Epithelial-to-Mesenchymal Transition Is a Mechanism of ALK Inhibitor Resistance in Lung Cancer Independent of ALK Mutation Status

ALK rearrangement, most commonly EML4-ALK, is detected in approximately 3%–5% of unselected non–small cell lung cancer (NSCLC; refs. 1, 2). Crizotinib, a first-generation ALK tyrosine kinase inhibitor (TKI), shows dramatic clinical efficacy against ALK-rearranged NSCLC. Recent clinical trials demonstrated the response rate of approximately 60% and progression-free survival (PFS) of approximately 12 months (3–5). However, almost all patients who strongly responded to crizotinib acquire resistance over time. The most defined mechanism of crizotinib resistance is secondary or acquired ALK mutations. Unlike the EGFR-TKI resistance in EGFR-mutated NSCLC, where one secondary resistance mutation (EGFR^T790M) is detected in approximately 60% of resistant patients (6), diverse ALK resistance mutations (e.g., L1196M, I1171T/N/S, L1152P/R, F1174C/L/V, C1156Y/T, I1171T/N/S, S1206C/Y, G1269A/S, V1180L, and G1202R) can be detected in about 20%–25% of resistant patients. The second-generation ALK-TKIs, alectinib (7–10) and ceritinib (11), which have different spectrum of sensitivity to these ALK resistance mutations, have been approved to treat crizotinib-resistant ALK-rearranged NSCLC. In addition, other mechanisms of ALK-TKI resistance have been reported, such as ALK gene amplification (12, 13), activation of bypass signaling (e.g., HGF-Met, EGFR, c-KIT, IGF-1R, and HER3; refs. 14–17), acquisition of other driver oncogenes (including mutated EGFR and BRAF; ref. 18), and limited drug penetration due to blood–brain barrier and overexpression of P-glycoprotein (19). In these cases, alternative treatment strategies have been proposed (14–17, 19).

Epithelial-to-mesenchymal transition (EMT) is an increasingly recognized driver of tumor progression, and a cause of both innate and acquired resistance to various cytotoxic and targeted drugs, including EGFR-TKIs (20). Previous studies reported that in patients whose EGFR-TKI resistance was associated with EMT, the prognosis was much worse than that of patients whose EGFR-TKI resistance was due to EGFR^T790M mutation in EGFR-mutated NSCLC (21). A recent study (13) reported that EMT component and ALK resistance mutation were simultaneously detected in a single tumor lesion in patients with ALK-rearranged lung cancer who were resistant to ALK-TKIs. However, it is still unknown whether ALK-TKI resistant tumor cells combine mesenchymal phenotype with ALK resistance mutation, or each of the mesenchymal type tumor cells and ALK resistance mutation–positive tumor cells coexist in a single lesion. In any of these cases, no therapy for EMT-associated targeted drug resistance has yet been established, being considered an urgent unmet need.

In this study, we intensively examined crizotinib-resistant tumor lesions obtained from a patient with ALK-rearranged lung cancer. We found that mesenchymal-type tumor cells and ALK resistance mutation can coexist in a single tumor lesion as independent mechanisms underlying the ALK inhibitor–resistant cancers. Utilizing human EML4-ALK lung cancer cell lines, we further explored the mechanism by which EMT was induced during acquisition of crizotinib resistance and show the therapeutic strategy to overcome EMT-mediated ALK-TKI resistance in vitro and in vivo.

Autopsy samples from a patient who acquired resistance to crizotinib were used for these tests. The lung tumor tissues were first stained with antivimentin for IHC. Epithelial regions (ALK⁺, vimentin⁺, and E-cadherin⁻) and mesenchymal regions (ALK⁺, vimentin⁻, and E-cadherin⁺) in a total of 29 slides (5 μm each) were collected by Micro Laser Systems MBIII (Carl Zeiss Microscopy).

DNA extraction was performed following the manufacturer's recommendations with slight modifications. Briefly, microdissected formalin-fixed, paraffin-embedded (FFPE) sample were treated with proteinase K (Promega) for 16 hours at 70°C. After adding lysis buffer, DNA isolation was performed using a Maxwell16 FFPE Tissue LEV DNA Purification Kit on a Maxwell16 Instrument (Promega) according to manufacturer's instructions. DNA was eluted in 50 μL of nuclease-free water, and the concentration was measured by a Quantus Fluorometor (Promega).

