Selective TRK Inhibitor CH7057288 against TRK Fusion-Driven Cancer

The concept of precision medicine is to efficiently treat and prevent diseases by using biomarkers to classify individuals according to their genetic and nongenetic characteristics (1, 2). This can assign the right therapy to the right patient at the right time. The realization of precision medicine satisfies not only medical needs but also social demands, given the recent increase in medical costs. In January 2015, U.S. President Barack Obama announced the Precision Medicine Initiative, which is designed to push forward the implementation of precision medicine by building a massive collection of clinical data and biomarkers. Cancer treatment is one area in which precision medicine is most advanced, because the genetic alterations that drive tumorigenesis and cancer progression are recognized by biomarkers, and appropriate therapy can be selected (3). For example, the anti-HER2 antibody trastuzumab was approved in 1998 with its companion diagnostic to select HER2-positive breast cancer (4), and erlotinib, gefitinib, and other EGFR inhibitors have been used on lung cancer harboring EGFR mutations (2).

One type of genetic alteration that is frequently observed in cancer and is known to drive tumorigenesis is chromosomal rearrangement, as represented by kinase fusions. Some kinase fusions, such as BCR-ABL, EML4-ALK, and ROS1 fusions, have been proved to act as strong oncogenic drivers, and molecular targeted therapies that inhibit these fusion kinases—the ABL inhibitor imatinib, ALK inhibitor alectinib, and ALK/ROS1 inhibitor crizotinib—provide dramatic survival benefits to patients harboring the target alterations (5–9). However, there remain unmet medical needs in patients suffering from cancer with other genetic alterations, including novel kinase fusions.

The tropomyosin receptor kinase (TRK) family of receptor tyrosine kinases consists of TRKA, TRKB, and TRKC, which are coded by neurotrophic tyrosine receptor kinase (NTRK) 1, NTRK2, and NTRK3 genes, respectively (10, 11). NTRK1, which was one of the first oncogenes identified in human cancer, was originally found as a fusion gene with tropomyosin in the 1980s (12, 13). Since then, TRK fusions have been reported in thyroid cancer, and also certain types of rare and/or pediatric cancer (14, 15), including congenital fibrosarcoma, secretory breast carcinoma, and mammary analogue secretory carcinoma (MASC) of salivary glands, were revealed to have TRK fusion in almost all cases. Recent advances in next-generation sequencing technology have uncovered cancer harboring NTRK genes that are fused to a variety of partners across cancer types, including NSCLC and colorectal cancer (14, 16). However, how TRK fusion induces tumorigenesis and cancer progression is largely unknown. Because preclinical analysis shows that cancer cells harboring a TRK fusion have strong dependency on the kinase activity, kinase inhibitors against TRK could be an effective therapeutic approach for patients with TRK fusion–positive cancer. Several TRK inhibitors are currently being evaluated in clinical studies, and entrectinib and larotrectinib have exhibited clinical efficacy against TRK fusion–positive cancer (17–20).

Mutation of a target protein is one of the most frequently observed reasons for acquired resistant mechanisms against molecular targeted therapies (21). For example, in EGFR-mutant non–small cell lung cancer (NSCLC) that progressed after treatment with the first-generation EGFR inhibitors, approximately half the patients acquired the T790M mutation that does not respond to those inhibitors. Recently, it has been reported that TRK inhibitors showed partial responses in patients harboring TRKA and TRKC fusions, but tumors in these patients eventually became resistant to the therapy because of the emergence in TRKA of G595R and G667C mutations and in TRKC of G623R (which corresponds to G595R in TRKA; refs. 18, 22, 23).

In this study, we evaluated the novel pan-TRK inhibitor CH7057288. It selectively inhibits TRKA, TRKB, and TRKC kinase activity and has potent efficacy against TRK fusion–driven cancer cells. CH7057288 has a benzofuran motif that is structurally distinct from other reported TRK inhibitors and is active against a mutation that is resistant to the compounds currently in clinical trial. We also used CH7057288 to study oncogenic signaling pathways for TRK fusions.

