Identification of Novel Ezrin Inhibitors Targeting Metastatic Osteosarcoma by Screening Open Access Malaria Box Free

This article is featured in Highlights of This Issue, p. 2409

The vast majority of deaths associated with cancer are due to the metastatic spread of cancer cells from a primary tumor to distant sites (1). Osteosarcoma is a highly metastatic cancer of bone that afflicts children, adolescents, and young adults, with the majority of patients having micrometastasis at the time of initial diagnosis (2). Although, the survival of patients with metastatic disease at diagnosis remains to be poor with 5-year survival rates being less than 20%, significant improvements have been achieved in the management of localized tumors through development of multimodality approaches that resulted in 5-year overall survival of 60% to 78% (3–7). However, recurrent osteosarcoma occurs in 30% to 40% of those patients initially diagnosed with localized disease (8). Development of metastasis to the lungs remains the most common cause of death in patients with osteosarcoma. Hence, a mechanistic understanding of the metastatic process and development of molecularly targeted therapeutics aimed at preventing such disseminated disease may provide additional improvements in disease outcomes for patients with metastatic disease (9).

Ezrin is a member of the ERM (ezrin, radixin, moesin) protein family that functions as a linker protein between F-actin in the cortical layer and membrane-associated proteins on the cell surface (10). Ezrin serves as a key regulator of diverse cellular processes such as formation and organization of cell-surface structures, maintenance and determination of cell shape, and modulation of cell adhesion, migration, and signaling pathways (10, 11). ERM proteins are characterized by the presence of three distinct regions, the including N-terminal membrane-associated domain [called N-terminal ERM-association domain (N-ERMAD)], followed by a long central α-helical region and a C-terminal [called C-terminal ERM-association domain (C-ERMAD)] domain, which is able to bind both actin filaments in the cytoskeleton and N-ERMAD. In the current understanding of ezrin regulation, the self-association of the protein by head-to-tail joining of the molecule is believed to mask the respective ligand binding sites, which leads to a dormant closed conformation. The activation of ezrin from its closed conformation to the open form has been proposed to occur in a two-step process that involves phosphorylation of a conserved threonine residue (Thr-567) in the C-ERMAD of ezrin and binding to membrane phosphatidylinositol 4,5-biphosphate via its N-terminal domain (12). This conformational change is regarded as critical for binding of ezrin to membrane proteins, including transmembrane receptors, small GTPase regulators, and adaptors (10–12). Thus, ezrin acts as a scaffolding protein and can assemble highly specific and regulated clusters of membrane proteins at the cell cortex. Independent from its canonical role as a cytoskeletal scaffolding protein at the plasma membrane, our recent findings suggest that ezrin has a likely role in early phases of protein translation initiation as part of a ribonucleoprotein complex (13). We also recently demonstrated that ezrin interacts with proteins involved in protein translation initiation and stress granule dynamics. Among those proteins, we identified DDX3, a DEAD-box RNA helicase involved in multiple aspects of RNA metabolism from transcription to translation, as a direct binding partner of ezrin (14).

The elevated levels of ezrin have been shown to confer advantageous metastatic behavior in a mouse model of osteosarcoma (15) and high ezrin expression has been linked to poor outcome in patients suffering from osteosarcoma (16). Many clinical studies have also implicated the role of ezrin in tumor progression of a variety of cancers, including carcinomas of the lung, breast, colon, pancreas, endometrium, and ovary, cutaneous and uveal melanomas, oral squamous cell carcinomas, rhabdomyosarcoma, brain tumors, and soft tissue sarcomas (17, 18). We have recently discovered a small molecule inhibitor of ezrin NSC305787, which directly binds to ezrin and inhibits its function associated with the metastatic phenotype in both in vitro and in vivo experimental models (19).

NSC305787 contains an aromatic quinoline core structure with a bulky aliphatic adamantyl group in its northeast quadrant and a piperidine unit in the southeast region (Supplementary Fig. S1). Examination of the structure of NSC305787 revealed that it shares significant structural features with commonly used quinoline-based antimalarial agents such as quinine, quinidine, mefloquine, choloroquine, and amodiaquine (Supplementary Fig. S1), which are mostly active against the inthraerythrocytic stages of the malaria parasite (20). The close structural resemblance between NSC305787 and antimalarial agents, therefore, has led us to screen the quinoline family of antimalarial drugs, including quinine, quinidine, mefloquine, chloroquine, and amodiaquine, as well as Medicines for Malaria Venture (MMV) Malaria Box content as ezrin inhibitors. The MMV Open Access Malaria Box constitutes 200 drug-like and 200 probe-like compounds selected from 19,000 structurally unique molecules with the confirmed activity against the blood stage of Plasmodium falciparum (21–24).

