Division of Neuroscience, Beckman Research Institute, City of Hope, Duarte, California, USA
Department of Hematology/Hematopoietic Cell Transplantation, City of Hope, Duarte, California, USA
Department of Hematology/Hematopoietic Cell Transplantation, Beckman Research Institute of the City of Hope, 1500 East Duarte Road, Duarte, California 91010, USA. Telephone: 626-471-7177; Fax: 626-301-8857
The utility of neural stem cells (NSCs) has extended beyond regenerative medicine to targeted gene delivery, as NSCs possess an inherent tropism to solid tumors, including invasive gliomas. However, for optimal clinical implementation, an understanding of the molecular events that regulate NSC tumor tropism is needed to ensure their safety and to maximize therapeutic efficacy. We show that human NSC lines responded to multiple tumor-derived growth factors and that hepatocyte growth factor (HGF) induced the strongest chemotactic response. Gliomatropism was critically dependent on c-Met signaling, as short hairpin RNA-mediated ablation of c-Met significantly attenuated the response. Furthermore, inhibition of Ras-phosphoinositide 3-kinase (PI3K) signaling impaired the migration of human neural stem cells (hNSCs) toward HGF and other growth factors. Migration toward tumor cells is a highly regulated process, in which multiple growth factor signals converge on Ras-PI3K, causing direct modification of the cytoskeleton. The signaling pathways that regulate hNSC migration are similar to those that promote unregulated glioma invasion, suggesting shared cellular mechanisms and responses.
Disclosure of potential conflicts of interest is found at the end of this article.
Intracranially or intravenously injected exogenous neural stem cells (NSCs) migrate toward neural pathologies in murine models of central nervous system injury. This innate chemotropism has been exploited for cell replacement therapies to treat diverse neurological diseases [1, , , , , , , –9]. These therapies offer hope for functional neurological recovery through the ability of the NSCs to differentiate and integrate appropriately into the host cytoarchitecture. However, the use of NSCs has not been restricted to cell replacement strategies. Murine and human neural stem cells (hNSCs) possess an inherent tumor tropism that supports their use as a reliable delivery vehicle to target therapeutic gene products to primary and secondary invasive glioma cells throughout the brain [10, , , , –15], as well as to other types of solid tumors, including melanoma brain metastases, medulloblastoma, and neuroblastoma [16, , , –20]. Recent advances in NSC biology have created immortalized hNSC lines, which retain their stem-like properties over time and passage. Both immortalized and primary NSC pools have demonstrated similar tumor-tropic properties, whereas immortalized hNSC lines, such as HB1.F3, are particularly well suited for the expansion of clones that stably express therapeutic genes designed to treat invasive gliomas [3, 21, –23].
Gliomas release numerous chemokines and growth factors that are capable of stimulating the directed migration of exogenous and endogenous NSCs into the tumor microenvironment. Agents such as stem cell factor-1, monocyte chemoattractant protein-1, and stromal cell-derived factor-1 are potent chemotactic molecules originally identified as inducers of hematopoietic cell migration [24, 25] and recently shown to stimulate NSC migration [26, , , –30]. However, growth factors, such as epidermal growth factor (EGF), hepatocyte growth factor (HGF), vascular endothelial growth factor (VEGF), components of the extracellular matrix, and metalloproteinase-2 fragment, are recognized as potent gliomatropic agents for NSCs [23, 30, , , , –35].
