A benzarone derivative inhibits EYA to suppress tumor growth in SHH medulloblastoma

Abstract

Medulloblastoma is one of the most common malignant brain tumors of children, and 30% of medulloblastomas are driven by gain-of-function genetic lesions in the Sonic hedgehog (SHH) signaling pathway. EYA1, a haloacid dehalogenase (HAD) phosphatase and transcription factor, is critical for tumorigenesis and proliferation of SHH medulloblastoma (SHH-MB). Benzarone and benzbromarone have been identified as allosteric inhibitors of EYA proteins. Using benzarone as a point of departure, we developed a panel of 35 derivatives and tested them in SHH-MB. Among these compounds, DS-1-38 functioned as an EYA antagonist and opposed SHH-signaling. DS-1-38 inhibited SHH-MB growth in vitro and in vivo, showed excellent brain penetrance, and increased the lifespan of genetically engineered mice predisposed to fatal SHH-MB. These data suggest that EYA inhibitors represent promising therapies for pediatric SHH-MB.

Introduction

Medulloblastoma is one of the most common pediatric malignant tumors, accounting for about 20% of central nervous system (CNS) tumors in children. The Sonic Hedgehog (SHH) subtype accounts for about 30% of all medulloblastomas (1,2). Standard of care therapeutic modalities (surgery, chemotherapy, radiation) are effective in 60–70% of patients, but have long-term sequelae on quality of life, and therefore less toxic and more targeted therapeutics are clearly needed. SHH subtype medulloblastomas (SHH-MB) are defined by constitutive SHH signaling due to mutations in PTCH1, SUFU, SMO, or overexpression of GLI2 (3–6). Small-molecule antagonists of SMO, including Sonidegib (LDE225; Novartis) and Vismodegib (GDC-0449; Roche) have shown some clinical efficacy in treating medulloblastoma and basal cell carcinoma (7–10). However, while initial success with these inhibitors is achieved, long term efficacy is limited due to resistance (11,12). Thus, it is critical to discover new targets and inhibitors that will effectively constrain the growth promoting effects of SHH pathway dysregulation.

We previously showed that the haloacid dehalogenase (HAD) phosphatase Eyes Absent 1 (EYA1) is essential to development and progression of SHH-medulloblastoma and could be a promising therapeutic target (13). EYA1 is highly expressed in SHH-medulloblastomas, and single cell sequencing indicates that EYA1 is expressed in virtually every individual cancer cell (13). Inhibition of EYA1 interrupts SHH pathway signaling, and reduced levels of EYA1 decrease medulloblastoma mortality rates in mouse models (13). EYA1 dephosphorylates the terminal tyrosine of histone variant H2AX, thereby affecting DNA repair, apoptosis, and ultimately cell survival, and so may have effects on tumor growth beyond its role in SHH signaling (14). In addition, EYA1, like other members of the EYA family (1–4) binds to SIX family transcription factors to regulate gene expression during embryogenesis and development (15,16). Both EYA and SIX family members are overexpressed and have been implicated in many different cancers including glioblastoma, leukemia, ovarian, breast and Wilms’ tumor. Moreover, overexpression of EYA and/or SIX have been shown to correlate with progression and metastasis of these cancers (17–23). Thus, the clinical impact of EYA antagonists might transcend cancers driven by SHH signaling.

While phosphatases are traditionally considered poor targets due to the indiscriminate nature of the catalytic domain, EYA proteins are a unique family of HAD phosphatases. The EYA proteins have a highly conserved C-terminal domain of 270 amino acids that contains the HAD motifs, commonly known as EYA domain, which interacts with the SIX domains (Fig. 1A). EYAs are the only enzymes with validated tyrosine phosphatase activity in the HAD domain (15,24,25)(Fig. 1A). Furthermore, among the 107 classical protein tyrosine phosphatases (PTPs) in humans, EYAs are the only ones that have an active aspartate residue instead of cysteine residue for dephosphorylation—a distinctive, and potentially druggable, topological feature (24,26).

Figure. 1. Benzarone derivatives are potential EYA inhibitors with on target effects in MB21 cells.

A) Functional domains of human EYA1: transactivation domain (TAD, blue), phosphothreonine phosphatase domain (pT-P)/ EYA domain (ED2, green), the conserved EYA domain (ED)/ phosphotyrosine phosphatase (pY-P, yellow) and embedded haloacid dehalogenase (HAD) motifs (gray). B) Chemical structure of benzarone and benzarone derivatives and corresponding modified moieties. C) SL2 cells were stimulated with SAG-conditioned media or vehicle control (DMSO) as well as 10 µM of the indicated compound. GLI-responsive Firefly luciferase activity normalized to constitutive Renilla luciferase activity was then measured. Fold change of GLI-luciferase activity (SAG/vehicle) are shown. All compounds inhibited SAG-induced GLI-luciferase activity significantly (p≤0.001, one-way ANOVA, Bonferroni’s multiple comparisons test); mean of 3 wells per n=3, error bars= SEM. D) MB21 cells were plated and after 24 hrs treated with 10 µM of compounds. The cells were spiked with 10 µM of compounds again at Day 4 and viable cells were counted at Day 7; mean, error bars = SEM. Statistical analysis: One-way ANOVA, Bonferroni’s multiple comparisons test, ** P ≤ 0.01, *** P ≤ 0.001, *** P ≤ 0.0001. F-H) DS38 dose-response curves generated through the Broad Institute PRISM assay at 8-point dose (3-fold dilution) with a 5-day treatment in Daoy (F), TOV112 (G), and LS513 (H) cell lines.

A successful strategy to develop selective kinase inhibitors is to target an allosteric site (27,28). Benzarone and its derivative benzbromarone, both uricosuric agents, have been previously identified as allosteric inhibitors for EYA2 and efficacy has been reported against the other EYA proteins as well (16,23,29). Using benzarone as a platform, here we develop a panel of 35 benzarone derivatives and assess their ability to impact the EYA1 phosphatase and inhibit SHH-medulloblastoma growth. Several of these new molecular compounds have enhanced inhibitory characteristics, particularly compound DS-01-38 (DS38). DS38 binds EYA1, inhibits EYA1 phosphotyrosine phosphatase activity in vitro and in a cell-based assay and interrupts SHH signaling. Compound DS38 inhibits SHH-MB cell growth in vitro in mouse and as well as human models. Furthermore, we determined that DS38 easily crosses the blood-brain barrier and increases by more than 40% the lifespan of mice genetically predisposed to developing fatal SHH-medulloblastoma. Taken together, our results show that this inhibitors of the EYA1 phosphatase provide a promising avenue for developing therapeutics for SHH-medulloblastoma.

Materials and Methods

Ethics statement

All mice were bred at The Jackson Laboratory (JAX) and Dana-Farber Cancer Institute (DFCI) in accordance with the National Institutes of Health and US Department of Agriculture criteria. Protocols for their care and use were approved by the Institutional Animal Care and Use Committees (IACUC) of JAX and DFCI.

Mice and genotyping

SmoM2 (Gt(ROSA)^{26Sortm1(Smo/EYFP)Amc}/J) mice (RRID:IMSR_JAX:005130) have been previously described (30) and were obtained from the Jackson Laboratory. The mice harbor the SmoM2/EYFP fusion gene, whose expression is blocked by a loxP-flanked STOP fragment placed between the Gt(ROSA)26Sor promoter and Smo/EYP sequence. Homozygous SmoM2 mice were bred to heterozygous mice harboring the Atoh-Cre transgene (B6.Cg-Tg(Atoh1-cre)1Bfri/J) (RRID:IMSR_JAX:011104) also obtained from the Jackson Laboratory, to derive SmoM2, Atoh-Cre^tg/0 mice. Genotyping was performed by Transnetyx using real-time PCR.

