M-COPA, a Golgi Disruptor, Inhibits Cell Surface Expression of MET Protein and Exhibits Antitumor Activity against MET-Addicted Gastric Cancers

Gastric cancer (GC) is the third leading cause of cancer-related death worldwide, and its incidence remains high especially in east Asia, including Japan and Korea (1). GC is often diagnosed at an advanced stage, and the prognosis of such patients is poor. Owing to its diversity of morphologic forms and considerable heterogeneity, the classification of GC has been complicated. Recently, The Cancer Genome Atlas (TCGA) Research Network proposed a new molecular classification of GC into four subtypes (2). Among these, approximately 50% of gastric tumors were categorized as the chromosomal unstable subtype, containing frequent gene amplification of receptor tyrosine kinases (RTK), such as human epidermal growth factor receptor 2 (HER2), MET (also called “hepatocyte growth factor receptor”) and fibroblast growth factor receptor 2 (FGFR2), which have been shown to drive cells to malignant proliferation and survival (3, 4).

Thanks to recent progress in understanding molecular pathways underlying carcinogenesis, new targeted treatment options have become available for treating cancer patients. In respect to the development of targeted therapies, monoclonal antibodies (mAb) and small-molecule inhibitors of tyrosine kinase (TKI) activity are ideal candidates that target tumor cells via binding to RTKs. For example, trastuzumab is an mAb developed for treating HER2-positive breast cancer; gefitinib is a small-molecule TKI-targeting EGFR developed for treating lung cancer. So far, six mAbs and 23 TKIs have been developed as approved drugs for treating cancer; however, trastuzumab and ramucirumab are the only two molecularly targeted drugs approved to date for treating GC patients. Therefore, the development of new targeted drugs against GCs is of keen interest. Because gene amplification of MET, HER2, and FGFR2 is often observed in GCs, many TKIs targeting these RTKs have been developed and several clinical trials in GC patients are in progress.

The Golgi apparatus plays an essential role in the transport, processing, and sorting of numerous proteins (5–8). Most cell-surface and secreted proteins in eukaryotic cells pass through the Golgi apparatus, allowing for posttranslational modification such as processing and glycosylation, and subsequently transport to plasma membrane (9–11). Aberration of Golgi function is associated with certain forms of inherited diseases, cancer, and diabetes (12). ADP ribosylation factor 1 (Arf1), a small GTPase and a member of Ras superfamily (13, 14), is required for maintenance of Golgi structure and function via formation of complex I (COPI) or clathrin-coated vesicles transported among the endoplasmic reticulum (ER), Golgi, and post-Golgi (15–19). We previously demonstrated that M-COPA (2-methylcoprophilinamide, also called "AMF-26") suppressed Arf1-mediated vesicle transport, disrupted the structure of the Golgi apparatus, and exerted antitumor activity in vivo against tumors xenografted into mice (20). In fact, M-COPA has been shown to inhibit secretion of intercellular adhesion molecule-1 (ICAM-1) and cell surface expression of VEGFR-2 (21). Therefore, we postulated that M-COPA could inhibit the processing and transport of RTKs to the cell surface and extracellular space, and thereby exert antitumor activity against RTK-addicted cancers.

In this study, using GC cell lines with or without MET amplification, we demonstrated that M-COPA inhibited the processing and transport of MET protein onto the cell surface, attenuated aberrant MET signaling and exerted preferential antitumor activity against MET-addicted GCs. Moreover, we obtained similar results in FGFR2-amplified signet ring GC cells KATO III. The present results suggest that a Golgi-targeted drug could be a novel therapeutic modality against MET-addicted GCs, as well as perhaps RTK-addicted cancers from different origins.

M-COPA (also called “AMF-26,” chemical name: (2E,4E)-5-[(1S,2S,4aR,6R,7S,8S,8aS)-7-hydroxy-2,6,8-trimethyl-1,2,4a,5,6,7,8,8a-octahydronaphthalen-1-yl)-2-methyl-N-(pyridin-3-ylmethyl)penta-2,4-dienamide] was totally synthesized by Eisai Co., Ltd. according to the methods established previously (22). 5-FU and paclitaxel were purchased from Sigma-Aldrich Co. LLC., crizotinib was purchased from Selleck Chemicals. For in vitro studies, these compounds were reconstituted to 20 mmol/L in DMSO (Sigma-Aldrich) and stored at −20°C. SN-38 was purchased from Sigma-Aldrich, and cisplatin was purchased from Nippon Kayaku Co., Ltd. For animal experiments, M-COPA was suspended in 0.15N hydrochloride acid (Wako Pure Chemical Industries). Antibodies used for immunostaining are listed in Supplementary Table S1.