Consequently, we screened the antimalarial compounds for their direct binding ability to ezrin. We followed this primary screen with secondary functional assays using in vitro, in vivo, and ex vivo experimental models. Here, we identify compounds from MMV400 library as novel small molecule inhibitors of ezrin blocking cell motility and invasion.

The Malaria Box was provided in 100% DMSO at 10 mmol/L concentration in a 96-well format, with 5 plates each containing 80 compounds. The Plate mapping corresponds to the May 2012 shipment. Detailed information regarding the Malaria Box is described in Spangenberg and colleagues (24). Antimalarial compounds that showed potent binding activity to recombinant ezrin protein were purchased from commercial sources (MMV666069 from ChemBridge Corp., #67378588; MMV665977 from Vitas-M Laboratory, Ltd., #STK701059; MMV020243 from ChemBridge Corp., #9290708; MMV020549 from ChemDiv Inc., #E950-0299; MMV396680 from ChemDiv Inc., #G420-0124; MMV667492 from Vitas-M Laboratory, Ltd., #STK529340; MMV000448 from Vitas-M Laboratory, Ltd., #STK589099; MMV006172 from Vitas-M Laboratory, Ltd., #STK746950; MMV666103 from Asinex Ltd., #BAS02538915).

Mouse K7M2 and K12 cell lines were kindly provided by Dr. Chand Khanna from NCI/NIH (MD, USA) in 2007; no authentication was done by the authors. Cells were maintained in DMEM supplemented with 10% FBS in a humidified atmosphere of 5% CO₂ at 37°C.

For cytotoxicity experiments, cells were plated on T-75 flasks and grown until they reached 70% to 80% confluency. Cell proliferation was monitored in the xCELLigence Real-Time Cell Analysis (RTCA; ACEA Biosciences Inc.) system using E-Plate 96 or E-Plate 16 (ACEA Biosciences Inc.) that are integrated with gold microelectrode arrays in the bottom of each well. A total of 100 μL of culture media was added into the plates, and the plates were inserted into the station for background measurement. Cells were then seeded into the wells (5,000 cells/well in 100 μL) and left to attach overnight. The medium was then removed and replaced with 100 μL of fresh medium containing indicated concentrations of the compounds. Impedance was measured every 10 minutes during 48 hours and is represented as cell index.

Recombinant human ezrin was purified as described previously (25). Escherichia coli strain M15 [pREP4] (Qiagen) was used for the expression of untagged, full-length ezrin from pQE16 expression vector. Ezrin was purified by using a hydroxyapatite (Bio-Rad) column followed by cation exchange on a SP-Sepharose fast flow (GE Healthcare Bio-Sciences) column in AKTA Explorer Chromatography System (GE Healthcare Bio-Sciences). Protein purity was determined by SDS-PAGE followed by Coomassie staining.

The primary screen of the MMV library was performed on a Biacore 3000 instrument using a CM5 chip coupled with ezrin on one flow cell and glideosome-associated protein GAP50 (26) as a negative control on another flow cell. Compounds from the MMV Malaria Box were dissolved to a final concentration of 10 μmol/L in HBS-P [10 mmol/L HEPES, pH 7.4, 150 mmol/L NaCl,% 0.05 (v/v) surfactant P20] and passed over all four flow cells. The running buffer contained 1% DMSO with HBS-P to match the DMSO level of the samples.

After the initial screening of antimalarial compounds, the small molecules that showed at least three-fold difference in relative binding to ezrin compared with GAP50 were further evaluated to identify their binding affinities using Biacore T200 instrument at room temperature. First flow cell was left empty for background signal subtraction and second flow cell of a CM5 sensor chip was used for immobilization of the purified recombinant ezrin protein (∼10,000 RU) by amine coupling method in sodium acetate buffer pH 4.5. NSC305787 was used as a positive control in these surface plasmon resonance (SPR) experiments. HBS-P [10 mmol/L HEPES, pH 7.4, 150 mmol/L NaCl,% 0.05 (v/v) surfactant P20] containing DMSO between 2% and 10% final concentration was used as the running buffer. Kinetic analysis was done by injecting seven different concentrations of compounds (0.78, 1.56, 3.13, 6.25, 12.5, 25, and 50 μmol/L) over ezrin-captured and control surfaces in triplicates. Each injection was done with 60 seconds association time and 180 seconds dissociation time. K_D values for compound–ezrin interactions were obtained using BiaEvaluation software (version 1.0).

