Cabozantinib Is Active against Human Gastrointestinal Stromal Tumor Xenografts Carrying Different KIT Mutations

This article is featured in Highlights of This Issue, p. 2823

Gastrointestinal stromal tumor (GIST) is the most common mesenchymal tumor of the digestive tract and in some regions of Europe constitutes the largest group of sarcomas (1, 2). The vast majority of GISTs express KIT (CD117), a receptor tyrosine kinase (RTK; ref. 1). In about 85% to 90% of cases, GIST is driven by activating mutations either in KIT (70–75% of cases) or in platelet-derived growth factor receptor alpha (PDGFRA) genes (10%–14% of tumors), resulting in a constitutive activation of the encoded RTK (1). This leads to continuous stimulation of important downstream signaling pathways, such as the phosphoinositide 3-kinase (PI3K)/AKT and mitogen-activated protein kinase (MAPK) pathways, ultimately causing uncontrolled cell proliferation and survival (1). The constitutive activation of these pathways via KIT is the basis for the treatment of GIST patients with tyrosine kinase inhibitors (TKI) targeting the mutated receptor (1). The type of KIT/PDGFRA mutation is considered to be a prognostic factor as well as to have a predictive value for response to treatment with a given TKI (3, 4).

Patients with advanced GIST are commonly treated with imatinib, a small-molecule TKI that competitively binds to and inhibits the activation of mutated KIT/PDGFRA (5). Despite initial success, with this drug in the majority of patients, with time in nearly all patients, the disease progresses due to emergence of resistance to imatinib, mainly caused by heterogeneous secondary mutations in the KIT gene (6, 7), which result in structural changes of the receptor that will hinder the binding of the TKI (8). Second- and third-line treatment standards (sunitinib and regorafenib, respectively) are effective but the quality and duration of response tends to be worse than with the first-line imatinib (9, 10). GIST patients who have failed three lines are left without further approved therapeutic options, while the general condition of many of these patients would still allow for further systemic therapy. Hence, there is a need for developing and testing novel treatment strategies for these patients.

Cabozantinib is a multitargeted TKI, with activity against several receptor tyrosine kinases, including KIT, MET, VEGFR2 (vascular endothelial growth factor receptor 2), and AXL (11, 12). In preclinical studies, cabozantinib has been shown to have potent antitumor and antiangiogenic activity in several in vivo and in vitro models (13). Currently, cabozantinib is approved for the treatment of progressive, metastatic medullary thyroid cancer (14). In a recent randomized phase III trial, cabozantinib resulted in improved progression-free survival and objective response rate in patients with advanced renal cell carcinoma as compared with everolimus (15).

In the current study, we investigated the in vivo efficacy of cabozantinib in GIST using patient-derived xenografts carrying heterogeneous KIT mutations.

Imatinib mesylate was purchased from Sequoia Research Products Ltd. and was dissolved in sterile water. Cabozantinib was provided by Exelixis Inc. and was prepared as a suspension in sterile water. The following primary antibodies were used to perform Western blotting (WB) and immunohistochemistry (IHC): phospho-KIT^Y719 (pKIT^Y719), phospho-KIT^Y703 (pKIT^Y703), phospho-AKT^S473(pAKT^S473), AKT, alpha-tubulin, phospho-MAPK^T202/Y204 (pMAPK), p42/44 MAPK (MAPK), 4E-binding protein 1 (4EBP1), phospho-4EBP1 (p4EBP1), phospho-histone H3 (pHH3), and cleaved PARP (poly ADP ribose polymerase) from Cell Signaling Technology, Ki67 from Thermo Scientific, KIT from Dako, DOG1 from Novocastra, beta-actin from Sigma-Aldrich, CD31 from Dianova. For WB secondary antibodies, conjugated with horse radish peroxidase (HRP) from Santa Cruz Biotech and Western Lightning Plus-ECL (PerkinElmer) were applied for band visualization. For IHC anti-rabbit EnVision+ System-HRP and 3′diaminobenzidine-tetrahydrochloride (DAB) both from Dako were used.