Four GC cell lines (St-4, MKN1, MKN45, and MKN74) are components of the JFCR39 panel of human cancer cell lines described previously (23, 24). Of these, St-4 was established in our foundation in 1990 (25). MKN1, MKN45, and MKN74 were purchased from Immuno-Biological Laboratories Co., Ltd. in 1991. Other two GC cell lines, Hs-746T, and SNU-5 were purchased from the ATCC in 2014. Cell lines were cultured in RPMI-1640 medium (Wako Pure Chemical Industries) supplemented with 5% (v/v) FBS (Moregate Biotech), penicillin (100 U/mL), streptomycin (100 μg/mL), and kanamycin (1 μg/mL) in a humidified atmosphere including 5% CO₂ at 37°C. Authentication of St-4, MKN1, MKN45, and MKN74 cell lines was performed by short tandem repeat (STR) analysis using PowerPlex16 Systems (Promega) according to the manufacturer's instructions by BEX CO., LTD., in 2009 and 2016. Details were described in Supplemental Materials and Methods. Finally, the STR profiles of MKN1, MKN45 and MKN74 were compared with those in the reference database of the Japanese Collection of Research Bioresources Cell Bank. Because St-4 was developed in our foundation and no reference data was deposited, we compared the profile determined in 2016 to that in 2009. The KATO III cell line was originally purchased from the ATCC in 2000. Authentication of the cell line was performed in 2016 and compared the profile with that in the ATCC STR database.

The inhibition of cell growth was assessed by measuring changes in total cellular protein in a culture of each GC cell line after 48 hours of drug treatment by use of a sulforhodamine B assay (26). Positive values represent net protein increase before and after drug exposure (% of control growth) and negative values represent cell death [protein amount after 48 hours-exposure (%) of control cells at the start of drug exposure]. The drug concentration required for 50% reduction in net protein increase (GI₅₀) was calculated as described previously (23, 27, 28).

Cells were incubated with M-COPA for 6 or 24 hours, and then washed with ice-cold PBS and stained with antibodies against human MET or human FGFR2 conjugated with phycoerythrin (PE). Then cells were washed three times with ice-cold PBS, and stained with propidium iodide (1 μg/mL; Sigma-Aldrich). The fluorescence intensity of cell surface MET or FGFR2 was measured by flow cytometric analysis (FACS Calibur or FACS Verse, Becton, Dickinson and Company). The data were analyzed by using FlowJo (FlowJo LLC.).

Cells were incubated with M-COPA for 1, 6, or 24 hours, and then lysed as described previously (29). Proteins in cell lysates were separated in 4% to 15% sodium dodecyl sulfate-polyacrylamide gel (Bio-Rad Laboratories) electrophoresis, followed by electroblotting onto a nitrocellulose membrane (Bio-Rad Laboratories). Immunoreactive bands were identified with an ODYSSEY CLx Infrared Imaging System (LI-COR Biosciences).

The antitumor effect of M-COPA was tested in vivo against MKN45-derived human GC xenografts in mice. Animal care and treatment was performed in accordance with the guidelines of the Animal Use and Care Committee of the Japanese Foundation for Cancer Research, and conformed to the NIH Guide for the Care and Use of Laboratory Animals. Female nude mice of BALB/c genetic background were purchased from Charles River Laboratories JAPAN, Inc., maintained under specific pathogen-free conditions and provided with sterile food and water ad libitum. Each nude mouse was subcutaneously inoculated with a generated tumor fragment of size 3 mm × 3 mm × 3 mm. When the tumors reached a volume of 100 to 300 mm³, animals were randomly divided into control and M-COPA groups (each group containing five or six mice). Then administration of M-COPA was started (day 0). The experimental group of mice was orally administered a given dose of M-COPA (50 mg/kg of BW) on a daily basis from day 0 to 4 (n = 6), or weekly (75 mg/kg of BW) on day 0, 7, and 14 (n = 5). The control group of mice (n = 6) was orally administered with 0.15N hydrochloride acid solution instead of M-COPA. The tumor volume of tumor-bearing mice was measured as described previously (29). To assess toxicity, the body weights of the tumor-bearing mice were measured.