Synthesis of compounds

2, 1-(5-Bromobenzofuran-2-yl)ethan-1-one

A mixture of 5-bromosalicylaldehyde (3.03 g, 15.1 mmol) and potassium hydroxide (0.84 g, 15 mmol) was heated in MeOH at reflux for 5 mins, then cooled. Chloroacetone (1.49 mL, 18.1 mmol) was added, and the reaction mixture was stirred at 80 °C for 2.5 hours. A further portion of chloroacetone (0.74 mL, 9.0 mmol) was added and the reaction was heated for 1 hour. The methanol was removed, and the residue was recrystallized from hot EtOH to give 2.25 g (62 %) of a white solid. ¹H NMR (500 MHz, DMSO-d₆) δ ppm 8.06 (d, 1H) 7.85 (d, 1H) 7.73 (d, 1H) 7.68 (dd, 1H) 2.58 (s, 3H).

3, 5-Bromo-2-ethylbenzofuran

Hydrazine hydrate (1.59 mL) was added to a solution of 1-(5-bromobenzofuran-2-yl)ethan-1-one (2.24 g, 9.37 mmol) in ethylene glycol (25 mL) at 120 °C, and the solution was heated at 180 °C for 30 mins. The reaction was allowed to cool to 120 °C, and potassium hydroxide (1.58 g, 28.1 mmol) was added. After 4.5 hours, the reaction was allowed to cool, poured into ice water and extracted with DCM. The organic layer was dried (MgSO₄), filtered and concentrated and the residue was purified by silica chromatography (0–20% EtOAc in Hex) to give 1.09g (47%) of a colorless oil. ¹H NMR (500 MHz, DMSO-d₆) δ ppm 7.75 (d, 1H) 7.48 (d, 1H) 7.35 (dd, 1H) 6.58 (s, 1H) 2.79 (q, 2H) 1.26 (t, 3H).

4, (5-Bromo-2-ethylbenzofuran-3-yl)(4-methoxyphenyl)methanone

4-Methoxybenzoyl chloride (1.07 g, 6.29 mmol) and SnCl₄ (0.736 mL, 6.29 mmol) were added to a solution of 5-bromo-2-ethylbenzofuran (1.09 g, 4.84 mmol) in CS₂ (10 mL) at 0 °C. After 30 mins, the cooling bath was removed, and the reaction was allowed to stir overnight at RT. The reaction was quenched with water and extracted with EtOAc. The organic layer was washed with 1N HCl, water, NaHCO₃ and brine, dried over MgSO₄, filtered and concentrated. The residue was purified by silica chromatography (10–30% EtOAc in Hex) to give 1.45 g yellow oil (83%), which formed a solid upon standing. ¹H NMR (500 MHz, DMSO-d₆) δ ppm 7.80 (d, 2H) 7.67 (d, 1H) 7.55 (d, 1H) 7.51 (dd, 1H) 7.11 (d, 2H) 3.88 (s, 3H) 2.79 (q, 2H) 1.24 (t, 3H).

5, (5-Bromo-2-ethylbenzofuran-3-yl)(4-hydroxyphenyl)methanone

A solution of BBr₃ (1M in DCM, 3.34 mL) was added to a solution of (5-bromo-2-ethylbenzofuran-3-yl)(4-methoxyphenyl)methanone (300 mg, 0.835 mmol) in DCM (5 mL) at −78 °C. The reaction was stirred for 1 hour at −78 °C, and 30 mins at RT, then cooled again to −78 °C and a further portion of BBr₃ in DCM (1M, 3.34 mL) added. The reaction was stirred for 1 hour at −78 °C, and 4 hours at RT, then cooled in an ice bath and diluted with DCM (20 mL). Water (20 mL) was added, and the organic layer was separated and washed with water and brine, dried (MgSO₄), filtered and concentrated. The residue was triturated with ~4:1 Hex:EtOAc, and filtered to give 154 mg of an off-white solid. The filtrate was purified by silica chromatography (10–40% EtOAc in Hex) to give a further 57 mg white solid (68% combined yield). ¹H NMR (500 MHz, DMSO-d₆) δ ppm 10.50 (s, 1H) 7.70 (d, 2H) 7.65 (d, 1H) 7.53 (d, 1H) 7.50 (dd, 1H) 6.91 (d, 2H) 2.80 (q, 2H) 1.24 (t, 3H).

DS-1-038, (2-Ethyl-5-(1-methyl-1H-pyrazol-4-yl)benzofuran-3-yl)(4-hydroxyphenyl)methanone

A mixture of (5-bromo-2-ethylbenzofuran-3-yl)(4-hydroxyphenyl)methanone (53 mg, 0.154 mmol), 1-methyl-4-(4,4,5,5-tetramethyl-1,3,2-dioxaborolan-2-yl)-1H-pyrazole (38 mg, 0.184 mmol), Cs₂CO₃ (60 mg, 0.184 mmol) and tetrakis(triphenylphosphine)palladium (9 mg, 0.008 mmol) in dioxane (2 mL) was heated at 90 °C. After 2.5 hours, further portions of 1-methyl-4-(4,4,5,5-tetramethyl-1,3,2-dioxaborolan-2-yl)-1H-pyrazole (32 mg, 0.154 mmol) and tetrakis(triphenylphosphine)palladium (9 mg, 0.008 mmol) were added and the reaction was stirred at 90 °C overnight. The mixture was cooled, diluted with EtOAc (30 mL) and washed with water, brine and dried (MgSO₄). The residue was purified by silica chromatography (20–80% EtOAc in Hex) to give 11 mg (21%) of the title compound. ¹H NMR (500 MHz, DMSO-d₆) δ ppm 10.47 (s, 1H) 8.05 (s, 1H) 7.74 (m, 3H) 7.62 (d, 1H) 7.54 (m, 2H) 6.91 (d, 2H) 3.84 (s, 3H) 2.77 (q, 2H) 1.24 (t, 3H).

LO-2-208, (5-Bromo-2-ethylbenzofuran-3-yl)(4-methoxy-2-methylphenyl)methanone

Two drops of DMF and oxalyl chloride (2.5 mL) were added to a suspension of 2-methyl-4-methoxybenzoic acid (498 mg, 3 mmol) in dichloroethane (7.5 mL) cooled to 0 °C. After 10 minutes, the reaction was allowed to warm to rt. After 45 minutes, the mixture was concentrated under reduced pressure, and the residue dried in a vacuum oven to give a dark brown solid used in the next step without further purification, as a 1M solution in DCM.

SnCl₄ (70 uL, 0.6 mmol) was added slowly to a solution of 5-bromo-2-ethylbenzofuran (113 mg, 0.5 mmol) and 4-methoxy-2-methylbenzoyl chloride (0.6 mL, 1M in DCM, 0.6 mmol) in DCM (1.5 mL) at 0 °C. The reaction was allowed to warm to rt and stirred overnight. The reaction was poured into water and extracted twice with EtOAc. Combined organic layers were washed with brine, dried (Na₂SO₄), filtered and concentrated, and the residue was purified by silica chromatography (0–15% EtOAc in Hex) to give 187mg of a colorless oil (100%). ¹H NMR (500 MHz, DMSO-d₆) δ ppm 7.65 (d, 1H) 7.55 (d, 1H) 7.50 (dd, 1H) 7.41 (d, 1H) 6.98 (d, 1H) 6.87 (dd, 2H) 3.84 (s, 3H) 2.68 (q, 2H) 2.36 (s, 3H) 1.18 (t, 3H).