Immunoprecipitation (IP) experiments were performed on lysates of K7M2 mouse osteosarcoma cells as described previously (14). Total cell lysates were prepared in phospho-lysis buffer (50 mmol/L HEPES pH 7.9, 100 mmol/L NaCl, 4.0 mmol/L sodium pyrophosphate, 10 mmol/L EDTA, 10 mmol/L sodium fluoride, and 1% Triton X-100) containing 2.0 mmol/L sodium vanadate, 1.0 mmol/L PMSF, 4.0 μg/mL aprotinin, 4.0 μg/mL leupeptin, and 1.0 μg/mL calyculin A. Lysates were incubated on ice for 30 minutes, and then cleared by centrifugation at 16,000 × g for 10 minutes at 4°C. An equal amount of protein lysate (∼1,000 to 1,500 μg/1.0 mL phospho-lysis buffer) was incubated with 2.0 μL of anti-ezrin antibody (Sigma Aldrich; #E8897) or the equal amount of control mouse IgG (Santa Cruz Biotechnology; #sc-2025) overnight at 4°C on a rotating axis, followed by incubation with BSA-blocked Protein A/G Plus-Agarose beads (Santa Cruz Biotechnology) for 1 hour. Subsequent to washing of the pelleted beads three times with 0.3 mL (each) IP-wash buffer (10 mmol/L Tris-HCl, pH 7.5, 150 mmol/L NaCl, 1.0 mmol/L EGTA, 1.0 mmol/L EDTA, 1% Triton X-100, 0.5% Nonidet P-40, 0.2 mmol/L sodium vanadate, and 0.2 mmol/L PMSF), the bound proteins were eluted by boiling samples in SDS-PAGE sample buffer for 5 minutes. The immune complexes were then analyzed by Western blot analyses. Immunoblotting (IB) experiments were performed on both K7M2 and K12 cells as previously described (14, 27). Equal amount of protein samples were separated by SDS-PAGE and then transferred to an Immobilon-P membrane (Millipore). In order to block nonspecific binding sites, the membrane was incubated in 5% nonfat dry milk diluted in TTBS (20 mmol/L Tris-HCl, pH 7.5, 150 mmol/L NaCl, 0.05% Tween 20) for 1 hour at room temperature followed by the addition of primary antibody for 2 hours at room temperature. Dilutions for primary antibodies were as follows: anti-ezrin antibody (Sigma Aldrich; #E8897) at 1:2,500, anti-phospho ezrin(Thr567)/Radixin(Thr564)/Moesin(Thr558) antibody (Cell Signaling Technology; #3141) at 1:1,000, and anti-actin-HRP (Santa Cruz Biotechnology; #sc-1615) at 1:5,000. The membrane was then rinsed three times with TTBS and incubated for 1 hour at room temperature in HRP-conjugated anti-mouse (GE Healthcare Bio Sciences; #NA931V) or HRP-conjugated anti-rabbit (GE Healthcare Bio-Sciences; #NA934V) secondary antibody diluted at 1:5,000 in TTBS. The blot was washed three times in TTBS and immunoreactive bands were revealed by using Millipore Immobilon Western Chemiluminescent HRP substrate according to the manufacturer's instructions (Millipore Corporation). Chemiluminescence was detected using a Fujifilm LAS-3000 imaging system.

Cell migration experiments were performed using the xCELLigence RTCA (ACEA Biosciences Inc.) system as described previously (28). This method allows label-free monitoring of cell migration continuously in real time via electrical impedance readout. Cell migration experiments were performed using 16-well plates (CIM-16, ACEA Biosciences Inc.), which consists of an upper and a lower chamber separated by a microporous membrane containing randomly distributed 8-μm pores. Briefly, 160 μL of DMEM media containing 10% FBS was loaded in the lower wells as chemoattractant and the upper chamber was attached. The serum-free media in the lower well was used as a negative control for cell motility. Upper wells were filled with 50 μL serum-free media and the plate was left in the RTCA-DP device at 37°C for 1 hour to pre-equilibrate. Following measurement of background signal generated by cell-free media, 50 μL of K7M2 or K12 cell suspension (1.5 × 10⁵ cells) in serum-free media was placed into the top wells. Next, 50 μL of serum-free media containing appropriate concentrations of NSC305787 and the antimalarial compounds were added into the wells. The final concentration of DMSO in cell suspension was 1%. The assembled plate was then transferred to the RTCA-DP machine and data were collected every 10 minutes over 21 hours. The electrical impedance measured by the device as a dimensionless parameter termed cell index was used to calculate% motility of compound-treated osteosarcoma cells compared with vehicle-treated control.