Formalin-fixed, paraffin-embedded tissue sections (4-μm-thick) were deparaffinized in xylene and taken through a series of graded alcohols to water. Then antigens were retrieved through wet autoclave pretreatment (20 min at 121°C) in 10 mmol/L citrate buffer (pH 6.0). Sections were blocked with 3% H₂O₂ and 1% goat serum before incubation with the primary antibody at 4°C overnight. By using Dako EnVision Detection System/HRP, Rabbit/Mouse (DAB+; Agilent Technologies Company), sections were incubated with horseradish peroxidase (HRP)–conjugated polymer secondary antibody, and thereafter peroxidase activity was visualized by DAB reaction according to the manufacturer's instructions. The sections were counterstained with hematoxylin (Dako). The immuno-stained specimens were imaged using a microscope BX41 (Olympus Corp.) with a 20x, NA 0.50, objective, and DP2-BSW Software (Olympus Corp.).

The two-sided Mann–Whitney U test was used to assess the statistical significance of the antitumor efficacy of M-COPA in terms of relative tumor growth ratio and body weight change on days 2, 4, 8, 11, 15 and 18. The number of samples is indicated in the description of each experiment. All statistical tests were two-sided.

First, we examined the effect of M-COPA on cell surface expression of MET protein in GC cell lines. To this end, we exploited three GC cell lines with or without MET amplification. As shown in Fig. 1A, the MET-amplified GC cell lines, MKN45, Hs-746T, and SNU-5, overexpressed total MET protein and its phosphorylated active form, as previously reported (30). The St-4 cell line also expressed total and phosphorylated form of MET, even though the intensities of those expression were weaker than those of MET-amplified GC cells. We also examined baseline expression of EGFR, HER2, HER3, and FGFR2, which have been reported to be amplified in GC cells (2). None of the phosphorylated forms of these proteins except HER3 in MKN45 were detected in any of the GC cell lines examined, although total EGFR and HER3 were detected in some cell lines.

We then examined cell surface expression of MET and the effect of M-COPA treatment in these GC cell lines. As shown in Fig. 1B, baseline MET expression on the cell surface was higher in MET-amplified GC cell lines (MKN45, Hs-746T, SNU-5) than that observed in other GC cell lines, in parallel with the levels of total MET protein determined by immunoblot analysis (Fig. 1A). Upon treatment with M-COPA for 24 hours, cell surface expression of MET efficiently decreased in a dose-dependent manner; it was significantly reduced at concentrations of 100 nmol/L or higher in SNU-5 and MKN45 cells, whereas ≥300 nmol/L was needed to mediate such reduction in Hs-746T cells. Next, we examined the expression profile of MET protein over time following treatment of MKN45 cells with M-COPA. After 6 hours treatment, MET protein expression was reduced to approximately 25% (1/4) at concentrations of 100 nmol/L or higher and was dramatically declined to negative control levels (IgG-isotype control antibody) within 24 hours at concentrations of 300 nmol/L or higher (Fig. 1C). From these data, we concluded that M-COPA efficiently inhibited cell surface expression of MET protein in MET-amplified GC cells, in a dose- and time-dependent manner.

We next evaluated the effect of M-COPA on the growth of GC cell lines with or without MET amplification. Expectedly, all three MET-amplified cell lines were highly sensitive to the MET inhibitor crizotinib (Fig. 2B), as compared with other GC cell lines, indicating that their growth was addicted to MET, in agreement with previous reports (31). Interestingly, MET-amplified cell lines also exhibited a better drug response to M-COPA than those without MET amplification, albeit the selectivity was not as marked as that observed with crizotinib (Fig. 2A). Moreover, the concentrations of M-COPA needed to inhibit cell growth was comparable with those needed to inhibit cell surface expression of MET in the MET-amplified cell lines; the GI₅₀ concentration for the MKN45, Hs-746T, and SNU-5 cell lines was 29, 40, and 19 nmol/L, respectively. These data indicated that M-COPA preferentially inhibited the growth of MET-addicted GC cells and the growth inhibition coincided with decreased expression of MET protein on the cell surface. For comparison, the anticancer effects of chemotherapeutic agents used in the clinic for GC patients, namely 5-FU, paclitaxel, SN-38, and cisplatin, were examined. However, neither 5-FU, paclitaxel nor cisplatin exhibited significant differences in terms of growth inhibitory activities in the GC cell lines with MET amplification with the only exception that SNU-5 tended to be highly sensitive to these agents. On the other hand, SN-38 exhibited preferential activity in these MET-amplified cell lines in a similar vein to M-COPA (Fig. 2C–F).