Cell lines were obtained from the ATCC, Asterand, Deutsche Sammlung von Mikroorganismen und Zellkulturen (DSMZ), European Collection of Authenticated Cell Cultures (ECACC), Immuno-Biological Laboratories (IBL), National Cancer Institute (NCI), and Japanese Collection of Research Bioresources (JCRB) Cell Bank (see Supplementary Table S1). The NSCLC cell line CUTO-3 (16) and its subclone CUTO-3.29 (19) were kindly provided by Dr. Robert C. Doebele (University of Colorado). The acute myelogenous leukemia cell line MO-91 (24) was a generous gift from Dr. Takeshi Kondo (Hokkaido University). All cell lines were cultured according to suppliers' instructions and confirmed as Mycoplasma negative with culture or PCR methods described elsewhere. Cellular experiments were performed within 20 passages after thawing.

NIH3T3 cells (ATCC) were infected with lentiviruses containing wild-type (WT) and mutant MPRIP–NTRK1 fusion genes that had been prepared using pCDH-CMV-MCS-EF1-Puro (System Biosciences). CUTO-3 cells were infected with lentiviruses containing a firefly luciferase gene (Luc2) that had been prepared using pReceiver-Lv105 (GeneCopoeia). Cells stably expressing the above genes were established through tolerance to puromycin following the lentivirus infection.

CH7057288 (compound 23; ref. 25), entrectinib, and larotrectinib were chemically synthesized.

The activity of TRKA, TRKB, TRKC, INSR, ALK, EGFR, FGFR2, HER2, JAK2, KDR, MET, ROS1, and SRC (Carna Biosciences), RAF1 (Thermo Fisher Scientific), and WT-, G667C-, and G595R-TRKA (SignalChem) towards substrate peptides was determined by homogeneous time-resolved fluorescence assay with LANCE Eu-W1024-labeled Anti-phosphotyrosine (PT66) Antibody (PerkinElmer), according to the standard methods described elsewhere (26). Time-resolved fluorescence was measured with an EnVision HTS microplate reader (PerkinElmer). The activity of AKT1, CDK, CHK1, ERK1, PKA, and PKCa (Carna Biosciences) towards substrate peptides was determined by IMAP FP Screening Express Progressive Binding System (Molecular Devices; ref. 27). Fluorescence polarization was measured with an EnVision HTS microplate reader. Enzyme-inhibitory activity was calculated by the formula (1 − T/C) × 100 (%), where T represents the measured value of the wells with a compound and C represents that of the wells without a compound. Values for 50% inhibitory concentration (IC₅₀) were calculated using Microsoft Office Excel 2013. KINOMEscan analysis was performed by DiscoveRx.

Cells were lysed with Cell Lysis Buffer (Cell Signaling Technology) containing cOmplete and PhosSTOP (Roche Diagnostics). The grafted tumors were homogenized using BioMasher (K.K. Ashisuto) before lysis. The lysates were denatured with Sample Buffer Solution with Reducing Reagent for SDS-PAGE (Nacalai Tesque) or NuPAGE LDS Sample Buffer (Thermo Fisher Scientific) and were then subjected to SDS-PAGE. After SDS-PAGE, Western blotting was performed as described previously (28).

The primary antibodies used against the indicated antigens were: pY674/675-TRKA (4621), pan-TRK (4609), pY783-PLCγ1 (2821), PLCγ1 (5690), pS473-Akt (9271), Akt (9272), pT202/Y204-ERK (9101), ERK (9102), pS780-Rb (8180), E2F1 (3742) and GAPDH (2118; Cell Signaling Technology), Rb (sc-50), CyclinD1 (sc-753), SPRY4 (sc-18607), and MKP3/DUSP6 (sc-137246; Santa Cruz Biotechnology), ETV1/ER81 (ab81086), TRKA (ab76291; Abcam), and Tubulin Alpha (MCA77G; AbD Serotec).

