Cabozantinib Overcomes Crizotinib Resistance in ROS1 Fusion–Positive Cancer

An increasing number of genetic alterations that aberrantly activate tyrosine kinases have been identified as oncogenic drivers of non–small cell lung cancer (NSCLC). Active mutations of epidermal growth factor receptor (EGFR), such as the L858R point mutation or deletion/insertion of several amino acids between exons 19 and 20, are more commonly observed in patients with NSCLC. The active mutation of KRAS is also predominantly found in patients with NSCLC. In addition to these active oncogene mutations, chromosomal rearrangements involving the tyrosine kinase domains of ALK, ROS1, and RET are observed in 1% to 5% of patients with NSCLC (1). The oncogenic fusion protein in NSCLC can be targeted by tyrosine kinase inhibitors (TKI), such as crizotinib; therefore, a number of specific TKIs targeting the fusion tyrosine kinase are currently under development. Although EGFR inhibitors (e.g., gefitinib or erlotinib) or the ALK inhibitor crizotinib show remarkable efficacy in most cases, the majority of patients will develop tumors resistant to targeted therapies in less than 1 year of treatment (2, 3). In cancers harboring the ALK fusion protein, several mechanisms of crizotinib resistance have been reported, including acquired secondary mutations in the kinase domain of ALK, genomic amplification of the ALK fusion gene, and amplification or activation of other kinases (3–7).

Recently, crizotinib was shown to be an effective inhibitor of ROS1 tyrosine kinase, and two case reports have described the activity of crizotinib in patients with ROS1-rearranged lung cancers (8, 9). Although crizotinib exhibited activity in a patient with NSCLC harboring the ROS1 fusion, a resistant tumor eventually emerged. Recently, the G2032R mutation in the ROS1 kinase domain was identified in a crizotinib-treated resistant tumor, which was not observed before treatment (10). The mutation was located in the solvent-front region of the ROS1 kinase domain and was analogous to the G1202R ALK mutation identified in crizotinib-resistant ALK-rearranged lung cancers. We previously reported that the ALK G1202R mutation confers high-level resistance to crizotinib compared with all next-generation ALK inhibitors that were examined (3). Therefore, it is important to identify novel compounds that can overcome the G2032R ROS1 mutation, which confers crizotinib resistance in these cancers.

In this study, we tested several ALK inhibitors to examine the potency of the sterically distinct ALK inhibitors, because the kinase domains of ALK and ROS1 are highly similar and grouped in the same kinase family (11). Subsequently, we identified a number of different crizotinib and/or ceritinib resistant mutations including G2032R mutation in the ROS1 kinase domain by N-ethyl-N-nitrosourea (ENU)–driven accelerated mutagenesis screening. High-throughput drug screening identified several kinase inhibitors as a potent ROS1 inhibitor, and identified that the cMET/RET/vascular endothelial growth factor (VEGFR) inhibitor cabozantinib can potently inhibit both wild-type (WT) and the resistant mutant CD74–ROS1. On the basis of these results, we propose the use of several inhibitors as alternative therapeutic strategies for ROS1-rearranged cancers and cabozantinib as a key drug for overcoming crizotinib resistance in ROS1 fusion–positive cancer cells lines, particularly those mediated by secondary mutations.

Crizotinib was obtained from ShangHai Biochempartner; alectinib, cabozantinib, and ceritinib (LDK378) were purchased from ActiveBiochem; NVP-TAE-684 and ASP3026 were purchased from ChemieTek; AP26113 was purchased from Selleck; and foretinib was purchased from AdooQ BioScience. Each compound was dissolved in dimethyl sulfoxide (DMSO) for cell culture experiments. For inhibitor screening, the SCADS Inhibitor Kit was provided by the Screening Committee of Anticancer Drugs supported by a Grant-in-Aid for Scientific Research on Innovative Areas, Scientific Support Programs for Cancer Research, from the Ministry of Education, Culture, Sports, Science, and Technology of Japan.

Genomic DNA was isolated from cell pellets after proteinase K treatment. The ROS1 kinase domain was amplified by polymerase chain reaction (PCR) from the genomic DNA and sequenced bidirectionally using Sanger sequencing.

Human embryonic kidney 293FT cells (Invitrogen) were cultured in Dulbecco's Modified Eagle Medium supplemented with 10% fetal bovine serum (FBS; D-10). Ba/F3 cells, which are immortalized murine bone marrow–derived pro-B cells, were cultured in D-10 media with or without 0.5 ng/mL of interleukin (IL)-3 (Invitrogen). Crizotinib-resistant ROS1 fusion–positive NSCLC patient-derived MGH047 cells were cultured in ACL-4 medium supplemented with 3% FBS (10).

