RET fusions have been found in lung adenocarcinoma, of which KIF5B–RET is the most prevalent. We established inducible KIF5B-RET transgenic mice and KIF5B–RET-dependent cell lines for preclinical modeling of KIF5B–RET-associated lung adenocarcinoma. Doxycycline-induced CCSP-rtTA/tetO-KIF5B-RET transgenic mice developed invasive lung adenocarcinoma with desmoplastic reaction. Tumors regressed upon suppression of KIF5B–RET expression. By culturing KIF5B–RET-dependent BaF3 (B/KR) cells with increasing concentrations of cabozantinib or vandetanib, we identified cabozantinib-resistant RETV804L mutation and vandetanib-resistant-RETG810A mutation. Among cabozantinib, lenvatinib, ponatinib, and vandetanib, ponatinib was identified as the most potent inhibitor against KIF5B–RET and its drug-resistant mutants. Interestingly, the vandetanib-resistant KIF5B-RETG810A mutant displayed gain-of-sensitivity (GOS) to ponatinib and lenvatinib. Treatment of doxycycline-induced CCSP-rtTA/tetO-KIF5B-RET bitransgenic mice with ponatinib effectively induced tumor regression. These results indicate that KIF5B-RET–associated lung tumors are addicted to the fusion oncogene and ponatinib is the most effective inhibitor for targeting KIF5B–RET in lung adenocarcinoma. Moreover, this study finds a novel vandetanib-resistant RETG810A mutation and identifies lenvatinib and ponatinib as the secondary drugs to overcome this vandetanib resistance mechanism. Mol Cancer Ther; 15(10); 2521–9. ©2016 AACR.
Lung adenocarcinoma is a major subtype of non–small cell lung cancer (NSCLC) with increasing incidence in many countries (1). Compared with other histologic types of lung cancer, lung adenocarcinoma frequently harbors KRAS or protein tyrosine kinase (PTK) aberrations (2, 3). Although mutant KRAS is difficult to target directly at present, PTKs are targetable, allowing precision treatment targeting individual patient's driver oncogene. Targeting PTK driver oncogenes with small-molecule inhibitors in lung adenocarcinoma has shown clinical success in managing NSCLC. This is exemplified by EGFR and ALK inhibitor therapies in EGFR mutation and ALK-fusion lung adenocarcinoma (4, 5). However, acquired resistance to EGFR and ALK kinase inhibitors usually occur within a year after the drug treatment, making it necessary to use secondary drugs to prolong the therapeutic response (6, 7). Moreover, except for EGFR mutations and ALK fusions, other PTK alterations in lung adenocarcinoma occur at <3% rates (8, 9), rendering clinical studies of these PTK oncogenes difficult (10).
Several laboratories have identified recurrent RET fusion genes in 1% to 2% of lung adenocarcinoma cases (8, 11–14). Among the fusion partners, KIF5B is the most prevalent. RET rearrangements in lung adenocarcinoma are mutually exclusive from other known driver oncogenic mutations, including KRAS, EGFR, ALK, BRAF, and ERBB2 (8). Moreover, ALK, RET, and ROS1 oncofusion-associated lung adenocarcinoma tissues harbor significantly fewer mutations than other lung adenocarcinoma tissues (10). This suggests a strong role of these fusion oncogenes in driving the tumors, implicating them as excellent therapeutic targets.
A number of PTK inhibitors (tyrosine kinase inhibitors, TKI) are known to cross-inhibit RET kinase activity (15). These include cabozantinib and vandetanib that have been approved by the FDA for the treatment of advanced medullary thyroid cancer, which carries RET point mutations in 30% to 50% of cases (15). So far, two reports have described the responses of 6 RET-fusion lung adenocarcinoma patients treated with cabozantinib: 4 patients had partial responses and 2 patients had stable disease (16, 17). One patient with RET-fusion lung adenocarcinoma who received vandetanib was reported to have decreased tumor size (18). These are encouraging initial findings but more studies are needed. Several clinical trials of RET kinase inhibitors are ongoing (8, 15). However, the rarity of this molecular subtype of the disease presents a barrier for the current and future clinical studies, as it will take a long time to enroll adequate number of patients into the clinical trials.