For the spheroid assay, the cell lines were added to the wells of PrimeSurface96U plates (Sumitomo Bakelite) containing serially diluted compound solutions and were incubated at 37°C for 4 or 7 days. To determine the amount of viable cells, we used CellTiter-Glo Luminescent Cell Viability Assay (Promega). Luminescence was measured by EnVision Xcite (PerkinElmer). Antiproliferative activity and IC₅₀ values were calculated in the same way as for the protein kinase assay. For the high-throughput 384-well assay, we used 384-well cell-culture microplates (Greiner Bio One) and performed the assay as above with a 4-day culture period.

All in vivo studies were approved by the Chugai Institutional Animal Care and Use Committee (IACUC). Female SCID mice (C.B-17/Icr-scid/scidJcl) for the MO-91 model or BALB/c nude mice (CAnN.Cg-Foxn1^nu/CrlCrlj) for the other models were obtained from CLEA or Charles River, respectively, and kept under specified pathogen-free conditions. Cells were suspended in 200 μL HBSS or PBS and injected subcutaneously into the right flank of the mice. For the CUTO-3 and MO-91 models, an equal amount of Matrigel (Corning) was mixed with the cell suspension. Tumor size was measured using a gauge twice per week, and tumor volume (TV) was calculated using the following formula: TV = ab²/2, where a is the length of the tumor, and b is the width (29). Once the tumors reached a volume of approximately 200 to 300 mm³, animals were randomized into groups (n = 5 or 7 in each group), and treatment was initiated. Compounds were orally administered once a day. For pharmacokinetic analysis, plasma samples were collected at pre-dose and 0.5, 2, 4, 7, and 24 hours post-dose after multiple (11 days) dosing (n = 2). Resected tumors were lysed for Western blotting or fixed with 10% formalin/PBS overnight and then embedded in paraffin for staining. Sections were stained with the Ki-67 antibody (M7240; Dako) and Vectastain ABC Kit (Vector Laboratories) and with in situ Apoptosis Detection Kit (TaKaRa).

CUTO-3-Luc cells were transplanted intracranially into nude mice as described previously (30). To assess tumor mass, VivoGlo Luciferin (Promega) was injected into the abdominal cavity. After anesthetizing the mice, bioluminescence imaging was taken by NightOWL II LB983 (Berthold Technologies) and the region of interest in the bioluminescence image was quantified by the total photons (ph/s) using IndiGO software (Berthold Technologies). Seventeen days after transplantation, animals were randomized into two groups (n = 7 in each group) based on intracranial bioluminescence, and treatment was initiated. CH7057288 was orally administered once a day for 30 days. Because two animals in the CH7057288 group died while being anesthetized, the remaining 12 animals were used for data analysis. Event-free survival was recorded in days after treatment initiation, and events (judged according to the policy of the IACUC) included death, more than 20% body weight loss, and a motility defect. The log-rank test was used to compare Kaplan–Meier curves.

Protein crystallography was performed by Proteros Biostructures. The kinase domain of human TRKA was purified using affinity and gel filtration chromatography. Crystals grew at 4°C from a 0.1:0.1 μL mixture of the protein (10 mg/mL in 20 mmol/L Tris (pH 8.5), 50 mmol/L NaCl, 0.5 mmol/L PMSF, 2 mmol/L DTT) including 2 mmol/L ligand with reservoir solution (0.05–0.15 mol/L KH₂PO₄, 0.1 mol/L MES (pH 6.25), 1.5–2.5 mol/L NaCl, 0.05–0.15 mol/L NaH₂PO₄, 0.01 mol/L TCEP HCl). Diffraction data were collected at 100 K in the beamline CMCF-08ID at the CANADIAN LIGHT SOURCE with a Rayonix 300 detector (Rayonix). The data set was processed with XDS and scaled with XSCALE (31) in the space group H 3 2. The structure of the TRKA and CH7057288 complex was solved by molecular replacement. The model was rebuilt with Coot (32) and refined with REFMAC5 (33) to a final resolution of 2.76 Å. TLS refinement was used to improve maps and models. The final model consisted of residues 489–792 with a break (residues 685–686 was disordered). For crystallographic data and refinement statistics, see Supplementary Table S2.

