Imatinib inhibits c-Kit-induced hypoxia-inducible factor-1α activity and vascular endothelial growth factor expression in small cell lung cancer cells Free

Angiogenesis is a critical process in tumor progression that not only provides the growing tumor with required oxygen, nutrients, and growth factors but also provides the circulatory access that allows tumor cells to metastasize (1, 2). Tumor angiogenesis is regulated by a variety of both positive and negative regulatory molecules (1, 3). When the net effects of the positive regulatory factors overwhelm the effects of the negative factors, the recruitment of preexisting blood vessels and growth of new vessels into tumors occurs, in a process known as the “angiogenic switch” (3). In experimental models this angiogenic switch is associated with a rapid increase in tumor size, local invasion, and the development of metastatic disease. Among the most important proangiogenic determinants of the angiogenic switch is the level of vascular endothelial growth factor (VEGF) expression by tumor cells. VEGF expression is regulated by two general types of stimuli: oxygen tension and polypeptide growth factors or cytokines, as well as oncogenically activated components of their downstream signaling pathways (1, 4).

Low oxygen tension regulates VEGF expression by enhancing transcription of the VEGF gene and by stabilization of its mRNA. It is now clear that the transcription factor hypoxia-inducible factor (HIF)-1α is the major regulator of VEGF transcription in response to hypoxia (5, 6). In some cells, hypoxia also results in a prolongation of VEGF mRNA half-life (7), which may also be ultimately attributable to the activity of HIF-1α (8). HIF-1α forms a heterodimer with constitutively expressed HIF-1β that binds to a hypoxia-responsive element (9, 10) present in the promoter or enhancer of many hypoxia-inducible genes (5, 6). HIF-1α levels are tightly regulated by oxygen tension through hydroxylation of prolyl residues, which occurs under normoxic conditions. This prolyl hydroxylation is required for interaction with the von Hippel-Lindau tumor suppressor protein, which is the recognition component of an E3 ubiquitin ligase that mediates the destruction of HIF-1α through the proteasomal degradation pathway. Hypoxic conditions retard the prolyl hydroxylation of HIF-1α and the protein is stabilized, enabling it to transactivate hypoxia-responsive genes, including VEGF (6).

Growth factors and cytokines mediate an increase in VEGF mRNA expression by regulatory mechanisms distinct from those involved in hypoxic regulation (1, 5). Whereas stabilization of HIF-1α protein is the primary mechanism by which hypoxia influences VEGF expression, growth factors enhance HIF-1α protein synthesis and transactivation domain function. Growth factors and cytokines such as epidermal growth factor, heregulin, insulin, insulin-like growth factor I, and interleukin-1β induce VEGF expression by enhancing expression of HIF-1α protein and its DNA-binding capability (5). This occurs under normoxic conditions and can result in an additive effect when growth factor stimulation occurs under hypoxic conditions (11, 12). Growth factor receptor tyrosine kinase stimulation seems to enhance HIF-1α mRNA translation via activation of the phosphatidylinositol 3-kinase-Akt-mammalian target of rapamycin signaling pathway and subsequent phosphorylation of translational regulatory proteins p70 S6 kinase and eukaryotic initiation factor 4E binding protein (11, 12). Activation of mitogen-activated protein kinase (MAPK) pathways correlates with phosphorylation of HIF-1α and enhancement of its transactivation function, as well as phosphorylation of eukaryotic initiation factor 4E (11, 13).