Briefly, digital PCR reactions were performed using a Bio-Rad QX200 Droplet Digital PCR (ddPCR) platform. Reactions were set up by ddPCR Supermix for Probes (no dUTP; Bio-Rad), LBx Probe ALK L1196M (A071) for the mutation analysis, and LBx Probe APOB CNV for the CNV analysis (RIKEN GENESIS Co., Ltd.). Thermal cycling was performed on a Veriti thermal cycler (Applied Biosystems) using the following thermal cycling protocol: 10 minutes 95°C, 40 × (30 seconds 94°C, 1 minute 58°C), 10 minutes 98°C with the 50% ramp rate. Results were analyzed and exported using the Quantasoft 1.6.6.0320 software.

A human lung adenocarcinoma cell line A925LPE3 with an EML4-ALK fusion protein (variant 5a, E2:A20), which is highly tumorigenic variant of A925L was established (22), and the resistant cells were derived and characterized as described previously (17). Original A925L were established from a surgical specimen obtained from a Japanese male patient (T2N2M0, stage IIIA). The human lung adenocarcinoma cell lines H2228 and H3122 (23) containing a EML4-ALK fusion protein were used in this research. The H2228 was purchased from the ATCC. The H3122 was kindly provided by Dr. Jeffrey A (Novartis Institutes for BioMedical Research, Cambridge, MA). The human lung embryonic fibroblast lines MRC-5 and IMR-90 were obtained from RIKEN Cell Bank. H1975 was kindly provided by Dr. John D. Minna (University of Texas Southwestern Medical Center, Dallas, TX). HCC827 cells were obtained from the ATCC. All cells were maintained in RPMI-1640 medium supplemented with 10% FBS, penicillin (100 U/mL), and streptomycin (10 μg/mL) in a humidified CO₂ incubator at 37°C. Cells were passaged for <3 months before renewal from frozen, early-passage stocks. Cells were regularly screened for Mycoplasma by using MycoAlert Mycoplasma Detection Kits (Lonza). Cell lines authentication was performed by a short tandem repeat analysis in a laboratory at the National Institute of Biomedical Innovation (Osaka, Japan). Crizotinib, alectinib, ceritinib, lorlatinib, vorinostat, and quisinostat were obtained from Selleck Chemicals. SB43152 was purchased from Cayman Chemical.

The primary antibodies, including anti-ALK (C26G7), anti-phospho-ALK (Tyr1604), anti-phospho-Akt (S473), anti-Akt, anti–E-cadherin, anti–Zo-1, anti-claudin1, antivimentin, anti–N-cadherin, anti-ZEB1, anti-Snail1, anti-Snai2, anti-Twist, anti-PAI1, anti-MED12, anti–phospho-EGFR, anti–phospho-c-KIT, anti–c-KIT, anti–phospho-SRC, anti-SRC, anti–phospho-SMAD2, anti-SMAD2, and anti–β-actin were obtained from Cell Signaling Technology. Anti-ERK1/ERK2, p-ERK1/ERK2 (T202/Y204), and anti-EGFR were obtained from R&D Systems. Anti-SMAD7 was purchased from Santa Cruz Biotechnology. Western blot analysis was as described previously (17). Phospho-RTK array Kit (ARY001B) was obtained from R&D Systems.

Cell viability was measured by the MTT dye (24) reduction method. Cells, plated at a density of 2 × 10³/100 μL RPMI-1640 plus 10% FBS per well in 96-well plates, were incubated for 24 hours. Crizotinib or alectinib were then added to each well, and incubation was continued for another 72 hours. Cell growth was measured with MTT solution (2 mg/mL; Sigma), as described previously (17).

Cells were cultured in complete medium without inhibitor for 48 hours, followed by treatment with the histone deacetylase (HDAC) inhibitor quisinostat in a 10-cm dish for 24 to 72 hours. Cells were placed in 96-well plates and treated with crizotinib, and then incubated for another 72 hours for the MTT assay.