To generate a preclinical model of RET-fusion lung adenocarcinoma in an immune competent environment, we generated tetO-KIF5B-RET transgenic mice and used CCSP-rtTA/tetO-KIF5B-RET bitransgenic mice to induce KIF5B–RET expression in the lungs by doxycycline (Dox). We found that KIF5B–RET induced invasive lung adenocarcinoma with desmoplastic reaction. After tumor development, continuous expression of the KIF5B–RET fusion gene is required to maintain the lung tumors. Using a parallel in vitro cell-based assay, we identified ponatinib as the most potent inhibitor against KIF5B–RET, its kinase gatekeeper mutations, and a novel vandetanib-resistant RETG810A mutant. Treatment of doxycycline-induced CCSP-rtTA/tetO-KIF5B-RET transgenic mice with ponatinib resulted in tumor regression.
Materials and Methods
Antibodies to phospho-RET(Tyr905) and cleaved-PARP were from Cell Signaling Technology, antibodies to Flag-tag and β-actin were from Sigma, anti-RET antibody was from Santa Cruz Biotechnology. Anti-TTF1 (ab137061) antibody was from Abcam. Anti-cytokeratin antibody was from Dako (cat. No. Z0622). Ponatinib was from LC Laboratories. Cabozantinib, vandetanib, and lenvatinib were from Selleckchem.
Generation of transgenic mice
The KIF5B–RET (K15;R12, RET51 long form; ref. 12) coding region with a Flag tag coding sequence at the C-terminus was synthesized by GeneArt Gene Synthesis (Life Technologies) and cloned into the L3/L2 loxP-tetO plasmid (19). The 6.6-kb BssHII tetO-KIF5B-RET transgene cassette (Fig. 1A) DNA fragment was used to generate tetO-KIF5B-RET transgenic mice in FVB/N strain as described in Supplementary Information.
Bitransgenic CCSP-rtTA/tetO-KIF5B-RET (C/KR) mice were produced by crossing tetO-KIF5B-RET mice with CCSP-rtTA mice (also in FVB/N strain; Fig. 1B; refs. 19–21). To induce KIF5B–RET expression, C/KR mice at 4-weeks-old were fed with rodent chow containing 200 mg/kg doxycycline (Dox diet, Bio-Serv). Both male and female mice were used in the experiments.
Genotyping of CCSP-rtTA transgenic mice was as described previously (19, 21). Genotyping of tetO-KIF5B-RET transgenic mice was performed using GoTaq Hot Start Green Master Mix (Promega) using primer pairs: K5R-3T, 5′-CACGAGAGCTGATGGCACTAACACT and TETO-3B, 5′-CACAGCCCAGCCCAGCCCTCTACT. The 25 μL PCR reaction protocol was: 3 minutes at 94°C, 30 cycles of 94°C for 30 seconds, 57°C for 30 seconds, and 72°C for 30 seconds with a final extension step of 72°C for 5 minutes, which yielded a 309-bp product. Animal experiments were approved by the Institutional Animal Care and Use Committee of University of South Florida.
Cell cultures and analyses
The mouse BaF3 cells were obtained from Dr. H.G. Wang (H. Lee Moffitt Cancer Center) in 2000 and have been stored in liquid nitrogen. BaF3 cells were cultured in RPMI-1640/10% FBS supplemented with 2 ng/mL interleukin-3 (IL-3) as described previously (22). The identity of BaF3 cells were evaluated on the basis of their dependency on mouse IL-3. The authors have not authenticated BaF3 cells using the DNA-based method. To generate KIF5B–RET-transformed BaF3 cells (B/KR), BaF3 cells were infected with lentiviruses containing a flag-tagged KIF5B–RET gene. Individual puromycin-resistant cell lines were isolated, screened for the presence of KIF5B–RET by immunoblotting, and evaluated for IL-3 independence.