The migratory properties of glioma cells and NSCs have similar behavioral characteristics. A thorough understanding of the molecular events that regulate NSC migration to glioma is necessary to optimize the use of NSCs as therapeutic delivery vehicles. We therefore undertook a comprehensive analysis of hNSC gliomatropism using the immortalized HB1.F3 hNSC line, which targets human gliomas in both in vitro and in vivo models [23, 31]. We demonstrate that the human glioma cell lines U251 and U87 produce HGF and VEGF, which act as potent chemoattractants for HB1.F3 cells. These growth factors stimulate receptor tyrosine kinase signaling that leads to the formation of a receptor-signaling complex that includes class I phosphoinositide 3-kinase (PI3K), which has previously been shown to be an important regulator of directed cell migration [36, , –39]. In particular, we found that HGF promoted the most NSC migration, which was reflected in elevated levels of Rac1-GTP. Inhibition of HGF signaling through short hairpin RNA (shRNA) knockdown of its cognate receptor, c-Met, significantly attenuated the chemotactic response of HB1.F3 cells toward U251 and U87 glioma cells in transwell migration assays. This reduction in cell migration was specific to HGF signaling, because NSC migration stimulated by other factors was preserved. However, inhibition of the PI3K pathway significantly inhibited the chemotactic migration toward all growth factors tested (HGF, VEGF, and EGF), as well as gliomas, suggesting that the growth factors produced by gliomas converge on the PI3K signaling pathway. Collectively, these results reveal that PI3K serves as a critical convergence point for growth factor-mediated directed migration of hNSCs. Deregulation of the PI3K checkpoint leads to the aberrant PI3K signaling seen in highly metastatic gliomas, suggesting that stem cells and malignant tumors use similar signaling pathways to enable cell motility.
Materials and Methods
Cell Culture Model
HB1.F3, F5, and A4 hNSCs were maintained in growth medium consisting of Dulbecco's modified Eagle's medium (DMEM) (Invitrogen, Carlsbad, CA, http://www.invitrogen.com), 10% heat-inactivated fetal bovine serum (FBS) (Omega Scientific, Tarzana, CA, http://omegascientific.com), and 10 ng/ml EGF (Invitrogen) and supplemented with 100 units/ml penicillin and 100 μg/ml streptomycin and grown at 37°C in 5% CO2. All other cell lines used (U251, U87, MRC5, and HEK293) were maintained in growth medium consisting of DMEM and 10% heat-inactivated FBS and supplemented with 100 units/ml penicillin and 100 μg/ml streptomycin and grown at 37°C in 5% CO2.
For chemotaxis, transwell migration assays were carried out as previously described . Briefly, HB1.F3 cells were serum-starved, and 5 × 104 cells were resuspended in DMEM containing 0.25% FBS and seeded into the upper chamber of a fibronectin-coated transwell insert (Fisher Scientific International, Hampton, NH, http://www.fisherscientific.com). Conditioned media (DMEM, 0.25% FBS) were collected from 1 × 105 cells of appropriate cell lines of interest (U251, U87, MRC5, and HEK293) for use in the lower chamber. Alternatively, recombinant growth factors (HGF, VEGF, EGF, and transforming growth factor α [TGFα]) (R&D Systems Inc., Minneapolis, http://www.rndsystems.com) were added to DMEM containing 0.25% FBS in the lower chamber at the indicated concentrations. HB1.F3 cells were then permitted to migrate for 2 hours followed by fixation in 4% paraformaldehyde (PFA) for 10 minutes at room temperature. The number of migrated cells was quantified using ImageJ (NIH). For chemokinetic or cell-scattering assays, 5 × 104 HB1.F3 cells were seeded on fibronectin-coated slides and allowed to adhere overnight. Cells were then serum-starved for 1 hour prior to the assay. Recombinant HGF (100 ng/ml) was added to DMEM containing 0.25% FBS and used to bathe the cells for 1 hour. Following stimulation, cells were fixed in 4% PFA (10 minutes) and then processed for immunocytochemistry. A cell was scored as migratory when F-actin stress fibers were redistributed to sites of membrane ruffles.
Adherent cultures were grown on fibronectin-coated glass slides and treated appropriately for each specific experimental paradigm. Cells were fixed in 4% PFA (10 minutes), rinsed in phosphate-buffered saline (PBS), and labeled with the appropriate primary antibody (anti-c-Met [Cell Signaling Technology, Beverly, MA, http://www.cellsignal.com], anti-p85 [Abcam, Cambridge, MA, http://www.abcam.com], anti-Gab1, anti-Ras [Upstate, Charlottesville, VA, http://www.upstate.com], anti-Rac1, or anti-Arp2 [Santa Cruz Biotechnology Inc., Santa Cruz, CA, http://www.scbt.com]) using standard immunocytochemical techniques. Appropriate species-specific secondary antibodies conjugated to Alexa-594 were used together with the DNA binding dye Hoechst 33258 (1 μg/ml) and phalloidin-Alexa-488 to identify F-actin (Molecular Probes, Eugene, OR, http://probes.invitrogen.com). Images were acquired on a Zeiss LSM510 Axioplan-2 upright confocal microscope using LSM imaging software (Carl Zeiss, Jena, Germany, http://www.zeiss.com).