The cell migration assay was performed using Transwell (Sigma-Aldrich). A total of 1 ×10⁵ cells in 200-μL medium were placed in the upper chamber and the lower chamber was filled with 1-mL serum-free media. Cells were incubated for 24 hours, nonmigrating cells were removed using cotton swabs, and the cells that migrated to the lower chamber were stained with 0.1% crystal violet for 5 minutes. Cells were counted under a microscope. The data presented are representative of three independent experiments.

siRNA oligonucleotides specific to ALK and E-cadherin were obtained from Santa Cruz Biotechnology. siRNA specific to ZEB1(#1, #2) and siRNA negative control were purchased from Thermo Fisher Scientific. Introduction of siRNA was performed using Lipofectamine2000 (Thermo Fisher Scientific) in accordance with the manufacturer's instructions.

The miR-200c-141 promoter (1057 bp) was amplified from #4 cells and ligated into the pNL2.2 [NlucP/Hygro] vector (4832 bp) to generate the pNL2.2-miR200c construct. The selection of correct clones was confirmed by restriction map analysis and sequencing. Cells were transfected with pNL2.2-miR200c by Lipofectamine 2000. Twenty-four hours after transfection, cells were treated with drugs and after 24 hours, the NanoLuc luciferase activity was measured using Nano-Glo Luciferase Assay System (Promega). Each transfection was carried out in triplicate in 96-well plates.

For RNA quantification, reverse transcription of the collected RNAs was performed using SuperScript VILO cDNA synthesis Kit and Master Mix (Invitrogen). TaqMan Gene Expression Assays (Applied Biosystems) for MED12 (#1726657) were used according to the manufacturer's protocol. qPCR analysis was performed with ViiA7 Real-Time PCR system (Applied Biosystems). The relative miRNA levels were calculated using the formula 2^−ΔΔC_t. The data were normalized with respect to the expression of ACTB (#4331182). For miRNA quantification, TaqMan MicroRNA Assays (Applied Biosystems) for miR-200c (#002300) and miR-141 (#000463) were used according to the manufacturer's protocol. qPCR analysis was performed with a 7900HT Fast Real-Time PCR System (Applied Biosystems). The relative miRNA levels were calculated using the formula 2^−ΔΔC_t. The data were normalized with respect to the expression of RNU6B (#001093).

We used 5-week-old female SCID mice (Clea) for the study. For the pleural carcinomatosis model (22), the skin and subcutaneous tissue on the right side of the chest were cut and the parietal pleura were exposed. Tumor cells (1 × 10⁶/100 μL) were then injected into the right thoracic cavity through the parietal pleura by using a 27-G needle. Subsequently, the incisions were sutured to close the wound. One week after inoculation, mice were treated orally with ALK-TKIs, 50 mg/kg of crizotinib, or 20 mg/kg of alectinib. Then, the mice were randomized into three arms for the ALK-TKI continuation, 10 mg/kg quisinostat for 5 days followed by no treatment, or 10 mg/kg quisinostat for 5 days followed by the ALK-TKI treatment. The luciferase expression of tumors and mouse body weights were measured twice per week. All surgeries were carried out under sodium pentobarbital anesthesia, and efforts were made to minimize the suffering of the animals. All animal experiments in this study were performed in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the Ministry of Education, Culture, Sports, Science, and Technology, Japan. The protocol was approved by the Committee of Ethics of Experimental Animals, and the Advanced Science Research Center, Kanazawa University (Kanazawa, Japan).

After inoculation, the quantity of tumors was tracked in live mice by repeated noninvasive optical imaging of tumor-specific luciferase activity using the IVIS Lumina XR Imaging System (PerkinElmer) as described previously (17, 22).

We first evaluated the presence of mesenchymal-type tumor cells, representing EMT, and ALK resistance mutation in clinical specimens obtained from a patient before and after crizotinib treatment. The female patient was diagnosed by biopsy specimen from a subclavian lymph node metastasis as stage IV primary lung adenocarcinoma with ALK rearrangement. She received crizotinib treatment and showed partial response. However, 4 months after the initiation of crizotinib, she died due to acute disease progression. Specimens from the primary lung tumor, as well as from the brain and subcutaneous metastases were obtained at autopsy. In all specimens, the presence of ALK-rearranged tumor cells was confirmed by IHC for ALK. In a pre-crizotinib tumor obtained from the diagnostic subclavian lymph node, we found that almost all tumor cells were E-cadherin–positive/vimentin-negative, indicating the epithelial phenotype (Fig. 1A). In this tumor, known ALK mutations associated with ALK-TKI resistance were not detected. On the other hand, in specimens from the crizotinib-treated primary lung tumor, brain and subcutaneous metastases, we found both the E-cadherin–positive/vimentin-negative tumor cells and the E-cadherin–negative/vimentin-positive tumor cells, indicating coexistence of epithelial and mesenchymal phenotype tumor cells in the same lesion (Fig. 1A; Supplementary Fig. S1). Unexpectedly, ALK^L1196M, the gatekeeper mutation, which confers resistance to crizotinib, was detected in all crizotinib-treated tumors obtained at autopsy (Fig. 1A). These results indicate that mesenchymal-type tumor cells may exist after crizotinib treatment even in the ALK resistance mutation–positive tumor lesion.