To identify drug-resistant mutations, established B/KR cells were incubated with increasing concentrations of cabozantinib or vandetanib in RPMI-1640/10% FBS as described in the Results. Individual drug-resistant cell lines were isolated from semi-solid methylcellulose cultures, which contained 4:6 ratio of MethoCult H4100 (Stemcell Technologies) and 20%FBS in RPMI-1640. Genomic DNA of drug-resistant cell lines was isolated and the RET kinase domain coding region was sequenced in both strands of DNA following PCR amplification as described previously (23).
To determine drug sensitivity, B/KR cells (1,500 cells/well) were incubated in RPMI-1640/10% FBS in the presence of indicated concentrations of drugs in 96-well plates in triplicates. On day 5, viable cells were measured using the CellTiter-Glo reagent (Promega). IC50 measurements were obtained from at least two independent experiments.
Mice were euthanized by CO2 asphyxiation. Lungs were flushed with 10 mL PBS and insufflated with 10% buffered formalin. After fixing overnight in 10% buffered formalin, paraffin blocks were prepared. Tissue sections were stained with hematoxylin and eosin (H&E). H&E stain tissue slides were examined by three pathologists and scanned using a ScanScope XT (Aperio). The Genie V1 histology pattern recognition software (Aperio) was used to segment hyperproliferative lesions/tumors from other lung tissue areas and background using the same parameters as described previously (21). Trichrome staining was performed by the Tissue Pathology Core at Stephenson Cancer Center.
Immunohistochemistry (IHC) was performed with a Leica Bond III automated IHC protocol using a Leica Bond-III Polymer Refine Detection reagents (Cat. No.: DS 9800). Epitope retrieval was citrate buffer (pH 6.0) at 100°C for 20 minutes. Anti-TTF1 and anti-cytokeratin antibodies were used at 1:400 dilution for 30 minutes.
Magnetic resonance imaging and computerized tomography protocols
For MRI, mice were anesthetized with 2% isoflurane and transferred to mouse cradle mounted on an insertion device and positioned within the RF coil of the magnet and kept under anesthesia for the duration of the experiment. The mice were physiologically monitored and maintained with a Model 1030 Monitoring and Gating System (SA Instruments, Inc.). A respirator sensor pad was placed under each animal to manually control anesthesia mixture whereas a fiber optic rectal thermometer was used for temperature feedback control, which was set to maintain a body temperature of 37 ± 1°C. Respiration rate was maintained at 50 to 60 breaths per minute and respiratory-gated MR images were acquired during the resting phase after exhalation. MRI was performed using a 7-T horizontal magnet (ASR 310, Agilent Technologies) equipped with nested 205/120/HDS gradient insert in a bore size of 310 mm. Two RF coils, a 35-mm Litzcage coil (Doty Scientific, Inc.) and a 24-mm Litzcage coil (Doty Scientific, Inc.), were used depending on the sizes and weights of the mice. Temperature control of the imaging gradients was achieved by means of a water chiller (Neslab Waters) and maintained at 12°C for all experiments.
Coronal multislice T2-weighted fast spin-echo (FSEMS) respiratory-gated sequences were acquired with TE/TR = 30.05/1497.60 ms, data matrix = 256 × 128, 20 slices, field of view (FOV) = 90 × 40 mm2, 16 averages and slice thickness of 1.20 mm over about 10 to 12 minutes. In T2-weighted images, tumor volumes and dimensions were quantified using manually drawn regions of interest (ROI) within lung on a slice-by-slice basis, where pixels above a given threshold were counted as part of diffuse lesions. Analysis was done with MATLAB using a script that employed multithresholding segmentation method from the image processing toolbox of MATLAB. Care was taken to exclude the heart, vessels, and mediastinum.
For μCT, mice were scanned on an Inveon μCT scanner (Inveon micro CT/PET/SPECT, Siemens) using a free-breathing protocol. Animals were anesthetized with a mixture of 2% isoflurane and oxygen. Breathing rate was controlled by a BioVet physiologic monitoring system (m2m Imaging Corporation) and kept at average 60 breaths per minute. The 440 projections were taken over 220° arc trough 12 minutes over an FOV of 2.7 × 1.5 cm. The X-ray tube scanning parameters were 80 kVp voltage, 480 μA current, 1,000 ms exposure, we also applied 1.5 mm Al filter.
