Purpose: Imatinib dramatically reduces gastrointestinal stromal tumor (GIST) 18F-FDG uptake, providing an early indicator of treatment response. Despite decreased glucose internalization, many GIST cells persist, suggesting that alternative metabolic pathways are used for survival. The role of mitochondria in imatinib-treated GIST is largely unknown.
Experimental Design: We quantified the metabolic activity of several human GIST cell lines. We treated human GIST xenografts and genetically engineered KitV558del/+ mice with the mitochondrial oxidative phosphorylation inhibitor VLX600 in combination with imatinib and analyzed tumor volume, weight, histology, molecular signaling, and cell cycle activity. In vitro assays on human GIST cell lines were also performed.
Results: Imatinib therapy decreased glucose uptake and downstream glycolytic activity in GIST-T1 and HG129 cells by approximately half and upregulated mitochondrial enzymes and improved mitochondrial respiratory capacity. Mitochondrial inhibition with VLX600 had a direct antitumor effect in vitro while appearing to promote glycolysis through increased AKT signaling and glucose transporter expression. When combined with imatinib, VLX600 prevented imatinib-induced cell cycle escape and reduced p27 expression, leading to increased apoptosis when compared to imatinib alone. In KitV558del/+ mice, VLX600 alone did not induce tumor cell death, but had a profound antitumor effect when combined with imatinib.
Conclusions: Our findings show that imatinib alters the metabolic phenotype of GIST, and this may contribute to imatinib resistance. Our work offers preclinical proof of concept of metabolic targeting as an effective strategy for the treatment of GIST. Clin Cancer Res; 24(4); 972–84. ©2017 AACR.
Metabolic reprogramming is an emerging hallmark of cancer with the potential for therapeutic targeting. Although imatinib's effect on GIST glucose uptake has been described, compensatory metabolic changes secondary to imatinib therapy are not well defined. Here, we demonstrate that imatinib reduces glucose uptake, which decreases tumor cell glycolytic activity and enhances mitochondrial respiratory capacity. Inhibiting mitochondrial oxidative phosphorylation augments the efficacy of imatinib, highlighting an important limitation of imatinib monotherapy and supporting the use of metabolic therapies in GIST.
Gastrointestinal stromal tumor (GIST) is the most common sarcoma (1), often arising from an oncogenic mutation in the KIT receptor tyrosine kinase (RTK; ref. 2). While imatinib has revolutionized the treatment of GIST through targeted KIT inhibition (3), resistance and disease progression often develop within 2 years (4, 5). The emergence of secondary KIT mutations and the activation of alternative oncogenic RTKs have been identified in the majority of resistant tumors (6–9). Promising second- and third-line therapies have only provided minimal benefit (10, 11), highlighting the need for novel therapeutic approaches.
Otto Warburg originally hypothesized that cancer is characterized by dysfunctional mitochondria, observing that even in the presence of oxygen, cancer cells exclusively use glucose to produce lactate (12–15). It was later discovered that cancer cells preferentially use glucose because it provides a unique growth advantage by more rapidly generating one-carbon intermediates, fatty acids, cholesterol, and nucleotides for DNA synthesis and cell division (16, 17). Studies in K-Ras–driven lung cancer and murine melanoma also established an integral role of cancer cell mitochondria in tumor initiation and growth (18, 19), ultimately disproving Warburg's original hypothesis. Nevertheless, the Warburg effect of cancer cells preferentially utilizing glucose in the presence of oxygen defined the cancer metabolic phenotype and led to 18F-fluorodeoxyglucose (FDG) positron emission tomography (PET) scanning as a reliable modality for tumor imaging.
GIST demonstrates intense metabolic activity and avid glucose uptake on 18F-FDG PET scanning (20). While imatinib-related changes in tumor volume can take months to occur (21), decreased glucose uptake on 18F-FDG-PET scan can be seen within 24 hours of imatinib initiation (22). Prospective studies have correlated an early metabolic response on 18F-FDG-PET scanning with improved progression-free survival, validating the relevance of this cancer metabolic phenotype (20, 21, 23). Interestingly, imatinib-induced decreases in glucose uptake coincide with the emergence of a quiescent cell phenotype characterized by reversible exit from the cell division cycle (24). While the cell-cycle regulating DREAM complex and the anaphase promoting complex (APC)–p27Kip1 signaling axis are known to be mediators of imatinib-induced quiescence (24, 25), reliance on mitochondrial function for energy is also a property of quiescent tumor cells (26) and may represent another mechanism of imatinib resistance in GIST.