CUTO-3 and KM12-Luc (JCRB) cell lines were treated with DMSO or 1 μmol/L CH7057288 for 4 and 24 hours. Cellular RNA was extracted using miRNeasy Mini Kit (QIAGEN). RNA samples were prepared for sequencing using TruSeq RNA Sample Prep Kit v2 (Illumina) to generate an mRNA library, and 100 bases were sequenced from both ends of the library using HiSeq 2500 Sequencing System (Illumina). RSEM software was used to align reads against RefSeq transcripts and calculate expression values (FPKM) for each gene (34). Fold changes in FPKM plus one were calculated to find upregulated genes (more than three-fold expression) and downregulated genes (less than 0.333-fold expression) compared to DMSO-treated cells in each of the two cell lines. A list of commonly upregulated or downregulated genes in the two cell lines was identified as the TRK fusion–regulated gene set (Supplementary Table S3). The upstream regulators of the gene set were generated through the use of IPA (Qiagen; https://www.qiagenbioinformatics.com/products/ingenuity-pathway-analysis; ref. 35).

RNA of the grafted tumors was extracted using miRNeasy Mini Kit and QIAshredder (Qiagen). Expression levels of the 770 cancer-related genes were measured using nCounter PanCancer Pathways Panel and nCounter Analysis System (NanoString Technologies).

In a kinase inhibitor screening, we identified the benzofuran derivative CH7057288 as a TRK inhibitor belonging to a novel chemical class (Supplementary Fig. S1). Our biochemical analysis of its enzyme inhibitory activity against TRK family members showed that CH7057288 potently inhibits TRKA, TRKB, and TRKC at nanomolar IC₅₀ (Table 1). To evaluate its kinase selectivity, we screened tyrosine and serine/threonine kinases and found that the compound did not inhibit eleven tyrosine and seven serine/threonine kinases as strongly as it inhibited the TRKs. To evaluate the selectivity further, we used the KINOMEscan panel (DiscoverX Corporation), which is a competitive binding assay using an ATP analogue for 403 nonmutant and 65 mutant kinases. The analysis revealed that at 100 nmol/L CH7057288 bound to only 6 kinases other than TRKA, TRKB, and TRKC (Supplementary Fig. S2). Therefore, CH7057288 is a potent and selective pan-TRK inhibitor.

Then, we investigated the compound's cellular inhibition of TRK activity using three TRK fusion–positive cancer cell lines. The NSCLC cell line CUTO-3, CRC cell line KM12-Luc, and acute myeloid leukemia cell line MO-91 harbor MPRIP-NTRK1, TPM3-NTRK1, and ETV6-NTRK3, respectively (16, 36, 37). First, we analyzed phosphorylation of TRK and possible downstream factors by Western blotting. CH7057288 potently inhibited autophosphorylation of TRK in a dose-dependent manner (Fig. 1A). As for downstream signaling of TRK fusion, we investigated phosphorylation of the PLCγ1, MAPK, and Akt pathways, since these three pathways are known to be involved in TRK signaling (11). The compound suppressed phosphorylation of PLCγ1 and ERK, although suppression levels varied somewhat in different cell lines, and Akt phosphorylation was slightly inhibited in CUTO-3 but strongly suppressed in KM12-Luc and MO-91. These observations indicate that CH7057288 has inhibitory activity to TRK and blocks TRK fusion–mediated signaling in cells.

We next confirmed the effects on TRK-dependent cell growth. Consistent with its inhibition of TRK phosphorylation, CH7057288 showed single-digit to double-digit nanomolar IC₅₀ on cell growth in TRK fusion–positive cell lines (Fig. 1B). In contrast, the TRK-negative NSCLC and CRC cell lines, NCI-H1299 and HCT-116, were insensitive to the TRK inhibitor. Then, we used high-throughput 384-well assay systems to expand the cell growth inhibition assay into a large-scale screening panel consisting of various types of cancer cell lines. As shown in Fig. 1C and Supplementary Table S1, TRK fusion–positive cell lines (black bars) were drastically enriched on the CH7057288-sensitive side in 272 cell lines. Notably, four of the most sensitive six cell lines, MO-91, KM12, KM12-Luc, and CUTO-3.29, have a TRK fusion. However, most of the remaining cell lines showed micromolar IC₅₀. This TRK fusion–selective activity of CH7057288 likely resulted from the high selectivity demonstrated in our kinase panels, mentioned above.