Zebrafish (Danio rerio) were raised and maintained in Georgetown University zebrafish facility using standard husbandry techniques as described previously (29). A translation-blocking anti-ezrin morpholino (MO), which has been previously described and validated by Link and colleagues (30), was synthesized by Gene-Tools, LCC. Embryos were collected and staged according to both hours post-fertilization (hpf) at 28.5°C and by morphologic criteria (31). For chemical screening, zebrafish embryos were arrayed in 96-well plates and compounds were added at 10 to 50 μmol/L concentrations at 3 hpf. Embryos were scored for epiboly defects and/or characterized for other morphologic defects at time points corresponding between 7.0 and 8.5 hpf and 1, 3, and 5 days post-fertilization (dpf), respectively. All animal procedures were approved by the Institutional Animal Care and Use Committee of Georgetown University.

The pulmonary metastasis assay (PuMA) was performed as previously described (32). GFP-expressing K7M2 cells (5 × 10⁵) were injected to three Balb/c mice through tail vein. The mice were euthanized by CO₂ inhalation after 15 minutes of tumor cell injection. The trachea was exposed and infused with a culture medium/agarose solution using a gravity perfusion apparatus. The inflated lungs were removed and placed in cold PBS containing 100 U/mL penicillin and 100 μg/mL streptomycin at 4°C for 20 minutes to solidify the agarose. Axial sections of 1 to 2 mm in thickness were cut from each lobe. Two lung sections from each of the three animals (in total six lung sections per treatment) were placed on a piece of sterile Gelfoam (Pfizer-Pharmacia & Upjohn Co.) that had been preincubated in culture medium. Compounds were added immediately to the culture media at day 0 of the culture period. Lung sections were incubated at 37°C in a humidified atmosphere with 5% CO₂. The medium was replaced with fresh culture medium with or without compounds every other day. Animal care and use was in accordance with the guidelines of the NIH Animal Care and Use Committee, and approved by the Institutional Animal Care and Use Committee of Georgetown University.

A Nikon ECLIPSE Ti fluorescence inverted microscope (Nikon Instruments) equipped with a DS-Fi1c high-definition cooled color camera head was used to capture images of GFP-positive tumor cells at 10× magnification. Images were processed and analyzed by NIS-Elements BR 3.0 imaging software (Nikon Instruments). Metastatic burden was quantified in the lung sections at day 8 by measuring the total fluorescent area of each lung section as μm² occupied by the metastatic cells. The fluorescent area for each lung section was normalized to 100 μm² for day 0 to allow quantitative evaluation of metastatic progression over time.

Lung tissues were fixed in 10% buffered formalin overnight at 4°C and embedded in paraffin using standard techniques. Serial sections of 5-μm thickness were cut from paraffin-embedded tissues and mounted on glass slides. Tissue sections were stained with hematoxylin and eosin (H&E). H&E-stained slides were then examined for the viability and structural integrity of pulmonary architecture. Images of H&E-stained lung sections were taken at 10× objective using Olympus BX40 microscope equipped with ProgRes C5 digital camera.

All data were reported as mean ± SD of n observations. Statistical evaluation was performed by an unpaired Student t test using the GraphPad Prism version 6.0c (GraphPad software), taking a probability P < 0.05 as statistically significant.

We used the SPR technology to screen the compounds that directly bind to purified recombinant ezrin protein that was immobilized on a sensor chip. We first screened antimalarial drugs, quinine, quinidine, mefloquine, amodiaquine, and chloroquine for their direct binding ability to ezrin due to their structural similarity to our lead compound, NSC305787 (Supplementary Fig. S1). However, none of the compounds produced any significant binding to ezrin compared with NSC305787.

We then extended our screen to the MMV400 library, of which 91 compounds demonstrated varying degrees of binding to ezrin above the background level (Fig. 1A). We used a negative control protein, GAP50, in order to eliminate nonspecific binders. The hit selection of Malaria Box screening was based on a threshold of three-fold difference in relative binding of a compound to ezrin with respect to GAP50. Based on this selection criterion, 12 hits were identified out of 91 compounds, of which 4 were “drug-like” and the remaining 8 were “probe-like” (Fig. 1B). Among the 12 primary hits, we subsequently determined the binding affinities of 9 commercially available compounds (Fig. 2A) for ezrin by detailed SPR studies in a Biacore T200 instrument. For this multiple concentration secondary analysis, we excluded MMV396680 and MMV000448 due to their poor solubility. This analysis yielded K_D values ranging from 2.1 ± 1.0 to 29.4 ± 4.4 μmol/L for seven hit compounds (Table 1, Fig. 2B; Supplementary Fig. S2).