To assess 72-hour drug treatment, 2,000 to 3,000 cells were plated in replicates of three to six in 96-well plates. Following drug treatments, the cells were incubated with the CellTiter-Glo Assay reagent (Promega) for 10 minutes. Luminescence was measured using a Centro LB 960 microplate luminometer (Berthold Technologies). The data were graphically displayed using GraphPad Prism version 5.0 (GraphPad Software). IC₅₀ values were determined using a nonlinear regression model with a sigmoidal dose response in GraphPad.

Lysates were prepared as previously described (3, 12). Equal volumes of lysate were electrophoresed and immunoblotted with antibodies against phospho-ROS1 (Tyr2274), ROS1 (69D6), phospho-p42/44 ERK/MAPK (Thr202/Tyr204), p42/44 ERK/MAPK, phospho-Akt (Ser473; D9E), panAkt (C67E7), phospho-S6 ribosomal protein (Ser240/244, D68F8), S6 ribosomal protein (54D2), STAT3 (79D7), phospho-STAT3 (Tyr705; Cell Signaling Technology), GAPDH (6C5, Millipore), and β-actin (Sigma).

cDNA encoding WT or mutant CD74–ROS1 was cloned into 1,520 retroviral expression vectors (pLenti), and viruses were replicated in 293FT cells by transfecting with packaging plasmids. After retroviral infection, Ba/F3 cells were selected by incubation with puromycin (0.7 μg/mL) for 2 weeks. For Ba/F3 cells infected by CD74–ROS1 variants, IL3 was withdrawn from the culture medium at least 2 weeks before the experiments.

The ENU mutagenesis screening protocol was based on procedures published by Bradeen and colleagues (13) and O'Hare and colleagues (14). Briefly, ENU (Sigma) was dissolved in DMSO at a concentration of 100 mg/mL. All materials that came in contact with ENU were decontaminated with 0.2 mol/L NaOH. For each resistance screen, approximately 1.5 × 10⁸ Ba/F3 CD74–ROS1 cells in a total of 160 mL of growth media were exposed to a final concentration of 100 μg/mL of ENU. After approximately 16 hours, the cells were collected by centrifugation, washed, and incubated for 24 hours. After a 24-hour recovery period, the cells were split into five aliquots of 3 × 10⁷ cells each. Crizotinib or ceritinib was added at a final concentration of 30, 50, 100, or 200 nmol/L, and cells of each aliquot were distributed into five 96-well plates (5 × 10⁴ cells in 200-μL media per well). Plates were incubated over a course of 4 weeks with regular inspection. When clear signs of cell growth were microscopically observed and a color change of the media occurred, the content of the respective well was transferred into 1 mL of growth media containing the original concentration of inhibitors in a 24-well plate. After approximately 1 week of expansion, the cell number was sufficient for further processing (see below).

Genomic DNA was prepared by lysing the cells with proteinase K buffer, which was heat inactivated at 95°C for 5 minutes. Then, the temperature was gradually decreased by 2°C/min. For sequence analysis, a DNA fragment covering the entire kinase domain of ROS1 was amplified using KOD Plus (TOYOBO). The PCR products were then purified with a gel purification kit (GE healthcare) and sequenced using standard Sanger sequencing.

Inhibitor screening was conducted using a subset of the modified SCADS library containing 282 compounds in three 96-well microplates. Parental, CD74–ROS1 WT-, or CD74–ROS1–G2032R–expressing Ba/F3 cells were seeded in triplicates in 96-well plates on day 1, and each inhibitor was added at 10 nmol/L, 100 nmol/L, 1 μmol/L, and 3 μmol/L on the same day. Cell viability was determined on day 4 using the CellTiter-Glo Assay. The cell viability from triplicate plates was averaged to determine relative cell growth compared with that of DMSO-treated controls.

All data are presented as mean ± SD. Statistical analysis was performed using the two-tailed Student t test. Significance was established for P values < 0.05.