The acquired data were reconstructed 3D volume with dimensions 1024 × 576 × 441 voxels with 27.2669 um effective isotropic voxel size. CCD detector binning was 2 × 2. Reconstructed data of absorption coefficients were converted to Hounsfield units (HU) in range of −1,000 Hu to 1,000 HU, where the value of −1,000 HU corresponds to the air and 0 HU corresponds to the water (we used 50 mL water phantom for calibration). Image data were evaluated by board certified radiologist using preclinical Inveon Research Workstation 3D Visualization module (Siemens). The lung μCT images were viewed in coronal, sagittal, and axial planes as well as in 3D. The 3D model of lung's air pathways was rendered from original HU data after applying intensity threshold procedure.
Statistical analyses were performed using the unpaired t test with Welch's correction, without assuming equal SDs. The statistical significance was set to 0.05.
Inducible KIF5B–RET transgenic mice
We generated a tetO-KIF5B-RET transgenic mouse line (A1) as described in the Supplementary Information. These monotransgenic mice had no detectable phenotype and no KIF5B–RET transcript in the tissues that we have examined (Fig. 1C, top). CCSP-rtTA/tetO-KIF5B-RET bitransgenic mice (C/KR) were generated by crossing tetO-KIF5B-RET mice with CCSP-rtTA mice (19, 20) and evaluated for doxycycline-induced expression of KIF5B–RET. As illustrated in Fig. 1C (bottom), doxycycline specifically induced KIF5B–RET mRNA in the lungs but not in the heart, intestine, liver, kidney, or spleen of C/KR mice. Immunoprecipitation of lung tissue lysates with an anti-Flag antibody followed by immunoblotting detected KIF5B–RET protein and phosphorylated KIF5B–RET in the lungs of doxycycline-induced C/KR mice (Fig. 1D).
In another microinjection experiment performed later, we obtained two other tetO-KIF5B-RET transgenic mouse lines (A2, G6). CCSP-rtTA/tetO-KIF5B-RET bitransgenic mice generated from A2 and G6 lines displayed the same phenotype as that observed in the A1 line (Supplementary Fig. S1). Because the A1 line was generated much earlier, most of our experiments were conducted using this line.
KIF5B–RET transgenic mice develop lung adenocarcinoma with desmoplastic reaction
As early as one month after doxycycline induction, the lungs of C/KR bitransgenic mice developed areas of hyperplastic and early tumor lesions that were often accompanied by thickening of pleural stroma (Fig. 2A). Pleural thickening was observed in human lung adenocarcinoma harboring RET fusion (Supplementary Fig. S2A). By 4 to 5 months after doxycycline induction of KIF5B–RET expression, extensive lung tumors were observed in these C/KR mice (Fig. 2B and C). These tumors displayed predominantly lepidic growth pattern with focal solid pattern, visceral pleural involvement, and desmoplastic reaction (Fig. 2B and C). Lymphocytes were often observed near the desmoplastic reaction areas and macrophages were observed in the tumor lesion areas. In total, we examined 12 lungs from C/KR mice induced with doxycycline for 4 to 7 months (4 four months, 6 five months, 1 six months, and 1 seven months). Invasive tumors were observed in all 12 lungs, which were confirmed by cytokeratin IHC in every case. We have not observed metastasis of tumors to distal organs. However, we have not ruled out the possibility that distal metastasis may occur when the doxycycline-induced C/KR mice are kept for a longer period of time.
Desmoplastic reaction is associated with invasive lung adenocarcinoma in human lung cancer patients (Supplementary Fig. S2B) and was reported in the first case of KIF5B–RET lung adenocarcinoma when KIF5B–RET fusion was discovered (14). In The Cancer Genome Atlas (TCGA) Research Network lung adenocarcinoma study (2), desmoplastic stroma are visible in the tissue sections of RET-fusion lung adenocarcinoma cases (Supplementary Fig. S2A).