MET is a transmembrane heterodimer that composed of two disulfide-linked chains of 50 kDa α-subunit and 145 kDa β-subunit (32). The molecule is originally synthesized as a single-chain 170-kDa precursor (Pr170), which is cotranslationally glycosylated. Terminal glycosylation and proteolytic cleavage generate the mature heterodimer (33). To clarify whether M-COPA could disturb processing of MET, expression levels of the precursor form and a mature β-subunit of MET were estimated in MKN45 cells by Western blot analysis. As shown in Fig. 3A, the amount of the Pr170 was increased upon treatment with M-COPA in a dose-dependent manner at 6 hours, whereas the active β-subunit and phosphorylated form of MET were decreased. Finally, the active phosphorylated form of MET were dramatically reduced at concentrations of 100 nmol/L or higher in 24 hours, in parallel to inhibition of cell surface expression of MET (Fig. 1C) and cell growth (Fig. 2A). These results indicated that the processing and the transport of MET protein onto the cell surface were prevented by M-COPA treatment.

We next examined the effect of M-COPA on the activation status of the downstream signaling pathway from MET. M-COPA decreased the levels of phosphorylated Gab2, Akt, and S6 within the same drug concentration range and time-course as those required for inhibition of MET processing and transport (Fig. 3B). The effectiveness of M-COPA was also demonstrated in other two MET-addicted GC cell lines, Hs-746T and SNU-5 (Supplementary Fig. S1A). M-COPA also inhibited the cell surface expression of MET in St-4 cells (Fig. 1B). Therefore, we investigated the activation status of MET and its downstream signaling pathway molecules. As shown in Supplementary Fig. S1B, M-COPA repressed the MET processing, MET-activation, and phosphorylation of Akt. However, phosphorylation of S6 ribosomal protein was still remained by M-COPA treatment, and concordantly St-4 cell line showed lower sensitivity to crizotinib and M-COPA (Fig. 2A). From these data, we concluded that M-COPA inhibited the MET-dependent signaling pathway via inhibition of MET processing and its cell surface expression as mediated by the Golgi apparatus, resulted in attenuating the abnormal cell proliferation in MET-amplified GC cell lines.

To examine the effect of other RTKs amplified in GCs, we exploited KATO III, which was known as a FGFR2-amplified signet ring cell GC cell line that exhibited a high sensitivity to FGFR2-TKI (34, 35). As we expected, M-COPA caused downregulation of cell surface expression, electrophoretic mobility shift probably due to inhibition of post-translational modification, and dephosphorylation of FGFR2 protein, and finally decreased phosphorylated form of downstream signaling molecules such as Akt and S6 (Fig. 4A and B). These effects were observed in the similar concentration range to that required for inhibition of cell growth (Fig. 4C). These results suggested that M-COPA suppressed cell surface expression of FGFR2 as a result of Golgi dysfunction, and thereby exerted antitumor effect in FGFR2-amplified cells, as well as MET-amplified cells.

Finally, we tested the antitumor efficacy of M-COPA against tumor xenografts derived from the MET-amplified GC cell line, MKN45, in vivo. M-COPA was orally administrated daily for the first 5 days, or weekly for 3 weeks. In both administration groups, treatment with M-COPA significantly decreased the tumor size than that observed in the control group (Fig. 5A, top). Slight but significant weight loss was observed in the daily administration group on days 2 and 4; however, after stopping administration, no significant difference was observed among the three groups on day 8 and later (Fig. 5A, bottom).

To validate the proof of concept that M-COPA administration exerts an antitumor effect via inhibition of cell surface expression of MET, we examined the expression of phosphorylated MET in tumor tissue sections by immunohistochemistry. The immunostaining intensities of phosphorylated MET on the cell membrane were markedly decreased in the M-COPA–treated group on day 2 compared with those in the control group (Fig. 5B and C). These data strongly suggested that M-COPA exerted an in vivo antitumor effect on MET-amplified MKN45-derived xenografts via downregulation of cell surface expression of active MET.

