Targeting EYA3 in Ewing Sarcoma Retards Tumor Growth and Angiogenesis

Ewing sarcoma is an aggressive but rare tumor of the bone and surrounding soft tissue that predominantly afflicts children and young adults. Although nonmetastatic disease is usually responsive to chemotherapy, radiation, and surgery, recurrences are common (1). Patients presenting with advanced, metastatic, or relapsed disease have a very poor prognosis (2). In over 85% of Ewing sarcomas, a somatic mutation results in a translocation between the EWSR1 gene on chromosome 22 and the FLI1 gene on chromosome 11 (t(11;22)), creating the EWSR1/FLI1 fusion gene. The protein product EWS/FLI1 combines the DNA-binding properties of FLI1 and the transcriptional regulation function of EWSR1, and leads to dysregulated transcription of genes associated with cell proliferation and survival. Other than this, driving translocation Ewing sarcoma is characterized by a low somatic mutational burden. From a therapeutic standpoint, it is of note that mutations in genes involved in kinase signaling pathways have not been reported. A few targeted therapeutics are currently in clinical trials including insulin growth factor-1 receptor inhibitors (NCT02306161), the PARP inhibitor olaparib (NCT01858168), and LSD1 inhibitors (NCT03514497 and NCT03600649). Immune-checkpoint inhibitors are generally unsuccessful in treating Ewing sarcoma family tumors (3), most likely because of the absence of PD-L1 expression (4). Although sarcomas are vascular, and tumor cells secrete proangiogenic factors, monotherapy with antiangiogenics has, at best, provided survival benefits on the order of weeks (5). Clinical trials of antiangiogenics in combination with cytotoxic chemotherapy are showing more promise (6, 7). Overall, relative to other malignancies, few targeted therapeutics exist, and there has been no major improvement in the treatment of relapsed or refractory Ewing sarcoma since the advent of multiagent adjuvant chemotherapy. Hence, the identification of new targetable pathways is of great interest.

Because the transcription factor fusion EWS/FLI1 is a master regulator of the tumorigenic phenotype in Ewing sarcoma, it is an obvious target for therapeutics. However, despite much effort, no EWS/FLI1-targeted therapeutic has yet reached a clinical trial. The general difficulty in targeting transcription factors that lack a defined active site and the intrinsic disorder of the EWS/FLI1 protein (8, 9) contribute to the challenge. Approaches currently in favor include functional modulation of EWS/FLI1 through the disruption of either EWS/FLI1–protein interactions or EWS/FLI1 fusion–associated epigenomic or transcriptomic changes. Among these are micro-RNAs (miRNA) that contribute to EWS/FLI1-initiated oncogenic pathways (10, 11). miR-708 is an EWS/FLI1-downregulated miR that binds to the 3′-untranslated region of the EYA3 gene, thus repressing EYA3 expression (schematized in Fig. 6N; ref. 8). In Ewing sarcoma cells, downregulation of miR-708 correlates with high levels of the EYA3 protein. EYA3 belongs to the Eyes Absent (EYA) family of proteins that are both protein tyrosine phosphatases (PTP) as well as transcriptional activators (12, 13). It has previously been shown that loss of EYA3 in Ewing sarcoma cells confers chemosensitivity in vitro through an impaired DNA damage repair process (14). In the present study, we demonstrate that loss of EYA3 in the Ewing sarcoma cell line A673 impairs tumor growth in vivo, inhibits cell survival and migration in vitro, and inhibits tumor angiogenesis through downregulation of vascular endothelial growth factor (VEGFA) levels. Furthermore, pharmacologic inhibition of EYA3 PTP activity retards both cell line–derived xenograft tumor growth and the growth and vascularization of a patient-derived Ewing sarcoma xenograft. Together, these observations demonstrate that EYA3 promotes Ewing sarcoma tumor growth and tumor angiogenesis, and that inhibiting EYA3-dependent molecular signaling could be beneficial in the treatment of Ewing sarcoma tumors.

Anti-EYA3 (Abcam; cat. #ab95876; RRID:AB_10681036), anti-Ki67 (Thermo Scientific; cat. #MA5-14520; RRID:AB_10979488), pan anti-actin C4 (Seven Hills Bioreagents; ref. 15), antiendomucin (Abcam; cat. #ab106100 RRID:AB_10859306), anti-γ-H2AX (Millipore; cat. #05-636 RRID:AB_309864), anti-H2AX (phospho-Y142; Abcam; cat. #ab94602, RRID:AB_10858263), anti-VEGFA (Thermo Scientific; cat. #MA1-16629; RRID:AB_2212682), anti-cleaved caspase-3 (Cell Signaling Technology; cat. #9664S, RRID:AB_2070042), biotinylated goat anti rabbit IgG (H+L; Vector Labs; cat. #BA1000, RRID:AB_2313606), biotinylated goat anti rat IgG (H+L; Vector Labs; cat. #BA9400, RRID:AB_2336202), biotinylated goat anti mouse IgG (H+L; Invitrogen; cat. #62-6540, RRID:AB_2533949), Alexa Fluor 488 anti-rat (H+L; Invitrogen; cat. #A21208, RRID:AB_2535794), Alexa Fluor 594 anti-rabbit (H+L; Invitrogen; cat. #A21207, RRID:AB_141637).