Desmoplastic reaction also occurred in the lung tumors in doxycycline-induced C/KR mice derived from Line A2 and Line G6 of tetO-KIF5B-RET transgenic mice (Supplementary Fig. S1A and S1B). In comparison, we have not observed desmoplastic reaction in lung adenomas and adenocarcinomas in KrasLA1 mice that express KrasG12D (∼ 6 months of age, Supplementary Fig. S1C; ref. 25), doxycycline-induced CCSP-rtTA/tetO-EGFRL858R transgenic mice (21, 26), or doxycycline-induced CCSP-rtTA/tetO-SHP2E76K transgenic mice (n > 20 lungs examined in each case; ref. 19).
The KIF5B–RET oncogene is required to maintain lung tumors
To determine whether lung tumors in doxycycline-induced C/KR mice were addicted to the KIF5B–RET oncogene, we identified 7 doxycycline-induced C/KR mice (4–5 months) with MRI-detectable lung tumors. These mice were then fed regular chow without doxycycline. One month after doxycycline withdrawal, these mice were examined by MRI again. Lung tissue sections were examined by histology and tissue lysates analyzed by immunoprecipitation and immunoblotting for the expression of RET.
As shown in Fig. 3A, doxycycline withdrawal resulted in regression of MRI-detected tumor lesions in all 7 mice. Lungs from 5 of these mice were examined by histology as illustrated in Fig. 3B. The remaining lesions in the lungs from doxycycline withdrawn mice were mostly of incompletely resolved fibrotic tissues. Using a histology pattern recognition program (Aperio) as described previously (21), the hyperproliferation/tumor lesion areas were semiquantified. The area of interest (AOI), which included some of bronchiolar epithelia in addition to hyperproliferation/tumor lesions, in these 5 lungs was 4.33 ± 1.76% (mean ± STD). In comparison, the AOI from 5 doxycycline-induced C/KR mice was 38.80% ± 25.34%, whereas the AOI from the 5 wild-type mice was 2.47% ± 1.17%. Thus, doxycycline withdrawal from C/KR mice resulted in a significant reduction (P = 0.008) in the AOI to a level similar to that of normal wild-type mice (P = 0.98; Fig. 3B).
Both KIF5B–RET protein and phosphorylated (pY905) KIF5B–RET were detected in doxycycline-induced C/KR mice (Fig. 3C). They were no longer detected in the lungs one month after doxycycline withdrawal. Taken together, these data show that after tumor formation, continuous expression of KIF5B–RET is required to maintain the lung tumors.
Ponatinib is the most potent KIF5B–RET kinase inhibitor
BaF3 cells depend on IL-3 for survival. We established stable BaF3 cells expressing KIF5B–RET (B/KR). When cells were incubated in medium without IL-3, the parental cells and vector-control cells underwent apoptosis in 6 hours as measured by PARP cleavage, whereas B/KR cells continued to grow (Fig. 4A and B). Thus, KIF5B–RET transformed BaF3 cells into cytokine independence. The KIF5B–RET transformation was inhibited by cabozantinib that suppressed the KIF5B–RET kinase activity (Fig. 4C and D).
Cabozantinib and vandetanib inhibited B/KR cells with IC50s of 0.175 and 0.90 μmol/L, respectively. To isolate cabozantinib- or vandetanib-resistant KIF5B–RET mutations, we incubated B/KR cells (2 × 107 cells) with sequentially increasing concentrations of cabozantinib (from 0.175 to 0.65 or 0.85 μmol/L in two independent experiments) or vandetanib (from 0.90 to 4.5 or 5.5 μmol/L in two independent experiments). Individual drug-resistant cells were cloned by methylcellulose culture. After verification of constitutive KIF5B–RET tyrosine kinase activity in these cells in the presence of cabozantinib or vandetanib in the culture medium (Supplementary Fig. S3A), the PTK domain coding regions of KIF5B–RET from these drug-resistant cell clones were sequenced. The RETV804L gatekeeper mutation was identified in all 10 cabozantinib-resistant cell lines. A novel RETG810A mutation was found in 8 of 10 vandetanib-resistant cell lines. We next immunoprecipitated KIF5B–RET, KIF5B–RETV804L, and KIF5B–RETG810A from these cells and assayed their PTK activity in vitro in the presence of increasing concentrations of cabozantinib or vandetanib. The results confirmed that KIF5B–RETV804L was resistant to cabozantinib and that KIF5B–RETG810A was resistant to vandetanib (Supplementary Fig. S3B).