We evaluated the nine primary hits based on their physicochemical properties to determine whether they conform to the Lipinski's “Rule of Five” for druggability (33). Molecular weight (≤500 daltons), partition coefficient (cLogP ≤5), and the number of hydrogen bond donor (≤5) and accepting groups (≤10) present in the compound are used to predict whether or not a compound is likely to be orally bioactive based on its absorption or permeation. Veber and colleagues (34) have also proposed additional rules, including the number of rotatable bonds (≤10) and the polar surface area (≤140), to predict the compounds with good oral bioavailability. Computational analysis of the compounds using the MolInspiration software (http://www.molinspiration.com) revealed that all the primary hits except MMV006172 and MMV396680 had structural features that satisfy all six conditions of the above methodology (Table 1). Thus, although six antimalarial compounds were initially categorized as “probe-like” in MMV Malaria Box (Fig. 2A), four of them hold therapeutic promise for further functional assays and lead optimization studies. The lead compound NSC305787 did not meet all of the Lipinski and Veber drug-likeness criteria due to its high computed partition coefficient (cLogP) of 6.30 (Table 1).

We first employed an osteosarcoma cell culture method to confirm the hits from the primary screen on the basis of their ability to inhibit ezrin-mediated cell motility. K7M2 murine osteosarcoma cells are highly aggressive and metastatic to the lungs and were derived from less aggressive K12 cells originating from spontaneously occurring osteosarcoma in BALB/c mouse (2). K7M2 cells express higher levels of ezrin protein than K12 (Fig. 2C), which leads to enhanced metastatic phenotype of K7M2 cells (15, 32). Initially, we determined the cytotoxicity profiles of each compound for both cell lines to identify the nontoxic dose limits prior to investigating their ability to inhibit ezrin-mediated cell motility (Supplementary Fig. S3). Because cell viability is not affected by the alterations in ezrin level, concentrations that reduce the cell viability were considered as nonspecific response and irrelevant to the ezrin inhibition. Therefore, all experiments were performed at appropriate concentrations of individual drugs and time intervals that do not affect cell viability. Compound MMV000448 was discarded at this stage because of its relatively high cytotoxicity toward both K12 and K7M2 cell lines (Supplementary Fig. S3).

We used a real-time cell monitoring technology based on electrical cell impedance in a label-free environment to explore the potential inhibitory effect of compounds on migration of K12 and K7M2 cells. We used a Boyden-like dual chamber system consisting of an upper chamber and a lower chamber separated by a 8-μm pore-sized membrane. As cells migrate from the upper chamber to the lower chamber in response to serum, they contact and adhere to the gold microelectrode sensors on the underside of the membrane, which results in an increase in electrical impedance. Figure 2C shows the percentage of motility of the K7M2 and K12 cells in the presence of primary hits. Data were normalized to control treatment, which was taken as 100%. The addition of anti-ezrin compound NSC305787 decreased the migration of the K12 and K7M2 cells to 70.2 ± 8.0% and 38.5 ± 10.8% of the control, respectively. All the primary hits except MMV020549 and MMV396680 exhibited the same pattern as NSC305787 with a more marked reduction in the motility of K7M2 cells compared with that of K12 cells at varying degrees (Fig. 2C; Supplementary Fig. S4).

The functional analysis of ezrin2 (the zebrafish homolog) protein in zebrafish developmental model revealed that ezrin2 is required for germ layer morphogenesis during gastrulation (30). Injection of antisense MO oligonucleotides designed to knockdown ezrin2 expression shows specific phenotypes in zebrafish characterized by the reduced epiboly movements and in the most severe cases, the complete detachment of the blastoderm from the yolk cell at early gastrulation (30). Figure 3A shows the ezrin MO-induced inhibition of epiboly. Compared with un-injected wild-type siblings at 60% epiboly, MO-injected embryos show reduced epiboly and a thickening of the blastoderm. The failure of gastrulation is seen at the initiation of involution as an exaggerated thickening at the blastoderm margins. Severe ezrin morphant phenotypes are characterized by strong epiboly defects resulting in cells piling up at the animal pole (19). This phenotype was used as a guide to screen small molecules by observing embryonal development in 96-well-plate format. Thus, in addition to the chemotaxis assays, we tested the potential biologic effects of the nine primary hits on a zebrafish embryonic phenotype assay. We initially screened the compounds at 10 and 30 μmol/L concentrations, evaluating phenotypes at germ rising stage, 1, 3, and 5 dpf. A brief summary of the findings from this initial screening is presented in Supplementary Table S1 and Supplementary Fig. S5. Treatment with 10 μmol/L NSC305787 mimicked the ezrin knockdown phenotype, which resulted in 100% epiboly failure with ultimate lysis of the embryos by 1 dpf. Among the hit compounds tested, embryos treated with MMV020549 and MMV667492 at 10 and 30 μmol/L concentrations developed the best ezrin MO-like phenotypes, which were characterized by morphologic defects, including delayed development, convergence extension defects, apc-mutant-like phenotypes, deformed embryos, and eventually the lysis of the embryos. Following treatment with MMV666069, MMV020243, and MMV396680, the embryos exhibited subtle defects with no swim bladder, curved embryo phenotypes, and slight increase in mortality, respectively, whereas MMV665977-, MMV006172-, MMV666103-, and MMV000448-treated embryos developed normally.