As mentioned, MET protein is produced as a 170 kDa single-chain precursor form, and the precursor is then posttranslationally glycosylated, cleaved to yield a mature form consisted of an α-chain (50 kDa) and a β-chain (145 kDa) via the Golgi apparatus (32, 33, 36). In MET-amplified cells, overexpression of MET protein triggers its autophosphorylation and recruitment of its effectors, resulting in hyperactivation of intracellular downstream signaling pathways, including the PI3K–AKT axis (37). This abnormal signal activation is known to drive cells to malignant proliferation, thereby entering a “MET-addicted” state (38, 39). In the present study, we clearly demonstrated that M-COPA decreased the level of the active β-subunit form and increased the precursor form in a dose-dependent manner, consistent with a previous report showing that a Golgi inhibitor brefeldin A (BFA) abrogated the processing of nascent MET protein (36). Moreover, we found for the first time that inhibition of MET processing by M-COPA coincided with downregulation of its cell surface expression and abrogation of its downstream oncogenic signals represented by reduction of phosphorylated forms of Gab, Akt, and S6, and ultimately suppressed tumor growth in MET-addicted GC cell lines. Downregulation of cell surface expression of activated MET protein in parallel to the observed antitumor effects was also confirmed in the case of MKN45-derived tumor xenografts after administration of M-COPA in vivo. These results strongly suggested that inhibition of processing and cell surface expression of MET protein is the main mechanism by which M-COPA exerts antitumor effects against MET-addicted GC cells. In addition to MET, phosphorylated HER3 was also detected in MKN45 cells (Fig. 1A), as previously reported (40). MET has been shown to interact with HER3 and activate HER3 signal in gastric and lung cancer cells (40, 41). In our study, M-COPA also repressed phosphorylation and cell surface expression of HER3 (Supplementary Fig. S2), as well as MET, suggesting that repression of HER3 signal could also be involved in antitumor effect of M-COPA in MKN45 cells.

We also demonstrated that cell surface expression of FGFR2 was also downregulated by M-COPA treatment in FGFR2-addicted KATO III, a human signet ring cell GC cell line (Fig. 4). FGFR2 is known as a target for cancer therapy, and now several FGFR2-TKIs, such as AZD4547, are in clinical study stage. M-COPA also repressed the maturation of FGFR2, phosphorylation of its signaling pathway molecules, and inhibited cell growth under the concentration of submicromole order. These data strongly suggested that M-COPA could exert antitumor effect in cancer cells addicted to not only MET but also FGFR2 and other RTKs, including HER2, mutant EGFR, and so on.

St-4 is a cell line that expresses small amount of MET protein (Fig. 1A). This cell line is not addicted to MET expression and it did not respond to crizotinib (Fig. 2B). Our results indicated that M-COPA abolished cell surface expression (Fig. 1B) and processing of MET, but it hardly affected phosphorylation level of S6 ribosomal protein (Supplementary Fig. S1B) and did not exhibit high sensitivity to M-COPA (Fig. 2A). In contrast, St-4 expresses substantial amount of EGFR at the cell surface, but M-COPA hardly attenuated cell surface expression of EGFR (Supplementary Fig. S3). The precise mechanism by which these differences in M-COPA response were achieved remains unclear. Interestingly, phosphorylated form of MET was detected but that of EGFR was hardly detected in the St-4 cell line before drug exposure (Fig. 1A), suggesting that cell surface expression of RTKs in its activated form could be selectively abolished by M-COPA treatment. Involvement of phosphorylated status of RTKs in M-COPA efficacy is under investigation.

Besides RTKs, we examined the effect of M-COPA on ABC transporters such as P-glycoprotein and BCRP. Although M-COPA reduced cell surface expression of BCRP in BCRP-expressing breast cancer HBC-5 cells, it did not affect cell surface expression of P-glycoprotein in adriamycin-resistant AD10 cells derived from human ovarian cancer A2780 cells (Supplementary Fig. S4). We have not yet determined the mechanism by which these differences caused, but we supposed that the differences in coated vesicles (e.g., COPI-coated or clathrin-coated vesicles) may cause different M-COPA response. Further studies are needed to clarify the mechanism by which M-COPA attenuates protein expression at the cell surface.