Benzarone (2-Ethyl-1-benzofuran-3-yl)-(4-hydroxyphenyl)methanone (BZ) was obtained from Toronto Research Chemicals, DMSO (D2650) from Sigma-Aldrich), and CCK-8 kit from Dojindo Molecular Technologies). A673 and RD-ES cells were obtained from ATCC (A673 ATCC Crl-1598 obtained in January 2016, RD-ES ATCC HTB-166 obtained in February 2016) tested for Mycoplasma contamination (ATCC Mycoplasma detection kit, 30-1012K), maintained in DMEM containing 1% vv^–1 penicillin (100 IU mL^–1), streptomycin 100 mg mL^–1, 10% vv⁻¹ fetal bovine serum (FBS), and used within 10 passages. CRISPR/Cas9 was used to disrupt Eya3 expression by causing a double-strand break within the gene using the Eya3 CRISPR/Cas9 KO system (SC-406221 and SC-406221-HDR, Santa Cruz Biotechnology) consisting of a pool of three plasmids encoding the Cas9 nuclease and one of three target-specific guide RNAs:

Guide RNA 1: TCGCTCATCCAATGATTATA

Guide RNA 2: AACGTATGGACTACCTCCT

Guide RNA 3: GAAATACTTACTGGTTGCTC

Lentiviral short hairpin constructs targeting EYA3 (shEya3#3 and shEya3#5) and control shRNA (scramble) were obtained from Sigma-Aldrich and Addgene.

pLKO.1 puro shEYA3#3 (TRC000005163): CCGGCCCTTCTAGAAGTCCATCTTTCTCGAGAAAGATGGACTTGTATAAGGGTTTTTG

pLKO.1 puro shEYA3 #5 (TRC0000051605)

CCGGCCGGAAAGTGAGAGAAATCTACTCGAGTAGATTTCTCTCACTTTCCGGTTTTTG

pLKO.1 puro nontarget shRNA control (scramble)

CCGGGCGCGATAGCGCTAATAATTTCTCGAGAAATTATTAGCGCTATCGCGCTTTTT

Puromycin selection was used in all instances. In addition, several CRISPR/Cas9-generated Eya3-ko colonies were selected and amplified over about 30 days.

For experiments using conditioned medium, 1 million cells were seeded in triplicate and cultured in DMEM with 10% FBS for 24 hours. Cells were washed with PBS and then cultured in 1 mL of EBM-2 (endothelial cell basal medium, Lonza) with 5% FBS for 24 hours. Medium was collected and centrifuged for the collection of cell-free supernatant (conditioned medium, CM).

For primary Ewing sarcoma tumor cell tissue from a recently diagnosed 11-year-old patient with aggressive, untreated Ewing sarcoma with an EWSR1-FLI1 (type 2) fusion was used in an initial xenograft in NSG mice (IRB 2008-0021). Xenografts were established and serially transplanted by intramuscular injection of minced tumor fragments in a 50% Matrigel (Corning) suspension. Tumors obtained after the third intramuscular implantation of tumor tissue (F3) were used for experiments. Tumor tissue was dissociated using 1 mg/mL collagenase (17018-029, Gibco) and patient-derived xenograft (PDX) cells were cultured in RPMI-1640 medium (Gibco Life Technology) with 5% FBS, 0.005 mg/mL insulin (Sigma-Aldrich), 0.01 mg/mL transferrin (Sigma-Aldrich), 30 nmol/L sodium selenite (Sigma-Aldrich), 10 nmol/L hydrocortisone (Sigma-Aldrich), 10 nmol/L β-estradiol (Sigma-Aldrich), and 10 mmol/L HEPES.

Cell survival in two-dimensional culture was measured using the CCK-8 kit as previously described (16, 17) and following the manufacturer's protocol.

Animal experiments were performed in accordance with the recommendation of the Institutional Animal Care and Use Committee at Cincinnati Children's Hospital Medical Center IACUC2016-0019. The use of deidentified patient tumor tissue was approved by the Institutional Review Board at Cincinnati Children's Hospital Medical Center IRB2008-0021.