Human GIST xenografts were established by subcutaneous transplantation of clinical GIST specimens from consenting patients undergoing surgical interventions or biopsies to female adult NMRI (nu/nu) mice (Janvier Laboratories) as previously described (16). Three PDX models, namely UZLX-GIST4 [passage (p) 21, KIT p.K558_G565delinsR; imatinib-sensitive model], -GIST2 (p. 17, KIT p.A502_Y503dup; dose-dependent resistance to imatinib), and -GIST9 (p. 6, KIT p.P577del;W557LfsX5;D820G; imatinib- and sunitinib-resistant) were used in the study. These models are characterized by human origin and stable morphological features and molecular characteristics similar to the original tumor as previously described (16–18).

A total of 52 mice bearing tumors, engrafted on both sides, were grouped as follows: control (untreated; UZLX-GIST4, n = 7 mice; UZLX-GIST9, n = 6; UZLX-GIST2, n = 4); imatinib 50 mg/kg/bid (UZLX-GIST4, n = 7; UZLX-GIST9, n = 6; UZLX-GIST2, n = 5) and cabozantinib 30 mg/kg/qd (UZLX-GIST4, n = 6; UZLX-GIST9, n = 6; UZLX-GIST2, n = 5).

The treatment was administered orally by gavage and lasted 15 days, and tumor volumes were measured three times per week. The relative values to the baseline tumor volume expressed as percentages were used for each time point. The general wellbeing of mice and their body weight was followed up daily. On day 15, all mice were euthanized 2 hours after the last treatment and tumors were partly snap-frozen and partly fixed in formalin for further histological and molecular assessments.

Xenografting of tumor samples collected from consenting patients was approved by the Medical Ethics Committee, University Hospitals Leuven, Belgium. In vivo experiments were approved by the Ethics Committee for Laboratory Animal Research, KU Leuven and performed according to their guidelines and Belgian regulations.

Formalin-fixed tumor specimens were embedded in paraffin, and 4-μm sections were cut for hematoxylin and eosin (H&E) and IHC stainings. Stained tumor fragments were used for the assessment of histologic response (HR) as previously reported (19). Mitotic and apoptotic activity was evaluated by counting mitotic and apoptotic cells on H&E-stained slides in 10 high power fields (HPF) at 400-fold magnification (400×). Phospho-histone H3 (pHH3) and cleaved PARP immunostainings were used to assess the proliferative and apoptotic activity by counting immunopositive cells in 10 HPF at 400×. The Ki67 labeling index was calculated as the average percentage of Ki67-stained tumor cells in 5 pictures (400×), to assess proliferation. The antiangiogenic activity was assessed by measuring the tumor vessel density in ex-mouse tumors using CD31 immunostaining, defined as the number of vessels counted in 5 HPF at 200-fold magnification (200×). KIT and DOG1 immunostaining was used to confirm the GIST diagnosis of the ex-mouse tumors and to follow KIT expression. For all evaluations we used a CH30 microscope equipped with Color View digital camera (Olympus), and images were analyzed using Cell D imaging software (Olympus).

For WB, tumor lysates were prepared from the snap-frozen tumor specimens as described previously (16). Levels of chemiluminescence were captured with the LAS mini 3000 system (Fuji).

Comparisons between the tumor volumes on day 1 versus day 15 and between treatment groups were done using Wilcoxon-matched pair (WMP) and the Mann–Whitney U (MWU) tests, respectively. Statistica 12.0 (StatSoft) was used for all statistical analyses, and a P value of <0.05 was considered statistically significant.

After two weeks, untreated tumors from all models showed a significant increase in tumor volume from baseline (P < 0.05 for all models; WMP; Fig. 1).

As expected, imatinib caused significant tumor regression compared with baseline in UZLX-GIST4 (to 37% of the baseline volume; P = 0.005, WMP). In the two other models, despite treatment with imatinib, the tumor volume increased, reaching 139% in UZLX-GIST2 and 199% in UZLX-GIST9, which was similar to the tumor growth seen in untreated controls (Fig. 1).