Stem cell factor (SCF), the ligand for the c-Kit receptor tyrosine kinase, has been implicated in the regulation of angiogenesis in hematopoietic and nonhematopoietic malignancies but the mechanisms that mediate this effect have not yet been elucidated. It has been shown that rat mammary tumor cells transfected with a SCF expression vector develop increased microvascular density (14). SCF has also been implicated in the mobilization of endothelial progenitor cells required for angiogenesis (15). Treatment of neuroblastoma cell lines and xenografts with imatinib (Gleevec), a specific inhibitor of c-Kit and platelet-derived growth factor receptor, resulted in decreased VEGF expression (16). Finally, we have recently observed that treatment of small cell lung cancer (SCLC) cell xenografts with SU5416, a multitargeted kinase inhibitor that inhibits both c-Kit and the VEGF receptor-2, resulted in decreased tumor microvascular density. Whereas, in vivo, it could not be determined the degree to which the effect on angiogenesis was due to inhibition of tumor cell c-Kit versus VEGF receptor 2 of endothelial precursors, in vitro studies showed that SU5416 directly inhibited VEGF expression by the tumor cells (17). Because at least 70% of SCLC cells coexpress SCF and c-Kit (18) and tumor microvessel density and VEGF expression have been shown to correlate with stage and prognosis in SCLC cell (19, 20), we have explored the regulation of VEGF expression by SCF in SCLC cells. In addition, we have also determined the effect of imatinib, a highly efficient inhibitor of c-Kit signal transduction (21), on VEGF expression.

SCF treatment of multiple c-Kit-expressing SCLC cell lines increased VEGF secretion by ∼2-fold, an effect efficiently inhibited by imatinib. An increase in VEGF mRNA in the H526 cell line occurred within the first 2 hours of SCF treatment and was not caused by alterations in mRNA stability. The phosphatidylinositol 3-kinase inhibitor LY294002 blocked the increase in VEGF mRNA, implicating c-Kit-mediated activation of phosphatidylinositol 3-kinase in the phenomenon. VEGF promoter-reporter transfections indicated that the SCF-mediated increase in VEGF mRNA occurred through a transcriptional mechanism partially dependent on an intact HIF-1 binding site. SCF enhanced nuclear HIF-1α levels, which correlated well with increased HIF-1α binding to a consensus hypoxia-responsive element. SCF-mediated effects on HIF-1α expression were additive with those produced by CoCl₂, a hypoxia-mimetic agent. Thus, activation of c-Kit by SCF has a significant effect on VEGF expression, and inhibition of c-Kit signaling with imatinib could have clinically relevant antiangiogenic effects by inhibiting VEGF expression.

SCLC cell lines (22) were cultured in complete medium consisting of 10% (v/v) fetal bovine serum (Life Technologies, Invitrogen Corporation, Carlsbad, CA), 2 mmol/L l-glutamine (BioWhittaker, Walkersville, MD), and 50 units/mL penicillin-streptomycin (BioWhittaker) in RPMI 1640 (Life Technologies, Invitrogen). Where indicated, when grown in the absence of serum, 0.1% bovine serum albumin (Sigma, St. Louis, MO) and recombinant SCF (100 ng/mL; Peprotech, Rocky Hill, NJ) were added to the medium. To mimic hypoxia, cells were grown in the presence of 100 μmol/L CoCl₂ (Sigma). To block c-Kit signaling, cells were grown in media containing imatinib mesylate (Novartis Pharma, Basel, Switzerland), LY294002, or PD98059 (Calbiochem, San Diego, CA) solubilized in DMSO, as indicated. The concentration of DMSO in all cultures, including controls, was 0.1%.

Secretion of VEGF by 5 × 10⁵ cells in 1 mL of culture medium was measured using an ELISA assay kit as directed by the supplier (R&D Systems, Minneapolis, MN). For Northern blotting, a 444-bp cDNA probe was generated by reverse transcription-PCR using commercial oligonucleotide primers (R&D Systems), following the directions supplied by the manufacturer. Radiolabeling was carried out by random priming using the Ready-To-Go kit (Amersham, Piscataway, NJ). RNA was isolated from the H526 cell line using the Trizol reagent (Invitrogen) and Northern blotting was done as previously described (23). For determination of VEGF mRNA half-life, quiescent H526 cells were incubated in serum-free medium containing 10 μg/mL actinomycin D (Sigma) in the presence or absence of 100 ng/mL SCF. RNA was isolated at serial times after addition of actinomycin D and Northern blotting was done as described above. VEGF mRNA was quantitated using a Raytest (Newcastle, DE) phosphoimaging station and the AIDA 2.0 software package. The VEGF signal intensity in each lane was normalized to that of β-actin and half-life was calculated as previously described (23).