7, (2-Ethylbenzofuran-3-yl)(4-methoxy-2-methylphenyl)methanoneALV-1-080

SnCl₄ (200 uL, 1.71 mmol) was added slowly to a solution of 2-ethylbenzofuran (192 mg, 1.31 mmol) and 4-methoxy-2-methylbenzoyl chloride (315 mg, 0.171 mmol) in benzene (3 mL). The reaction was heated at 80 °C and stirred overnight, then quenched with water and extracted into DCM. The organic layer was washed with sat NaHCO₃, 1M HCl and brine, and dried (Na₂CO₃), filtered and the residue purified by silica chromatography (0–15% EtOAc in Hex) to give the title compound (157 mg, 40%). ¹H NMR (500 MHz, CDCl₃-d) δ ppm 7.38 (d, 1H) 7.33 (d, 1H) 7.27 (dd, 1H) 7.19 (m, 1H) 7.11 (m, 1H) 6.76 (d, 1H) 6.66 (dd, 1H) 3.80 (s, 3H) 2.77 (q, 2H) 2.38 (s, 3H) 1.22 (t, 3H).

ALV-1-087, (2-Ethylbenzofuran-3-yl)(4-hydroxy-2-methylphenyl)methanone ALV-1-087

BBr₃ (1.39 mL, 1M in DCM) was added to a solution of (2-ethylbenzofuran-3-yl)(4-methoxy-2-methylphenyl)methanone (136 mg, 0.46 mmol) at −78 °C, then allowed to warm to rt and stirred overnight. The reaction was quenched with water and extracted with DCM. The organic layer was washed with sat NaHCO₃, 1M HCl and brine, and dried (Na₂CO₃), filtered and the residue purified by silica chromatography (0–50% EtOAc in Hex) to give the title compound (101 mg, 78%). ¹H NMR (500 MHz, CDCl₃-d) δ ppm 7.38 (d, 1H) 7.27 (m, 2H) 7.20 (m, 1H) 7.11 (m, 1H) 6.71 (d, 1H) 6.62 (dd, 1H) 5.03 (s, 1H) 2.78 (q, 2H) 2.35 (s, 3H) 1.22 (t, 3H).

ALV-01-088, (2-Ethylbenzofuran-3-yl)(4-hydroxy-3-methylphenyl)methanone

ALV-1-088 was prepared in the same manner as ALV-1-087 from 2-ethylbenzofuran and 4-methoxy-3-methylbenzoyl chloride. ¹H NMR (500 MHz, CDCl₃-d) δ ppm 7.74 (d, 1H) 7.64 (dd, 1H) 7.50 (d, 1H) 7.45 (d, 1H) 7.30 (m, 1H) 7.22 (m, 1H) 6.86 (d, 1H) 5.91 (s, 1H) 2.92 (q, 2H) 2.31 (s, 3H) 1.35 (t, 3H).

Compounds were reconstituted in DMSO to 10 mM concentration, and then diluted into the appropriate vehicle. For all other compounds, please reach out to David A. Scott, (DavidA_Scott@DFCI.HARVARD.EDU).

Cell culture

SL2 cells were obtained from American Type Culture Collection and cultured according to their recommendations.

MB21, SMO (D477G) MB21, MB55, and MB56 cells were prepared from medulloblastoma tumors of Ptch^+/− mice as previously described (31). Non-adherent cells were cultured in DMEM/F12 media, 2% B27, 1% Pen/Strep. For short term in vitro experiments, the MED172FH PDX cells were incubated in the same media together with final concentration of 10 µM of the compounds or the equivalent volume of DMSO as a control. MB cells were tested for Mycoplasma testing using rapid NAT-based/RT-PCR assay in 2017. Between collection or thawing, less than 30 passages were performed for experiments.

Luciferase assays

For Shh activation, SL2 cells were plated in 96-well plates (5 × 10⁴ cells/well) with 3 wells per condition. After 24 hours, the cells were cultured for 24 h in DMEM 1% P/S with reduced serum (0.5%) in the presence of 500 nM SAG and 10 µM of compounds (or varying concentrations for dose-response assays) in the same media for 48 hours. Luciferase assays were done using a dual luciferase reagent (Promega #1960) according to manufacturer’s protocol. Duplicate plates were tested as well. Plates were read using a CLARIOstar Plus plate reader and MARS software (BMG Labtech).

Firefly/Renilla averages from duplicate wells were determined for each plate. To determine fold change of GLI1 luciferase activity, averages from duplicate wells from SAG conditions were divided by DMSO (vehicle) conditions. Statistics were done by unpaired t-test using GraphPad.

Cell growth and viability assays

MB21 or MED1712FH were plated on 96-well plates (5 × 10⁴ cells/well) in 50 µL of culturing media. After 24 hours, 10 mM of corresponding compounds diluted in 50 µL were added to the wells with the final concentration of 10 µM (or varying concentrations for dose-response assays). At day 4, this was repeated.

Cell viability and growth was tested by CellTiter-Glo^® 96 Aqueous One Solution (Promega G358) or CellTiter-Glo^® 2.0 (Promega G9243) at different time points according to manufacturer’s instructions and fluorescent or luminescent signals were measured with CLARIOstar Plus plate reader and MARS software (BMG Labtech). Cell number ratio was determined by the ratio of the treated sample relative to the control (DMSO treated).

PDX model and tumor dissociation

MED17FH PDX models, which were first generated by the Olson lab (32), were obtained from the Center for Patient-Derived Models at Dana-Farber Cancer Institute. PDX tissue was implanted into NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NGS)(RRID:IMSR_JAX:005557) mice directly into the mouse cerebellum using protocols previously published (32). When mice showed signs of morbidity, mice were humanely euthanized, and tumors were extracted for dissociation.

Patient-tumor cells were isolated from allocated tumor tissue using an adapted previously published protocol (33). Tissues were minced using a sterile feather blade. Red blood cells were then removed by incubating the minced tissue in 4:1 ratio of RBC lysis buffer (Invitrogen #00433357) and serum-free media (DMEM/F12 media, 2% B27, 1% Pen/Strep) for 10 minutes on ice. Tissue was then treated with with 1X Hanks buffered saline solution (HBSS) containing 200 units/mL DNase I (Worthington #LK003172) and 200 μg/mL Liberase Blendzyme-1 (Sigma Aldrich # 5401054001) for 90 minutes at 37 °C. Digested pieces were mechanically dissociated using a P-1000 pipette and then allowed to separate by gravity filtration. The supernatant, containing the live cells were then moved into a new tube. Cells were then spun down and resuspended in the serum-free media. Total number of cells were counted using Trypan Blue.

Western blotting

For western blot assays, protein was extracted using RIPA Lysis and Extraction Buffer (Thermo). Protein lysates were run on NuPAGE 4–12% gradient Bis-Tris gels (Thermo) and transferred to PVDF membranes (Millipore) using standard procedure. Optimal primary antibodies dilutions were evaluated empirically. Goat anti-rabbit or anti-mouse peroxidase-conjugated antibodies (Bio-Rad) were used as secondary antibodies. Protein bands were visualized using SuperSignal West Dura solution (Thermo Fisher). Image acquisition was obtained using GE Amersham Imager AI680.

Band densities were quantified using ImageJ, RRID:SCR_003070 and normalized to the loading control. Statistical comparisons were determined using unpaired t-tests using GraphPad Prism, RRID:SCR_002798.