Immunoblotting and Immunoprecipitation
The following antibodies were used for biochemical analysis: anti-c-Met, anti-p-ERK1/2, anti-ERK1/2, anti-p-AKT, anti-AKT (Cell Signaling Technology), anti-p85 (Abcam), anti-Gab1, anti-Ras (Upstate), and anti-Rac1 (Santa Cruz Biotechnology). For biochemical analyses, samples were processed as previously described .
Rac1-GTP Binding Assay
Cells were grown in 100-mm dishes, were serum-starved for 1 hour, and either were treated with 100 ng/ml HGF or remained in DMEM for an additional hour. Cells were quickly rinsed in PBS, and 500 μl of lysis buffer was added. Cells were scraped, and lysates were centrifuged for 5 minutes at 14,000 rpm. To determine the levels of Rac1-guanosine-5′-triphosphate (GTP), 50 μg of cell lysate was incubated with either PAK1 (Cdc42/Rac-interactive binding)-glutathione S-transferase (GST) beads or GST beads for 1 hour with rocking at 4°C. Samples were centrifuged for 2 minutes at 2,000 rpm, and supernatant was discarded. The beads were washed three times with lysis buffer. The levels of active Rac1 (Rac1-GTP) were detected by immunoblotting using specific anti-Rac1 (Santa Cruz Biotechnology). Total Rac1 levels were determined by immunoblotting for Rac1 in 50 μg of cell lysate.
Reverse Transcription-Polymerase Chain Reaction
Total RNA was isolated with the RNeasy Mini Kit (Qiagen, Hilden, Germany, http://www1.qiagen.com) using the manufacturer's recommended protocol, performed as previously described .
Sandwich Enzyme-Linked Immunosorbent Assay
Microtiter plates (96-well; Nunc, Rochester, NY, http://www.nuncbrand.com) were coated with the specific monoclonal capture antibodies (R&D Systems) at the indicated concentrations in PBS overnight at 4°C. Conditioned media derived from separate independent samples were added in quadruplicate for each cell type and incubated for 2 hours at room temperature. Appropriate biotinylated detection antibodies (R&D Systems) were used and incubated for 2 hours at room temperature. Plates were washed and then incubated with streptavidin-horseradish peroxidase for 1 hour at room temperature. Finally, plates were incubated with substrate (3,3′,5,5′-tetramethylbenzidine liquid substrate system; Sigma-Aldrich, St. Louis, http://www.sigmaaldrich.com) for 10 minutes at room temperature in the dark. The reaction was stopped by addition of 2 N H2SO4. Absorbance was measured at 450 nm, with wavelength correction set to 540 nm in a SpectraMAX 250 microplate reader (GMI, Ramsey, MN, http://gmi-inc.com). All experiments were performed nine separate times.
Intracranial Implantation of Glioma Cells
U251cells were prepared for injection into the brains of experimental mice and resuspended in PBS at a density of 4 × 104 cells per microliter. Athymic female nu/nu mice, 7–8 weeks old, were anesthetized with ketamine/xylazine. A stereotactic apparatus was used to drill a burr hole into the skull, and 8 × 104 U251 glioma cells were injected into the right frontal lobe 2 mm lateral, 0.5 mm anterior to bregma and 2.5 mm deep from the surface of the dura mater with a 30-gauge, 5-μl Hamilton syringe. Intracranial injection of neural stem cells occurred 14 days after intracranial implantation of U251 cells. NSCs were labeled with chloromethylbenzamido-1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocaibocyanine perchlorate (CM-DiI) (Molecular Probes, Eugene, OR, http://probes.invitrogen.com), and 2 × 105 cells were injected intracranially in a posterior-lateral position with regard to the glioma injection site, 2.75 mm lateral, 0.5 mm posterior to bregma and 2.5 mm deep from the surface of the dura mater with a 30-gauge, 5-μl Hamilton syringe. Five days following the injection of NSCs, animals were euthanized by CO2 asphyxiation. Mice were perfused transcardially with ice-cold 4% (wt/vol) PFA. Brains were harvested and postfixed in 4% PFA/PBS for 48 hours at 4°C. Brains were then sectioned horizontally into 100-μm-thick slices on a microtome (Vibratome 1000; Pelco, St. Louis, http://www.vibratome.com). Sections containing tumor were identified using a Nikon Eclipse TE 2000U inverted epifluorescence microscope (Nikon, Tokyo, http://www.nikon.com) and mounted on slides in Prolong-Gold confocal mounting solution (Molecular Probes) containing 70 μg/ml 4,6-diamidino-2-phenylindole (Sigma-Aldrich) for visualization of cell nuclei.