H526 cells were transfected with 25 μg of pGL2-VEGF (pVEGF-KpnI, ATCC MB-4, American Type Culture Collection, Manassas, VA; ref. 24), pGL2-Basic (Promega, Madison, WI), or pGL2-Promoter (Promega) along with 5 μg pSV2-β-Gal using the Lipofectamine 2000 transfection reagent as directed by the manufacturer (Invitrogen). A 4-bp mutation (TACGTG→GAATTC) was introduced into the core HIF-1 binding site (−975) in pGL2-VEGF using the QuickChange Mutagenesis kit (Stratagene, La Jolla, CA) to derive the pGL2-VEGFmutant plasmid. Twenty-four hours after transfection, the cells were made quiescent by incubation overnight in serum-free medium and then stimulated with 100 ng/mL SCF or left unstimulated for an additional 24 hours. Cells were lysed and luciferase and β-galactosidase activities were determined with the Luciferase and Beta-Glo Assay systems (Promega) using a single-photon luminometer. Relative luciferase activity in the presence and absence of SCF was calculated by normalizing luciferase activity to β-galactosidase activity.

Nuclear proteins were isolated using the NE-PER kit (Pierce Biotechnology, Inc., Rockford, IL) as directed by the manufacturer. Double-stranded oligonucleotides encompassing the consensus hypoxia-responsive element and an oligonucleotide with a CGT to AAA mutation within the core binding site were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). Oligonucleotides were end-labeled using polynucleotide kinase and electrophoretic mobility shift assays were done using 10 μg of nuclear protein as previously described (25).

Whole-cell lysates were prepared by resuspending the cells in cold SDS sample buffer [1% SDS, 0.04 mol/L Tris-HCl (pH 6.8), 5% glycerol]. The lysates were boiled and sheared through a 25-gauge needle and protein concentrations were determined using a commercial assay kit (BCA, Pierce Biotechnology). Nuclear protein extracts were added in a 1:1 ratio to 2× SDS sample buffer and boiled. Fifty micrograms of protein were resolved on a 10% polyacrylamide gel and transferred to polyvinylidene difluoride membranes. The membranes were probed with either anti-HIF-1α rabbit polyclonal antiserum (NB 100-134) or affinity-purified antibody (NB 100-449, Novus Biologicals, Inc., Littleton, CO), β-actin monoclonal antibody (Sigma), or N-myc monoclonal antibody (Santa Cruz Biotechnology). Signals were detected and quantitated using the West Pico chemiluminescent system (Pierce Biotechnology) with the aid of a Raytest cooled CCD camera imaging system and the AIDA 2.0 software package. Total cellular HIF-1α was normalized relative to β-actin and nuclear HIF-1α was normalized relative to N-myc, which is autonomously expressed from a highly amplified locus in the H526 cell line.

To determine the effects of SCF on VEGF expression, quiescent SCLC cell lines that express c-Kit (H526, H69, and WBA) and one that does not (H146) were stimulated with SCF for 24 hours in serum-free medium. The conditioned medium was collected and assayed for VEGF using an ELISA (Fig. 1A). The H146 cell line did not express basal levels of VEGF and could not be stimulated to express VEGF by SCF. VEGF secretion doubled in c-Kit-expressing cells treated with SCF relative to controls. Imatinib (STI571) completely blocked the increase in VEGF, confirming that this phenomenon is mediated by c-Kit activation because the other targets of imatinib (platelet-derived growth factor receptor and Abl) are not expressed or activated by SCF in SCLC cell (21). To further determine the manner in which SCF regulated VEGF expression, RNA was isolated from H526 cells at successive time intervals following SCF stimulation. Figure 1B illustrates that SCF induced a 2-fold increase in VEGF mRNA expression by 2 hours after SCF addition, which persisted at 24 hours after exposure. As illustrated in Fig. 1C, VEGF mRNA levels did drop slightly in the 6- to 10-hour period after the SCF exposure in two of four replicates. The reasons for this inconsistent transient drop in VEGF mRNA are unclear at present. On the whole, however, the increase in VEGF mRNA observed correlated very well with the increase in VEGF secretion and was also inhibited by imatinib.