A total of 5 × 10⁶ A673 wild-type or Eya3-ko cells mixed with 30% Matrigel in 100 μL PBS were injected subcutaneously into the flank of NSG mice (6-week-old, female), and tumor size was monitored by caliper measurements [volume (mm³) = (length × width²)/2]. At the indicated times, mice were randomly divided into vehicle and treatment groups and were administered either vehicle (5% DMSO in phosphate-buffered saline PBS) or BZ in 5% DMSO/PBS (25 μg/g body weight) via intraperitoneal injection once every 2 days.

For PDXs, 1×10⁶ PDX cells were mixed with 30% Matrigel in 100 μL PBS and injected intramuscularly into the right thigh of NSG mice (6-week-old, male). Mice body weight was monitored every two days. After one month, mice were randomly divided into vehicle and treatment groups and treated as for the A673 xenografts.

All drug studies were blinded. Animal welfare and weight was monitored every 2 days. At the end of the experiment, tumors were harvested and weighed.

Xenograft tumor tissue specimens were fixed in 4% PFA and embedded in paraffin. Five-micrometer sections were deparaffinized in xylene and treated with a graded series of alcohol and rehydrated in PBS. Antigen retrieval was done using 10 mmol/L sodium citrate buffer. Samples were incubated with 3% H₂O₂ for 10 minutes followed by blocking in 2% bovine serum albumin in PBS for 1 hour. The slides were incubated with primary antibody overnight at 4°C. For IHC, sections were incubated with secondary antibody followed by amplification using the ABC kit (Vector Laboratories), then developed with diaminobenzidine (Vector Laboratories) reagent and counterstained with fast red nuclear stain. For immunofluorescence, sections were incubated with secondary antibody conjugated with Alexa Fluor (Invitrogen) for 1 hour at room temperature, and DAPI was used as counterstain.

A673 cells (500) were seeded with 2mL DMEM containing 10% FBS in each well of a six-well plate and cultured for 24 hours. The existing medium was replaced with fresh DMEM containing 5% FBS and 0.1% DMSO or the indicated concentrations of BZ once every three days and cultured for 14 days. Colonies were fixed with 100% methanol for 15 minutes at room temperature and stained with 0.05% of crystal violet for 30 minutes. Colonies were counted to investigate the effect of BZ.

Cells (2,000) were seeded per transwell (cat. #P18P01250, Millipore) in 100 μL DMEM containing 5% FBS and different concentrations of BZ in 0.1% DMSO, or CM as indicated. Both chambers of the transwell contained the same medium. Transwells were incubated for 18 hours. Cells from upper well were cleaned with a cotton swab, migrated cells were fixed in 100% methanol at room temperature for 30 minutes, and then stained with giemsa for 30 minutes. The number of migrated cells was counted after three washes. All experiments were performed three times.

Cells (10,000) were suspended in 1% low melting temperature agarose diluted in Tris-borate EDTA (TBE; 90 mmol/L Tris, 90 mmol/L boric acid, 2 mmol/L EDTA, pH 8.5) and spread on coverslips previously coated in 1% low melting temperature agarose and allowed to solidify for 10 minutes at 4°C. Cells were lysed overnight at 4°C in 2.5 mol/L NaCl, 100 mmol/L EDTA, 10 mmol/L Tris pH 10, 1% wv^–1 sarkosyl, 1% Triton X-100. Lysed cells were neutralized with TBE. Lysates were incubated in alkaline buffer incubation (300 mmol/L NaOH, 1 mmol/L EDTA, pH 12.3) and then electropheresed at 25 V for 40 minutes at room temperature. Slides were dehydrated in 70% ethanol and stained with propidium iodide (1 mg/mL). Slides were imaged on a Zeiss confocal microscope. Data were analyzed using OpenCOMET (18).

Multicellular tumor spheroids (MCTS) were formed using the liquid overlay method (19), collected, and fixed in 4% paraformaldehyde/PBS. For cryopreservation, fixed MCTS were dehydrated with sucrose and snap-frozen with optimal cutting temperature media (Tissue Tek). Cryo-sections (5 μm) were blocked in PBS pH 7.4, 0.15% Triton X 100 vv^–1, 10% FBS vv^–1, and 2% BSA wv^–1 in a humidified chamber, incubated with the indicated antibodies. DNA was labeled with Hoechst dye. Sections were mounted using Fluorogel with DABCO (17985-04; Electron Microscopy Science). Fluorescently stained MCTS sections were imaged on a Zeiss confocal microscope. Nuclei were counted as Hoechst-positive cells by using watershed separation and quantification with particle analysis in ImageJ software (ImageJ, RRID:SCR_003070) followed by manual counting of the stained nuclei.

PDX cells (4×10⁴) were seed in 96-well plates precoated with 1% agarose in 100 μL DMEM with 10% FBS. Spheroids were monitored and measured. After treated with BZ for 96 hours, spheroids were fixed in 4% PFA for 30 minutes and cryoprotected in 15%, 30% sucrose and embedded in OCT and 5-μm cryostat sections were cut.