Replication-incompetent lentiviruses were produced by transient cotransfection of a four plasmid-based system into the packaging 293FT cell line (Invitrogen). Following transfection, cells were maintained in DMEM containing 2.5% FBS for 48 hours. Virus-containing media were centrifuged at 3,000 rpm for 5 minutes and filtered through a 0.2-μm filter. Virus was concentrated by ultracentrifugation at 25,000 rpm at 4°C for 1.5 hours using a Beckman SW28 swinging-bucket rotor (Beckman Coulter, Fullerton, CA, http://www.beckmancoulter.com). Virus was resuspended to a titer of ∼1 × 107 colony-forming units, divided into aliquots, and stored at −80°C until use. HB1.F3 cells were transduced with a multiplicity of infection of 10 and cultured for a period of 48 hours. Cells were then used for migration assays. To validate target gene knockdown, RNA was isolated and reverse-transcribed, real-time polymerase chain reaction (PCR) was performed on an iQ5 real-time PCR detection system (Bio-Rad, Hercules, CA, http://www.bio-rad.com), and Western blot analysis was used to validate target gene overexpression.
Data obtained from independent experiments were subjected to appropriate statistical analysis: one-way analysis of variance and post hoc Tukey's test or Student's t test using GraphPad Prism 4 (GraphPad Software, Inc., San Diego, http://www.graphpad.com). Final graphed values are represented as normalized means ± SD or raw data mean ± SEM. All experimental data were obtained from a minimum of three independent experiments.
hNSCs Activate Receptor Tyrosine Kinase Signaling in Response to Glioma-Produced Growth Factors
We selected two well-established human glioma cell lines, U251-H80 and U87-MG, to assay for putative soluble chemotactic factors. We also chose the nontumorigenic MRC5 cell line as a positive migratory control and the human embryonic kidney cell line HEK293 as a low migratory control for HB1.F3 cells. Reverse transcription-PCR was carried out as an initial high-throughput screen to identify potential chemotactic ligands that were expressed by gliomas and the candidate ligand's cognate receptor(s) on HB1.F3 cells. Our analysis showed that gliomas express mRNAs for numerous ligands. We narrowed these ligands down to likely candidates, on the basis of hNSCs receptor expression to the following growth factors: EGF, TGFα, VEGF, and HGF (Fig. 1A).
We used quantitative sandwich enzyme-linked immunosorbent assay to validate our gene expression data and to ensure that the glioma cells were secreting the candidate growth factors. Conditioned media were collected from our cell lines of interest and assayed for the presence of each of the following growth factors: EGF, TGFα, VEGF, and HGF (Fig. 1B). The U251 and U87 glioma cell lines produced significant concentrations of HGF (1,677 ± 31 and 906 ± 34 pg/ml, respectively; p < .001). Furthermore, these glioma cell lines also produced high levels of VEGF (5,332 ± 31 and 1,220 ± 52 pg/ml, respectively; p < .001). Neither EGF nor TGFα was produced in any significant quantity by any of the cell lines assayed, suggesting that these growth factors may not have a role in hNSC gliomatropism. Collectively, these results suggest that HGF and VEGF may function as the primary chemotactic growth factors produced by gliomas.