To confirm that the novel RETG810A mutation renders vandetanib resistance and to evaluate if drug resistance may be attributed to overexpression of KIF5B–RET protein, we recreated a BaF3/KIF5B-RETG810A cell line (B/KRG810A-REC) using lentivirus encoding the KIF5B–RETG810A. We then compared expression of KIF5B–RET and KIF5B–RETG810A protein in these cells and vandetanib resistance between B/KRG810A and B/KRG810A-REC. As shown in Supplementary Fig. S3C, KIF5B–RET and KIF5B–RETG810A protein levels were similar between B/KR and the B/KR-derived B/KRG810A cells. The independently generated B/KRG810A-REC cells had a 20% higher level of KIF5B–RETG180A protein. The vandetanib IC50s were 5.310 and 5.675 μmol/L for B/KRG810A and B/KRG810A-REC cells (Supplementary Fig. S3D). These data confirm that the RETG810A mutant is resistant to vandetanib. Although we do not rule out the possibility that increased KIF5B–RET expression may confer some extent of drug tolerance, our results indicate that vandetanib resistance of B/KRG810A cells is mainly due to the G810A mutation.
Cabozantinib, vandetanib, ponatinib, and lenvatinib are multikinase TKIs that have been reported in on-going clinical trials of RET-fusion NSCLC (15). We compared these four drugs in the KIF5B–RET-dependent growth of B/KR cells. As shown in Fig. 4D, ponatinib was the most potent inhibitor with an IC50 of 0.07 μmol/L for B/KR cells. B/KRV804L gatekeeper mutation cells were pan-resistant to all four TKIs, resulting in 6.2, 8.7, 15.7, and 76-fold increase in IC50s for ponatinib, vandetanib, cabozantinib, and lenvatinib, respectively (Fig. 4D). However, ponatinib had the least fold increase in IC50 and remained the most potent inhibitor for B/KRV804L cells with a sub-micromolar IC50.
Vandetanib IC50 for B/KRG810A cells was increased 5.7-fold to 5.1 μmol/L compared with B/KR cells. Interestingly, although cabozantinib had similar IC50s in B/KRG810A cells and B/KR cells, ponatinib and lenvatinib had lower IC50s in B/KRG810A cells by 8.7- and 3.6-folds, respectively. Thus, although the RETG810A mutation causes vandetanib resistance, it becomes hypersensitive to ponatinib and lenvatinib. We hereby term this type of mutation as a gain-of-sensitivity (GOS) mutation.
To further compare the potencies of these four drugs for RET and RETV804L and also to determine the potencies of these drugs for another known RET gatekeeper mutant RETV804M, IC50s for inhibition of recombinant RET, RETV804L, and RET V804M in an in vitro kinase assay were measured in parallel via a commercial service (Reaction Biology). As shown in Supplementary Fig. S4, ponatinib again displayed the most potent inhibitory activities against RET, RETV804L, and RETV804M among these four compounds.
Ponatinib induces lung tumor regression in doxycycline-induced C/KR bitransgenic mice
Because ponatinib was the most potent TKI for KIF5B–RET and its gatekeeper mutations, we chose it for the in vivo study in our KIF5B–RET transgenic animal model of NSCLC. Doxycycline-induced C/KR mice (4–5 months) were examined by MRI and μCT. Eight mice with MRI-detected lung tumor lesions were divided into two groups and treated with ponatinib or vehicle (4 mice/group) by oral gavage (30 mg/kg/d, 5 d/wk; see Supplementary Fig. S5 for treatment schedule and body weight monitoring data). After 4 weeks of drug treatment, the lungs of these mice were examined again by MRI and μCT, and lung tissue sections were analyzed by histology.