To investigate whether the mesenchymal type tumor cells acquire ALK^L1196M mutation, we sought to measure the copy number of ALK^L1196M in epithelial and mesenchymal-type tumor lesions, separately. We first stained the primary lung tumor for E-cadherin and vimentin, respectively, by IHC (Fig. 1B), followed by isolated the E-cadherin–positive/vimentin-negative lesions and the E-cadherin–negative/vimentin-positive lesions, respectively, by laser capture microdissection (LCM; Fig. 1C). Pathologists carefully confirmed that almost all cells in the lesions selected for LCM were tumor cells. We then evaluated the copy number of ALK^L1196M and wild-type ALK in the epithelial and mesenchymal type tumor cell lesions, separately. Very interestingly, more than 12 copies of ALK^L1196M mutation were detected in 1 μg of total DNA from the epithelial-type tumor cell lesion (accounting for 27% of all ALK alleles including the wild-type ALK), by sharp contrast, ALK^L1196M mutation was hardly detected (only less than 2 copies in 1 μg DNA and accounting for 5% of all ALK alleles including the wild-type ALK) in the mesenchymal type tumor cell lesion (Fig. 1D and E). EML4-ALK fusion gene has been known to be heterozygous on the chromosome, and we found that ALK gene copy number was not increased in the crizotinib-resistant specimens based on the digital PCR analysis. If hypothesizing that the lesion consisted of only tumor cells, only 10% of the mesenchymal-type tumor cells had the ALK^L1196M mutation, indicating that about 90% of tumor cells acquired crizotinib resistance in the absence of ALK mutation in the mesenchymal-type lesion. Furthermore, in the subcutaneous metastasis, we isolated the E-cadherin–positive/vimentin-negative lesions and the E-cadherin–negative/vimentin-positive lesions, respectively, by LCM. We found that ALK^L1196M mutation was detected only in the epithelial tumor cell lesions, not in the mesenchymal type tumor lesions (Fig. 1E). These results strongly show that EMT is an independent resistant mechanism against ALK-targeted therapy.

To investigate the mechanisms by which EMT cause ALK-TKI resistance, we sought to establish human ALK-rearranged lung cancer cell lines, which acquire EMT-dependent crizotinib resistance. We first induced crizotinib resistance in a pleural effusion mouse model inoculated with an EML4-ALK lung cancer cell line, A925LPE3 (22). After continuous treatment of the recipient mice with crizotinib, they experienced the relapse of pleural carcinomatosis, which acquired resistance to crizotinib. We then generated a cell line, named APE-CR, from the pleural effusion, and also established 15 clones by limiting dilution. Of these, we focused on clones #3, #4, and #8 because they showed a spindle-like mesenchymal morphology, whereas the parental A925LPE3 cells had the cobblestone-like epithelial morphology (Fig. 2A). Next, we examined the sensitivity of selective ALK inhibitors to the generated cells. As shown in Fig. 2B, the three clones were over six times more resistant to crizotinib compared with parental A925LPE3 cells, and cross-resistant to the next-generation ALK-TKIs alectinib, ceritinib, and lorlatinib. Furthermore, crizotinib effectively inhibited ALK phosphorylation in both the parental A925LPE3 cells and these three clones, whereas downstream phospho-AKT and phospho-ERK were inhibited in parental A925LPE3 cells, but not in the three clones (Fig. 2C). Consistently, knockdown of ALK by specific siRNA inhibited the viability of parental A925LPE3 cells, but not of the three clones (Fig. 2D). In addition, known resistance mutations in ALK were not detected in the parental A925LPE3 cells or these three clones. These findings indicate that these three clones acquire resistance by an ALK-independent resistance mechanism.