Because mRNA stabilization has been shown to be one way in which VEGF expression may be regulated, we determined whether SCF treatment altered VEGF mRNA half-life. H526 cells in serum-free medium were treated with actinomycin D (at a concentration that blocked RNA synthesis by >99%) in the presence or absence of SCF and the decay of VEGF mRNA was followed over time by Northern blotting (Fig. 2A). The decay of VEGF mRNA was not significantly altered by SCF treatment, with a densitometrically calculated mRNA half-life of ∼130 minutes (Fig. 2B). This value is generally consistent with previous reports that have determined the VEGF mRNA half-life to be in the range of 40 minutes to 4 hours under normoxic conditions, with a high degree of cell type dependence (7, 8, 26).

These results suggested that the increase in VEGF mRNA on SCF treatment was transcriptionally mediated. To determine if this was the case, the VEGF promoter cloned upstream of a luciferase reporter gene was transiently transfected into H526 cells, which were then stimulated with SCF for 24 hours in serum-free medium. SCF treatment led to a 70% increase in relative luciferase activity compared with unstimulated control cells (Fig. 3). The activity produced by a control plasmid containing the SV40 promoter was not affected by SCF treatment. The SCF-mediated increase in promoter activity was completely inhibited by 5 μmol/L imatinib. Furthermore, mutation of the HIF-1 binding site at the −975 position in the VEGF promoter (24) also largely eliminated the SCF-mediated increase in promoter activity. These results indicate that SCF enhances VEGF expression via its effects on the VEGF promoter, predominantly through the HIF-1 binding site. However, because significant stimulation of the VEGF promoter did occur even in the absence of a functional hypoxia-responsive element, c-Kit activation may also enhance VEGF promoter activity through non-HIF-1-dependent mechanisms (27, 28).

Expression of HIF-1α, the major transcriptional regulator of VEGF, can be enhanced by numerous polypeptide growth factors. To initially determine if SCF partially mediates its effect on VEGF expression by enhancing HIF-1α expression, H526 cells were incubated for 24 hours in serum-free medium containing or lacking SCF and the expression of HIF-1α was assessed by Western blotting of whole-cell lysates (Fig. 4A). HIF-1α was difficult to detect in whole-cell lysates made from unstimulated cells and, surprisingly, we observed no detectable increase in HIF-1α in lysates from cells treated with SCF. However, in the presence of CoCl₂, a hypoxia-mimetic agent known to stimulate HIF-1α levels by inhibiting its proteasomal degradation (5), a clear increase in HIF-1α levels was seen in SCF-treated cells relative to cells treated with CoCl₂ alone. This SCF-mediated increase in cellular HIF-1α levels was inhibited by imatinib, which had no effect on the CoCl₂-mediated increase in HIF-1α. To enhance sensitivity in detection of SCF-mediated changes in HIF-1α expression, we isolated the nuclear protein fraction, which should be enriched in HIF-1α. As illustrated in Fig. 4B and C, SCF induced a nearly 2-fold increase in the amount of nuclear HIF-1α, as opposed to CoCl₂, which induced an ∼10-fold increase. Imatinib completely blocked the SCF-mediated increase in HIF-1α. Pretreatment with the phosphatidylinositol 3-kinase inhibitor LY294002 also resulted in nearly complete inhibition of nuclear HIF-1α accumulation whereas treatment with the MAPK/extracellular signal–regulated kinase kinase inhibitor PD98059, which inhibits MAPK activation, reduced the increase in nuclear HIF-1α by only about half. It should be noted that after extended electrophoresis, HIF-1α appeared as multiple bands on staining with several different antibodies (not shown), likely the result of multiple posttranslational modifications that the protein is known to undergo (29). In addition, SCF did not affect nuclear HIF-1β levels whereas CoCl₂ induced the same modest increase in HIF-1β in the absence or presence of SCF (data not shown).