To test this hypothesis and to determine the extent of intracellular signaling stimulated by the candidate growth factors, we used a proteomics approach via an Src homology (SH2) domain protein array (supplemental online Table 2). Control hNSCs were cultured in the absence of serum and growth factors for 1 hour; cell lysates were then collected and assayed for tyrosine-phosphorylation on an SH2 domain protein array. Densitometric scanning of the SH2 protein array revealed a low basal level of PI3K, Ras, and phospholipase C gamma (PLCγ) activity (Fig. 1C). This basal level of intracellular signaling was used to normalize all subsequent assays. The hNSCs cultured in the absence of serum or growth factors were subsequently stimulated with EGF, HGF, or VEGF for 1 hour. All three growth factors induced robust activation of intracellular signaling (Fig. 1C). There was no difference in the ability of any individual growth factor to preferentially activate PI3K or Ras signaling. However, the overall level of intracellular signaling was dependent on the specific growth factor. EGF induced the lowest level of intracellular signaling (q= 12.9; p < .001), whereas HGF (q= 30.1; p < .001) and VEGF (q= 35.6; p < .001) induced the strongest activation of PI3K and Ras signaling. These data establish that hNSCs express a variety of growth factor receptors and are capable of responding to their cognate ligand by activating PI3K and Ras signaling.
HGF Stimulates Robust hNSC Migration
Having established that hNSCs respond to a variety of glioma-produced growth factors through activation of intracellular signaling events, we exposed HB1.F3 hNSCs to increasing concentrations of HGF, VEGF, EGF, and TGFα and used a transwell migration assay to measure the differential effects on hNSC migration elicited by these ligands. All growth factors produced a significant chemotactic response (F= 87.7, p < .0001) that was dose-dependent (F= 162.5, p < .0001) (Fig. 2A, 2B). HGF had the highest efficacy, yielding significant migration at all concentrations tested (p < .001). VEGF produced the second-greatest dose-dependent response, where deviation from HGF occurred at concentrations of 20 ng/ml and higher (p < .01). Both EGF and TGFα stimulated migration at concentrations of 10 ng/ml and higher (p < .001), but there was no significant difference in their ability to stimulate chemotaxis (p > .05). These results demonstrate that hNSCs possess an intrinsic migratory capacity toward multiple growth factors but that the overall induction of migration depends on the efficacy of the specific growth factor, with HGF having the highest potency.
To verify that growth factor-enhanced chemotaxis involved changes in the cytoskeleton, we assayed monolayer cultures of HB1.F3 hNSC scattering, or chemokinesis, on a fibronectin substratum. Cell protrusions were visualized by F-actin staining of stimulated and nonstimulated cells, and images were captured through laser scanning confocal microscopy. The incidence of static cells (detected by the presence of stress fibers) and motile cells (detected by the presence of membrane ruffles) was quantified. In the absence of growth factors, numerous F-actin stress fibers were present, but when cells were stimulated, these fibers dissociated, resulting in the redistribution of F-actin to membrane ruffles at the cell surface (Fig. 2C). The incidence of scattering and membrane ruffling recapitulated the trends seen in growth factor-stimulated cell chemotaxis, confirming that growth factors promote robust actin cytoskeletal reorganization during hNSC motility (p < .001) and that HGF elicits the strongest induction of chemokinesis (Fig. 2C). Collectively, these results demonstrate that hNSCs have a tremendous capacity to respond to growth factors in their environment that promote cell motility and that this response is reflected in cytoskeletal reorganization, a critical requirement for cell migration.
HGF Stimulation Promotes the Formation of a c-Met Receptor-Signaling Complex at the hNSC Plasma Membrane
Since HGF promotes a strong intracellular signaling response in hNSCs, which results in the stimulation of migration, we investigated the biochemical events induced by HGF treatment. We performed immunoprecipitation assays to identify endogenous intracellular binding partners of the HGF receptor (c-Met) and the p85α regulatory subunit of the PI3K signaling complex, both of which are critical components of stem cell migration . Treatment of HB1.F3 cells with HGF, followed by immunoprecipitation of c-Met or p85α, revealed associations between these two proteins (Fig. 3A). Furthermore the scaffold protein Gab1, which serves as a docking site for downstream adaptor proteins, was bound to c-Met and p85α during receptor activation (Fig. 3A). Ras, a member of the superfamily of small GTPases that regulates multiple signaling pathways, was also bound to c-Met and p85α (Fig. 3A). Activation of PI3K signaling requires Ras and another small GTPase, Rac1 . In support of these findings, Ras and Rac1 also coimmunoprecipitated with the active c-Met receptor and p85α PI3K subunit (Fig. 3A).