In contrast, treatment with cabozantinib decreased the tumor volume in all three models. In the imatinib-sensitive model, cabozantinib treatment resulted in a significant regression of the tumor volume on day 15 (to 30% of baseline; P = 0.002, WMP). In the KIT exon 9-mutant model, cabozantinib significantly reduced the tumor volume to 51% of baseline (P = 0.005, WMP) and the effect was more pronounced than in the imatinib-treated tumors (P < 0.001, MWU). In the imatinib-resistant model, cabozantinib slowed tumor growth (141%), and the volume of the xenograft was significantly lower than the tumor volume of the control group (P < 0.001, MWU) and imatinib (P = 0.01, MWU), on day 15 (Fig. 1). Of note, no relevant treatment-related toxicity was observed during the two weeks of treatment.

Control tumors from all models showed stable morphological and immunohistochemical characteristics in terms of KIT and DOG1 immunopositivity (Supplementary Fig. S1), resembling the features observed in the original patient sample used for xenografting and previous passages. Moreover, ex-mouse tumor samples maintained the same KIT mutations as previously found.

In all models, the HR was mainly characterized by the induction of necrosis, as assessed on H&E. In the UZLX-GIST9 model, cabozantinib resulted in strong (grade 2 and 3) HR, observed in about 50% of tumors, while only a minimal HR was found in UZLX-GIST4 and UZLX-GIST2 (Fig. 2).

Subsequently, we evaluated the effect of the different treatments on the proliferative and apoptotic activity of the tumors. Regardless of the model, control tumors showed high mitotic activity with the average of 36 mitotic figures per 10 HPF. Imatinib treatment resulted in a significant reduction of mitotic activity only in UZLX-GIST4 (>50-fold reduction compared with control, P = 0.004; MWU). In contrast, cabozantinib significantly reduced the mitotic activity in all three models compared with control (P = 0.002 in UZLX-GIST4, P < 0.001 in UZLX-GIST2, and P = 0.01 in UZLX-GIST9; MWU), with a more pronounced effect in the first model (Table 1). In the other two models, -GIST2 and -GIST9, cabozantinib significantly reduced the mitotic activity also when comparing with imatinib treatment. The assessment of apoptotic activity showed that cabozantinib significantly increased the apoptosis only in the UZLX-GIST4 model, when compared with control (>50-fold, P = 0.002; MWU) and imatinib (15-fold, P = 0.004; MWU). These results were confirmed using immunohistochemical markers (pHH3 and Ki67) for proliferation and cleaved PARP for apoptotic activity (Table 1).

Assessment of MVD using CD31 staining showed imatinib treatment did not have a significant impact on the vessel count in any of the three models. In contrast, compared with both control and imatinib, cabozantinib significantly reduced the average number of CD31-positive vessels in all models, regardless of their molecular background (Table 1; Fig. 3).

The WB analysis showed that KIT and its main downstream intermediate signaling proteins were expressed and activated in all three models (Fig. 4). As expected, in UZLX-GIST4, imatinib inhibited phosphorylation of pKIT^Y719 and pKIT^Y703 as well as pAKT^S473 compared with control. Cabozantinib inhibited the phosphorylation of KIT at both phosphorylation sites (pKIT^Y719 and pKIT^Y703), with a more pronounced inhibition of the latter. In the UZLX-GIST2 model, cabozantinib resulted in strong inhibition of KIT, AKT, MAPK, and 4EBP1 phosphorylation when compared with the untreated controls. On the other hand, no remarkable inhibition of KIT or its downstream intermediates was observed in UZLX-GIST9 (Fig. 4).

In spite of the tremendous advances in the treatment of metastatic GIST with the development of several TKI therapies, the problem of resistance to these agents presents a considerable clinical challenge. To date, many patients with advanced GIST are left without any approved therapy options within a relatively short period of time after starting systemic treatment, mainly due to the emergence of resistance to the approved agents imatinib, sunitinib, and regorafenib. The development of novel therapeutic options that can target possible resistance mechanisms is a key research priority.

Our current study reveals that cabozantinib is active in PDX models of GIST carrying heterogeneous KIT mutations and different levels of sensitivity to imatinib. These models have already proven their utility for preclinical drug testing and supported the initiation of clinical trials (16, 17).