In light of the initial findings, we tested the effect of MMV020549 and MMV667492 in detail at different time points. MMV020549 exhibited a dose-dependent inhibition of epiboly and gastrulation between 7.0 and 8.5 hpf (Fig. 3B and C). MMV667492 showed a strong anti-ezrin-like phenotype in a dose range of 10 to 30 μmol/L (Fig. 3B and C). We observed no evidence of gastrulation occurring in MMV667492-treated embryos.

Osteosarcoma has a high propensity for early metastasis, with tumor cells being more prone to metastasize to the lungs than the other tissues (16, 35). We tested the potential anti-ezrin activity of our hits in a PuMA utilizing ex vivo cultures of lung tissue. This method allows the assessment of metastatic progression of the cancer cells reaching the lung (32). GFP-expressing K7M2 mouse osteosarcoma cells were injected into the tail-vein of Balb/c mice. Following injection of the tumor cells, the mice were euthanized and the lungs were infused with a mixture of culture medium and agarose to maintain the lung structure. The lungs were then allowed to cool to solidify the agarose medium solution, and axial sections were made from each lung lobe. Lung sections were placed on sterile Gelfoam sections bathing in culture media and immediately treated with either vehicle or compounds. Metastatic foci growing in lung slices were then visualized and quantitated based on GFP intensity using a fluorescence microscope (Figs. 4 and 5).

Exposure of lung tissue slices to the anti-ezrin compound NSC305787 significantly inhibited the metastatic growth of high-ezrin expressing K7M2 cells compared with the vehicle-treated group. When the lung tissues were treated with MMV667492 and MMV020549, the most active hits found in the zebrafish embryonic phenotype assays, we observed a significant decrease in the number of GFP-expressing metastatic foci compared with vehicle-treated control group (Figs. 4 and 5). We also tested the potential antimetastatic activities of MMV666069 and MMV020243, which demonstrated subtle to moderate effects in zebrafish embryonic developmental assays. MMV666069 significantly inhibited the metastatic growth of the K7M2 osteosarcoma cells and produced a distinguishing phenotype in the live fluorescent imaging of lung sections (Fig. 5). On the other hand, a large number of metastatic colonies containing high-ezrin expressing K7M2 osteosarcoma cells were evident in the lung tissues treated with MV020243 (Figs. 4 and 5). Interestingly, MMV665977, an ineffective compound found in zebrafish embryonic development assays, significantly inhibited the metastatic growth of the K7M2 cells in the lung tissues as well (Figs. 4 and 5).

We performed histologic examinations of lung tissue slices treated with compounds to confirm the maintenance of lung architecture over a 8-day period of drug treatment in ex vivo PuMA. We observed structurally intact pulmonary architecture (Fig. 5), which confirms that the observed inhibition of metastatic growth by compounds was not due to the deterioration of lung tissue.

We performed structural and shape comparison analysis of ezrin inhibitors using the OpenEye software Rapid Overlay of Compound Structures (ROCS; ref. 36). A heat map was generated based on the combined shape complementarity and chemical Tanimoto score as calculated by the ROCS. This analysis revealed that the identified ezrin inhibitors from the MMV400 library fall into two shape classes (Fig. 6). It is likely that the two binding sites exist on ezrin, which are occupied by these compounds. However, extended compounds such as MMV020549 can adopt a very compact form and fit into the shape constraints of NSC305787 as well. These computational predictions will have to be experimentally validated in crystal structure or nuclear magnetic resonance studies.

Metastasis remains the major cause of cancer mortality. Even though our understanding of molecular events that drive metastasis has improved for the past two decades, there is still lack of efficient therapeutics targeting metastatic progression. Extensive research on metastasis biology has led to the identification of a number of molecular pathways and cellular mechanisms with druggable targets in several cancers (37–39). Ezrin is one such protein with a cytoskeleton–membrane linker function that has been identified as a major driver of osteosarcoma metastasis (15, 16). Earlier work in our laboratory has led to the discovery of a quinoline-based small molecule inhibitor of ezrin, NSC305787, which possesses a very close structural similarity to the quinoline family of antimalarial agents. Here, we present the findings of our screening of the antimalarial compounds in an attempt to find novel scaffolds with better activity and improved physicochemical properties for targeting ezrin-dependent metastatic phenotype. The overall comparative results for screening assays are summarized in Table 1.