When we evaluated the in vivo potency using mouse xenograft models, CH7057288 induced potent tumor growth inhibition against all three models, with remarkable tumor regression in CUTO-3 and MO-91 (Fig. 2A). Importantly, no significant body weight loss was observed after treatment with CH7057288 in these experiments. In the KM12-Luc model, body weight steadily decreased in the vehicle-treated group but recovered in the CH7057288-treated groups accompanied by tumor growth inhibition. Moreover, when we treated the CUTO-3 model for a longer term, tumor regression was durable for at least 4 weeks (Fig. 2B).

Then, we investigated the pharmacokinetics in plasma and pharmacodynamic responses in tumors. CH7057288 exhibited dose-dependent exposure (Supplementary Fig. S3). Because of relatively short terminal half-life (3–5 hours), the plasma concentration 24 hours after dose dropped to approximately a few tenths to a hundredth of T_max. The plasma concentration at 4 hours postdose in 1 mg/kg–treated animals was sub-micromolar range which can strongly suppress TRK in Fig. 1A. Consistently, Western blot analysis showed substantial suppression of TRK and PLC phosphorylation in KM12-Luc tumors 4 hours after single administration (Fig. 2C). Moreover, staining the sections taken from CUTO-3 xenograft tumors with Ki-67 (a proliferation marker) and TUNEL (an apoptosis marker) revealed that Ki-67 decreased and TUNEL increased after the treatment (Supplementary Fig. S4). Taken together, CH7057288 inhibited in vivo TRK activity and induced both of cell-cycle arrest and apoptosis.

In addition to the subcutaneous implantation models, we tested CH7057288 in an intracranial implantation model of CUTO-3-Luc that mimics metastasis to the central nervous system (CNS). As shown in Fig. 2D, luminescence was reduced after CH7057288 treatment but was enhanced in the vehicle control group, indicating regression of intracranial tumors by the compound. When event-free survival was assessed, survival was significantly prolonged by 30-day CH7057288 treatment compared with vehicle treatment (Fig. 2E). The CH7057288-treated animals survived throughout the treatment without events, in contrast to the control mice, of which more than half were removed from the experiment by Day 26. Mean survival of the treated group reached 67 days after treatment initiation. These tests collectively show that CH7057288 demonstrated potent in vivo antitumor activity, with reasonable pharmacodynamic response in subcutaneous xenograft models and prolonged event-free survival in an intracranial model.

We then investigated the activity against specific mutations in TRKA; namely G595R and G667C that are resistant to entrectinib and larotrectinib. First, in biochemical kinase assays we confirmed that, consistent with public information, the inhibitory activity of entrectinib and larotrectinib was lower against the two mutants than against the WT protein (Fig. 3A), and that of CH7057288 was also lower against the G595R mutant. In contrast, CH7057288 inhibited G667C-TRKA at a comparable IC₅₀ to its inhibition of WT-TRKA. In the cellular analysis, the IC₅₀ values for entrectinib and larotrectinib against NIH3T3 cells expressing G667C-TRKA fused to MPRIP were more than 10- and 400-fold higher, respectively, than against those expressing MPRIP-WT-TRKA (Fig. 3B). As in the biochemical assay, CH7057288 potency against cells transfected with the G667C-TRKA fusion was similar to its potency against WT-TRKA cells. Suppression of TRK phosphorylation in these cells (Fig. 3C) was also consistent with the results shown in Fig. 3A and B. In the in vivo studies, the degree of tumor regression induced by CH7057288 against WT- and G667C-TRKA fusion transfectants was similar (Fig. 3D), whereas G595R tumors were resistant to all compounds.