To determine if the SCF-mediated increase in HIF-1α resulted in enhanced binding to its consensus recognition sequence, we isolated nuclear proteins from H526 cells treated with SCF and/or CoCl₂. The nuclear extracts were analyzed by electrophoretic mobility shift assay for their ability to form specific HIF-1α/DNA complexes with an oligonucleotide containing a consensus hypoxia-responsive element. Figure 5 illustrates that SCF enhanced the ability of nuclear extracts to generate specific HIF-1α/DNA complexes. This effect seemed to be additive with CoCl₂ in inducing DNA complex formation. Specificity of the complex formation was documented by showing that wild-type, but not mutant, unlabeled oligonucleotide could compete for HIF-1α binding whereas both mutant and wild-type oligonucleotides eliminated the intense nonspecific bands. In addition, a polyclonal HIF-1α antiserum disrupted the HIF-1α complex but not the nonspecific band, whereas the control rabbit immunoglobulin G failed to disrupt either band.

Activation of c-Kit by SCF is known to initiate signal transduction down the MAPK and phosphatidylinositol 3-kinase pathways, among others (30). To determine the importance of these signaling pathways in the regulation of VEGF expression, we treated H526 cells with SCF over a 24-hour period in the presence of the MAPK/extracellular signal–regulated kinase kinase inhibitor PD98059, the phosphatidylinositol 3-kinase inhibitor LY294002, or DMSO vehicle. Figure 6A illustrates that whereas inhibition of MAPK signaling had little effect on VEGF mRNA expression, phosphatidylinositol 3-kinase inhibition resulted in complete inhibition of SCF-induced expression. These results are in general agreement with those illustrated in Fig. 4, which documented that LY294002 effectively inhibited the SCF-mediated increase in nuclear HIF-1α. To determine whether changes in VEGF expression correlated with hypoxia-responsive element binding, nuclear proteins were extracted from H526 cells treated with SCF in the absence or presence of LY294002 and PD98059. As previously noted, nuclear extracts from CoCl₂-treated cells contained a greater amount of HIF-1α capable of binding its cognate DNA sequence than extracts from SCF-treated cells, which had greater binding activity than extracts from unstimulated cells (Fig. 6B). Treatment with LY294002 resulted in a dramatic and nearly complete loss of HIF-1α DNA binding whereas treatment with PD98059 resulted in a more modest reduction in binding. Thus, inhibition of SCF-mediated VEGF expression by phosphatidylinositol 3-kinase inhibition correlated with a decrease in nuclear HIF-1α and a loss in the ability to form complexes with the hypoxia-responsive transcriptional element.

Angiogenesis has been shown to be critical for growth and progression of numerous tumor types and VEGF has been shown to be a critical mediator of tumor angiogenesis. In SCLC cell, microvessel density and VEGF expression correlate well with both stage and prognosis (19, 20). To better define the regulation of SCLC cell angiogenesis, we explored the role of SCF and c-Kit, components of a common autocrine loop in SCLC cell, in modulating VEGF expression. Activation of c-Kit by SCF treatment of multiple SCLC cell lines resulted in a doubling of VEGF secretion, with a similar increase in VEGF mRNA in the H526 cell line (Fig. 1). SCF treatment resulted in a 70% increase in luciferase reporter expression mediated by a VEGF promoter-reporter construct without a change in VEGF mRNA stability (Figs. 2 and 3). These results are all very consistent with transcriptional enhancement of VEGF expression through activation of c-Kit by SCF, similar to modulation of VEGF expression by other ligands for receptor tyrosine kinases such as insulin-like growth factor I, heregulin, and epidermal growth factor (11, 12, 31). The fact that this process could be interrupted by imatinib (Figs. 1 and 3) confirms its c-Kit dependence and suggests that these findings could have clinical applicability, especially in light of previous observations on imatinib sensitivity of VEGF expression in neuroblastoma (16). In addition to neuroblastoma, angiogenesis has been shown to be an important factor in the biology of gastrointestinal stromal tumors (32). The success of imatinib in treating this tumor may in part be a consequence of inhibition of VEGF expression, a possibility that could also be relevant to the treatment of other malignancies in which c-Kit plays a pathogenic role (33). In addition to inhibiting VEGF expression, imatinib may also inhibit mobilization of angiogenic precursors (15) and pericyte-induced angiogenic maturation mediated by platelet-derived growth factor receptor (3, 16).