EGFR activation is a known mechanism of crizotinib resistance (12); however, we did not observe EGFR activation in these three clones (Supplementary Fig. S2A). Hence, combined use of EGFR inhibitors did not restore crizotinib sensitivity of these clones (Supplementary Fig. S2B). Furthermore, other receptor tyrosine kinases reported as ALK-TKI resistance mechanisms including FGFR, HER3, MET, IGF-1R, c-KIT, and SRC were not activated in these clones (Supplementary Fig. S2A and S2C). On the other hand, these clones showed typical mesenchymal phenotype, such as decreased epithelial marker (E-cadherin, Zo-1, and claudin1) expression, increased mesenchymal marker (vimentin, N-cadherin, and ZEB1) expression (Fig. 3A and B). Microarray analysis confirmed that the expression of several epithelial and mesenchymal phenotype–related genes decreased and increased, respectively, in the most resistant #4 cells, compared with the parental A925LPE3 cells (Fig. 3C). In addition, the resistant cells increased cell motility than the parental A925LPE3 cells (Fig. 3D and E). These findings indicate that the three clones, #3, #4, and #8, acquired crizotinib resistance predominantly by EMT.

We next sought to clarify the mechanism by which these three clones developed a mesenchymal phenotype. A recent study reported that the classical EMT inducer, TGF-β, caused ALK inhibitor resistance in an ALK-rearranged NSCLC cell line, H2228 (25). We observed that TGF-β1 treatment reduced E-cadherin expression and slightly induced crizotinib resistance in H2228 cells, but not in other ALK-rearranged NSCLC cell lines, such as A925LPE3 or H3122 (Supplementary Fig. S3A and S3B). In addition, treatment of TGF-β1 inhibitor SB431542 did not revert EMT and restore the sensitivity against crizotinib in these clone cells (Supplementary Fig. S3C and S3D). As shown in Fig. 3A and Supplementary Fig. S3E, the expression of MED12, whose suppression was reported to induce mesenchymal phenotype through activation of TGF-β signaling in ALK-rearranged lung cancer cells (26), was not suppressed in the resistant clones, compared with the parental A925LPE3 cells. Furthermore, these three clones did not produce detectable levels of TGF-β, and the expression of multiple TGF-β target genes including PAI-1, Smad7 in these clones were not activated (Fig. 3A). These data suggest that EMT in these three clones was induced by mechanisms other than activation by TGF-β.

ZEB1 is known to be an E-cadherin transcriptional repressor and plays a crucial role in EMT (27), and we found that the expression of ZEB1 was increased in the three clones, whereas Snail1, Snail2, and Twist, which have been reported as transpritional factors associated with EMT were not increased in these clones (Fig. 3A). Thus, we first transfected cells from clone #4 with ZEB1-specific siRNA and analyzed the expression of E-cadherin. As we expected, ZEB1-specific siRNA reduced the expression of ZEB1 and increased the expression of E-cadherin (Fig. 4A). Interestingly, we found that the knockdown of ZEB1 restored the sensitivity to crizotinib (Fig. 4B; Supplementary Fig. S4A and S4B) in these three clones. Because ZEB1 expression can be regulated by miRNAs (28, 29), we then conducted a miRNA profile analysis of the parental and #4 cells. As shown in Fig. 4C, the expression of miR-200 family members including miR-141 and miR-200c was remarkably decreased in #4 cells compared with the parental cells. miR-200 family is known to play an important role in regulation of EMT (28). In addition, overexpression of miR-200c was reported to cause mesenchymal-to-epithelial transition (MET; ref. 30). Consistently, we observed that overexpression of miR-141 or miR-200c by transfection decreased the ZEB1 expression and increased the E-cadherin expression in #4 cells (Supplementary Fig. S5A). These findings indicate that these clones acquired crizotinib resistance via EMT mediated by decreased expression of miR-200 family.