There are three approaches to downregulation of HGF/MET signaling in human clinical studies: anti-HGF mAbs, anti-MET mAbs, and small-molecule MET TKIs (42). Class I TKIs, such as crizotinib, bind to the MET ATP–binding pocket and show specificity against MET and some other tyrosine kinases, including ALK (43). Crizotinib seemed to exert better antitumor activities to MET-amplified GC xenografts (44), in other words, M-COPA response was modest as compared with crizotinib (Fig. 5). However, crizotinib usage has been reported to induce a second mutation at the gatekeeper position of the ATP-binding pocket of the targeted kinases, resulting in acquired resistance (45, 46). Class II TKIs such as cabozantinib, show broad specificity as compared with class I inhibitors, binding to a region past the gatekeeper position and occupying a hydrophobic pocket at a deeper location (47). The present results imply that Golgi-targeted drugs such as M-COPA could be a novel therapeutic option in addition to mAbs and TKIs for treating GCs addicted to MET or FGFR2 via inhibition of the processing and the transport of MET/FGFR2 onto the cell surface. Moreover, this class of drug could also be useful for treating cancers from different tissues of origin whose growth is dependent on RTK expression, especially those harboring a secondary mutation exhibiting acquired resistance to pretreated TKIs. In this situation, the Golgi-targeted drugs are also expected to overcome the TKI resistance by inhibiting cell surface expression of the mutated RTKs. The effect of M-COPA on other RTK-addicted cancer cells (e.g., EGFR-mutated lung cancer) and TKI-resistant cells is under investigation.

We demonstrated that M-COPA exhibited higher sensitivity to MET-addicted cell lines than MET non-addicted cell lines (Fig. 2A); however, the selectivity for MET-addicted cells was not as marked as that seen with crizotinib. In other words, M-COPA did partially interfere with the growth of MET non-addicted GC cells. Although we demonstrated that neither EGFR, HER2, HER3, nor FGFR2 were activated in these cell lines, involvement of other RTKs or other cell surface proteins should be considered. BFA, another Golgi disruptor, is known to trigger ER stress and unfolded protein response (48), and similar efficacy of M-COPA is expected. Involvement of this pathway in antitumor effect of M-COPA in both MET-addicted and non-addicted cancer cells is under investigation. On the other hand, a selection of chemotherapeutic drugs, apart from SN-38, did not display any evidence of MET status-specific sensitivity in the panel of MET amplified and unamplified cell lines tested (Fig. 2C–F). The reason why SN-38 was more sensitive toward MET-addicted cell lines than nonaddicted cell lines remains unclear.

In M-COPA–treated MET-amplified cell lines, processing of MET protein was suppressed, but the loss of mature MET β-chain was not accompanied by a corresponding increase of MET precursor, especially in SNU-5 cells. At present, we have not yet elucidated the precise mechanism by which this occurred, and the fate of MET protein in M-COPA treated cells remains unclear. Unfolded protein response is known to attenuate translation initiation via phosphorylation of eIF2 alpha by PERK (48). Therefore, one possibility is that attenuation of total MET protein might be occurred as a consequence of general translation suppression.

We assessed M-COPA–induced toxicity by measuring body weight loss. As described before, slight weight loss was observed in the daily administration group, whereas weight loss was alleviated after stopping administration. Concurrently, loss of weight was observed in tumor bearing control mice, as well as those weekly administered, in accordance with the previous report that inoculation of MKN45 cells into nude mice caused cachexia and body weight loss (49), which resulted in no significant difference among the three groups on day 8 and later.

In conclusion, we demonstrated that M-COPA inhibited the processing and the transport of MET protein onto the cell surface, attenuated aberrant MET signaling, and exerted a preferential antitumor activity of M-COPA against MET-addicted GCs. The present results suggested that a Golgi-targeted drug could be a novel therapeutic modality that has a unique mode of action for targeting RTK, in addition to mAb and TKI therapies.

No potential conflicts of interest were disclosed.

Conception and design: Y. Ohashi, S. Dan

Development of methodology: Y. Ohashi, M. Okamura, I. Shiina

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Ohashi, M. Okamura, A. Hirosawa, N. Tamaki

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y. Ohashi, M. Okamura

Writing, review, and/or revision of the manuscript: Y. Ohashi, H.-W. Choi, I. Shiina, S. Dan

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. Akatsuka, H.-W. Choi, I. Shiina

Study supervision: K. Yoshimatsu, T. Yamori, S. Dan

Other (compound synthesis): K.-M. Wu

The authors thank Kanami Yamazaki and Yumiko Nishimura for their technical assistance.

This work is supported by Adaptable & Seamless Technology Transfer Program through Target-driven R&D (A-STEP; AS2614144Q) in 2014 from the Japan Science and Technology Agency (JST) and in 2015 from the Japan Agency for Medical Research and Development (AMED), a grant from the Vehicle Racing Commemorative Foundation, and a grant from National Cancer Center Research Development Fund (#26-A-5).

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.