In addition to being sensitive to imatinib, the SCF-mediated increase in VEGF promoter activity was also largely dependent on an intact HIF-1 binding site (Fig. 3). Surprisingly, however, we could not show an increase in total cellular HIF-1α by Western blotting after SCF treatment alone. However, in the presence of the hypoxia-mimetic agent CoCl₂, SCF treatment did clearly enhance levels of HIF-1α relative to cells treated with CoCl₂ alone (Fig. 4A). This additive response has previously been seen when the mechanism of insulin-like growth factor I–mediated VEGF expression was studied (11). One possible reason for the failure to detect an increase in HIF-1α in cells treated with SCF alone is that HIF-1α levels may have been below the limit of detectability by Western blotting of whole-cell lysates in the absence of CoCl₂. To improve sensitivity and to allow for direct correlation with hypoxia-responsive element binding, we extracted nuclear proteins and documented that SCF induced a nearly 2-fold increase in nuclear HIF-1α (Fig. 4B and C), which correlates well with the observed increase in VEGF expression (Fig. 1). Because of our difficulty in detecting HIF-1α in whole-cell lysates, we cannot completely exclude the possibility that SCF enhances transport of HIF-1α to the nucleus. However, in the presence of a hypoxia-mimetic agent, it is clear that SCF enhances total cellular as well as nuclear HIF-1α levels (Fig. 4), suggesting that the predominant effect of SCF is mediated by enhanced HIF-1α expression. HIF-1-independent mechanisms may also be contributory, based on the attenuated stimulation of a hypoxia-responsive element mutant promoter-reporter construct by SCF (Fig. 3). In general, however, the ability to form HIF-1α-containing complexes with the hypoxia-responsive element correlated very well with SCF effects on VEGF expression. However, the marked decrease in hypoxia-responsive element binding induced by LY294002 treatment to levels well below that seen using extracts from quiescent cells (Fig. 6) does seem to be slightly out of proportion to the more modest decline in nuclear HIF-1α content after LY294002 treatment (Fig. 4). This observation suggests that SCF-induced phosphatidylinositol 3-kinase signaling, in addition to enhancing the quantity of HIF-1α, may also enhance its ability to bind its cognate DNA element. This could be caused by phosphatidylinositol 3-kinase–mediated posttranslational modification of HIF-1α itself or another component necessary for the formation of a stable complex with the hypoxia-responsive element. Multiple posttranslational modifications have been described for HIF-1α, including prolyl hydroxylation, ubiquitination, SUMOylation, acetylation, S-nitrosation, and phosphorylation (29). Of these, however, S-nitrosation and phosphorylation have been the only modifications where enhancement of transcriptional activation, the effect mediated by SCF, has been documented.

In summary, we have documented for the first time that SCF induces a doubling in VEGF expression in c-Kit-expressing SCLC cell lines and that this increase is due to enhanced VEGF transcription, which can be completely inhibited by imatinib. This enhanced VEGF transcription correlated well with a SCF-mediated increase in nuclear HIF-1α levels and binding to its cognate DNA element and was additive with the effects of the hypoxia-mimetic agent CoCl₂. The SCF-mediated increase in VEGF expression seemed to be primarily dependent on activation of phosphatidylinositol 3-kinase because the phosphatidylinositol 3-kinase inhibitor LY294002 completely blocked hypoxia-responsive element binding and SCF-mediated VEGF expression. These data are consistent with a model in which c-Kit-mediated phosphatidylinositol 3-kinase activation leads to enhanced HIF-1α expression and transactivation potential and suggest that direct inhibition of c-Kit or its activation of the phosphatidylinositol 3-kinase-Akt pathway could result in inhibition of angiogenesis, which would be clinically exploitable in SCLC cell and other malignancies in which c-Kit activation occurs.