The effect of imatinib on GIST mitochondrial dynamics is largely unknown. Because imatinib treatment dramatically reduces glucose uptake and promotes tumor cell quiescence in GIST, we hypothesized that alterations in mitochondrial metabolism associated with tumor cell dormancy contribute to imatinib resistance. Herein, we investigate the role of mitochondrial bioenergetics in GIST and show that imatinib treatment shifts the metabolic phenotype of GIST from glycolysis to oxidative phosphorylation (OxPhos). Furthermore, inducing mitochondrial dysfunction in the presence of imatinib with a mitochondrial OxPhos inhibitor increased tumor cell dependence on glycolysis, enhancing the apoptotic and anti-glycolytic effects of imatinib in vitro and in vivo.
Materials and Methods
Mice and treatments
Age- and sex-matched KitV558del/+ mice (27) and NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG) mice (The Jackson Laboratory) were used. Our Institutional Animal Care and Use Committee approved all procedures and treatments. For xenograft experiments, 106 GIST-T1 cells in PBS were mixed 1:1 with BD Matrigel Matrix Growth Factor Reduced (BD Biosciences) and injected subcutaneously into the right flanks of NSG mice. Xenograft tumor volume was measured using the modified ellipsoid formula of 0.5 × length × width × width, and treatments were started when mean tumor volume reached approximately 100 mm3. Imatinib (600 mg/L in drinking water) was provided by Novartis and VLX600 (4.5 mg/kg dissolved in 4.23 mmol/L HCl and diluted with 0.9% NaCl, administered intravenously) was purchased from EMD Millipore. Control mice received regular drinking water and vehicle 4.23 mmol/L HCl in 0.9% NaCl intravenously.
Cell lines and cell viability assay
The human GIST-T1 cell line (KIT exon 11, provided by Dr. Takahiro Taguchi, Kochi Medical School; ref. 28), the human HG129 cell line (KIT exon 11, which we established in 2012; ref. 8), and the human GIST-882 cell line (KIT exon 13, kindly provided by Dr. Jonathan Fletcher, Brigham and Woman's Hospital) have been described previously (29). Kit expression and mutation status was confirmed in all cell lines by Western blot and sequencing. Mycoplasma testing was performed in GIST-T1 and HG129 cell lines prior to use, most recently in 2015. Experiments were performed within two months of thawing early passage cells. Cells were cultured in complete medium containing RPMI-1640 with 10% fetal bovine serum, 2 g/L glucose, 2 mmol/L l-glutamine, 50 U/mL penicillin–streptomycin, and 10 mmol/L HEPES. To measure cell viability, 104 GIST-T1, GIST-882, and HG129 cells were seeded in replicates of 5 in a 96-well flat-bottomed plate (Falcon) and cultured for 48 to 72 hours with imatinib, VLX600, or H20- and/or DMSO-only solvent controls. Viability was measured using a colorimetric tetrazolium salt assay (Cell Counting Kit-8, Dojindo Molecular Technologies) per the manufacturer's instructions, and values were normalized to control. For low glucose cell viability assays, identical media were used except the glucose concentration was 0.25 g/L.
Glucose consumption and lactate production measurements
Glucose uptake and lactate production of live cells were measured enzymatically using a YSI 7100 enzymatic analyzer (Life Sciences). Cells were plated in 6-well plates and allowed to achieve 80% to 90% confluency. Cells were then exposed to drugs for 12 hours after seeding. At the end of the assay, media were collected and kept at −20C prior to measurement. Metabolite concentrations were normalized to cell number.
Real-time metabolic analyses
Metabolic analysis of live cells was performed using the Seahorse XFe96 extracellular flux analyzer (Seahorse Bioscience). Cells were seeded in XF96 well plates (20,000–50,000 cells per well) and incubated overnight at 37°C in 5% CO2. Cells were then exposed to drugs of interest for 24 hours prior to the assay, unless otherwise indicated. For Mito Stress and Fuel tests, cell culture medium was replaced 1 hour before the assay with minimal DMEM without phenol red supplemented with 1 mmol/L sodium pyruvate, 10 mmol/L glucose, and 2 mmol/L glutamine. Oxygen consumption rate (OCR) was analyzed at basal conditions and after sequential injections of oligomycin (1 μmol/L), carbonyl cyanide-4-(trifluoromethoxy)phenylhydrazone (FCCP) (2 μmol/L), and antimycin/rotenone (0.5 μmol/L). For analysis of mitochondrial fuel oxidation pathways, a Mito Fuel Flex test (Seahorse Bioscience) was performed according to the manufacturer's instructions at the recommended drug concentrations, and the Seahorse XF Mito Fuel Report Generator was utilized to produce results. For Glycolysis Stress tests and Glycolytic Rate assays, cell culture medium was replaced 1h before the assay with minimal DMEM without phenol red supplemented with 2 mmol/L glutamine or minimal DMEM without phenol red supplemented with 1 mmol/L sodium pyruvate, 10 mmol/L glucose, 2 mmol/L glutamine, and 5.0 mmol/L HEPES, respectively. Extracellular acidification rate (ECAR) was measured under basal conditions and after sequential injections of glucose (10 mmol/L), oligomycin (1 μmol/L), and 2-deoxyglucose (100 mmol/L). For characterization of glycolytic rate, a glycolytic rate assay (Seahorse Biosciences) was performed according to the manufacturer's instructions and the Seahorse XF Glycolytic Rate Assay Report Generator was utilized to calculate proton efflux rate. All metabolic assays were normalized to cell number or total protein content as previously described by Bradford assay. Relative changes in maximum mitochondrial respiration and spare respiratory capacity (SRC) were calculated by normalizing to basal OCR when indicated.