Lung tumor lesions in the vehicle-treated mice were similar or worse based on MRI and μCT examination (Fig. 5A and B, Supplementary Figs. S6 and S7, Supplementary Information S2, Supplementary Movie). In contrast, the lung lesions were greatly reduced in ponatinib-treated mice. Semiquantitative measurement of tumor lesions by MRI showed that the mean VOI was significantly (P = 0.0286) decreased from 494 ± 83 mm3 (before treatment) to 68 ± 51 mm3 (after treatment) in the ponatinib-treated cohort (Fig. 5B). In comparison, the mean MRI VOI was not significantly changed (P = 0.6571) in the vehicle–control cohort before (580 ± 227 mm3) and after (480 ± 176 mm3) treatment. Analysis of H&E-stained lung tissue sections (Supplementary Fig. S6) by the histology pattern recognition program show that vehicle-treated cohort had a mean 23.25 ± 3.20% of AOI with hyperplasia or neoplasia, whereas ponatinib-treated mice had significantly (P = 0.0286) less AOI (mean = 2.79 ± 0.63%) similar to that of the wild-type mice (2.47% ± 1.17%, Figs. 3B and 5C). In another experiment, lung tissue lysates from doxycycline-induced C/KR mice treated with vehicle or ponatinib for 1 month were analyzed for tyrosine phosphorylation of KIF5B–RET. As shown in Fig. 5D, KIF5B–RET protein was decreased and its tyrosine phosphorylation was suppressed in the lungs of ponatinib-treated animals. Thus, ponatinib is effective in inhibiting KIF5B–RET kinase activity and reducing the lung tumors of C/KR mice.
RET oncogene fusions occur in 1% to 2% of lung adenocarcinoma. The rarity of the molecular lesion presents a barrier for clinical studies of RET fusion–associated lung adenocarcinoma (10). To facilitate in vivo studies of RET-associated lung adenocarcinoma, we have established an inducible transgenic mouse model of KIF5B–RET-associated lung adenocarcinoma.
KIF5B–RET displayed a constitutively active PTK activity as measured by pRET when it was expressed in the lungs of doxycycline-induced C/KR transgenic mice. Hyperproliferative lesions and early tumors in the lungs were observed as early as one month after doxycycline induction of the C/KR mice, and progressed to invasive lung tumors in 4 to 5 months after doxycycline induction of KIF5B–RET expression in the C/KR mice. Lung tumors formed in this doxycycline-induced C/KR model resemble human invasive lung adenocarcinoma and are associated with desmoplastic reaction and visceral pleural involvement. Desmoplastic stroma was also evident in the RET fusion–positive lung adenocarcinoma cases in TCGA. Various histologic patterns of RET fusion–positive lung adenocarcinoma have been observed, including the prevalence of poorly differentiated, solid-predominant histologic features in RET fusion cases of lung adenocarcinoma (17, 27–29). Lung tumors in our C/KR mouse model recapitulate the invasive desmoplastic feature of RET fusion–associated tumors.
Similar to the absence of desmoplastic stroma in the lung tumors of KrasLA1 mice that we observed in this study, a previous study using the adenovirus-Cre recombinase/LSL-RasG12D mouse model also found that lung adenoma and adenocarcinoma developed in that mouse model lack stromal reaction (30). Knockout of tgfbr2 in the LSL-KrasG12D model allows the KrasG12D-driven lung tumors to progress to invasive lung adenocarcinoma with desmoplastic reaction (30). Human NSCLC often can cause desmoplasia when they invade into stroma. Among 230 cases of human lung adenocarcinoma characterized in TCGA (2), there are 1 case with TGFBR2 deletion and 1 case with TGFBR2 point mutation (E257D), both of them occur in KRASG12C tumors (www.cbioportal.org). Although many other lung adenocarcinoma cases in the TCGA cohort show the desmoplastic histology, such as RET-fusion cases, they are not associated with TGFBR2 deficiency. Thus, although TGFBR2 deficiency could promote desmoplastic stroma deposition, other factors are also likely to contribute to the stromal response to the invasive tumors.