The tyrosine kinase domains of ALK and ROS1 shared 70% identity, and both kinases belong to the same branch in a kinase phylogenic tree (11). To identify inhibitors capable of inhibiting the kinase activities of the ROS1 fusion protein, we tested the potency of various ALK inhibitors to CD74–ROS1. First, we established IL3 independently growing Ba/F3 cells by transformation with CD74–ROS1, which is the most frequently observed ROS1 fusion gene in NSCLC. From the polyclonal CD74–ROS1–addicted Ba/F3 cells, we picked up the clone with a high expression of CD74–ROS1 and similar crizotinib sensitivity to the polyclonal cells (clone #6; Supplementary Fig. S1), which was propagated to examine the sensitivity to various ALK inhibitors currently being clinically evaluated. Our results showed that crizotinib, ceritinib (LDK378), and AP26113 exhibited remarkable growth suppression of CD74–ROS1 Ba/F3 cells, ASP3026 showed moderate inhibitory activity, and alectinib (CH5424802) showed none (Fig. 1A and B). Corresponding to the cell growth-inhibiting activity, crizotinib, ceritinib, AP26113, and ASP3026, but not alectinib, inhibited phospho-ROS1 and its downstream phospho-STAT3 in a dose-dependent manner (Fig. 1C). Among these compounds, crizotinib and ceritinib are clinically available for ALK fusion–positive NSCLC. Furthermore, ceritinib, AP26113, and ASP3026 were shown to be active against the ALK gatekeeper mutation (L1196M), which is most frequently observed in crizotinib-resistant ALK-rearranged NSCLC (15). Therefore, we decided to identify potential resistance mechanisms to crizotinib or ceritinib in CD74–ROS1 mediated by a resistance mutation in the ROS1 kinase domain.

To identify ROS1 mutations responsible for resistance to crizotinib or ceritinib in ROS1 fusion–positive cancers, we performed random mutagenesis screening by exposing the CD74–ROS1 Ba/F3 cells to the alkylating agent ENU, followed by selection using various concentration of crizotinib or ceritinib. After culturing with the inhibitors for 3 to 4 weeks, we observed an inhibitor dose-dependent reduction in the number of wells with growing cells. The resistant cells were recovered and the ROS1 kinase domains were sequenced. The resistant clones were selected by treatment with 200 nmol/L of crizotinib or ceritinib, and all carried the G2032R mutation (Fig. 2A). Of the clones selected with 100 nmol/L crizotinib, one clone harbored the K2003I mutation in the CD74–ROS1 and a second clone harbored no mutation. After expanding the isolated clone from ENU mutagenesis screening harboring K2003I mutated CD74–ROS1, we tested the sensitivity to crizotinib and ceritinib. We found that K2003I-mutated ROS1 did not confer resistance to crizotinib or ceritinib. On the other hand, G2032R-mutated CD74–ROS1 conferred high resistance to both crizotinib and ceritinib (Fig. 2B and C). When the cells were selected using a lower concentration of crizotinib (50 nmol/L) or ceritinib (100 nmol/L), various mutations in the clones were identified (Supplementary Fig. S2). Next, we tested the isolated clones from ENU mutagenesis for crizotinib or ceritinib sensitivity. The recovered Ba/F3 cells harboring the mutations E1990G with M2128V, L1951R, G2032R, or L2026M with K2003I in ROS1 showed IC₅₀ values against crizotinib that were more than 3-fold higher than that of WT CD74–ROS1–expressing BaF3 cells (Supplementary Fig. S3A and S3B). On the other hand, Ba/F3 CD74–ROS1 cells harboring the L2026M mutation, which is a gatekeeper mutation corresponding to L1196M in ALK, were sensitive to ceritinib. Likewise, the mutations E1990G with M2128V, L1951R, or G2032R conferred resistance to ceritinib. In particular, the CD74–ROS1–expressing Ba/F3 cells harboring the G2032R mutation were extremely resistant to both crizotinib and ceritinib. Then, we conducted immunoblot analysis of the recovered Ba/F3 cells by treating the cells with various concentrations of crizotinib or ceritinib. The results were consistent with those of the cell viability assay, in which phosphorylation of CD74–ROS1 harboring the G2032R mutation was not completely attenuated even following treatment with 1 μmol/L of crizotinib or ceritinib (Fig. 2B). The G2032R mutation was recently identified in a patient with crizotinib refractory CD74–ROS1 fusion-positive NSCLC (10). In contrast, cells carrying the L1951R and E1990G with M2128V mutations exhibited resistance to crizotinib or ceritinib, consistent with the results of the cell viability assay. Cells harboring the L2026M/K2003I double mutant exhibited resistance to crizotinib but not to ceritinib (Supplementary Fig. S3C).