RNA sequencing of vehicle and imatinib-treated KitV558del/+ mouse tumors was performed by our Integrated Genomics Core facility and normalized using the software package DESeq. The expression data for genes involved in the following metabolic pathways (glycolysis, pyruvate metabolism, and oxidative phosphorylation) were visualized with heatmap.2 (www.cran.r-project.org).
Protein from flash frozen GIST tissue or cell lines was analyzed as before (30), with the exception of mitochondrial OxPhos protein, which was not boiled and transferred to a PVDF membrane as recommended by Abcam. Antibodies used against glyceraldehyde-3-phosphate dehydrogenase (GAPDH; Clone D16H11), p-Kit (y719), KIT (D13A2), p-AKT (S473), AKT (11E7), p-ERK (T202/Y204), ERK (ERK1/2), GLUT-1 (D3T3A), and β-actin (13E5) were purchased from Cell Signaling Technologies. Antibody to p27 (57/Kip1/p27) was purchased from BD Biosciences. Antibody cocktail to mitochondrial OxPhos proteins (ab110413) was purchased from Abcam with rat heart mitochondria used as a positive control (Abcam).
Quantitative real-time PCR
Total RNA was extracted from KitV558del/+ mouse bulk tumor, reverse transcribed, and amplified with PCR TaqMan probes for mouse Hif1a (Mm00468869_m1), COX I (Mm04225243_g1), COX II (Mm03294838_g1), COXIVi1 (Mm01250094_m1), and TFAM (Mm00447485_m1). Quantitative PCR was performed using a ViiATM7 real-time PCR system (Applied Biosystems). Data were calculated by the 2−ΔΔCt method as described by the manufacturer's protocol and were expressed as fold increase over the indicated controls.
Flow cytometric analysis was performed on KitV558del/+ mouse tumors as before (8). All cells were analyzed on a BD LSR Fortessa (BD Biosciences). Mouse-specific antibodies for CD45 (Clone: 30F11), F480 (BM8), and CD4 (RM4-5) were obtained from Biolegend. Mouse-specific antibodies for CD11b (M1-70), Ly6G (1A8), Ly6C (AL-21), CD3 (17A2), NK1.1 (PK136), and Kit (2B8) were obtained from BD Biosciences. Mouse antibodies for MHCII (M5/114.15.12), CD8 (53.6.7), and FoxP3 (FJK-16S) were obtained from eBioscience. Intracellular Ki-67 staining of live cells was performed using an eBioscience Fixation and Permeabilization Buffer kit, Zombie Aqua fixable viability dye (Biolegend), and human antibody to Ki-67 (Ki-67) or isotype (MOPC-21). For ROS staining, CellROX Green Flow Cytometry Assay Kit was purchased from Thermo-Fisher Scientific and utilized per the manufacturer's instructions. For Annexin V staining, an Annexin V staining kit was purchased from BD Biosciences and used as instructed. For cell cycle analysis, cells were treated in 6-well plates for 24 hours and fixed in 70% ethanol, digested in 0.1 mg/mL RNase A, and stained with propidium iodide. G0–G1, S, and G2–M phases were calculated by Watson Pragmatic using FlowJo software (FlowJo, LLC).
Tumors were fixed in 4% paraformaldehyde, embedded in paraffin, and cut into 5-μm sections. Hematoxylin and eosin (H&E) was performed using standard methods. p27-Kip1 (Clone K25020) and Ki67 (Vector Laboratories) staining was performed by our institutional Molecular Cytology Core. Slides were scanned with MIRAX scan (Zeiss) and analyzed with Pannoramic Viewer.
Data were analyzed using Prism 6.0 (GraphPad Software). An unpaired two-tailed Student t test was performed when applicable. Data were considered statistically significant when P < 0.05.