The association of PI3K, Ras, and Rac1 is important because Rac1 is a critical molecular switch that regulates membrane ruffling and cell migration. Since Rac1 can directly regulate the cytoskeleton, we next tested whether HGF treatment affects Rac1 activation in hNSCs. Using an affinity pull-down assay to detect GTP-loaded Rac1, we observed that stimulation with HGF considerably increased the level of GTP-Rac1, compared with untreated cells (Fig. 3B). These results demonstrate that HGF can promote the formation of an active PI3K-Ras-Rac1 signaling complex.
Because activation of c-Met and the PI3K-Ras-Rac1 signaling complex leads to the activation of Rac1, and these events typically take place at the plasma membrane, we examined the possibility that HGF can stimulate the production of an active signaling complex at the cell surface. Monolayer cultures of HGF-stimulated and nonstimulated HB1.F3 cells were seeded on a fibronectin substratum and stained for F-actin to identify membrane ruffles, which were colocalized with HGF effector protein distribution. In untreated cells, the c-Met receptor did not colocalize with the cytoskeleton, whereas HGF treatment resulted in the redistribution of F-actin to the plasma membrane, where it colocalized with the c-Met receptor (Fig. 3C). In a similar manner, F-actin redistributed and colocalized with Gab1, Ras, and p85α in HGF-stimulated hNSCs (Fig. 3C), suggesting that the HGF-induced biochemical association occurs at the plasma membrane of migrating cells. Plasma membrane-associated colocalization of Rac1 with F-actin was observed in HGF-stimulated cells. However, Rac1 was also identified beyond F-actin staining in membrane protrusions, suggesting that Rac1 is functioning at the leading edge of actin nucleation (Fig. 3C). To verify that the membrane protrusions are sites of actin elongation, we examined the cellular distribution of the actin nucleation protein Arp2 in HGF-treated and untreated cells. The distribution of Arp2 in HGF-stimulated cells was similar to that of Rac1, such that Arp2 colocalized with F-actin but extended beyond the actin filaments into the membrane protrusions (Fig. 3C). Taken together, these results illustrate that when HGF signals through its cognate receptor, c-Met, a Ras-PI3K-Rac1 intracellular signaling complex forms, which leads to the activation of actin redistribution at the leading edge of the cell.
Ras and PI3K Signaling Are Critical for hNSC Motility
We next investigated whether there were any specific requirements for HGF signaling components in mediating hNSC migration. Since Ras plays a central role in the transduction of intracellular signals to multiple pathways, which include PI3K and mitogen-activated protein kinase (MAPK), we used a transwell assay to determine the roles for Ras in regulating hNSC migration. Twenty-four hours prior to HGF stimulation, HB1.F3 cells were transduced with either wild-type Ras or dominant-negative Ras G15A. HGF-induced cell migration was significantly decreased by inhibition of Ras signaling (F= 105.8, p < .0001) (Fig. 4A), suggesting that Ras activation and redistribution to the plasma membrane are critical steps in the motility of hNSCs. Western blots were performed to determine changes in downstream cell signaling events. HB1.F3 cells displayed constitutive activation of MAPK activity (p-ERK 1/2), even in the absence of growth factors, and MAPK pathway activation was marginally reduced through inhibition of Ras (Fig. 4B). Furthermore, stimulation with HGF induced pronounced PI3K pathway activation (p-AKT) that was significantly attenuated in the absence of Ras signaling (Fig. 4B). These data suggest that HGF-stimulated cell migration preferentially used PI3K signaling instead of MAPK. To elucidate downstream targets of Ras, HB1.F3 cells were transduced with wild-type or dominant-negative MEK1 A217/A221 to block activation of MAPK. Inhibition of MEK1 failed to inhibit HGF-induced cell migration (F= 0.59, p > .05) (Fig. 4C), and Western analysis demonstrated that inhibition of MEK1 was sufficient to prevent activation of MAPK (Fig. 4D). These data demonstrate that MAPK signaling is not required for hNSC motility and suggest that Ras-dependent signaling in migrating cells occurs through an alterative pathway. Pharmacological inhibition of additional MAPK signaling cascades, p38, or JNK also failed to antagonize cell migration (data not shown).