Antibodies

Antibodies used were Gli1 (Cell Signaling Technology Cat# 2534, RRID:AB_2294745), CyclinD1 (Cell signaling Cat# 55506, RRID:AB_2827374), SIX1 (Cell Signaling Technology Cat# 12891, RRID:AB_2753209), Actin (Cell Signaling Cat# 8456, RRID:AB_10998774), pH2AX (Cell Signaling Cat# 5438, RRID:AB_10707494), Tubulin (Sigma-Aldrich Cat# T9026, RRID:AB_477593), and PARP (Cell Signaling Cat# 9542S, RRID:AB_2160739)

Protein Expression and Purification

Full-length EYA1 wild-type protein and full length D327A EYA1 in GST-fusion format was generated using Bac-to-Bac Baculovirus Expression system as previously described in (34). A GST-tag construct of catalytic domain of EYA1 (residues 322–592) was overexpressed in E. coli BL21 (DE3) and purified using affinity chromatography and size-exclusion chromatography. Briefly, cells were grown at 37°C in TB medium in the presence of 100 μg/mL of ampicillin to an OD of 0.8, cooled to 17°C, induced with 400 μM isopropyl-1-thio-D-galactopyranoside (IPTG), incubated overnight for 20 hrs at 17°C, collected by centrifugation, and stored at −80°C. Cell pellets were lysed in buffer A (25 mM HEPES, pH 7.5, 200 mM NaCl, 1 mM MgCl₂, and 1mM DTT) using Microfluidizer (Microfluidics), and the resulting lysate was centrifuged at 30,000g for 40 min. Glutathione beads (GSH-sepharose, Cytiva) were mixed with cleared lysate for 90 min and washed with buffer A. Beads were transferred to an FPLC-compatible column, and the bound protein was washed further with buffer A for 10 column volumes and eluted with buffer B (25 mM HEPES, pH 7.5, 200 mM NaCl, 1 mM MgCl₂, 1mM DTT, and 15 mM glutathione). The eluted sample was concentrated and purified further using a Superdex 200 16/600 column (Cytiva) in buffer A. 3C protease was added to SEC fractions, containing GST-EYA1, and GST-tag was removed by applying to a second glutathione bead column. The eluant containing cleaved EYA1 catalytic domain was concentrated to ~2–3 mg/mL and stored in −80°C.

Phosphatase Assay

In 96 half-well plates, 400 ng of full length EYA1 protein was incubated in a final volume of 20 µL with 1 mM of peptide (pY-H2AX: APSGGKKATQAJQEO), and 10 µM of compounds in a buffer (60 mM HEPES (pH7.5), 75 mM NaCl, 75mM KCl, 5 mM MgCl2, 1 mM EDTA, and 1 mM DTT) at 37C for 3 hours. The released phosphate was quantified using the Malachite Green Phosphate Assay Kit (Sigma-Aldrich, MAK307). Statistical comparisons were determined using one-way ANOVA using GraphPad software.

FIDA analysis

Flow-Induced Dispersion Analysis (FIDA): FIDA was carried out in the Fida 1 instrument (Fida Biosystems ApS, Denmark) using the standard protocol recommended by the manufacturer (35). In short, full length EYA1 (10 μM) or Eya1 catalytic domain (10 μM)was titrated with compound solution at various concentrations (0 – 300 μM) by injecting a 40 nL plug of protein solution into a 75 µm ID 1 m L capillary prefilled with compound solution and allowing binding to take place during the run. The complex formation could then be observed using intrinsic fluorescence of EYA1. The resulting hydrodynamic radius and fluorescence signal of EYA1 was calculated for each concentration of the compound by converting the measured diffusivity of EYA1 using the Stokes-Einstein relation (35). The Kd was calculated by fitting the binding curve to a 1:1 model using the Fida software 2.3 (Fida Biosystems ApS, Denmark). Analyses were performed in a PBS buffer, pH 7.4, containing 3% DMSO, 1mM DTT, and 0.01% Pluronic acid. Analyses were done in triplicate or as indicated.

RNA-sequencing

MB21 were seeded in 6-well plates and 10 µM of DS38 or vehicle (DMSO) was added to the cells 24 hrs after. After an additional 24 hrs or 5 days of treatment, cells were spun down (300 g) and RNA was harvested using a Qiagen RNeasy Mini Kit (#74004) according to the manufacturer’s protocol. Purified RNA was then submitted for sequencing to the Dana-Farber Molecular Biology Core Facility. Library preparation was performed using the Roche Kapa mRNA Hyper Prep kit (#KK8580, automated using Biomek i7 platform) and final libraries were sequenced on an Illumina NovaSeq 6000 sequencer to produce raw FASTQ files for each biological replicate. Sequencing analyses were completed using a previously published protocol (33).

Cancer Cell Profiling

DS-38, Benzbromarone, and DS-60 were screened as part of the Broad Institute consortium-style screens using the institute’s PRISM assay at 8-point dose (3-fold dilution) with a 5-day treatment for 852 cancer cell lines passing QC. The cell lines include both adherent and suspension cell line and mirror the diversity of the Cancer Cell Line Encyclopedia (CCLE) cell lines (see https://portals.broadinstitute.org/ccle). The PRISM assay workflow is highly standardized; detailed operating procedures of the PRISM assay as well as datasets used for analyses are described at https://www.theprismlab.org and https://depmap.org. To summarize briefly, the cell lines are stably transfected with unique DNA barcodes and pooled in a standardized manner. A 5-day viability assay is then completed whereby compounds are screened at 8-point dose in triplicate against pooled cell lines. Barcoded sequences are amplified by PCR and quantified from the mRNA isolated from the cells to generate cell line sensitivity profiles. These profiles are compared to deep multi-omic cell line characterization information in univariate analyses as well as multivariate predictive modeling algorithms to identify features that correlate with sensitivity.

Synergy experiments

MB21 or Smo (D447G) MB21 cells were plated (1 × 10⁴ cells/well) on 384-well plates. After 24 hours, DS38 and Sonidegib were added to the wells using a HP D300e Digital Dispenser. This was repeated at day 4 of drug incubation. After 7 days cell viability was determined using CellTiter-Glo^® 2.0 Cell Viability (Promega G9241) and luminescent signals were measured with CLARIOstar Plus plate reader and MARS software (BMG Labtech). Each condition was repeated in 3 wells, and duplicate plates were read.

Synergy scores were determined using SynergyFinder 2.0, RRID:SCR_019318 (https://synergyfinder.fimm.fi). From −10 to 10, the interaction between two drugs is likely to be additive, while larger than 10: the interaction between two drugs is likely to be synergistic (36).

Tissue preparation for MALDI MSI and microscopy

SmoM2 mice were injected with DS38 or DMSO (vehicle) at postnatal day 14 (40 mg/kg) and brains were harvested from 1 to 6 hours after injection. The brains were then flash frozen by immersion in liquid nitrogen and stored at −80°C prior to analysis. Tissues were sectioned at 10 μm thickness using a cryostat (Micro hm 550, Thermo, Waltham, MA) and thaw mounted onto indium tin oxide (ITO) coated slides (Bruker Daltonics, Billerica, MA) for MALDI MSI. A matrix composition of 2,5 dihydroxybenzoic acid (160 mg/mL) consisting of 70:30 methanol: water ( 0.1% TFA with 1% DMSO) was sprayed onto the tissue sections using a TM sprayer (HTX Technologies, Chapel Hill, NC). The following spraying parameters were used: flow rate (0.18 mL/min), spray nozzle temperature (75 °C), nitrogen gas pressure (10 psi), spray nozzle velocity (1200 mm/min), track spacing (2 mm), and two cycles. Consecutive serial sections were mounted onto microscopy slides for hematoxylin and eosin (H&E) staining and imaged using bright field microscopy (Zeiss Observer Z.1, Oberkochen, Germany) with a 10× objective.