Imatinib decreases glucose uptake and the glycolytic capacity of GIST cell lines
Using a genetically engineered KitV558del/+ mouse model in which a single intestinal GIST develops (27), we previously showed that 1 week of imatinib dramatically reduced tumor glucose uptake by 18F-FDG PET (31). It has also been shown that imatinib impaired the recruitment of glucose transporters to the GIST cell surface leading to decreased 18F-FDG uptake in vitro (32, 33). In order to quantify the metabolic changes, we treated GIST-T1, HG129, and GIST-882 cell lines with various doses of imatinib below their respective IC50 values of 50 nmol/L, 50 nmol/L, and 100 nmol/L, and enzymatically measured glucose consumption and lactate production of live cells. After 12 hours of 40 nmol/L imatinib, glucose uptake decreased by about half in both GIST-T1 and HG129 cells (Fig. 1A and B). At baseline, GIST-882 cells consumed very little glucose, which was unchanged with imatinib (Fig. 1C). Similarly, 12 hours of 40 nmol/L imatinib treatment significantly decreased lactate production by 53% in GIST-T1 and 61% in HG129 cells (Fig. 1A and B), while lactate production in GIST-882 cells was reduced but minimal at baseline (Fig. 1C).
To further define the metabolic impact of the imatinib-induced reduction in glucose uptake by tumor cells, we performed a glycolysis stress test using a Seahorse extracellular flux analyzer, which can measure glycolysis as a function of ECAR. Analysis of the rapidly growing GIST-T1 and HG129 cell lines showed that glycolysis was reduced in live cells by 70% and 79% within 24 hours of 100 nmol/L imatinib treatment, respectively, while glycolytic reserve was decreased by 85% and 80% (Fig. 1D and E). Conversely, slowly growing GIST-882 cells had a low glycolytic activity and glycolytic reserve at baseline (Fig. 1F). Because it has recently been criticized that Seahorse glycolysis stress tests may include non-glycolytic sources of media acidification in glycolysis measurement, we also performed a glycolytic rate assay, which accounts for mitochondrial sources of acidification and directly measures glycolysis as a proton efflux rate. Analysis of GIST-T1 and HG129 cell lines showed that 60% of proton efflux was attributed to glycolysis, which decreased significantly with imatinib therapy (Fig. 1G and H). In GIST-882, only 30% of proton efflux was attributed to glycolysis, which was significantly reduced with higher concentrations of imatinib (Fig. 1I). Together, these data confirm that GIST-882 is the least glycolytic and imatinib treatment significantly reduced glycolytic activity in all three human GIST cell lines.
Purely glycolytic metabolism results in two lactate molecules produced per glucose molecule consumed. Interestingly, baseline lactate production per glucose molecule was less than two in all three GIST cell lines (Fig.1J–L), indicating utilization of lactate by tumor cell mitochondria (34). In HG129 cells, lactate production per glucose molecule decreased significantly with 40 nmol/L imatinib (Fig. 1K), supporting our findings that HG129 cells become non-glycolytic with imatinib therapy. However, the ratio of the number of lactate molecules produced per glucose molecule consumed during imatinib treatment did not significantly change in GIST-T1 or GIST-882 cells (Fig. 1J and L), despite an overall decrease in glucose consumption and lactate production (Fig. 1A–C), indicating maintenance of mitochondrial function in these imatinib-treated tumor cells. Thus, imatinib treatment decreases lactate production but mitochondrial lactate utilization remains consistent in GIST-T1 and GIST-882 cell lines.
Imatinib therapy is associated with alterations in mitochondrial dynamics
We sought to identify the impact of imatinib treatment on mitochondrial dynamics. We discovered that prolonged imatinib treatment increased the expression of key mitochondrial enzymes in our KitV558del/+ mouse model of GIST. After 3 weeks of treatment with imatinib, RNA sequencing of bulk tumor revealed that the relative RNA expression of enzymes associated with glycolysis was coordinately downregulated while mitochondrial enzymes associated with the tricarboxylic acid cycle (TCA) were coordinately upregulated (Fig. 2A). Similarly, the RNA expression levels of key mitochondrial electron transport chain enzymes detected by quantitative real-time PCR, including cytochrome c oxidase (Cox) 1, 2, and 4, and mitochondrial transcription factor A (Tfam), were significantly upregulated within 2 weeks of imatinib treatment in our mouse model, while Cox 1, 2, and 4 protein expression in the GIST-T1 cell line was higher within 24 hours of 100 nmol/L imatinib treatment (Fig. 2B).
To determine the functional significance of these transcriptional changes, we performed a mitochondrial stress test using a Seahorse extracellular flux analyzer to measure mitochondrial bioenergetics. Mitochondrial SRC and maximum respiratory capacity (MRC) are mediated by mitochondrial Cox 2 and have been shown to enhance cellular viability in stressful environments (35). After 24 hours of treatment, imatinib significantly increased the SRC of remaining GIST-T1 and HG129 cells by 63% and 50%, respectively, while the SRC of the slowly growing GIST-882 cells remained unchanged (Fig. 2C). Real-time metabolic analysis also showed that imatinib treated GIST-T1 and HG129 cells had a significantly higher MRC (Fig. 2D).