Therefore, we next investigated the role of PI3K in Ras-dependent signaling. HB1.F3 cells were transduced with wild-type or dominant-negative p85α that lacked the p110-binding domain between the two C-terminal SH2 domains (iSH2Δ). When PI3K signaling was blocked, a strong and significant inhibition of HGF-induced cell motility was observed (F= 49.5, p < .0001) (Fig. 5A). Moreover, this inhibition was reflected in attenuated PI3K activation (Fig. 5B). These data demonstrate that Ras and PI3K are required for cell migration that is mediated through HGF receptor activation. Having shown that HGF activates the Ras and PI3K target effector protein Rac1, we used our cell migration assay to determine whether Rac1 is a critical HGF-associated target that regulates the switch between a static state and a motile state. HB1.F3 cells were transduced with either wild-type or dominant-negative Rac1 (N17). Inhibition of Rac1 activation (Fig. 5D) significantly inhibited HGF-stimulated cell migration (F= 51.8; p < .0001) (Fig. 5C), confirming that HGF signaling regulates the activity of Rac1, which has a direct role in cytoskeletal reorganization.
Growth Factor Signaling Converges on PI3K to Promote hNSC Migration
We investigated the tropism of hNSCs to gliomas through a highly efficient lentivirus-mediated transduction of HB1.F3 cells with shRNAs to directly target and knock down components of the HGF signaling cascade. We reasoned that if the primary chemotactic growth factor produced by gliomas is HGF, then a reduction in the expression of c-Met would have a dramatic effect on the tumor-targeting ability of hNSCs. Lentivirus-mediated transduction of c-Met shRNA significantly reduced c-Met expression levels relative to cells transduced with scrambled c-Met shRNA and untransduced cells (F= 6.3, p < .01) (Fig. 6B). As anticipated, hNSC migration toward HGF was significantly attenuated in transwell migration assays in which c-Met receptor expression was reduced (F= 64.3, p < .001) (Fig. 6A). Furthermore, this effect was specific to HGF, as neither VEGF- nor EGF-induced cell migration was affected (F= 1.05, p > .05; F= 0.76, p > .05) (Fig. 6A). To assess the requirement for c-Met in hNSC gliomatropism, control hNSCs cells deficient in c-Met were challenged with conditioned media collected from tumor and control cell lines. Loss of the c-Met receptor abolished hNSC migration toward U251 and U87 glioma-conditioned media, and a similar reduction was observed for the MRC5 fibroblast cell line, which also produced high levels of HGF (F= 85.2, p < .001; F= 6.66, p < .01; F= 10.9, p < .01) (Fig. 6A). These results demonstrate the critical requirement for c-Met signaling in mediating chemotaxis to sources of HGF, such as gliomas.
To examine whether chemotaxis is also dependent on PI3K, we transduced HB1.F3 cells with p85α shRNA to diminish PI3K activity. The PI3K knockdown response was statistically different from the response of scrambled p85α shRNA and untransduced cells (F= 5.9, p < .01) (Fig. 6D). Cells that lacked p85α showed a significant decrease in migration toward HGF, VEGF, and EGF (F= 69.9, p < .001; F= 15.4, p < .001; F= 29.6, p < .001) (Fig. 6C), suggesting that multiple growth factor-mediated chemotaxis pathways converge on and require PI3K. Similar transwell experiments were performed using conditioned media collected from tumor and control cell lines. Loss of p85α attenuated the migration of hNSCs toward conditioned media from the U251, U87, and MRC5 cell lines (F= 75.2, p < .001; F= 6.70, p < .01; F= 45.8, p < .001) (Fig. 6C). Therefore, hNSC tumor tropism is also critically dependent upon PI3K. Collectively, these results reveal that hNSCs possess the ability to respond to multiple extracellular factors and that these signals converge on PI3K, which in turn regulates cytoskeletal reorganization through Rac1.