Sprout formation and cell migration from endothelial spheroids were assessed as previously described (20) with minor modifications. Briefly, 8×10⁴ HUVEC cells (between passages 5 and 7; cat. #D2650, Sigma; obtained January 20, 2016) were resuspended in 4.8 mL EGM2. Methyl cellulose (200 μL) was added to the suspension. The entire mixture was pipetted in 25 μL hanging droplets on the lid of a nontissue culture treated petri-dish and incubated for 24 hours for spheroid formation. The spheroids were collected in PBS and washed twice with PBS and resuspended in 500 μL FBS, 1.5 mL of EBM-2 and 500 μL 2.5% methyl cellulose. 10× M199 was diluted in freshly prepared rat tail collagen (prepared as described in ref. 20) and titrated on ice to pH 7.4 using 0.1N NaOH to a final 1× M199 concentration. 2.5 mL of this mixture was added to the spheroid suspension, mixed, and immediately pipetted into a 48-well plate, incubated at 37°C/5% CO₂ for two hours and 100 μL of the indicated CM was overlaid on top of the solidified collagen. The plates were incubated for a further 24 hours and imaged on an EVOS FL microscope with a 10× objective. The length of individual sprouts was measured from the spheroid margin to sprout apex using ImageJ and the number of sprouts were manually counted after image acquisition.

The mRNA levels of Eya1, Eya2, Eya3, and Eya4 in Ewing sarcoma tumor cells were analyzed by RT-PCR. Total RNA was purified via PureLink RNA Mini Kit (Life Technologies). cDNA was synthesized with the PrimeScript reagent Kit. The sequences of primers used are:

EYA1

Sense: ATGGAAATGCAGGATCT

Antisense: GGTAGCTGTATGGTG

EYA2

Sense: GGACAATGAGATTGAGCGTGT

Antisense: ATGTCCCCGTGAGAGTAAGGAGT

EYA3

Sense: ATGGAAGAAGAGCAAGA

Antisense: GTTTGGGTTGCCTGAGG

EYA4

Sense: GGAGGTGCTTTCCCCATTGA

Antisense: CAGAAGGGCATGTTGTGCTTT

GAPDH

Sense: TTCATTGACCTCAACTAC

Antisense: CATGGACTGTGGTCATGAG

Results are presented as the mean ± SD. Statistical analyses were performed using GraphPad PRISM version 9.0 for Mac OSX, www.graphpad.com (GraphPad Prism, RRID:SCR_002798). A t test was used when two samples/conditions were compared and ANOVA for more than two groups.

EYA3 is present at elevated levels in Ewing sarcoma cells (14). RT-PCR performed with the A673 cell line shows Eya3 as the predominant Eya transcript, with a significantly fainter band for Eya4 (Supplementary Fig. S1A). We used CRISPR/Cas9 editing to generate Eya3-ko A673 cells (Eya3-ko; Supplementary Fig. S1B). Subcutaneous implantations of wild-type A673 and Eya3-ko cells were conducted on the two flanks of immune-deficient NSG mice to compare tumor growth in the same host. When tumors reached an average volume of 400 mm³, they were administered either vehicle (once every 2 days intraperitoneally) or the previously characterized EYA3 PTP inhibitor BZ (refs. 16, 17, 21–23; Supplementary Fig. S1C; BZ 25 mg/kg once every 2 days intraperitoneally). Tumor growth was monitored by caliper measurement, and animal welfare was monitored through observation and body weight measurements. Experiments were terminated when the vehicle-treated tumors approached 10% of body weight. Both Eya3-ko and BZ-treated tumors grew at a slower rate than control tumors and were 62% and 43% smaller, respectively, at the end of the experiments (Fig. 1A and B). Notably, BZ treatment has no effect on the growth of Eya3-ko tumors (Supplementary Fig. S1D). Histologic analysis of tumor sections showed increased areas of necrosis in the Eya3-ko and BZ-treated tumors relative to vehicle-treated controls (Fig. 1C). The proliferation index, estimated by quantifying the percentage of Ki67-positive cells, showed that loss or inhibition of EYA3 negatively affects tumor cell proliferation (Fig. 1D). A modest increase in apoptosis as measured by cleaved caspase-3 (CC3) staining was seen in Eya3-ko tumors, while a much more robust increase in CC3 staining was present in BZ-treated tumors (Fig. 1E). Because BZ treatment affects both tumor and stromal cells, this difference could be a consequence of EYA3 inhibition in both tumor and nontumor cells. Together, these observations are consistent with a role for the EYA3 PTP activity in the survival and proliferation of Ewing sarcoma tumor cells.