Cabozantinib is a novel compound, with an inhibitory effect on KIT as well as on other receptors such as MET, VEGFR2, and AXL (12), some of which have been implicated in resistance to imatinib (11). Using several cellular assays and in vivo experiments, Yakes and colleagues have shown the kinase inhibition and antiproliferative effects of cabozantinib in several preclinical solid tumor models (13). In GIST models, cabozantinib demonstrated profound antitumor effects in vivo (20). In our experiments, treatment with cabozantinib resulted in substantial tumor regression in two models with primary KIT mutations, and tumor growth delay in an imatinib-resistant model. The reduction in tumor burden was accompanied by a significant reduction in proliferation of the tumor cells; the mitotic activity was completely suppressed in UZLX-GIST4 and significantly decreased in the other two models. Furthermore, cabozantinib substantially decreased KIT and AKT activation in two out of three models. In UZLX-GIST4, the efficacy of cabozantinib on proliferation and KIT signaling was similar to that of imatinib, suggesting that cabozantinib has a direct effect on GIST cells through inhibition of KIT. On the other hand, the tumor growth delay in the UZLX-GIST9 model was associated with induction of necrosis which was not observed in the other models. The presence of necrosis may sometimes lead to an underestimation of the antitumor effect of the treatment as the necrotic tissue may replace the tumor and result in a temporary increase in tumor volume instead of tumor regression due to osmotic changes within the necrotic lesions (21). In refractory disease, tumor growth delay resulting from treatment can be considered a beneficial effect and hence prolong time to progression and potentially improve survival (22, 23).

Cabozantinib induced apoptotic activity in UZLX-GIST4, which was much more pronounced than the effect of imatinib. It has been suggested that imatinib as a cytostatic drug is not able to eradicate all GIST cells and that a subset of tumor cells remain viable and enter into a state of reversible cell quiescence rather than going into apoptosis (24, 25). The remarkable proapoptotic effect of cabozantinib can possibly be linked its antiangiogenic effects. Using PDX models of colorectal cancer, Song and colleagues showed that cabozantinib inhibited tumor angiogenesis which was accompanied by increased apoptotic activity (26). Interestingly, in all xenografts tested, we observed a significant reduction in MVD in cabozantinib treated-tumors compared with the control and imatinib. Moreover, in previous studies published by our group, sunitinib, another multitarget TKI with anti-angiogenic properties, also resulted in a prominent increase in apoptotic activity in the UZLX-GIST4 model as compared with imatinib (27, 28). In the UZLX-GIST2 and -GIST9 models, however, the antiangiogenic effect was not accompanied by a significant increase in apoptotic activity. This could be due to the presence of a high rate of spontaneous apoptosis among the untreated control mice; therefore, the treatment effect could not be observed. Moreover, in experiments performed earlier by our group, treatment of UZLX-GIST9 or -GIST2 models with the angiogenesis inhibitor sunitinib showed an effect similar to that of cabozantinib on tumor volume and MVD but did not result in an increase of apoptotic activity, suggesting the antiangiogenic effects in these models may not induce apoptosis but rather necrosis and/or inhibition of proliferation (18, 27). These observations support the investigation of cabozantinib in patients with GIST and suggest a potential advantage over imatinib.

Cabozantinib is known to be a potent antiangiogenic compound, which is likely a result of its inhibitory effect on kinases, including VEGFR2 MET, and AXL (12, 13). In GIST, a high MVD is correlated with tumor burden, VEGF expression, and cell proliferation and was shown to be a poor prognostic factor (29). In our models, cabozantinib led to a significant reduction in microvascularization, as assessed by CD31 immunostaining, and the effect was more pronounced than that in imatinib-treated tumors. Similar results were shown previously in Kit^V588del/−, which is a genetically modified GIST mouse model (20), but to our knowledge this is the first report of cabozantinib's activity in PDX models of GIST established by immediate transplantation of tumor material. The level of antiangiogenic activity was similar in all three models in our study, suggesting that this effect may be independent of the tumor's molecular background, which may be an important advantage of this compound, as other drugs used in this setting tend to work genotype specific.