To confirm whether these mutations confer resistance to crizotinib and ceritinib, we introduced each mutated CD74–ROS1 in Ba/F3 cells. All of the resistant mutated CD74–ROS1 (L1951R, L1982F, E1990G, F1994L, K2003I, L2026M, and G2032R) maintained transforming activity. Then, we tested the sensitivity of Ba/F3 cells expressing CD74–ROS1 mutants to crizotinib or ceritinib. Similar to the result of the recovered Ba/F3 cells from ENU mutagenesis screening, L1951R and G2032R mutated CD74–ROS1–induced Ba/F3 cells showed marked resistance to crizotinib, ceritinib, and AP26113. L2026M mutant–induced Ba/F3 cells were resistant to crizotinib but not to ceritinib or AP26113. L1982F, E1990G, or F1994L mutants showed slight resistance to crizotinib and ceritinib. (Fig. 3A–C and Supplementary Fig. S4A–S4C).

Next, we mapped these mutations on the crystal structure data of crizotinib: ROS1 to elucidate the location of mutations that confer resistance (Fig. 3D and Supplementary Fig. S5). L1951 and G2032 mutations were located in the solvent-front region (entrance of the crizotinib-binding pocket), and L2026, which correspond to the L1196 mutation in ALK, is a gatekeeper mutation of ROS1. All of the identified mutations that conferred higher crizotinib resistance were located close to the crizotinib-binding domain of the ROS1 kinase (Fig. 3D).

To identify potent ROS1 kinase inhibitors that selectively suppress the growth of Ba/F3 cells expressing either WT or the crizotinib-resistance mutant CD74–ROS1, we performed cell-based high-throughput screening with a series of kinase inhibitors and anticancer agents used in clinical practice or under current clinical evaluation. IL3-independent Ba/F3 cells expressing either WT or G2032R-mutated CD74–ROS1 were treated for 72 hours with serial dilutions of 282 kinase inhibitors and anticancer drugs in the SCAD inhibitor library. Potential ROS1 kinase inhibitors were selected for further evaluation using the following criteria: selective growth-inhibitory effect (<40% cell viability) against WT or G2032R-mutated Ba/F3 CD74–ROS1 cells at an inhibitor concentration of ≤100 nmol/L and ≥10-fold lower IC₅₀ value compared with that for Ba/F3 parental cells. Using this assay, we newly demonstrated that cabozantinib (XL184), foretinib, TAE684, SB218078, and CEP701, in addition to the ALK inhibitors under clinical evaluation or in clinic, are potent inhibitors of CD74–ROS1 Ba/F3 cell growth (Table 1; Fig. 4A and B; Supplementary Table S1). Furthermore, among these inhibitors, cabozantinib (XL184), foretinib, and TAE684 effectively inhibited the growth of both WT and G2032R-mutated CD74–ROS1 Ba/F3 cells, and the autophosphorylation of both WT and CD74–ROS1 (Fig. 4B). Of note, CEP701 showed intermediate selectivity to the growth of CD74–ROS1 Ba/F3 cells, and CEP701 only inhibited the autophosphorylation of WT CD74–ROS1 but not the autophosphorylation of G2032R-mutated CD74–ROS1. And to inhibit the phospho-ROS1 of G2032R-mutated CD74–ROS1, higher concentration of TAE684, foretinib, or cabozantinib, compared with that for CD74–ROS1 (WT)–expressing Ba/F3 cells was needed (Fig. 4A and B).

To examine the effect of cabozantinib on cells harboring these mutations, each of the Ba/F3 cells harboring the various CD74–ROS1 mutations (both ENU recovered Ba/F3 clones and CD74–ROS1 mutants–transformed Ba/F3 cells) were treated with cabozantinib, and the cell growth and phosphorylation of ROS1 were examined. The results showed that cabozantinib dose-dependently inhibited phospho-ROS1 in all crizotinib-resistant mutant strains and inhibited the growth of all Ba/F3 cells harboring crizotinib-resistant CD74–ROS1 mutations. IC₅₀ values of all crizotinib-resistant mutants against cabozantinib were less than 25 nmol/L, although the IC₅₀ values of crizotinib-resistant mutant (G2032R and L1951R) Ba/F3 cells were approximately 5- to 10-fold higher than that of WT CD74–ROS1 harboring Ba/F3 cells (Fig. 5 and Supplementary Figs. S6 and S7).