Tissue mimetic sample preparation for quantitative MALDI MSI

Control mouse brain tissue was homogenized and spiked with varying concentrations of DS38 ranging from 0–50 μM. The nine spiked calibration mixtures were dispensed into a pre-casted 40% gelatin tissue microarray consisting of 1.5 mm core diameter wells and the whole array was frozen until analysis. The quantitative tissue mimetic array was then cryo-sectioned at 10 μm thickness and placed onto the same ITO slide containing the DS38 and vehicle tissue sections.

MALDI MRM Mass Spectrometry Imaging

The MSI data was acquired using a timsTOF flex mass spectrometer (Bruker Daltonics, Billerica, MA) in positive ion mode (m/z 50–650) with multiple reaction monitoring (MRM) scanning. A standard of DS38 mixed in ESI solution was infused into the ESI source of the mass spectrometer at a flow rate of 2 μL/min. The MS targeted method was tuned by optimizing the quadrupole, MRM settings, collision cell, and focus pre-TOF parameters to monitor the product ion. The precursor to product ion transition for DS38 was 347.139 ✇ 121.02 with a 3 m/z precursor isolation width and a collision energy of 45 eV. Prior to MSI data acquisition, the method was mass calibrated using Agilent tune mix solution (Agilent Technologies, Santa Clara, CA) infused into the ESI. A MALDI-MSI method was then created using the ESI tuning conditions mentioned above with additional parameters including a 10,000 Hz laser repetition rate and 50 μm pixel step size consisting of 1,000 laser shots.

The data was visualized using SCiLS lab software (version 2023a core, Bruker Daltonics, Billerica, MA) without normalization. The tissue mimetic was used to correlate ion intensity with spiked DS38 concentration ranging from 0.0 – 50 μM. A linear relationship between these variables was established with a correlation coefficient of 0.9977 (Fig 1. B–C). A limit of quantification (LOQ) of 5.2 μM (S/N ratio of > 10) and limit of detection (LOD) of 1.6 μM (S/N ration of >3) were calculated.

SmoM2 intraperitoneal (IP) Injection and Kaplan-Meier curve

Rosa-26-SmoM2, Atoh-Cre (SmoM2) mice were injected via IP with DS-01-38 reconstituted in DMSO at 20 or 40 mg/kg starting from P7 once a week until euthanized. Controls were given DMSO alone with the equivalent volume. Mice were humanely euthanized when any signs of morbidity were seen and endpoint was recorded. Statistical significance was determined using Log-rank (Mantel-Cox) test using Graphpad software.

Data and Materials Availability

The RNA-seq data generated in this study are publicly available in GEO (GSE229194). Raw data for the RNA-seq assay in this study was generated at Dana-Farber Molecular Biology Core Facility. Derived RNA-seq data supporting the findings of this study are available from the corresponding author upon request. Other raw data generated in this study are available from the corresponding author upon request. Cancer cell profiling data generated by the Broad Institute PRISM platform can be found at https://portals.broadinstitute.org/ccle.

Results

Benzarone derivatives inhibit SHH signaling and proliferation of an SHH-MB model cell line

Uricosuric agents benzarone and its derivative benzbromarone (BB) have previously been identified as EYA protein inhibitors (29,37). To see if benzarone can be further modified to develop enhanced EYA inhibitors, we developed 35 new compounds by modifying primary chemical moieties of the benzarone scaffold (Fig. 1B and Supplementary Table 1). To assess the potential of these new derivatives as SHH pathway antagonists we used SHH Light II (SL2) cells, mouse fibroblast cells that stably express a GLI-dependent Firefly luciferase reporter gene and a Renilla luciferase expression vector as an internal control. Exposure to Shh protein or to the Smoothened agonist (SAG) activates SHH signaling and induces the GLI-dependent Firefly luciferase expression in SL2 cells (13). Previously, we found that EYA1 knockdown inhibits this increase of GLI activity (13). Therefore, the compounds’ ability to inhibit SHH signaling can be monitored in the same fashion. In control cells, SAG stimulation causes GLI transcriptional activity to increase by approximately 18-fold; all benzarone compounds tested at 10 µM suppressed this activation (Fig. 1C and Fig. S1A–B). A few of the novel compounds (ALV-1-48, DS-1-38, and DS-1-52) suppressed the SAG-induced GLI activation with greater efficacy than benzbromarone (Fig. 1C). While Benzbromarone and one of the new compounds, DS-1-52, caused significant decreases in Renilla luciferase activity, a measure of overall cell number, many of the new compounds did not (Fig. 1C–E and Fig. S1A–C). These data indicate that many of these new Benzarone derivatives inhibit SHH signaling, while Benzbromarone and DS-1-52 may be broadly toxic.

MB21 cells are model SHH-medulloblastoma (SHH-MB) cells derived from Ptch^+/− mice tumors that have constitutively activated SHH signaling (31). To determine if benzbromarone or any of the novel benzarone derivatives have effects on SHH-MB cell growth, MB21 cells were treated with 10 µM of the compounds over 7 days, with compounds being replenished on day 4 (Fig. 1D). Viable cells were measured at day 7. We found a majority of benzarone derivatives inhibited growth of MB21 cells (Fig. 1E).

The compounds that most effectively inhibited MB21 cell growth (ALV-1-87, ALV-1-88, DS-1-38, LO-1-208) as well as benzbromarone (BB) were chosen for further analysis (Fig. S2A–D). All selected compounds inhibited SHH signaling in a dose-dependent fashion in SL2 cells (Fig. S2A), and all, except the parental benzbromarone, exhibited a dose-dependent effect on cell growth in MB21 (Fig. S2B), as well as in two other SHH-MB cell lines MB55 and MB56 (Fig. S2C–D).

To address the cellular features that are associated with cancer susceptibility to DS38, we screened DS38, benzbromarone, and DS60 using the Broad PRISM platform (38) at 8-point dose with a 5-day treatment across 852 cancer cell lines. DS38 is not broadly toxic to cell lines but exhibited efficacy in some of these lines. Notably, DS38 inhibited cell viability of Daoy, a human medulloblastoma cell line derived from a SHH-medulloblastoma, as well as TOV112, a human ovarian cell line with highly activated SHH signaling (EC50 of 11.12 μM and 5.98 μM respectively) (Fig. 1F–G). However, a large intestine cancer cell line, LS513, with low SHH signaling, did not respond to DS38 (Fig. 1H). Since sequences of each of the 852 cell lines have been cataloged, the PRISM platform enables identification of mutations that correlate with altered sensitivity to DS38. Mutations in key SHH-signaling components Gli1, Gli2 and Ptch2 impact susceptibility to DS38 (Q value .074, .044 and .006 respectively; see Supplementary Table 2). The platform also allows us to compare results of drug susceptibility with gene expression data, with a CRISPR screen, and with copy number variability. Susceptibility to DS38 correlated with gene expression level of Eya2 and Atoh7, which are in a shared signaling pathway (39). There was a correlation between the effects of DS38 and CRISPR results for Ptch1 and NRP1, which are both critical components of SHH signaling in the cerebellum (13,40), while copy number of GNA12, which is implicated in SHH signaling and medulloblastoma (41), contributed to susceptibility (Feature importance for each of these elements 0.00468, 0.00242, 0.00388, 0.00316 and 0.00285 respectively; see Supplementary Table 3). Overall data from the screen indicate that DS38 impacted multiple cancer cell types that exhibit gene expression profiles associated with neuronal maturation, DNA damage repair, and microtubule binding, features typical of SHH-medulloblastoma (Supplementary Table. 4). Together these data indicate that DS38 is not broadly toxic to cancer cells, and mutations or copy number variations that affect the SHH signaling pathway are associated with susceptibility to this compound.