As expected, cabozantinib strongly inhibited the KIT signaling pathway in two models with primary mutations, while in the double-mutant, imatinib-resistant model, the inhibition was only marginal. Nevertheless, also in the latter model, cabozantinib led to a delay in tumor growth and strong inhibition of proliferation, which was more pronounced than in the imatinib-treated group. We postulate that in this case cabozantinib's activity is mainly attributable to its effects on the tumor vasculature rather than directly affecting the GIST cells. Since cancer cell survival and proliferation requires nutrient and oxygen supply through angiogenesis, the blockade of neovascularization could lead to the inhibition of proliferation (30). Recently, a similar observation was reported by our group in PDX models of liposarcoma when treated with the antiangiogenic multikinase inhibitor pazopanib, which showed antitumor activity mainly through inhibition of angiogenesis and proliferation (31).

As a multitargeted kinase inhibitor, cabozantinib can act through blocking several RTKs, including KIT, but also other kinases some authors postulate a potential role in GIST as shown in several preclinical models and patient samples (20, 32, 33). One of cabozantinib's targets is MET, the tyrosine kinase receptor for hepatocyte growth factor which is implicated in the resistance to different TKIs or monoclonal antibodies in several preclinical models and human tumors (34–36). In GIST, Cohen and colleagues showed that tumor samples from imatinib-resistant tumors can have acquired MET activation. In their study, treatment with MET inhibitors led to enhanced efficacy compared with imatinib in both sensitive and resistant models (20). AXL is another target inhibited by cabozantinib, which is possibly overexpressed in imatinib-resistant tumors, taking over the activity of KIT when inhibited by imatinib (“tyrosine kinase switch” hypothesis; refs. 37, 38). In our study, however, we did not observe MET or AXL expression on both transcript and protein levels in the ex-mouse tumors (data not shown); therefore, we could not attribute the antitumor effect of cabozantinib to the inhibition of these alternative RTKs in the GIST cells. We postulate that in our GIST PDX models, cabozantinib is acting mainly through the direct inhibition of KIT and its profound antiangiogenic effects on the tumor vasculature.

In conclusion, cabozantinib demonstrated antitumor efficacy in patient-derived GIST xenograft models characterized by different KIT genotypes and different sensitivities to imatinib. The kinase inhibitor decreased the tumor burden and reduced tumor growth mainly through a significant decrease in cell proliferation and antiangiogenic effects, and to some extent through inhibition of KIT/AKT signaling. In imatinib-resistant models, cabozantinib was more active than the standard treatment. This observation warrants further clinical assessment of the compound in GIST patients refractory to imatinib. Based on our xenograft findings, the European Organization for Research and Treatment of Cancer (EORTC) is now activating an early phase II trial testing cabozantinib in patients with metastatic GIST, who progressed during treatment with imatinib and sunitinib (EORTC 1317 “CABOGIST,” NCT00216578).

For all conflicts related to institutional activities, P. Schöffski's institution was compensated for some of his activities but not him personally. T. Van Looy is currently a clinical research coordinator at Janssen Pharmaceutical companies of Johnson and Johnson. D.T. Aftab is EVP, Business Operations at Exelixis and has ownership interest (including patents) in the same. No potential conflicts of interest were disclosed by the other authors.

Conception and design: P. Schöffski, T. Van Looy, D.T. Aftab, R. Sciot, A. Wozniak

Development of methodology: Y.K. Gebreyohannes, P. Schöffski, T. Van Looy, J. Wellens, L. Vreys, M. Debiec-Rychter, R. Sciot, A. Wozniak

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y.K. Gebreyohannes, P. Schöffski, T. Van Looy, J. Wellens, L. Vreys, D.T. Aftab, M. Debiec-Rychter, A. Wozniak

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y.K. Gebreyohannes, P. Schöffski, T. Van Looy, R. Sciot, A. Wozniak

Writing, review, and/or revision of the manuscript: Y.K. Gebreyohannes, P. Schöffski, T. Van Looy, J. Cornillie, D.T. Aftab, M. Debiec-Rychter, R. Sciot, A. Wozniak

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): P. Schöffski, T. Van Looy, J. Wellens, L. Vreys, U. Vanleeuw

Study supervision: P. Schöffski, T. Van Looy, A. Wozniak

Other (development of the xenograft platform that was the basis for this research): P. Schöffski

Exelixis Inc., South San Francisco, California provided cabozantinib and financial support for the presented studies. P. Schöffski received honoraria for educational and advisory functions from Exelixis.

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.