We thus hypothesized that chemical restoration of miR-200c/141 may reverse EMT and resensitize the tumor cells to ALK-TKIs. The miR-200c/141 is transcribed as one RNA molecule and then cleaved as miR-200c and miR-141, respectively (29). Therefore, to identify compounds that stimulate the miR-200c/141 promotor activity, we first transfected #4 cells with a nano-luc (Nluc) expression vector (Supplementary Fig. S5B and S5C) and performed a reporter assay on a 200 kinase inhibitor library (Supplementary Table S1). As shown in Fig. 4D, CUDC-101, an inhibitor of EGFR, HER2, and class I/class II HDACs (31, 32), had the highest potential to increase miR-200c/141 promotor activity. HDAC2, a member of class I HDACs, has been reported to play a critical role in suppression of the miR-200 family in breast cancer cells (33). In addition, mocetinostat, an inhibitor of class I HDACs, was reported to cancel ZEB1-associated drug resistance in several types of tumor cell lines (34). Thus, we further investigated the effect of various class I HDAC inhibitors on the miR-200c/141 promotor activity. As shown in Fig. 4E, quisinostat most potently increased the miR-200c/141 promotor activity among the tested HDAC inhibitors, including CUDC-101. Quisinostat, a “second-generation” HDAC inhibitor, has a high potency toward class I HDACs with a more prolonged pharmacodynamics response in vivo than “first-generation” HDAC inhibitors (35–37). We indeed confirmed that the treatment of quisinostat remarkably increased the expression of miR-200c and miR-141 (Fig. 4F) and decreased that of ZEB1 (Fig. 4G). Furthermore, transfection of antagomiR-200c, but not antagomiR-141, canceled the downregulation of ZEB1 by quisinostat in the #4 cells (Fig. 4G). These findings suggest that quisinostat decreases ZEB1 expression predominantly by upregulating miR-200c.

We next tested the hypothesis that quisinostat may restore ALK-TKI sensitivity by reverting EMT in the #4 cells. To determine the optimum concentration for quisinostat, we first used the normal lung fibroblast lines, MRC-5 and IMR-90. Because 0.03 μmol/L quisinostat had almost no effect on the viability of the fibroblast lines, we utilized this concentration of quisinostat in the in vitro experiments (Supplementary Fig. S6A). As shown in Fig. 5A and B, 0.03 μmol/L quisinostat upregulated the expression of E-cadherin and downregulated the expression of N-cadherin, vimentin, and ZEB1 in a time-dependent manner. Accordingly, treatment of #4 cells with quisinostat for 3 days reduced their motility and induced epithelial morphology (Supplementary Fig. S6B and S6C). To investigate whether quisinostat could reverse the sensitivity to crizotinib, we pretreated #4 cells with quisinostat for 3 days, and then treated them with crizotinib for another 3 days (Fig. 5C). Importantly, pretreatment of quisinostat restored sensitivity to crizotinib in these three clones to the level observed in the parental A925LPE3 cells. Furthermore, pretreatment of quisinostat also restored sensitivity to the second-generation ALK-TKIs, alectinib, ceritinib, and lorlatinib (Fig. 5C; Supplementary Fig. S7). As mentioned in Fig. 2C, crizotinib alone did not inhibit the phosphorylation of AKT or ERK in the resistant cells. However, when we pretreated #4 cells with quisinostat, the phosphorylation of AKT was strongly inhibited by crizotinib (Supplementary Fig. S8A), suggesting that AKT signaling plays an important role in the acquisition of crizotinib resistance in the cells with a mesenchymal phenotype.

To further investigate whether resensitization to ALK-TKIs by quisinostat depends on the expression of EMT marker proteins, #4 cells were transfected with E-cadherin-specific siRNA and then treated with quisinostat for 72 hours. We found that siRNA successfully blocked the quisinostat-induced E-cadherin expression (Fig. 5D), and almost completely canceled the effect of quisinostat on restoring the sensitivity to crizotinib (Fig. 5E). These results indicate that quisinostat can resensitize ALK-rearranged lung cancer cells with mesenchymal phonotype to ALK-TKIs by reverting the EMT.

The same phenomenon was observed in the EGFR-mutated NSCLC cell lines, HCC827KGR and H1975OR, which acquired resistance to gefitinib and osimertinib (AZD9291), respectively, associated with EMT. We found that pretreatment of HCC827KGR and H1975OR cells with quisinostat reversed the EMT and sensitized the cells to gefitinib and osimertinib, respectively (Supplementary Fig. S8B and S8C). These results suggest that quisinostat may overcome EMT-associated TKI resistance in several types of lung cancer cells with different driver oncogenes.