DS38 inhibit EYA Tyr phosphatase activity and binds to EYA1

Following DNA damage, EYA1 or its family members EYA2 and EYA3 can dephosphorylate histone H2AX at the phosphorylated Y142 site, thus promoting DNA repair and suppressing apoptosis (14,42). To determine if the compounds directly inhibit EYA1 Tyr phosphatase activity, we used an in vitro phosphatase assay in which full length EYA1 protein dephosphorylates a peptide corresponding to pYp-H2AX, and release of free phosphate is measured after 3 hours (34). A phosphatase-dead mutant D327A had no activity in this assay. Compound DS38 (10 μM), added to the assay system, reduced the phosphatase activity of full length EYA1 in vitro (Fig. 2A–B), while ALV87 and ALV88 were less efficacious. While DS38 reduced phosphatase activity of EYA1, it did not affect the interaction of EYA1 with the co-transcription factor SIX1, nor did DS38 alter the subcellular localization of EYA1 (Fig. S3). As pYp-H2AX represents a relevant substrate of EYA1, inhibition of EYA1 Tyr phosphatase should therefore increase the levels of pYp-H2AX in cells treated with effective inhibitors (Fig. 2A). To test the compounds’ ability to inhibit EYA Tyr phosphatase activity in cells, pYp-H2AX levels were analyzed after 48 hours of treatment with 10 µM of the select compounds in SL2 cells exposed to DMSO or SAG. Following treatment with DMSO or the Smo agonist (SAG), cells treated with benzbromarone (BB), and DS-01-38 (DS38) had significantly increased pYp-H2AX expression relative to control cells, while ALV-1-87 also increased pYp-H2AX levels in SAG treated cells (Fig. 2C–D). Together these data indicate that DS-38 is able to inhibit the EYA1 phosphatase activity.

Figure 2. DS38 inhibit EYA Tyr phosphatase activity in SL2 cells and binds to full length EYA1.

A) Schematic of the effects of inhibition of DS38 on EYA1 on pYp-H2AX (Y142) and free phosphate. B) Quantitative analyses of in vitro phosphatase assay showing levels of free phosphate released by wild-type (WT) full length EYA1 with pYp-H2AX peptide, and 10 µM of compounds. WT full length EYA1 protein alone with the peptide is used as a positive control and phosphatase-dead mutant D327A is used as a negative control; n=3, P values indicated or, **** P ≤ 0.001, One-way ANOVA, Bonferroni’s multiple comparison test. C) Representative western blot showing pYp-H2AX (phosphorylated at Ser139/Tyr142) expression before and after 48 hrs of SAG stimulation and compound treatment. DM: DMSO, Actin= loading control. D) Quantitative analysis of pYp-H2AX expression before GLI-responsive luciferase activity; P values as shown, or ** P ≤ 0.01, Multiple t-tests, n=3. E-F) Flow-induce Dispersion Analysis (FIDA) of full length EYA1 with DS38 (E) and DS60 (F). DS38 binds to EYA1 while DS60 showed no binding; n=3, error bars =SEM.

To determine whether DS38 binds directly to EYA1 (Fig. S4A), we used Flow-Induced Dispersion Analysis (FIDA) to assess conformational changes of full length EYA1, or of the Eya1 catalytic domain, when there is direct binding of the purified protein to individual small molecule compounds. The FIDA assay measures hydrodynamic radius (Rh) and intrinsic fluorescence of the protein in either the unbound or bound state. We used FIDA for these studies, as this methodology is very sensitive and useful for assessing binding in a quantitative manner using small amounts of auto-fluorescent proteins (35,43–45). Using this method, we can observe clear changes in apparent hydrodynamic radius (Rh) and in fluorescence signal of EYA1 as a function of increasing concentrations of peptides or of small molecules (Fig. 2E–F and Fig. S4B–D). Full length EYA1 and the catalytic domain of EYA1 bind directly to DS38 (Fig. 2E and Fig. S4B), albeit with micromolar affinity, while the negative control, DS-1-60, a benzarone derivative that did not inhibit SHH signaling or MB21 cell growth, did not show any binding to full length EYA1 or to the catalytic domain (Fig. 2F and Fig. S4D). The EYA1 catalytic domain binds to the pH2AX-D peptide, validating that it is the functional domain that dephosphorylates this known substrate (Fig. S4C), and DS38 also interacts with the catalytic domain of EYA1 with low affinity. DS38 does not bind promiscuously to other proteins, as we found no binding with purified domains of Brd4, ORFN, KRAS, Acat, or IDH1 (Fig. S4E). Taken together these data suggest that DS38 constitutes a benzarone derivative capable of selectively affecting EYA proteins.

DS38 inhibits SHH activation and proliferation while activating a DNA damage response

To evaluate the overall response of SHH-MB cells to these benzarone derivatives in an unbiased fashion, we extracted RNA from MB21 cells after 24 hours, or 5 days of treatment with 10 µM DS38 or vehicle control and performed RNA-seq to identify differentially expressed genes (Fig. 3 and Fig.S5A and Supplementary Table 5–6). The notable genes whose expression changes one day after treatment with DS38 include Cyclin Dependent Kinase 6 (CDK6) and Hedgehog Interacting Protein (HHIP), both of which decrease with treatment (Fig. S5 and Supplementary Table 5). As CDK6 is involved in cell cycle progression (46), while HHIP is a target of SHH signaling that enables a feedback loop for the signaling pathway (47–51), these data suggest that DS38 is affecting proliferation and SHH signaling in these cells. After 5 days of DS38 treatment, we found more extensive changes in gene expression, with an upregulation of genes involved in axonogenesis, neurogenesis, and forebrain development (Fig. 3, Supplementary Table 6), indicating that inhibition of Eya1 increases the expression of genesets involved in neuronal differentiation. This is consistent with prior data that Eya1 maintains symmetric division of undifferentiated cerebellar granule cell precursors, the cells of origin for SHH-MB (13). Genesets implicated in DNA replication and double strand break repair were also downregulated after treatment with DS38, which is consistent with the known functions of EYA1 in regulating proliferation and apoptosis (14) (Fig.3). Expression of gli1, the canonical indicator of SHH-signaling, decreased by 34%-fold after 5 days of DS38 treatment (Supplementary Table 6).

Figure 3. Differential gene expression following treatment with DS38.

RNA sequencing results shown in a volcano plot of differentially expression genes after 5 days of 10 µM of DS38 in MB21 cells. Downregulated genes: blue, Upregulated genes: red. GO enrichment analysis of upregulated and down regulated genes.

To further assess DS38 effects on MB21 at the molecular level across time point, we analyzed GLI1 protein levels and pYp-H2AX levels on day 1, 3, 5, and 7 of treatment with 10 µM DS38 (Fig. 4A–C). GLI1 expression progressively decreased in DS38 treated cells compared to vehicle treated cells, indicating that DS38 inhibits SHH signaling in MB21 cells over time (Fig. 4A–B and Fig. S5B). Activation of GLI1 has been shown to promote cancer and tumorigenic properties, including cell proliferation and survival (52). We find that CYCLIND1 expression significantly decreased on day 5 and 7, while cleaved PARP expression significantly increased on day 7 in DS38 treated cells compared to DMSO treated cells (Fig. 4A, C). At day 7, MB21 cells also exhibited increased pYp-H2AX (Ser139/Tyr142) expression compared to vehicle treated cells (Fig. 4A, C). Together these changes indicate that DS38 inhibits proliferation while activating a DNA damage and apoptotic response in MB21 cells, presumably by inhibition of EYA1.