To better understand how EYA3 and its tyrosine phosphatase activity affect A673 cell proliferation and survival in vitro, we used the tetrazolium salt WST-8 and colony formation assays. Upon either loss of EYA3 or when EYA3 PTP activity was pharmacologically inhibited, there was a substantial reduction in colony formation (Fig. 2A and B). Interestingly, loss of EYA3 does not have a significant and consistent negative effect on cell proliferation in 2D culture, while BZ treatment impairs proliferation in a dose-dependent manner as measured by the WST-8 reagent (IC₅₀ of 7 μmol/L; Fig. 2C). In line with previous reports showing that EYA3 PTP activity promotes cell migration (21, 24), transwell assays showed that both Eya3-ko cells and BZ-treated A673 cells displayed reduced cell migration (Fig. 2D). Similar results to those reported for these CRISPR/Cas9-generated Eya3-ko A673 cells were obtained when Eya3 expression was silenced using two different shEya3 (Supplementary Fig. S1B, S1E, and S1F). Furthermore, an additional A673 Eya3-ko clone (Eya3-ko#2; Supplementary Fig. S2A, S2B, S2D, and S2F) as well as the effect of Eya3 silencing in another Ewing sarcoma cell line RD-ES (Supplementary Fig. S2A, S2C, S2E, and S2G) was analyzed with similar results in terms of cell proliferation, migration, and colony formation.

Tumor cell proliferation and survival in a 3D context was examined using MCTS and previously optimized protocols (19, 25). BZ treatment resulted in a dose-dependent inhibition of MCTS growth (Fig. 3A) commensurate with its ability to inhibit cell proliferation. MCTS treated with 5 μmol/L BZ for 96 hours were visually damaged, while 2.5 μmol/L BZ-treated MCTS were physically intact but displayed a larger dark center typical of a developing hypoxic, necrotic core (Fig. 3B; ref. 26). Interestingly, Eya3-ko did not impede MCTS growth, rather Eya3-ko MCTS consistently expanded in diameter and became more necrotic and fragile (Fig. 3C and D). MCTS were sectioned and stained with either H&E, Ki-67, or cleaved caspase-3 (CC3) to better understand how cell proliferation and survival were affected by loss of EYA3 or its PTP activity. BZ treatment resulted in a dose-dependent reduction in the number of Ki67-positive cells (Fig. 3E) and a large increase in apoptotic cells (Fig. 3F). EYA3-ko modestly reduced proliferation (Fig. 3E) in this 3D context and increased apoptosis (Fig. 3F). Notably, BZ had negligible effect on the growth of Eya3-ko MCTS (Supplementary Fig. S3A). The consequences of EYA3 loss on MCTS growth, proliferation, and apoptosis in a 3D context were corroborated using an independent Eya3-ko clone (Eya3ko#2) in A673 cells and using RD-ES cells (Supplementary Fig. S3B–S3G).

To extend observations made using the Ewing sarcoma cell lines into a more clinically relevant model, we used a PDX. Tissue from a recently diagnosed 11-year-old patient with aggressive, untreated Ewing sarcoma was used in an initial xenograft in immune-compromised NSG mice. The patient had a large pelvic tumor with metastases to bone (but not lung). On H&E stains, the primary intraosseous tumor (Fig. 4A) and its soft-tissue extensions (Fig. 4B) displayed the typical small round cell morphology of Ewing sarcomas. Tumor cells were relatively uniform with round nuclei, delicate chromatin, inconspicuous nucleoli, and scant cytoplasm. IHC staining for CD99 showed a classic sharp, crisp membranous staining pattern (Fig. 4C). FISH demonstrated rearrangement of the EWSR1 gene supportive of the diagnosis of Ewing's sarcoma. P1 (first-generation) tumors were excised 3 months after the initial implantation, characterized by H&E staining and reimplanted into NSG mice to expand the tumor tissue. A third intramuscular implantation of tumor tissue (P3) was used for experiments (Fig. 4D). Treatment with either vehicle or BZ was initiated 30 days after xeno-transplantation. Intraperitoneal vehicle or 25 mg/kg BZ was administered on alternate days for a total of 18 doses. Animal well-being and weight was monitored, but because of the site of implantation tumor size could not be readily monitored during the experiment. No significant difference in weight gain over the treatment period was observed between vehicle-treated and BZ-treated animals. The experiment was terminated when vehicle-treated mice began showing signs of distress. Tumors were excised and weighed; tumors from BZ-treated animals were on average 50% smaller than from vehicle-treated controls (Fig. 4E). Tumor sections were stained for the proliferation marker Ki67 and showed that BZ treatment resulted in a large reduction in proliferation index (Fig. 4F and G). Staining for the apoptosis marker cleaved caspase-3 showed increased apoptosis in BZ-treated tissue (Fig. 4H), as was observed in A673 xenografts.