Given an apparent enhancement of mitochondrial respiratory activity in the presence of imatinib-induced decreased glucose uptake, we performed additional assays to further characterize GIST mitochondrial activity. First, we performed a Seahorse mito fuel assay, which measures the mitochondria's potential to oxidize three different mitochondrial fuels. Interestingly, imatinib treatment did not reduce GIST mitochondrial dependency or capacity to oxidize glucose (Fig. 2E). Imatinib treatment, however, did enhance the capacity for mitochondrial glutamine and fatty acid oxidation while increasing overall dependency on fatty acids for energy (Fig. 2E). We also observed that imatinib treatment decreased intracellular reactive oxygen species (ROS) in the remaining tumor cells (Fig. 2F), and decreasing tumor cell ROS has been shown to reduce GIST proliferation and invasiveness (36). Importantly, ROS production by immune cells in imatinib-treated tumors remained unchanged (Supplementary Fig. S1), indicating a tumor cell–specific effect. Together, these data show that imatinib significantly altered the metabolic phenotype of GIST under stress to enhance mitochondrial capacity and promote cellular survival, supporting our hypothesis that imatinib induces a quiescent phenotype reliant on mitochondrial energy (26).
Mitochondrial inhibition exerts an antiproliferative effect on human GIST cell lines and xenografts
Given our findings of increased mitochondrial capacity and function in cells surviving imatinib treatment, we sought to inhibit mitochondrial function as an adjunct to imatinib therapy. VLX600 is a mitochondrial OxPhos inhibitor shown to have an antitumor effect on quiescent HCT116 colon carcinoma tumor cells in vitro at a concentration of 6 μmol/L (37). Its primary mechanism was identified to be inhibition of Cox 1, 2, and 4 of the electron transport chain. Because imatinib treatment upregulated transcription and expression of Cox 1, 2, and 4 (Fig. 2B), we hypothesized that treatment with VLX600 would synergize with imatinib treatment. In vitro, imatinib plus VLX600 at doses higher than 1 μmol/L decreased GIST-T1 cell viability when compared with imatinib or VLX600 alone (Fig. 3A). Interestingly, VLX600 monotherapy was efficacious in multiple GIST cell lines in vitro (Fig. 3B; Supplementary Fig. S2), likely as a result of baseline mitochondrial utilization by cell lines (Fig. 1J–L).
To better understand the effect of combination VLX600 and imatinib on mitochondrial function, we again performed a mitochondrial stress assay using the Seahorse extracellular flux analyzer. We primarily utilized the GIST-T1 cell line because metabolic analyses showed a significant glycolytic capacity with maintenance of mitochondrial function (Fig. 1D and J). As monotherapy, VLX600 immediately reduced the OCR and the MRC of GIST-T1 cells via OxPhos inhibition (Fig. 3C), which is opposite of the effect we observed with imatinib (Fig. 2C and D). Similar to what has been previously reported, VLX600 did not reduce the mRNA or protein expression of Cox 2 and 4 with or without imatinib (data not shown; ref. 26). More importantly, however, moderate- (50 nmol/L imatinib + 6 μmol/L VLX) and high-dose (100 nmol/L imatinib + 12 μmol/L VLX) combination therapy significantly decreased functional mitochondrial OCR and ECAR within 12 hours of treatment (Fig. 3D; Supplementary Fig. S3A), while low-dose combination therapy (50 nmol/L imatinib + 1 μmol/L VLX600) did not significantly alter mitochondrial OCR or ECAR (Supplementary Fig. S3B). Thus, mitochondrial inhibition synergized with imatinib in vitro to decrease GIST viability and mitochondrial oxygen consumption via OxPhos inhibition.
Next, we performed xenograft experiments to explore the potential for mitochondrial inhibition in vivo. VLX600 treatment alone resulted in a significant decrease in GIST-T1 tumor volume compared with vehicle controls (Fig. 3E), likely reflective of the utilization of mitochondria we observed in vitro. While there were occasional areas of necrosis in untreated xenograft tumors, particularly in large tumors, only combination therapy induced consistent histologic effects (Fig. 3F). Importantly, we did not observe any significant hematologic abnormalities from VLX600 in our mouse models (Supplementary Fig. S4), suggesting minimal systemic toxicity.