To understand the molecular recognition of CH7057288 against WT- and mutant TRKA, we determined the X-ray crystal structure of the kinase domain of the WT-TRKA protein in complex with CH7057288 (Fig. 4A, PDB ID: 5WR7). In the X-ray structure, we noticed that a hydrogen bond is formed between Met592 in the hinge region and carbonyl oxygen of CH7057288 and that another hydrogen bond occurs between pyridine nitrogen and Lys544. In complex, the structure takes the DFG-out/αC-helix out conformation, which allows the ligand to interact with the DFG motif (Phe669) and the back pocket. This structural information can explain why CH7057288 showed equal or slightly stronger activity to the G667C mutant. When the Gly667 residue is substituted with cysteine, Cys667 locates in the optimal position to form a sulfur-π interaction that increases the affinity to the ligand (Fig. 4B). In contrast, bulky arginine would create steric repulsion in the G595R-TRKA and CH7057288 complex (Fig. 4C).

Next, we built 3D models of TRKA in complex with entrectinib or larotrectinib by referring to the DFG-in conformation of TRKA (Fig. 4D, PDB ID: 4AOJ). We chose this conformation because entrectinib is reported to bind with ALK in the DFG-in conformation (PDB ID: 5FTO), and since no crystal structural information regarding larotrectinib has been reported so far, we speculated that, given its similarity to other inhibitors such as entrectinib, it most likely also binds in the DFG-in conformation. In the model for entrectinib and larotrectinib, substituting Gly667 with cysteine would most likely generate steric repulsion with the aromatic rings and thus reduce the binding affinity. The effect of G595R substitution is the same as it was in the complex with CH7057288. Taken together, these structural insights support our understanding of the preferential sensitivity of the G667C mutant to CH7057288, in comparison with entrectinib and larotrectinib.

Although the findings shown in Figure 1A confirmed the previously reported activation of PLCγ1, ERK, and Akt by TRK fusion (11), the detailed downstream pathways of TRK fusion remained to be elucidated. Therefore, we analyzed the mRNA expression of cells treated with CH7057288 for 4 and 24 hours (GEO accession number: GSE119900). Expression of 28 (4 hours) or 148 (24 hours) genes in CUTO-3 and 127 (4 hours) or 770 (24 hours) genes in KM12-Luc cells was three times greater than in DMSO-treated cells. We picked out 8 (4 hours) or 45 (24 hours) genes common in the two cell lines (Supplementary Table S3) and analyzed them by IPA software (Qiagen) to identify the upstream regulators (Supplementary Table S4). Consistent with the findings in Fig. 1A, the results showed significant MEK inhibition both at 4 and 24 hours. E2F was inhibited and the E2F family repressor, E2F6 (38), was activated with the most significant P-values at 24 hours but not at 4 hours. p53 was also inhibited at 24 hours, and it is well known that E2F and p53 closely cooperate in cell cycle and apoptosis. These data indicated that TRK inhibition by CH7057288 suppressed the MAPK pathway first, resulting in abrogation of the E2F pathway. IPA suggested ERBB2 inhibition that was likely because the E2F target genes and ERBB2 target genes largely overlapped in the dataset. Then, we examined protein expression and phosphorylation of E2F and MAPK pathway–related genes in CH7057288-treated CUTO-3 cells (Fig. 5A). In concordance with the IPA analysis, expression of CyclinD1 and E2F1 and phosphorylation of Rb was suppressed along with TRK inhibition. In addition to the decrease in ERK phosphorylation observed in Fig. 1A, MAPK target genes DUPS6, SPRY4, and ETV1 (39) were downregulated by CH7057288 treatment. Moreover, in in vivo tumors treated with CH7057288, gene expression changes in E2F and MAPK pathway genes were consistent with in vitro results (Fig. 5B). These results suggest that the E2F and MAPK pathways are important downstream pathways of TRK fusion.