In vitro, using both WST-8 and trypan blue assays in 2D culture, the IC₅₀ with BZ for these PDX cells was 6.7 μmol/L. We analyzed the effect of BZ treatment on PDX cells in 3D culture. MCTS were generated and treated with either vehicle or BZ when they reached an average diameter of 400 μm. Growth rate was then followed until control MCTS stopped growing further. Treatment of PDX cell MCTS with BZ resulted in dose-dependent growth inhibition (Fig. 4I). MCTS were then fixed, sectioned, and stained with markers of proliferation and apoptosis; BZ treatment led to reduced Ki67 staining and increased apoptosis (Fig. 4J, K, and L). Together, these results corroborate the relevance of the EYA3 PTP activity in progression of a patient-derived Ewing sarcoma tumor.

The EYA3 PTP activity has previously been shown to promote DNA damage repair (27, 28). In various pathologies, loss of EYA3-PTP activity results in impaired DNA damage repair as evidenced by smaller percentages of cells that stain positively for the DNA damage and repair marker γH2AX (21–23). In order to evaluate the effect of EYA3 PTP inhibition on DNA damage repair in Ewing sarcoma, we stained both A673 and PDX tumor tissue for γH2AX. Positivity for γH2AX (Ser139 phosphorylated form of H2AX) has been ascribed with prognostic value in a variety of cancers including breast, endometrial, lung, cervical, and ovarian cancers (29, 30). DNA damage and γH2AX positivity are commonly induced by genotoxic chemotherapy, but is also present in untreated tumors as a result of replicative and oxidative stress. We examined γH2AX staining to evaluate the levels of constitutive DNA damage and ongoing repair. A673 tumors show high levels and broadly distributed γH2AX positivity (Fig. 5A). In contrast, PDX tumors have less γH2AX staining overall, and the staining was often present in distinct patches (Fig. 5E and F). In both A673-Eya3-ko tumors (Fig. 5C and D) and in BZ-treated A673 tumors (Fig. 5B and D), γH2AX staining was substantially reduced. BZ treatment did not reduce the percentage of γH2AX-positive cells in Eya3-ko tumors (Supplementary Fig. S4B) attesting to target specificity of BZ.

Similarly, BZ treatment reduced γH2AX staining in PDX tumors (Fig. 5G and H). Along with the evidence of increased apoptosis presented in Figs. 1–4, these observations can be interpreted as a failure to induce an adequate DNA damage repair response, thus tipping the balance toward cell death.

Both A673 and PDX cell MCTS were evaluated for the effect of EYA3 loss or BZ treatment on γH2AX levels. As in the case of tumor tissue, either loss of EYA3 or BZ treatment reduces γH2AX levels in A673 MCTS (Fig. 5I and J), and BZ treatment reduced γH2AX levels in PDX cell MCTS (Fig. 5K and L). An alkaline COMET assay was used to evaluate DNA damage and showed that both loss of EYA3 and BZ treatment resulted in increased DNA damage (Fig. 5M; Supplementary Fig. S4A).

H2AX-pY₁₄₂ is an established substrate of the EYA3 PTP activity (27). To assess target engagement, we stained tumor tissue with anti-H2AX-pY₁₄₂. Both A673 and PDX tumor sections showed low levels of constitutive H2AX-Tyr142 phosphorylation (Fig. 5N and R). Interestingly, this was not uniform but rather concentrated in areas with smaller more dense nuclei suggestive of cell death. Either loss of EYA3 (Fig. 5O and Q) or inhibition of EYA3-PTP activity with BZ (Fig. 5P, Q, R, S, and T) led to a substantial increase in positivity for H2AX-pY₁₄₂. This observation provides evidence that EYA3 dephosphorylates H2AX-pY₁₄₂ in Ewing sarcoma tumor cells and that BZ directly impairs this dephosphorylation reaction in vivo.

We evaluated vascular hotspots within the tumors (maximal vascular density MVD), a widely used method for evaluating vascular density. In both A673(Eya3-ko) xenografts and A673 xenografts treated with BZ, we see a significant reduction in MVD (Fig. 6A and B), while BZ did not affect the vascularity of Eya3-ko tumors (Supplementary Fig. S4C). BZ treatment also reduced vascularity in PDX tumors (Fig. 6E and F). Inhibition of tumor angiogenesis upon BZ treatment is consistent with our previous studies showing that endothelial EYA3-PTP activity promotes angiogenesis (17) and that loss of endothelial Eya3 inhibits tumor angiogenesis in Lewis lung carcinoma (LLC) xenografts (21). However, the antiangiogenic effect upon loss of Ewing sarcoma tumor cell EYA3 differs from our results with LLC tumors and suggests the presence of an EYA3-dependent proangiogenic property in Ewing sarcoma cells.