In our previous study of clinical crizotinib resistance in ROS1-rearranged NSCLC, we established the MGH047 cell line harboring the CD74–ROS1–G2032R mutation directly isolated from the pleural effusion of a crizotinib-resistant patient. Using this cell line, we compared the activities of cabozantinib and crizotinib and found that crizotinib did not inhibit the growth of MGH047 cells harboring the G2032R mutation, whereas cabozantinib potently inhibited the growth of MGH047 cells (Fig. 6A). Furthermore, as exhibited by the Ba/F3 cell line models, cabozantinib effectively suppressed phospho-ROS1 and downstream phospho-Akt, phospho-ERK, and phospho-ribosomal S6 proteins in MGH047 cells (Fig. 6B). These results suggest that cabozantinib presents an alternative therapeutic strategy to treat ROS1-rearranged NSCLC in both crizotinib-naïve patients and resistant cases caused by resistance mutations in the kinase domain.

Recently, the cMET/ALK/ROS1 inhibitor crizotinib has been clinically evaluated for treatment of ROS1-rearranged NSCLC and has shown remarkable activity (8, 9). Because of the similarity between ROS1 and ALK kinase domains, we examined the sensitivity of various ALK inhibitors on CD74–ROS1 fusion and found that all the tested ALK inhibitors, except for alectinib, effectively inhibited ROS1 fusion. Among those ALK inhibitors, ceritinib was recently approved by the U.S. FDA for ALK-positive patients with crizotinib-resistant or crizotinib-intolerant disease, because high response rate in crizotinib-resistant disease was observed in phase I study (16). Ceritinib has also been shown to be active in both WT and gatekeeper-mutated ALK (L1196M), which causes crizotinib resistance (17). In EGFR mutant–positive lung cancers that become resistant to EGFR TKIs, the secondary T790M gatekeeper mutation is detected in roughly one-half of all cases (2). In contrast, in crizotinib-resistant ALK-positive lung cancers, many types of resistance mutations in the ALK kinase domain were identified in various cell lines as well as crizotinib-resistant patients (3). Because the tyrosine kinase domains of ALK and ROS1 share approximately 70% homology, it is possible that many kinds of crizotinib-resistance mutations will also occur in ROS1-rearranged NSCLC.

To prospectively identify resistance mutations affecting the ALK/ROS1 inhibitors crizotinib and ceritinib, we performed a cellular drug resistance screen in CD74–ROS1–transformed Ba/F3 cells and identified several resistant mutations, including G2032R, as the most pronounced mutations that confer crizotinib resistance. So far, only the G2032R mutation in ROS1 has been identified in crizotinib-treated resistant patients with ROS1-rearranged NSCLC (10). In this study, the newly identified resistance mutations of L1982F, E1990G, F1994L, and L2026M were less frequent and conveyed milder resistance to crizotinib. In addition, a screen with the structurally distinct ALK/ROS1 inhibitor ceritinib revealed a slightly different mutation profile; however, the most pronounced resistant mutations were L1951R and G2032R. In addition to ENU-induced accelerated mutagenesis screening, we also performed saturated mutagenesis screening (18–20) and identified different unique mutations (E2020K and P2021L), but no G2032R mutation was observed (data not shown). Similarly, a previous study using the same methods to identify crizotinib-resistance mutations in EML4–ALK identified unique mutations but did not recapitulate the clinically relevant mutations. These results suggest that induced mutation profiles in mismatch repair-deficient Escherichia coli strains might be slightly different from plausible mutations in mammalian cells. However, the mutagenesis screening study with imatinib in Ba/F3 cells harboring BCR–ABL fusion showed that the saturated screening assay is useful to identify various resistance mutations including those with clinical relevance (18).

The mutations identified from ENU-accelerated mutagenesis screening can be categorized into three types. The first type includes solvent-front mutations (e.g., L1951R and G2032R), which are located in the solvent-front region of the kinase domain adjacent to the crizotinib-binding site. An amino acid change of the conserved glycine to arginine at position 2032 or leucine to arginine at position 1951 of the ROS1 kinase domain confers considerable resistance to multiple ALK/ROS1 kinase inhibitors, such as crizotinib, ceritinib, and AP26113. The G2032R ROS1 mutation is analogous to the G1202R ALK mutation, which has been identified in ALK-rearranged lung cancers that have become resistant to crizotinib, alectinib, and ceritinib (3, 17). It is likely that these solvent-front mutations decrease the affinity of the mutant ROS1 for crizotinib because of steric hindrance (10).

The second type includes the gatekeeper mutation L2026M, which is equivalent to gatekeeper mutations observed in EGFR (T790M), ALK fusion (L1196M), and BCR–ABL (T315I). The third type is characterized by helix αC (L1982F or V), which is a homologous residue of L1152 in ALK, previously identified in patients with crizotinib-treated ALK-positive NSCLC (6).