Our findings revealed a naphtoquinone-based antimalarial compound, MMV667492 (Fig. 2A), as the most potent inhibitor of ezrin-mediated cell motility and invasive phenotype in all functional assays. The remaining compounds were found to be inactive in at least one of the functional assays. Among those compounds, MMV020549, which is composed of piperidine and a pyrroloquinoline carboxamide (Fig. 2A), was active in all but the cell motility assay of osteosarcoma cells, whereas MMV666069 gave significant responses in all biologic activity assays with a moderate/subtle morphogenic effect in zebrafish embryonic developmental assays (Table 1). Both compounds have been assigned to drug-like category in the MMV400 library. MMV020243, which is composed of a pyrimidine ring, a phenyl group, and a second phenylurea ring (Fig. 2A), exhibited more potent inhibition of cell motility toward high-ezrin expressing cells and produced curved embryo phenotypes in zebrafish developmental assays, whereas it was not active in preventing ex vivo lung metastasis (Table 1). MMV665977 (an oxoquinoline derivative), MMV666103 (an oxoquinazoline-derivative), and MMV006172 (an acridine-derivative; Fig. 2A) were all inactive in mimicking the ezrin MO phenotype in zebrafish developmental assays but showed a more preferential inhibition of the motility of high-ezrin expressing K7M2 cells. MMV665977 also significantly prevented ex vivo lung metastasis of high-ezrin expressing the K7M2 osteosarcoma cells. MMV396680, a pyrazolo-pyrazine derivative, had a slight toxic effect on embryonic development of zebrafish and exhibited only a subtle increase in preferential inhibition of the motility of the K7M2 cells compared with that of the K12 cells. MMV000448, an acridine derivative, was excluded from cell motility assays due to its high cytotoxicity, and was inactive in zebrafish embryonic developmental assays (Table 1).

Although a clear structure–activity relationship was not discernible among compounds, MMV667492, as the most potent anti-ezrin compound found in this study, displayed a high level of molecular similarity to NSC305787 with a two fused-ring structure in the center and with a bulky aliphatic group in its north quadrant and hydrophilic groups capable of hydrogen bonding in the south region (Fig. 2A). Our primary screening based on the SPR analysis produced two acridine-containing antimalarial compounds with very close structures, MMV000448 and MMV006172 (Fig. 2A), and none of which was confirmed as a potent ezrin inhibitor through subsequent functional assays (Table 1).

K_D values of antimalarial agents for binding to ezrin were found comparable in the low μmol/L ranges, however, no correlation was found between binding affinity and ability to inhibit ezrin function in different assays (Table 1). It should be kept in mind that binding of these small molecules might occur at multiple sites on ezrin, therefore, a high binding affinity to a particular site might not reflect a strong inhibition of ezrin function. Because ezrin exerts its diverse biologic functions through protein–protein interactions, binding of compounds to different sites on ezrin might lead to distinct local conformational changes, which in turn may result in higher or lower preference of ezrin in binding to different set of proteins. This might also account for the differences in biologic activities of compounds in different functional assays, as there might be potential differences in binding partners of ezrin in mouse and zebrafish. As we tested the binding affinities of our compounds using recombinant human ezrin expressed in E. coli, they may also exhibit differences in binding to the mouse ezrin or the zebrafish homolog of ezrin, ezrin2 (30). In addition, these compounds might differentially affect the phosphorylation status of ezrin particularly on Thr567, Tyr146, Tyr353, and Tyr477 residues, each of which might be associated with specific signaling pathways leading to different cellular responses. Due to the high level of homology existing between ERM family proteins, the observed biologic effects might also partially result from the inhibition of radixin and moesin function. Importantly, for the experiments involving culturing of the cells or tissues, high serum protein binding may also have an impact on efficacy of the compounds, because it can affect the free fraction that is responsible for the biologic action. Finally, the differential response to compounds may reflect their differential permeability and/or uptake across the cell membranes and their intracellular metabolism.

The most notable compounds identified through our screen were MMV667492, MMV020549, and MMV666069, all of which exhibited superior activity over the original lead compound NSC305787 in inhibiting the pulmonary metastatic growth (Figs. 4 and 5). These compounds successfully satisfied all of the Lipinski's and Veber's drug-likeness criteria, whereas NSC305787 failed to meet the criteria of low cLogP (Table 1). Of those compounds, MMV666069 produced subtle but not a strong ezrin MO-like phenotype in zebrafish embryonic developmental assays. Apart from the above-mentioned reasons accounting for the differences in biologic activities of compounds between different assays, this might also be related to the limitations of the assay. Zebrafish embryos develop at least up to 48 hpf within a protective chorion, which is a 1.5- to 2.5-μm thick acellular envelope consisting of three layers; outer, middle, and innermost layers. Thus, chorion may provide a potential barrier for the uptake of some particular chemicals and particulates (40).