Figure 4. DS38 inhibits SHH activation and proliferation while activating a DNA damage and apoptotic response.

A) Representative western blot showing protein expression levels of GLI1, PARP, CYCLIND1, and pYp-H2AX (Ser139/Tyr142) after 1, 3, 5, 7 days of treatment of 10 µM of DS38 in MB21 cells. TUBULIN= loading control. B) Quantitative analysis of GLI1 protein expression normalized to vehicle control (DMSO) significantly decreased over time; n=3, *P= 0.0127, *** P ≤ 0.001, **** P ≤ 0.0001. C) Quantitative analysis of protein expression showing DS38 increases pYp-H2AX expression after 7 days and inhibits DS38 proliferation (CYCLIND1) while activating a DNA damage response (cleaved PARP) in MB21 cells. n=3, P=0.0351, **** P ≤ 0.0001. D) Representative western blot showing protein expression levels of GLI1, PARP, CYCLIND1 after 1, 3, and 5 days of in vitro DS38 treatment in MED1712FH cells. Quantitative analysis shows decrease of GLI1 (E) and CYCLIND1 (F) and increase of cleaved (clvd) PARP (G) after DS38 over time; n=3, error bars= SEM, ** P ≤ 0.01, *** P ≤ 0.001.

To assess the effects of DS38 on human SHH-MB cells, tumors were dissected from MED1712FH PDX models, and the dissociated tumor cells were treated with 10 µM DS38 in vitro for 5 days. At 3 and 5 days of treatment, GLI1 protein expression was significantly decreased compared to cells treated with vehicle (DMSO), indicating DS38 inhibits SHH signaling in these human tumor cells (Fig. 4D–E). CYCLIND1 expression levels also decreased significantly, while cleaved PARP expression increased after 5 days of treatment (Fig. 4D, F–G). Thus, MED172FH cells respond in a similar manner to the murine medulloblastoma cells, and these results indicate that an EYA1 inhibitor DS38 reduces proliferation and promotes a DNA damage response, resulting in apoptosis in both human and mouse SHH-medulloblastoma cells.

DS38 shows therapeutic potential in cells resistant to SMO receptor antagonists

Sonidegib (LDE225; Novartis) is a small molecule drug currently used in the clinic as a chemotherapeutic that inhibits the SMO receptor, hence preventing downstream SHH pathway signaling and tumor growth of basal cell carcinoma and medulloblastoma. The drug has previously been shown to reduce cell number in SMB lines, including MB21, and to increase apoptosis (31). As DS38 inhibits EYA1 activity and SHH signaling we asked if DS38 and Sonidegib in combination exhibit enhanced efficacy for SHH-MB cells. MB21 cells were treated with both Sonidegib (1 nM to 5 µM) and DS38 (0.5 µM to 15 µM) over 7 days. The combined treatment of DS38 and Sonidegib demonstrated an additive effect on the inhibition of MB21 cell growth with a ZIP synergy score of 5.365, indicating that the effects of DS38 and Sonidegib are neither redundant nor synergistic (Fig. 5A and Fig. S6A)(36). This is consistent with the idea that EYA1 promotes SHH signaling and has additional contributions to SHH-MB tumor growth. A key problem with Smo inhibitors clinically is the frequent emergence of resistance; a common resistance mechanism involves SMO mutations. MB21 cells stably expressing the SMO mutant (D477G) that have been shown to be resistant to Sonidegib (31), were tested for sensitivity to DS38 alone or in combination with Sonidegib. While the cells were no longer sensitive to Sonidegib as expected, MB21 D477G cells remained sensitive to DS38 (ZIP synergy score −5.789) (Fig. 5B and Fig. S6B), confirming that DS38 can bypass Smo mutant driven resistance and continue to inhibit SHH-MB cell growth.

Figure 5. DS38 and Sonidegib have additive effects in the inhibition of MB21 cell growth.

A) Synergy distribution of DS38 and Sonidegib in MB21 showing the Zero Interaction Potency (ZIP) score of 5.365. B) DS38 inhibits Smo mutant (D477G) MB21 cells which are resistant to Sonidegib. Additive effect of DS38 and Sonidegib confirmed with a synergy score of −5.789. A-B) Software: SynergyFinder 2.0, n=9.

DS38 crosses blood brain barrier to reduce tumor growth and increase survival of SmoM2 mice

A major impediment to treating brain tumors is the difficulty of generating compounds that cross the blood brain barrier and access the tumor itself. Matrix-assisted laser desorption/ionization mass spectrometry imaging (MALDI MSI) can be used to measure the spatial distributions of small molecules and drugs in tissue sections and to evaluate the ability of compounds to cross the blood brain barrier and reach tumor tissue (53,54). To determine whether DS38 can cross the blood brain and blood tumor barriers, we used a robust mouse model of SHH-MB in which a constitutively active mutation of Smo is expressed selectively in cerebellar granule cell precursors (Atoh1-Cre, Rosa-26-flx SmoM2mice), referred to as SmoM2. These mice develop SHH-MB tumors within one week of birth and die within 4 weeks of age (55). MALDI MSI images and optical H&E images were taken from sagittal sections of these SmoM2 mice, injected at postnatal day 14 with DS38 (40 mg/kg) or vehicle, and sacrificed 1–6 hours after injection (Fig. 6A and Fig. S7). Using a quantitative analysis of DS38 from a tissue mimetic (Fig. 6A) and corresponding calibration curve for the MALDI MSI (Fig. S7A), DS38 was found at high concentrations beginning 1 hour post injection in tumor (8.1 µM) and non-tumor regions (4.3 µM) (Fig. 6B). This demonstrates that DS38 transits the blood brain barrier with subsequent tumor penetration that is rapid and efficient. After two hours, the concentration more than doubled in non-tumor regions (17.6 µM) and increased about 3-fold in tumor regions (12.6 µM). The difference in DS38 concentration between the tumor and non-tumor brain regions may reflect inefficient perfusion of the tumor compared to normal brain (Fig. S7B). By 6 hours, DS38 was not detected in either the tumor or non-tumor brain regions, indicating the compound is metabolized quickly and has a short duration of action in vivo (Fig. 6A–B).

Figure 6. DS38 crosses blood brain barrier to increase the survival of SmoM2 mice.

