma cells acting through downregulation of the mTOR pathway (50). This combination also enhanced cell death in Ewing sarcoma cells. Together, these findings suggest the possibility of a more general therapeutic opportunity in Ewing sarcoma for targeting tyrosine kinase regulators of mTOR in combination with inhibitors of cell cycle, a hypothesis currently undergoing further investigation. With numerous drugs targeting the regulation of mTOR activity in clinical investigation, including FAK and IGF1R inhibitors, and Aurora kinase B, pan-Aurora kinase, and other cell-cycle inhibitors in various stages of clinical development, these classes of compounds are exciting therapeutic agents warranting further investigation for patients with Ewing sarcoma.

K. Stegmaier reports receiving commercial research grants from Novartis and is a consultant/advisory board member for Novartis and Rigel. No potential conflicts of interest were disclosed by the other authors.

Conception and design: S. Wang, M.I. Davis, K. Stegmaier, B.D. Crompton

Development of methodology: S. Wang, C.J. Veinotte, A. Wang, M.B. Boxer, A.L. Kung, J.N. Berman, B.D. Crompton

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Wang, E.E. Hwang, A.F. O'Neill, N. Melong, C.J. Veinotte, A.S. Conway, K. Wuerthele, A. Wang, E. Hughes, A.L. Kung, M.I. Davis, K. Stegmaier, B.D. Crompton

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Wang, E.E. Hwang, R. Guha, N. Melong, C.J. Veinotte, A.S. Conway, K. Wuerthele, M. Shen, G. Alexe, M.E. Lemieux, A. Wang, E. Hughes, A.L. Kung, J.N. Berman, M.I. Davis, B.D. Crompton

Writing, review, and/or revision of the manuscript: S. Wang, E.E. Hwang, A.F. O'Neill, N. Melong, C.J. Veinotte, M. Shen, M.E. Lemieux, A. Wang, X. Xu, M.D. Hall, A.L. Kung, J.N. Berman, M.I. Davis, K. Stegmaier, B.D. Crompton

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): N. Melong, A.S. Conway, K. Wuerthele, B.D. Crompton

Study supervision: M.D. Hall, J.N. Berman, K. Stegmaier, B.D. Crompton

Other (performed the matrix screening for the testing of the small- molecule library synergistic with FAK inhibitors in impairing Ewing cell growth): C. McKnight

This work was supported by the NCI 1 K08 CA188073-01A1, Team Clarkie St. Baldrick's Scholar Award, Rally Foundation, The Truth 365, Pediatric Cancer Research Foundation, Nathaniel Cavallo Fund, Team Fernando's Fight (to B.D. Crompton); R01 CA204915, Brian MacIsaac Sarcoma Foundation and Cookies for Kids Cancer (to K. Stegmaier); and the Intramural Research Program of the National Center for Advancing Translational Sciences, (Bethesda, MD). The authors thank Stephanie Meyer and Brigit McDannell for their technical assistance and the NCATS matrix team, including Sam Michael, Carleen Klumpp-Thomas, Paul Shinn, Tim Mierzwa, Lesley Griner, and Xiaohu Zhang, for their technical contributions.

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.