Mitochondrial inhibition reverses the phenotype of imatinib-treated cells
GIST is an excellent model of oncogene addiction because a single KIT mutation often drives tumor proliferation (38), and inactivation of that mutation is detrimental to tumor cell survival (38, 39). Having shown that mitochondrial inhibition combined with imatinib significantly decreased tumor cell viability in vitro and induced histologic effects in vivo, we wanted to explore the effect of mitochondrial inhibition on the KIT signaling pathway. Surprisingly, VLX600 treatment affected multiple elements of Kit signaling. The PI3K/AKT pathway is downstream of KIT and has been shown to stimulate aerobic glycolysis and glucose uptake via glucose transporters (33, 40). After 6 and 24 hours of VLX600 monotherapy in GIST-T1 cells, AKT signaling was increased while ERK signaling was decreased (Fig. 4A, left). Furthermore, while GLUT1 expression was relatively low in GIST-T1 cells at baseline, treatment with VLX600 dramatically increased GLUT1 expression 6 and 24 hours after treatment. Notably, imatinib countered the increases in AKT and GLUT-1 while further decreasing ERK (Fig. 4A).
Together, these data support a previously identified AKT-dependent mechanism of glucose uptake (33) and suggest that mitochondrial inhibition with VLX600 therapy may be associated with a shift towards glycolytic metabolism via AKT signaling in GIST-T1 in vitro. Despite what has been previously reported for VLX600 (37), induction of glycolysis was noted to be independent of HIF1a upregulation in our mouse model (Supplementary Fig. S5A). More recently, iron chelation was identified as another mechanism of VLX600 (41). However, iron supplementation did not abrogate the cytotoxic effect of VLX600 in GIST-T1 cells (Supplementary Fig. S5B), indicating that the primary mechanism for VLX600 in GIST is OxPhos inhibition.
We hypothesized that inducing glycolysis via OxPhos inhibition with VLX600 would make imatinib more effective in a low glucose environment, because glucose availability is limited. As expected, GIST-T1, HG129, and GIST-882 cells proliferated more slowly when cultured in low glucose media (Fig. 4B, top). Remarkably, imatinib was less potent in all three human cell lines when cultured in low glucose media, but the addition of low dose VLX600 restored the potency of imatinib in GIST-T1 and HG129 (Fig. 4B, bottom), suggesting that VLX600 improves imatinib's efficacy on slowly proliferating cells. It is also well described that imatinib induces tumor cell quiescence by two mechanisms: the APC-p27Kip1 signaling axis and the cell-cycle regulating DREAM complex (24, 25). Not surprisingly, we identified upregulation of p27Kip1 after 6 and 24 hours of imatinib treatment in vitro (Fig. 4A, right). Combination therapy with VLX600, however, decreased p27 expression in vitro (Fig. 4A, right). Together, these results support that imatinib is more effective on highly proliferating cells and decreased proliferation through a low glucose environment or imatinib-induced quiescence renders GIST cells susceptible to mitochondrial inhibition.
Finally, cell cycle analysis showed that imatinib therapy induced cellular quiescence, as evidenced by a significant reduction of GIST-T1 and HG129 cells in S-phase after 24 hours of treatment. However, with combination therapy, this shift to quiescence was prevented (Fig. 4C). Annexin V staining showed that combination therapy led to greater apoptosis than either treatment alone (Fig. 4D). Together, these data suggested that imatinib-treated cells became quiescent to avoid imatinib-induced death, yet VLX600 prevented this escape by driving glycolysis and cell cycle progression, enhancing their susceptibility to imatinib.
Mitochondrial inhibition enhances the antitumor effects of imatinib in a murine model of GIST
Thus far, we have shown that an imatinib-induced decrease in glucose uptake reduced glycolytic activity and shifted GIST metabolism toward a quiescent and more energy efficient phenotype, while mitochondrial inhibition forced a return to a glycolytic phenotype. Given these findings, we hypothesized that VLX600 and imatinib have competing effects on tumor cell metabolism, and we sought to explore these effects in a germline murine model of GIST. We treated KitV558del/+ mice with VLX600 and imatinib (Fig. 5A). After 2 weeks of combination treatment, tumor weight was significantly decreased compared with imatinib therapy alone (Fig. 5B). Tumor cells also expressed significantly lower levels of KIT by percentage of CD45− (i.e., nonimmune) cells and by median fluorescence intensity (Fig. 5C). Combination treatment was associated with significant reductions in KIT, AKT, and ERK signaling by Western blot at 6 hours, 1 week, and 2 weeks when compared with imatinib alone (Fig. 5D), which was consistent with the additive therapeutic effect of imatinib and VLX600 observed in vitro. Histology demonstrated a dramatic tumor effect in the combination group only (Fig. 5E), with an increased stromal response noted at the center of the tumors on H&E staining and a reduction in Ki67 staining. Importantly, the observed tumor effect in vivo appeared to be dependent on inhibition of KIT signaling, as no effect was observed in the VLX600 monotherapy group (Fig. 5E). Notably, our in vitro finding of decreased p27 signaling with combination therapy compared with imatinib alone was recapitulated on histologic staining in KitV558del/+ tumors after 2 weeks of therapy (Fig. 5E), suggesting that quiescent tumor cells were preferentially targeted.