Previous studies of the angiogenic growth factor expression profile in Ewing sarcoma cells revealed that VEGFA expression is most significantly associated with the development of microvessels (31). Both Eya3-ko tumors and BZ-treated A673 tumors had lower levels of VEGF as detected by IHC (Fig. 6C and D). Analysis of PDX tumor tissue corroborated both the antiangiogenic effect of BZ as measured by endomucin staining (Fig. 6E and F) and the reduction in VEGF levels in tumor tissue (Fig. 6G).

We then investigated whether EYA3 in A673 tumor cells contributes to the release of proangiogenic factors by testing the effect of tumor cell CM on endothelial cell proliferation and migration. Significantly reduced endothelial cell migration and proliferation were recorded in experiments conducted with CM from A673 cells lacking EYA3 relative to CM from control A673 cells (Fig. 6H and I). Similar results were obtained upon using CM from RD-ES cells (Supplementary Fig. S4D and S4E). To further explore the effect of tumor cell EYA3 on angiogenesis we used an endothelial spheroid sprouting assay. HUVEC spheroids were embedded in collagen and maintained in either A673-conditioned basal medium or A673 Eya3-ko CM. Sprouting angiogenesis was observed with A673 CM, but negligible sprouting was observed with Eya3-ko CM (Fig. 6J–L). Similar results were obtained when EYA3 levels in A673 were reduced using shEya3 (Fig. 6K and L). Together, these observations support a role for EYA3 in Ewing sarcoma cells as a positive regulator of a secreted angiogenic factor, and our in vivo data suggest that VEGFA is regulated by EYA3 in Ewing sarcoma cells. The molecular mechanisms through which EYA3 regulates VEGF levels in Ewing sarcoma cells remain to be determined. Because there is evidence, in other contexts, that EYA1 could regulate VEGFA through the HIF1α–PI3k–Akt pathway (32), we evaluated the effect of either EYA3 loss or BZ treatment on activated Akt (pS474). In our studies, we do not see a significant impact on either HIF1α or Akt activation (Fig. 6M). The mechanism of VEGFA regulation by EYA3 is the subject of ongoing investigation.

In this study, we show that EYA3 in Ewing sarcoma tumor cells promotes both tumor growth and tumor angiogenesis. Further, pharmacologic targeting of the EYA3 PTP activity inhibits Ewing sarcoma tumor growth and vascularization. Hence, EYA3 PTP activity can contribute to Ewing sarcoma progression through multiple cellular mechanisms. Although the retardation of tumor growth by EYA3-PTP inhibition or EYA3 loss is consistent with our previous observations using a murine lung carcinoma model (LLC in C57/Bl6 mice; ref. 21), this is the first report of a role for tumor cell EYA3 in Ewing sarcoma tumor angiogenesis.

The EYA proteins are known to contribute to cell-cycle progression and DNA damage repair (27, 28), both of which affect tumor growth. DNA damage repair pathways are commonly upregulated in cancers, but in Ewing sarcoma cell lines and primary tumors, several DNA damage repair genes, including BRCA1, GEN1, and ATM, are downregulated (33). In contrast, EYA3 levels are elevated via EWS/FLI1-mediated downregulation of miR-708 (Fig. 6N; ref. 14). In growing tumors, repair of replicative and oxidative stress–induced DNA damage permits unregulated tumor cell proliferation and genomic instability. The EYA-PTP activity promotes DNA damage repair through dephosphorylation of the C-terminal Y142 residue of H2AX (Fig. 6N; ref. 27). H2AX is a mediator of the DNA damage response acting as a scaffold for either repair-mediating or apoptosis-mediating factors. The phosphorylation state of H2AX-Tyr142 acts as a switch in this repair versus apoptosis decision, with the EYA-PTP tipping the balance toward repair (Fig. 6N; ref. 27). Upon loss of EYA3 or its PTP activity, we consistently see less staining for γH2AX (a marker for DNA damage and repair), and increased apoptosis. Interestingly, loss of EYA3 does not significantly affect tumor cell proliferation in 2D cell culture or the growth of MCTS derived from either A673, RD-ES, or patient-derived primary tumor cells. But there is a significant increase in the size of the necrotic zone and the percentage of apoptotic cells in MCTS lacking EYA3. Together, these observations are consistent with a role for EYA3 in the survival of tumor cells under hypoxic stress. Loss or inhibition of EYA3-PTP activity also inhibited cell motility consistent with previous studies showing a role for the EYA3 PTP activity in cell migration (24). We were unable to evaluate the effect of EYA3 inhibition on metastasis here since neither the A673 nor the PDXs yielded metastatic foci.