To examine the effect of quisinostat in vivo in the pleural carcinomatosis model, we inoculated #4 cells into the thoracic cavities, where this clone acquired crizotinib resistance, in the SCID mice. The mice were continuously treated with crizotinib or alectinib, and we confirmed the progression of pleural carcinomatosis, indicating ALK-TKI resistance. On the basis of the results of the in vitro experiments, we then pretreated the mice with or without quisinostat (10 mg/kg) for 5 days, and thereafter, treated them with crizotinib or alectinib. As shown in Fig. 6A, the mice treated with quisinostat for 5 days without crizotinib retreatment became moribund by day 32, indicating that quisinostat treatment alone was not enough for controlling crizotinib-resistant pleural carcinomatosis. Interestingly, pretreatment with quisinostat followed by crizotinib or alectinib successfully regressed pleural carcinomatosis, whereas exclusive continuous treatment with crizotinib or alectinib failed to control disease progression (Fig. 6A and B). During the treatment with quisinostat, no macroscopic changes or loss in body weight were observed in the mice compared with control. Furthermore, we confirmed that quisinostat treatment for 5 days increased the expression of E-cadherin and decreased that of vimentin in tumor cells obtained from the thoracic cavity (Fig. 6C; Supplementary Fig. S9A). Similar results were also observed by IHC: E-cadherin was restored to the levels observed in parental A925LPE3 tumors after the quisinostat treatment (Fig. 6D). In addition, repeated CT scans confirmed disappearance of pleural effusion in mice pretreated with quisinostat followed by crizotinib (Supplementary Fig. S9B). These findings strongly indicate that quisinostat is capable of reverting EMT and restoring the sensitivity of mesenchymal-type tumor cells to ALK-TKIs even in vivo.

In this study, we demonstrated that although mesenchymal-type tumor cells and the ALK^L1196M mutation coexisted in a single crizotinib-resistant lesion, the ALK^L1196M mutation was predominantly detected in epithelial-type tumor cells. These new findings indicate that the mesenchymal phenotype caused by EMT is an independent mechanism that can induce crizotinib resistance in patients with ALK-rearranged lung cancer. We further found that the tumor cells with mesenchymal phenotype were cross-resistant to the new-generation ALK-TKIs, such as alectinib, ceritinib, and lorlatinib. Very importantly, pretreatment with the HDAC inhibitor quisinostat, which upregulated the miR-200c/141 promotor activity, restored miR-200c expression, and reverted EMT, followed by treatment with an ALK inhibitor could circumvent the resistance due to EMT in ALK-rearranged lung cancer cells with mesenchymal phenotype.

A recent study by Gainor and colleagues reported that EMT could be observed in the tumors of patients with ALK- rearranged NSCLC treated with ALK-TKIs, including ceritinib. They detected mesenchymal tumor cells, defined by loss of E-cadherin and expression of vimentin, in five of 12 clinical specimens (13). Of the five specimens, ALK-resistant mutations were detected in three. Thus, they suggested that EMT may be not the sole driver of resistance. However, our digital PCR analyses combined with LCM after IHC staining for E-cadherin and vimentin demonstrated that the ALK^L1196M mutation was not detected in about 90% of tumor cells with mesenchymal phenotype in a crizotinib-treated tumor, indicating that EMT could be an independent crizotinib-resistant mechanism in ALK-TKI–resistant tumors. Gainor and colleagues also suggested that EMT, which was not observed before ALK-TKI treatment, might have been the predominant mechanism of resistance in a tumor also exhibiting a ALK^L1196M mutation, which was heavily treated with crizotinib, alectinib, and ceritinib (13). In our preclinical finding, ALK-rearranged lung cancer cells with mesenchymal phenotype without ALK-resistant mutations induced cross-resistance to crizotinib and these second-generation ALK-TKIs, suggesting that only EMT could play a role in conferring the resistance, thus supporting our hypothesis.

Circumvention of EMT-associated resistance, based on understanding the underlying mechanisms, is an unmet need of targeted drug therapy. We found that EMT could be induced by reducing the expression levels of miR-200 family members, including miR-200c and miR-141, which resulted in increasing ZEB1 and decreasing E-cadherin expression in ALK-TKI–resistant ALK-rearranged NSCLC cells. These observations are in the line with the evidence reported in pancreatic, colorectal, and breast cancers (38). Moreover, we provided the first demonstration that the second-generation HDAC inhibitor quisinostat could overcome EMT-associated ALK-TKI resistance by reverting EMT, which was mediated by enhanced expression of miR-200c but not miR-141, via upregulated miR200c/141 promotor activity. We further showed that pretreatment with quisinostat could overcome resistance to both crizotinib and alectinib in the pleural carcinomatosis model. Very importantly, our findings also suggest that pretreatment with quisinostat may overcome EGFR-TKI resistance in EGFR-mutated NSCLC. Quisinostat is a second-generation HDAC inhibitor, with potent activity to HDAC 1 and 2. However, the precise mechanism by which quisinostat increases miR-200c promotor activity is still unknown. Identification of such mechanism may allow the development of compounds that overcome EMT-associated targeted drug resistance much more efficiently and selectively.