Tyrosine kinase inhibition has been associated with dramatic changes in the metabolic phenotypes of primary and imatinib-resistant GIST (22, 42). Glucose uptake is significantly reduced on 18F-FDG PET scanning with imatinib and sunitinib therapy, and metabolic response correlated with an improvement in progression-free survival in both primary and resistant disease (21, 23, 42, 43). Interestingly, the development of imatinib resistance coincided with the reemergence of the glycolytic phenotype within the site of the original tumor (22). Cessation of imatinib therapy in patients exhibiting imatinib resistance results in a “flare” phenomenon in which the remaining GIST lesions exhibit a significant increase in glucose metabolism within days of stopping imatinib (44). This suggests that in addition to developing resistance, a population of imatinib-sensitive cells remain dormant in the tumor, and discontinuation of imatinib therapy allows their metabolic reactivation. These findings expose a therapeutic limitation of imatinib monotherapy and justify further research into the use of metabolic therapies to target imatinib-sensitive but dormant tumor cells.
Here, we show that imatinib therapy altered the metabolic phenotype of GIST by decreasing glucose uptake, and consequently glycolytic activity and capacity, while promoting mitochondrial energy metabolism in remaining tumor cells, consistent with imatinib's effects in chronic myelogenous leukemia (45, 46). We also show that this imatinib-induced shift in metabolic phenotype is associated with a significant enrichment in mitochondrial SRC and a decrease in intracellular ROS, both of which have been shown to promote cell survival in stressful environments (35, 36). Therefore, targeting mitochondrial respiration in imatinib-treated tumor cells seems like a logical approach to enhance the efficacy of imatinib in GIST.
We observed that imatinib treatment significantly increased the transcription and expression of key mitochondrial electron transport enzymes, specifically Cox I, II, and IV. Therefore, we hypothesized that the mitochondrial OxPhos inhibitor VLX600, which was shown to inhibit enzymes Cox I, II, and IV in colorectal cancer spheroids (37), would enhance the effect of imatinib. In GIST-T1, which we observed to have a high respiratory capacity at baseline, VLX600 decreased cell viability and proliferation as a monotherapy in addition to synergizing with imatinib. This effect was recapitulated in a xenograft model of GIST, as evidenced by a decreased tumor volume in VLX600-treated tumors, with a histologic effect only present in combination therapy. Similarly, in our genetically engineered mouse model, VLX600 required concurrent imatinib therapy to generate a response, as treatment with VLX600 alone did not result in a tumor weight change or histologic effect. This suggests that the histologic effects we observed with combination therapy in both xenografts and genetically engineered murine tumors are a result of VLX600 enhancing the effect of imatinib.
VLX600 has been shown to induce an HIF1a-mediated glycolytic response in tumor cells (37), and this effect is partially dependent on iron chelation (41). In our murine model of GIST, HIF1a was not upregulated with VLX600 treatment alone, or in combination imatinib/VLX600 therapy. Furthermore, iron supplementation failed to rescue GIST-T1 cells from the cytotoxic effects of VLX600, indicating that the primary mechanism of VLX600 in GIST is OxPhos inhibition. OxPhos inhibition by VLX600 upregulated GLUT1 expression and enhanced AKT signaling in vitro, suggesting a shift toward glycolytic metabolism. Importantly, imatinib-induced reduction in glucose uptake overcame the glycolytic effect induced by VLX600. When combined with imatinib, VLX600 reduced p27 expression in vitro and in vivo and prevented imatinib-induced cell cycle escape, likely potentiating the antitumoral effects of Kit inhibition. In fact, the decreased KIT signaling and profound histologic effects we observed with combination VLX600/imatinib therapy likely reflect an increased response to imatinib in the setting of forced glycolysis by mitochondrial inhibition, a competitive metabolic effect which has also recently been observed with combined AKT inhibition/paclitaxel therapy in a preclinical lung cancer model (47).