BZ has been extensively characterized in its role as an EYA-PTP inhibitor (13, 16, 17, 21–23). But BZ and the related compound benzbromarone have a long and checkered history as therapeutics (34, 35). Both compounds are uricosurics with exceptional efficacy in the treatment of gout. They were withdrawn by the manufacturer from European markets after over 20 years of use due to some reports of hepatotoxicity. However, because of their effectiveness as anti-gout agents, there remains a debate about their clinical utility (34, 35). Relevant to this report, with the dosage regimen used in this study we saw no detectable toxicity (22), and there is a case to be made that acceptable toxicity for an anticancer agent is different from that for the chronic treatment of gout. The issues regarding the safety profile of BZ notwithstanding, the results presented here provide strong validation of the EYA3-PTP activity as a target for the development of Ewing sarcoma therapeutics, supported by the use of BZ as a pharmacologic tool. In addition to the efficacy readout obtained through monitoring tumor growth and angiogenesis, we also examined a pharmacodynamic readout through target-engagement assays. We measured changes in the level of the substrate H2AX-pY142 to evaluate the molecular events triggered by BZ treatment, and the assembly of DNA damage repair complexes (γH2AX staining) as a functional consequence (Fig. 5). Strong upregulation of the EYA3-PTP substrate H2AX-pY142 was seen upon BZ treatment. Importantly, these results phenocopied observations made using a genetic strategy (loss of EYA3 in tumor cells, Eya3-ko). Although off-target effects cannot be ruled out for BZ, or indeed any chemical compound, our results are strongly supportive of EYA3 as a biologically relevant target of BZ in vitro and in vivo.

We anticipated a reduction in tumor vascularity upon treatment with BZ, based upon our previous studies showing that loss of EYA3-PTP activity in host endothelial cells can inhibit tumor angiogenesis (21). Unexpectedly, loss of EYA3 in A673 cells also reduced tumor vascularity and reduced VEGF levels in tumor tissue (Fig. 6). Previous studies have shown that VEGFA is the single most important regulator of angiogenesis in Ewing sarcoma (31). Positive regulation of VEGFA (and HIF1α) by EYA1 has previously been reported to occur via activation of the PI3K–Akt pathway in colorectal tumor cells (32). A more recent report demonstrates EYA1-dependent regulation of VEGFA levels in adipose-derived stem cells (36). This is the first evidence that EYA3 also regulates VEGFA levels. However, our studies do not support a role for the PI3k–Akt pathway in EYA3-mediated VEGFA regulation in the present context. The precise mechanisms through which EYA3 regulates the VEGFA pathway in Ewing sarcoma tumor cells, and the possibility that other angiogenic cytokines or growth factors are regulated by EYA3, remain the subject of ongoing investigations. Furthermore, we have previously shown that DNA damage repair promoted by EYA3 can play a role in pathologic angiogenesis (23). Hence loss of EYA3 PTP activity could modulate tumor vascularization through multiple mechanisms.

In Ewing sarcoma, microvessel density is a prognostic marker that is significantly correlated with the expression of VEGF-A (31, 37–39). Accordingly, anti-VEGF agents inhibit the growth of Ewing sarcoma family tumors (39). Although only modest success has been reported for VEGF-targeting antiangiogenics as monotherapy in the treatment of Ewing sarcoma, there is evidence that combination of conventional chemotherapy with antiangiogenic agents may be more effective (7). EYA3 could represent a unique target in this respect. Previous studies demonstrated chemosensitization in vitro upon loss of EYA3 in Ewing sarcoma cell lines (14), consistent with the ability of EYA3 to promote DNA damage repair. Exploiting the ability of EYA3-PTP inhibitors to act as antiangiogenics as well as chemosensitizers would allow the use of lower doses of cytotoxic chemotherapy in such combination regimens. Evaluation of the efficacy and safety of such treatment strategies is ongoing. The present report provides proof-of-principle that EYA3 PTP in Ewing sarcoma tumor cells contributes to tumor growth and angiogenesis (Fig. 6N) and is pharmacologically targetable in a patient-derived tumor model.

J.G. Pressey reports personal fees from the Department of Defense outside the submitted work. R.S. Hegde reports grants from NIH and grants from B+ Foundation during the conduct of the study; in addition, R.S. Hegde has a patent for US-9962362-B2 issued. No disclosures were reported by the other authors.

Y. Wang: Formal analysis, validation, investigation, methodology, writing–review and editing. R.N. Pandey: Formal analysis, validation, investigation, methodology, writing–review and editing. K. Roychoudhury: Investigation, writing–review and editing. D. Milewski: Investigation, writing–review and editing. T.V. Kalin: Resources. S. Szabo: Resources, visualization, writing–review and editing. J.G. Pressey: Resources, writing–review and editing. R.S. Hegde: Conceptualization, resources, formal analysis, supervision, funding acquisition, visualization, writing–original draft, project administration, writing–review and editing.

This work was supported through grants from the Andrew McDonough B+ Foundation and NIH NCI RO1-CA207068 to R.S. Hegde.

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.