The TRK family receptor tyrosine kinases respond to the neurotrophin family ligands that comprise nerve growth factor (NGF), brain-derived neurotrophic factor (BDNF), neurotrophin (NT) 3, and NT4 (11). The NGF–TRK axis plays various physiologic roles in neural functions and pathologic roles in neurologic diseases and cancer. Consistent with their roles in neural development and functions, isoform-selective activation and inactivation of TRK in neurogenic neoplasms, such as neuroblastoma and medulloblastoma, have been reported (10). The NTRK1 gene was originally identified as a fusion gene from a CRC patient in the 1980s (13). Massive parallel sequencing based on tremendous technologic advances revealed that there are patients harboring TRK fusions across a wide variety of cancer types at low but detectable frequencies (14, 40). Cases of TRKA, TRKB, or TRKC fusion have been found in approximately 1% of NSCLC, CRC, papillary thyroid carcinoma, glioblastoma multiforme, and head and neck squamous cell carcinoma. The TRK fusion is present in particular segments of glioma and melanocytic neoplasms with high frequency, 40% of non-brainstem high-grade glioma in infants, and 16% of Spitz tumor/spitzoid melanoma (41, 42). Recurrent TRK fusion in pediatric and young adult cancer patients might reflect its aggressive oncogenic activity. In this study, we identified CH7057288 as a novel TRK inhibitor that shows potent and selective antitumor activity against cancer driven by TRK fusions. The compound has the potential to be a therapeutic agent to meet medical needs in cancer patients harboring TRK fusions.

After treatment with entrectinib and larotrectinib failed in the clinical trials, sequencing analysis of the tumor samples identified G667C and G595R mutations in TRKA and a G623R mutation in TRKC. As shown in Fig. 4, CH7057288 seems to be capable of interacting to the Cys667 residue in the G667C-TRKA protein in the optimal configuration to create a sulfur-π interaction. In contrast, entrectinib and larotrectinib suffer from steric hindrances to this G667C mutation. This speculation is consistent with the observation that CH7057288 inhibited G667C-TRKA equally or slightly more strongly than it inhibited WT-TRKA. The G667C mutation in TRKA corresponds to G1269A in a clinically reported ALK mutation resistant to crizotinib (43). In addition to G667C, the G667A mutation in TRKA could happen by a single nucleotide substitution. However, alanine substitution would result in less steric repulsion to entrectinib than cysteine substitution, so that it is reasonable that the G667C, not G667A, mutation was observed in the entrectinib-resistant patient. Although entrectinib showed less activity to G667C-TRKA than WT, this mutant would not be a substantial problem due to high exposure of entrectinib in the clinical setting. The ratio of the G667C allele in the colorectal cancer patient was significantly lower than that of G595R, which is the more robustly resistant mutant (18). In our experiment, entrectinib induced antitumor activity to G667C-TRKA–expressing tumors. Just recently, a TRK inhibitor, LOXO-195, has been reported to show clinical efficacy to several patients harboring G595R-TRKA or G623R-TRKC as acquired resistant to larotrectinib (23). This compound is almost equally effective to the G595R mutant but less effective to the G667C mutant, and minimum plasma concentration in patients was lower than IC₅₀ value of G667C-TRKA. Therefore, clinical efficacy of LOXO-195 against tumors with the G667C mutation might not be sufficient compared to that against tumors with the WT- or G595R-TRKA fusion.

CNS metastasis is one of the most lethal complications in patients with cancer. It occurs in approximately 30% to 40% of lung and breast cancer during the course of disease development (44, 45). In terms of therapy response, the blood–brain barrier (BBB) would limit the efficacy of some anticancer drugs to CNS metastasis. ALK fusion–driven NSCLC initially responds to the first-generation ALK inhibitor crizotinib, but eventually disease progresses and patients often develop CNS metastases (8, 46). Alectinib, a second-generation ALK inhibitor active in an intracranial tumor model (30), showed superior antitumor response than crizotinib against patients harboring brain metastasis (5, 7). This evidence led us to evaluate the efficacy of CH7057288 to CNS lesions. CH7057288 achieved tumor regression against intracranial lesions and prolonged event-free survival, suggesting the possibility that our compound may benefit patients suffering from CNS metastasis. CH7052788 is a potential substrate of P-glycoprotein (P-gp) based on the result of the Caco-2 permeability assay (Supplementary Table S5), which is thought to be one of substantial components of BBB. However, generally speaking, P-gp substrates penetrate BBB to some extent, and there are multiple reports that BBB integrity is reduced during formation of brain metastasis (47). Although both of MEK inhibitor trametinib and BRAF inhibitor dabrafenib are P-gp substrates (48), combination of them showed efficacy against CNS lesions comparable to non-CNS lesions in melanoma patients harboring BRAF mutations (49).