To overcome resistance to crizotinib and ceritinib caused by the G2032R ROS1 mutation, we performed high-throughput drug screening, which subsequently identified cabozantinib (XL-184) as a potent ROS1 inhibitor that effectively inhibited both the WT CD74–ROS1 kinase as well as those harboring resistance mutations including G2032R. Furthermore, cabozantinib effectively inhibited the growth of the crizotinib-resistant patient-derived MGH047 cells harboring G2032R-mutated CD74–ROS1. Cabozantinib is a small molecule that inhibits the activity of multiple tyrosine kinases, including RET, MET, and VEGFR2. Currently, cabozantinib is clinically available for treatment of refractory medullary thyroid cancer. Data from previous clinical trials revealed a peak plasma concentration of cabozantinib after repeated oral administration (175 mg) of around 1,410 ng/mL (2810 nmol/L). Even in the much lower dose (0.64 mg/kg, which is corresponding to about 40-mg oral administration) treated patient, an average peak plasma concentration of cabozantinib after repeated oral administration was around 322 ng/mL (643 nmol/L; ref. 21). On the basis of the data from our study, we found that cabozantinib at concentrations less than 30 nmol/L inhibited all of the identified crizotinib-resistance mutations, which was much lower than clinically achievable levels. During the preparation of this manuscript, Davare and colleagues (22) identified that foretinib, which is an oral multikinase inhibitor targeting MET, VEGFR-2, RON, KIT, and AXL kinases and currently being clinically evaluated, is a potent inhibitor against ROS1 and overcomes resistance mutations including G2032R. Although we confirmed that foretinib also inhibited WT and all mutated crizotinib-resistance ROS1 fusions, our results suggest that cabozantinib is slightly more potent than foretinib. Furthermore, previously reported mean plasma concentrations of foretinib in two clinical trials were 72 and 340 nmol/L (23, 24). Although it is impossible to simply compare the plasma concentrations and expected efficacy in humans, cabozantinib is likely to be more potent and effective than foretinib.

In conclusion, our study clearly demonstrated that patients with crizotinib-resistant cancers due to an acquired mutation, such as G2032R, may benefit from more potent and effective ROS1 TKI. Notably, although solvent-front mutations are occasionally observed in patients with crizotinib-resistant ALK fusion-positive NSCLC, the frequency of G2032R mutations in ROS1-positive NSCLC has yet to be established. Because secondary mutations, such as the gatekeeper mutation, may not represent the predominant mechanism of acquired crizotinib resistance, additional studies are needed to elucidate other mechanisms of resistance. The results of these studies will be critical to selecting the best therapeutic strategies for targeting TKI resistance in clinical practice. Although, crizotinib is currently a key agent used to treat cancers harboring ROS1 translocations, cabozantinib may be able to prevent or overcome resistance to ROS1 inhibitors.

A.T. Shaw is a consultant/advisory board member for Genentech, Ignyta, Novartis, and Pfizer. J.A. Engelman reports receiving a commercial research grant from AstraZeneca, Novartis, and Sanofi-Aventis, and is a consultant/advisory board member for Novartis and Sanofi-Aventis. No potential conflicts of interest were disclosed by the other authors.

Conception and design: R. Katayama, J.A. Engelman, N. Fujita

Development of methodology: R. Katayama, Y. Kobayashi

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R. Katayama, Y. Kobayashi, L. Friboulet, E.L. Lockerman, S. Koike, A.T. Shaw

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): R. Katayama, Y. Kobayashi, S. Koike

Writing, review, and/or revision of the manuscript: R. Katayama, A.T. Shaw, J.A. Engelman, N. Fujita

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y. Kobayashi, S. Koike, N. Fujita

Study supervision: R. Katayama, J.A. Engelman, N. Fujita

The authors thank Dr. Mark M. Awad at Massachusetts General Hospital (MGH; Boston, MA) for helping with the establishment of MGH047 cells and Ms. Sidra Mahmood at MGH for helping with the cell survival assay experiments of MGH047 cells.

The study was supported, in part, by Japan Society for the Promotion of Science (JSPS) KAKENHI grant numbers 24300344 and 22112008 (to N. Fujita) and 25710015 (to R. Katayama), by NIH grant R01CA164273 (to A.T. Shaw and J.A. Engelman), and by a research grant from the Princess Takamatsu Cancer Research Fund (to N. Fujita).

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.