The infection of erythrocytes by the malaria parasite Plasmodium falciparum results in extensive remodeling of cytoskeleton and plasma membrane of the host cell by the parasite-coded proteins. Although the function of the majority of the exported proteins of parasite are not known, recent studies demonstrate that some of which interact with the components of host cytoskeleton (41). Parish and colleagues (42) have demonstrated that a member of the P. falciparum PHIST family proteins binds to the erythrocyte cytoskeleton-associated protein Band 4.1, also known as 4.1R. The protein Band 4.1 presents significant structural homology with ezrin, both of which share a N-terminal FERM domain followed by a long α-helical region (43). Recently, it has been demonstrated through phosphoproteome analysis that the malaria parasite P. vivax inhibits erythroid cell growth through downregulating the phosphorylation of ezrin, α-actinin-1, and Rho kinase, which indicates the involvement of ezrin-dependent cellular mechanisms in suppression of erythropoiesis by P. vivax (44). These findings suggest that our MMV400 library hits might exert their antimalarial effects via inhibition of erythrocytic ezrin function required for parasite–cell association but not through a direct inhibitory effect on the parasite growth.

We examined the effects of hits from the MMV400 library, including MMV667492, MMV020549, and MMV666069 and NSC305787 on phosphorylation status of ezrin in K7M2 osteosarcoma cells. Both MMV20549 and MMV666069 significantly increased ezrin T567 phosphorylation, whereas treatment of cells with MMV667492 produced only a moderate increase in phospho-ezrin levels (Supplementary Fig. S6). None of the small molecules altered total ezrin protein levels in the cell. We did not observe any change in phospho-ezrin levels with NSC305787 treatment. Although the current understanding of ezrin function is based on its activation by phosphorylation on T567 residue followed by subsequent membrane localization, our recent findings highlights a novel function of ezrin in regulation of gene translation that is distinct from its known functions at the plasma membrane as a cytoskeletal scaffolding protein (13, 14). On the basis of the aforementioned study demonstrating that the malaria parasite P. vivax proteins inhibit erythroid cell growth by preventing the phosphorylation of ezrin T567 (44), our present findings suggest that the MMV400 library hits might exert their antimalarial effects through increasing phospho-ezrin levels in the cell.

In summary, we have identified novel small molecule inhibitors of ezrin from the MMV Malaria Box that selectively target ezrin function as anti-invasive agents. Although more studies need to be undertaken to determine the mechanism of action of these antimalarial compounds, they hold promise as a novel class of antimetastatic agents for the treatment of metastatic osteosarcoma and raise the possibility that some of the antimalarial compounds might function through regulating ezrin activity in the host cells.

No potential conflicts of interest were disclosed.

Conception and design: H. Çelik, T.Z. Minas, M. Paige, J.A. Toretsky, A. Üren

Development of methodology: H. Çelik, S.-H. Hong, T.Z. Minas, M. Paige, E. Glasgow, J. Bosch, A. Üren

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H. Çelik, S.-H. Hong, D.D. Colón-López, J. Han, M. Swift, E. Glasgow, J. Bosch

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H. Çelik, S.-H. Hong, E. Glasgow, J.A. Toretsky, J. Bosch, A. Üren

Writing, review, and/or revision of the manuscript: H. Çelik, D.D. Colón-López, Y.S. Kont, M. Paige, E. Glasgow, J.A. Toretsky, J. Bosch, A. Üren

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y.S. Kont

Study supervision: H. Çelik, A. Üren

The authors thank Dr. Chand Khanna from the NCI/NIH (MD, USA) for osteosarcoma cell lines and Dr. Abraham T. Kallarakal for help with the Biacore experiments in determination of the binding affinities of compounds.

This work was supported by the U.S. Department of Defense grant W81XWH-10-1-0137 (A. Uren). Additional support was provided by the Johns Hopkins Malaria Research Institute and the Bloomberg Family Foundation (J. Bosch) and the National Science Foundation Graduate Research Fellowship Program under grant DGE-1232825 (D.D. Colón-López). The authors thank the Biacore Molecular Interaction Shared Resource and the Animal Model Shared Resource at the Lombardi Comprehensive Cancer Center (Georgetown University), which are supported by a grant P30 CA51008 from the NCI.

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.