Recent clinical trials, J-ALEX (9) and ALEX (10), demonstrated that alectinib given as a first-line treatment remarkably prolonged the PFS of advanced NSCLC with ALK rearrangement, as compared with crizotinib. Nevertheless, although alectinib has activity over a spectrum of crizotinib-resistant mutations, crizotinib does not inhibit the majority of alectinib-resistant mutations (13). Thus, the best sequence of ALK-TKIs for ALK-rearranged NSCLC has not yet been determined (39, 40). If the coexistence of EMT components and ALK mutations induces resistance to first-line crizotinib, the same would be expected in relation to resistance to second-generation ALK-TKIs with activity against crizotinib-resistant ALK mutations. In this study, we demonstrated that pretreatment with quisinostat could overcome alectinib resistance caused by EMT-associated crizotinib resistance in tumor cells from a pleural carcinomatosis model. Therefore, the therapeutic strategy of pretreatment with quisinostat followed by second-generation ALK-TKIs may be effective for overcoming targeted drug resistance where ALK resistance mutations and EMT components coexist (Fig. 6E). Further studies are warranted to clarify the involvement of EMT in resistance to second-generation ALK-TKIs given as the first-line treatment in NSCLC with ALK- rearrangement.

In conclusion, we demonstrated that the mesenchymal phenotype caused by EMT can coexist with ALK mutation as an independent resistance mechanism in ALK inhibitor–resistant tumors. Moreover, pretreatment with the HDAC inhibitor quisinostat, which upregulates the miR-200c/141 promotor activity, restores miR-200c expression, and reverts EMT, followed by an ALK inhibitor, can circumvent the resistance due to EMT. These findings illustrate the importance of developing compounds that selectively and efficiently increase miR-200c expression, and provide a rationale for their sequential use with new-generation ALK inhibitors to treat TKI-resistant ALK-rearranged NSCLC with ALK-resistant mutation and EMT.

S. Nanjo reports receiving a commercial research grant and has received speakers bureau honoraria from AstraZeneca. T. Nakagawa is an associate director at Eisai Co., Ltd. Y. Ishikawa is a consultant/advisory board member for Fujirebio Inc. K. Takeuchi has ownership interest (including patents) as royalty from Nichirei and is a consultant/advisory board member for Nichirei. S. Yano has received a commercial research grant from Chugai and has received speakers bureau honoraria from Pfizer and Chugai. No potential conflicts of interest were disclosed by the other authors.

Conception and design: K. Fukuda, S. Takeuchi, S. Yano

Development of methodology: K. Fukuda, S. Takeuchi, R. Katayama

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K. Fukuda, S. Takeuchi, S. Arai, R. Katayama, S. Nanjo, T. Nakagawa, H. Taniguchi, T. Suzuki, T. Yamada, H. Nishihara, H. Ninomiya, Y. Ishikawa, S. Baba, A. Horiike, N. Yanagitani, M. Nishio

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K. Fukuda, S. Takeuchi, R. Katayama, S. Nanjo, A. Tanimoto, K. Takeuchi, M. Nishio

Writing, review, and/or revision of the manuscript: K. Fukuda, S. Takeuchi, S. Nanjo, M. Nishio, S. Yano

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Takeuchi, R. Katayama, A. Tanimoto, A. Nishiyama, K. Takeuchi, M. Nishio, S. Yano

Study supervision: S. Takeuchi, S. Yano

This work was supported by grants from the Japanese Society for the Promotion of Science (JSPS) KAKENHI (grant numbers 16K19447 and 18K07261 to K. Fukuda; 17K09649 to S. Takeuchi; 16H05308 to S. Yano), and the Project for Cancer Research and Therapeutic Evolution (P-CREATE) grant number 16cm0106513h0001 to S. Yano.

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.