We have also shown that human GIST cell lines proliferate more slowly when cultured in low glucose media. Surprisingly, imatinib became less potent in this setting. This suggests that the therapeutic effect of imatinib is partially dependent on the proliferative state of the tumor cell. Remarkably, we also observed that lower doses of imatinib actually appeared to enhance GIST-T1 and HG129 cell viability in low glucose media. Because it has been shown that cells cultured in low glucose media become more reliant on mitochondria (48, 49), an imatinib-induced increase in SRC, as we have shown here, may promote cell survival in this setting. Notably, the addition of low dose VLX600 was sufficient to restore the potency of imatinib to normal levels in GIST-T1 and HG129 cell lines. GIST-882 cells, however, remained largely insensitive to lower doses of VLX600 in normal and low glucose media. In combination with the lack of reentry into the cell cycle of GIST-882 observed in cell cycle analysis, our findings suggest that the OxPhos inhibitory capacity of VLX600 is limited in GIST-882 cells.
Similarly, a small percentage of GIST cells are able to survive despite combined glycolytic and mitochondrial inhibition. One explanation is that VLX600 is only a partial mitochondrial OxPhos inhibitor. GIST-882 cells, the least glycolytic of our human cell lines, demonstrated a significant mitochondrial SRC on extracellular flux analysis, indicating high utilization of OxPhos. Yet, only treatments with higher doses of VLX600 were able to significantly inhibit GIST-882 cell proliferation, suggesting more potent OxPhos inhibition may be needed. Future studies should examine the relationship between the potency of mitochondrial inhibition and in vivo systemic toxicity. Though not observed in either of our murine models, systemic toxicity likely increases as the potency of mitochondrial inhibition increases. One approach to limit systemic toxicity of mitochondrial inhibition is to combine it with targeted molecular therapy, as we have done here. Oncogene-addicted cancers may have less metabolic flexibility in certain fuel pathways, as we have shown with an insignificant change in glucose dependency and capacity in our Mito Fuel test, potentially creating a therapeutic window for mitochondrial therapy in this setting.
Another hypothesis is that imatinib appears to induce other potentially targetable metabolic effects in GIST. After 3 weeks of imatinib treatment, mRNA expression of lactate dehydrogenase A (LDHA) and monocarboxylate transporter 1 (SLC16A1) was significantly upregulated in our murine KitV558del/+ tumors. Because SLC16A1 and LDHA are bidirectional enzymes and can shuttle lactate into a cell, one hypothesis is that prolonged glycolytic suppression by imatinib forces a reverse Warburg effect, where stressed tumor cells incite neighboring stromal cells to export lactate for use by the tumor (50). By obtaining lactate from the tumor environment for input into the TCA cycle (or NADH regeneration in the presence of mitochondrial inhibition), GIST cells may be able to survive despite the dramatic anti-glycolytic effect of imatinib. Similarly, an increased dependence on fatty acid oxidation after imatinib therapy suggests additional metabolic pathways may be activated, introducing new potential targets for metabolic therapy in GIST.
In conclusion, the combination of mitochondrial inhibition with imatinib therapy is a promising therapeutic strategy in GIST. Imatinib significantly alters the metabolic phenotype of remaining GIST cells by decreasing glucose uptake and glycolytic activity while promoting reliance on mitochondrial function. Inhibiting mitochondrial function not only compromises the metabolism of remaining tumor cells, but also drives tumor cells toward a glycolytic phenotype which ultimately increases their sensitivity to the anti-glycolytic and apoptotic effects of imatinib.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: G.A. Vitiello, B.D. Medina, A.J. Moral, F. Rossi, J.R. Cross, R.P. DeMatteo
Development of methodology: G.A. Vitiello, B.D. Medina, S. Zeng, J.Q. Zhang, J.R. Cross, R.P. DeMatteo
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): G.A. Vitiello, J.K. Loo, N.J. Param, J.N. Zhao, C.R. Antonescu
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): G.A. Vitiello, B.D. Medina, T.G. Bowler, J.Q. Zhang, M. Liu, A.J. Moral, F. Rossi, C.R. Antonescu, V.P. Balachandran, J.R. Cross, R.P. DeMatteo
Writing, review, and/or revision of the manuscript: G.A. Vitiello, B.D. Medina, T.G. Bowler, J.Q. Zhang, J.K. Loo, M. Liu, A.J. Moral, F. Rossi, C.R. Antonescu, R.P. DeMatteo
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): G.A. Vitiello, J.N. Zhao, R.P. DeMatteo
Study supervision: R.P. DeMatteo
The investigators were supported by NIH grants R01 CA102613 and T32 CA09501, Betsy Levine-Brown and Marc Brown, David and Monica Gorin (R.P. DeMatteo); NIH grant F32 CA186534 (J.Q. Zhang); and GIST Cancer Research Fund (R.P. DeMatteo and C.R. Antonescu). We thank members of the Sloan Kettering Institute Laboratory of Comparative Pathology, Molecular Cytology, The Donald B. and Catherine C. Marron Cancer Metabolism Center, Antitumor Assessment core facility, and Research Animal Resource Center. We thank Russell Holmes for logistical and administrative support.
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.