MAPK, PI3K, and PLCγ pathways have been reported to take roles in downstream signal transduction of TRK in physiological conditions (11). In TRK fusion–positive cells, these pathways have been shown to be activated by TRK inhibitors (16). However, there is no report of a genome-wide analysis of gene expression in TRK fusion–driven cancer cells. Consistent with previous reports, CH7057288 attenuated phosphorylation of ERK, Akt, and PLCγ in all three TRK fusion–positive cell lines that we tested (Fig. 1A). Inhibition of Akt phosphorylation in CUTO-3 cells was marginal, as was reported in larotrectinib-treated CUTO-3.29 (19). Moreover, we examined how CH7057288 affects gene expression in TRK fusion–driven cells in the whole transcriptome. As a result, we found that CH7057288 suppressed the E2F pathway most significantly at 24 hours, and the MAPK pathway was also strongly inhibited from the earlier time point than E2F. It is well known that Cyclin D–CDK4/6 complexes phosphorylate Rb to activate E2F transcription factors leading to G₁–S transition in the cell cycle. Cyclin D1 is one of the target genes in ERK-mediated transcriptional activation (39). Our mRNA- and protein-level analysis showed reduced expression of Cyclin D1 and E2F and reduced phosphorylation of Rb. Moreover, cell-cycle arrest was observed in CH7057288-treated tumors. Because the tested two cell lines have the p53 mutations that are R280I in CUTO-3 (Supplementary Fig. S5) and H179R in KM12-Luc (50), the observed cell-cycle arrest was a p53-independent manner. Collectively, in cancer cells harboring TRK fusion, activation of the MAPK-Cyclin D-E2F axis accelerates cell-cycle progression. Inhibition of TRK kinase activity that shuts the axis down resulted in remarkable antitumor effects.

Hiroshi Tanaka has ownership interest (including stock, patents, etc.) in Chugai Pharmaceutical Co., Ltd., Bristol-Myers Squibb Company, and Gilead Sciences, Inc. No potential conflicts of interest were disclosed by the other authors.

Conception and design: H. Tanaka, H. Sase, M. Yoshida, N. Oikawa

Development of methodology: M. Hasegawa, H. Tanimura, K. Sakata, T. Fujii, Y. Tachibana

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H. Tanaka, H. Sase, T. Tsukaguchi, M. Hasegawa, K. Sakata, Y. Tachibana, K. Takanashi, K. Hasegawa

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H. Tanaka, H. Sase, M. Hasegawa, A. Higashida, K. Hasegawa, N. Oikawa

Writing, review, and/or revision of the manuscript: H. Tanaka, H. Sase, M. Hasegawa, N. Oikawa, T. Mio

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H. Tanaka

Study supervision: H. Tanaka

Other (chemistry expertise): Y. Ono, T. Mio

We thank Tomoaki Hayashi, Maiko Izawa, Yumiko Hashimoto, Yasue Nagata, Yusuke Ide, Moon Hyung-Jo, and Lee Dong-Hee for in vitro experiments and Waka Nakai, Kenji Nakagawa, Yuuji Sakurai, and Shino Kuramoto for in vivo experiments. We are grateful to Hideaki Mizuno, Junichi Muroya, Takaaki Fukami, Takaaki Miura, and Sally Matsuura for their help in RNA-Seq, X-ray crystal structure, and English writing. This study was funded by Chugai Pharmaceutical Co., Ltd.

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.