A) MALDI MSI ion images and optical H&E-stained images from sagittal Rosa-26-SmoM2, Atoh-Cre (SmoM2) brain sections of mice treated with 40 mg/kg DS38 or DMSO vehicle and sacrificed 1–6 hours post treatment. Quantitative analysis of DS38 from the tissue mimetic using corresponding (Fig.S6A) calibration curve for MALDI MSI ranging from 0.0–50 μM. B) Measured concentrations of DS38 in mouse brain tissue sections based on tumor and non-tumor regions. C) SmoM2 mice were injected intraperitoneally with 20 or 40 mg/kg every 7 days starting at P7. D) Kaplan-Meier survival curve SmoM2 mice injected with DS38 or vehicle (DMSO), showing DS38-treated mice survived significantly longer. Control n=12, DMSO n=12, *** P ≤ 0.0001

To determine whether DS38 can alter the course of the disease for mice with SHH-MB, IP injections of DS38 (20 mg/kg or 40 mg/kg) were given once a week to SmoM2 mice starting at postnatal day 7 (Fig. 6C–D and Fig. S8A–B). Mice injected with DS38 survived at least one week longer than those given the vehicle (average survival 4.93 weeks in DS38 treated mice versus 3.43 weeks in control mice) (Fig. 6D). The survival of the SmoM2 mice increased significantly with either 40 mg/kg or with 20 mg/kg DS38 (Fig. S8A–B). Analysis of tumors at the time of death (3 weeks of age for control and 5 weeks of age for mice treated with 20 mg/kg DS38) indicated that many cells within the tumors were GLI1 positive, and almost all cells were proliferative as indicated by KI67 expression (Fig. S8C). Apoptotic cell death visualized with activated Caspase-3 (Fig. S8C) could be readily observed in the tumors. Moreover, mice tolerated compound DS38 well, and no side effects were seen. While juvenile mice treated with Sonidegib exhibit impaired growth due to premature closure of the epiphyseal plates, and similar toxic effects on overall growth have been reported in people (56), no growth defects were seen in mice injected with DS38. Together these results indicate that DS38, given just once a week, can limit tumor progression and increase animal survival, although it does not permanently alter the nature of the tumors.

Discussion

Brain cancers are currently the most frequent cause of cancer death in children. Moreover, even children that are successfully treated experience major long-term toxicity with current treatments, and so their quality of life is affected throughout their lives. One of the most common malignant brain tumors are the SHH-medulloblastomas, which are characterized by activation of the SHH pathway in cerebellar precursors due to mutations along the canonical signaling pathway, such as gain of function of SMO (9%), loss of function of SUFU (10%), and loss of function of PTCH1 (43%) (57) or more rarely due to Gli2 mutations (58). These insights initially led to optimism that therapies targeting SHH-signaling might provide effective and less toxic treatments for these children, and SMO inhibitors (sonidegib and vismodegib) are FDA approved for BCC (10,59). However, tumors rapidly and frequently develop resistance to SMO inhibitors and so the cancer relapses. Moreover, SMO inhibitors impede normal bone growth, and so are contra-indicated in young children (56). Here we generated a new benzarone derivative, DS38, that binds EYA1 and inhibits phosphatase activity, decreases SHH signaling and proliferation while activating DNA damage and apoptotic response in SHH-MB cells. Furthermore, DS38 crosses the blood brain barrier efficiently and reduces tumor progression in vivo with little toxicity, ultimately leading to an increase in the survival rate of SHH-medulloblastoma model mice. Collectively, these data provide proof-of-concept for inhibitors of EYA1 as treatment for pediatric SHH-MB.

Several studies have tried to identify additional ways of targeting the SHH-signaling pathway in medulloblastoma to overcome concerns with SMO inhibitors. These include HDAC inhibitors, sodium butyrate (NaB), trichostatin A (TSA), valproic acid (VPA), suberoyl anilide hydroxamic acid (SAHA, vorinostat), panobinostat, belinostat, and romidepsin (60). As mutations in SMO are a major driver for resistance, therapies that function downstream of SMO have the potential of bypassing the resistance. Not only do EYA proteins and its binding partners of the SIX family function downstream of SMO, they are consistently expressed in SHH-medulloblastoma cells, have additional oncogenic functions beyond the SHH signaling pathway (61,62), and are rapidly downregulated postnatally (63). These data suggest that inhibiting EYA1 may be effective in treating SHH-medulloblastomas and have limited side effects in children.

Among the benzarone derivatives tested, we found that efficacy in inhibition of SHH signaling did not directly correlate with decreases in SHH-MB cell growth (Fig. S1B). This suggests that EYA1 contributes to SHH-medulloblastoma progression through multiple mechanisms that extend beyond inhibition of canonical SHH signaling. The RNA seq analysis also suggests that treatment with DS38 has additional effects beyond the changes in SHH signaling (Fig.3). For example, the EYA family regulates the DNA repair and survival pathway in part by dephosphorylation of C-terminal pY142 on H2AX (14). Thus, EYA activity may be recruiting the MDC1/MRN complex instead of the apoptotic pathway when DNA damage occurs—directing tumor cells to DNA repair rather than apoptosis (14,42). In addition, through aPKC, EYAs regulate symmetric division, which has been long linked to tumorigenesis and delayed differentiation (34). A recent study shows that inhibition of the EYA2 tyrosine phosphatase activity decreases expression of the MYC gene in Group 3 MB tumors growing in vivo, including when xenografts are implanted in the flank and when implanted intracranially (23). In breast cancer, EYAs affect Erβ signaling and contribute to the migration, invasion, and metastasis of tumor cells (61). Taken together it seems likely that EYA phosphatase family have functions that extend beyond SHH signaling and contribute more broadly to cerebellar development, medulloblastoma growth and cancer progression.

As EYA proteins are multifunctional, the contribution of the phosphotyrosine phosphatase activity to proliferation, oncogenic behavior and SHH signaling has been controversial (15). In this study, we find that DS38 binds EYA1 and reduces phosphatase activity of full length EYA1 protein in vitro, while DS38 treatment alters the phosphorylation state of Eya1 substrates such as H2AX in cells. These data indicate that DS38 inhibits Eya1 phosphatase activity, and that the phosphatase activity is implicated in tumor growth. Results from the PRISM assay platform indicate that mutations in Gli1, Gli2 and Ptch2 that activate the SHH signaling pathway are predictive of cancer cell sensitivity to DS38; findings consistent with previous studies that implicate the EYA1 phosphatase in SHH signaling and oncogenesis (13). However, as benzbromarone can potentially impact additional distinct functions such as organic acid transport (64) and mitochondrial metabolism (65), and the PRISM assay identifies multiple cancer cells that respond to this inhibitor, we cannot exclude the possibility that additional targets could also contribute to the observed anti-oncogenic efficacy.

We find that DS38 is remarkably efficient in crossing the blood brain barrier and reaches a high concentration in tumor and non-tumor regions of the brain within an hour after a single IP injection (Fig. 6). Strikingly we find that DS38 stalls tumor progression and increases the life span in a SHH-MB mouse model. Together these findings indicate that in the future improved inhibitors of EYA1 phosphatase activity are promising approaches for treating SHH-medulloblastoma.

Scheme 1 –

synthesis of DS-1-038 and LO-1-208

(i) Chloroacetone, KOH, MeOH, 80°C, 62%; (ii) NH₂NH₂.H₂O, (CH₂OH)₂, 180°C, then KOH, 47%; (iii) 4-methoxybenzoyl chloride (4) SnCl₄, CS₂, 83%, or 4-methoxy-2-methylbenzoyl chloride (LO-1-208), SnCl₄, DCM, 100%; (iv) BBr₃, DCM, −78°C, 68%; (v) 1-methyl-4-(4,4,5,5-tetramethyl-1,3,2-dioxaborolan-2-yl)-1H-pyrazole, Pd(PPh₃)₄, Cs₂CO₃, dioxane, 90°C, 21%.

Scheme 2 –

synthesis of ALV-1-087

(i) 4-methoxy-2-methylbenzoyl chloride, SnCl₄, benzene, 80°C, 40%; (iv) BBr₃, DCM, −78°C, 78%.

Statement of Significance.

Development of a benzarone derivative that inhibits EYA1 and impedes the growth of SHH medulloblastoma provides an avenue for improving treatment of this malignant pediatric brain cancer.