Genomic analysis has suggested the exclusive dependency of lung adenocarcinoma on ALK, RET, or ROS fusion oncogenes (10). Consistently, data from our doxycycline withdrawal experiment indicate that lung tumors developed in doxycycline-induced C/KR mice require the continuing expression of the KIF5B–RET fusion oncogene to maintain the malignancy. This property suggests that KIF5B–RET is an effective target for drug treatment to eliminate KIF5B–RET-associated lung adenocarcinoma.
By culturing KIF5B–RET-dependent B/KR cells with RET kinase inhibitors, we identified cabozantinib-resistant RETV804L gatekeeper mutation and a novel vandetanib-resistant RETG810A mutation. Interestingly, although the RETG810A mutant remained sensitive to cabozantinib, it acquired GOS to ponatinib and lenvatinib. Recent clinical experience in EGFR and ALK inhibitor therapy of lung adenocarcinoma has shown that a major mechanism of developing drug resistance is secondary mutations in the targeted PTK kinase domain. Mutation-sensitive secondary TKIs can be developed and used to prolong the therapeutic response (31–33). Our finding here suggests that ponatinib or lenvatinib are the secondary drugs to use when RETG810A mutation occurs in vandetanib-treated RET-associated cancer, including lung adenocarcinoma and metastatic thyroid cancer.
Among cabozantinib, lenvatinib, ponatinib, and vandetanib, our data show that ponatinib is the most potent inhibitor of RET PTK. Treatment of lung tumors induced by KIF5B–RET in the transgenic animals resulted in tumor regression. Thus, ponatinib is an effective drug for treating KIF5B–RET-associated lung adenocarcinoma in this genetically engineered mouse model and is predicted to show antitumor efficacy in RET fusion lung adenocarcinoma patients. Although the gatekeeper RETV804L and RETV804M mutations are less sensitive to ponatinib than the wild-type RET, ponatinib remains the most potent inhibitor (Fig. 4; Supplementary Fig. S4). Thus, ponatinib appears to be the drug of choice for treating RET fusion–associated lung adenocarcinoma.
It is envisioned that more drugs will be developed, re-purposed, and evaluated for therapy of RET fusion–associated lung adenocarcinoma, either as novel single agent or in combination treatment. The immune competent C/KR mouse model that we developed here provides a resource allowing the direct side-by-side comparison of two drugs in preclinical studies or co-clinical trials (34, 35) of RET fusion–associated lung adenocarcinoma.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: J. Wu
Development of methodology: Q. Huang, C. Jin, G.V. Martinez, J. Wu
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Q. Huang, V.E. Schneeberger, C. Jin, R. Afzal, M.M. Budzevich, G.V. Martinez, E.B. Haura, D. Coppola, J. Wu
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Q. Huang, R. Afzal, M.M. Budzevich, R.J. Makanji, G.V. Martinez, T. Shen, L. Zhao, K.-M. Fung, D. Coppola, J. Wu
Writing, review, and/or revision of the manuscript: V.E. Schneeberger, N. Luetteke, C. Jin, G.V. Martinez, K.-M. Fung, E.B. Haura, D. Coppola, J. Wu
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Q. Huang, R. Afzal, E.B. Haura, J. Wu
Study supervision: J. Wu
Other (consultant for design of transgenic vector): N. Luetteke
We thank H. Lee Moffitt Cancer Center Core facility staff, including Animal, DNA sequencing, Histology, Microscopy Core facilities, and Stephenson Cancer Center/University of Oklahoma Health Sciences Center Histology Core facility staff for assistance. We also thank J.A. Whitsett for the CCSP-rtTA transgenic mice, D.C. Radisky and A.P. Fields for advice and assistance, and K. Politi and G. Felsenfeld for reagents.
This study was supported by National Institutes of Health grants R21CA175603 (to J. Wu), R01CA178456 (to J. Wu), P30CA076292 (to T. Sellers), P50CA119997 (to E. B. Haura), and P20GM103639 (to K.M. Fung).