Migration of NSCs Toward Gliomas Is a Broad Phenomenon
We noted that attenuation of the expression of the c-Met receptor in HB1.F3 cells effectively abolished all migratory responses to glioma cells (Fig. 6A). This suggests that HGF is the principal tumor-derived growth factor capable of stimulating stem cell chemoattraction. To determine whether other hNSC lines also migrate toward tumor-derived HGF and use PI3K signaling to do so, we generated two additional clonal v-myc-immortalized human fetal NSC lines, F5 and A4. These hNSC lines were challenged using transwell experiments containing recombinant HGF or conditioned media collected from U251 and U87 glioma tumor cells. As described above, H1B.F3 cells migrate toward media that contain recombinant HGF or conditioned media from glioma cell lines. A similar chemotactic migration was observed in F5 and A4 cell lines, suggesting shared tropic signaling (Fig. 7A–7C). To examine whether this gliomatropism is dependent on PI3K, hNSCs were pretreated with the selective PI3K inhibitor LY294002. Inhibition of PI3K significantly attenuated migration of all three hNSC cell lines assessed, demonstrating that the targeting of tumors by hNSCs has a common signaling pathway that uses PI3K as a critical convergence point (p < .001) (Fig. 7A–7C).
To validate the inherent tumor-targeting properties of hNSCs, U251 gliomas were implanted into the right posterior cortex of immune compromised mice and permitted to establish for a period of 2 weeks, at which point CM-DiI-labeled H1B.F3 cells were injected into the right anterior cortex. Five days after the injection of HB1.F3 cells, mice were euthanized and perfused with 4% paraformaldehyde, and brains were sectioned and subsequently examined for the location of labeled hNSCs. CM-DiI-labeled H1B.F3 cells were observed in the glioma tumor area, demonstrating their inherent tumor-targeting properties in vivo (Fig. 7D). In a similar manner, CM-DiI-labeled F5 or primary nonimmortalized hNSCs were injected into the right anterior cortex of glioma-bearing mice and permitted to migrate over a period of 5 days. The rates of migration of F5 cells and primary hNSCs were comparable to the rates of migration of H1B.F3 cell toward glioma (Fig. 7E). Collectively, our results demonstrate that hNSCs, both primary and immortalized lines, possess an inherent tumor-targeting ability. This broad finding increases the general applicability of hNSCs to the development of novel cancer therapies.
The ability of HGF to regulate positional information within the cortex of the central nervous system is consistent with the developmental expression patterns of HGF and its receptor c-Met, where HGF appears to be an important tissue organizer [43, –45]. The strongest chemotactic and chemokinetic response in the H1B.F3 hNSC line was induced by HGF, and this finding is in agreement with the developmental role of HGF. It is likely that isolated hNSCs will retain their inherent developmental capabilities, such as cell migration and attraction to sources of HGF.
The absence of a significant penetrance of c-Met mutations in gliomas initially overshadowed any potential involvement of HGF and c-Met in gliomagenesis, but recent observations underscore the importance of these proteins as crucial factors in the oncogenic process through dysregulation of their expression [46, , , , –51]. The autocrine and paracrine actions that are generated through HGF and c-Met overexpression appear to be a critical step in the initiation of tumor metastasis, and the extent of overexpression is a strong prognostic factor of the tumor grade [51, 52]. Indeed, the glioma cell lines selected for our study strongly overexpressed HGF, and this HGF contributed greatly, in a dose-dependent manner, to hNSC gliomatropism. These findings are in agreement with the observed chemotaxis of the immortalized murine NSC line C17.2, which shows a strong chemoattraction to both recombinant HGF and glioma-produced HGF [30, 53].
These combined data suggest that NSCs can use HGF as an instructive migratory cue. In addition, a functional c-Met receptor is essential, because ablation of this receptor significantly inhibits the capacity for hNSCs to migrate toward sources of HGF. Stimulation of hNSCs with HGF facilitates the association of the scaffold protein Gab1 with c-Met, which subsequently leads to the activation of multiple signaling pathways, including Ras, PLCγ, and PI3K. However, compared with other growth factors that are primarily mitogenic, HGF can elicit a strong motogenic response, likely because of prolonged Gab1 phosphorylation [41, 54]. Similarly, we found that growth factors such as VEGF, EGF, and TGFα could stimulate migration, but the strongest motogenic response was elicited by HGF. The activation of Ras appears to be a critical step in the downstream activation of PI3K, which is crucial for cell migration and invasion of transformed and nontransformed cells [55, 56]. Likewise, ablation of PI3K signaling through shRNA targeting of the p85α subunit in hNSCs caused a similar and significant attenuation of cell migration toward HGF and all other growth factors assayed. Collectively, these results demonstrate that multiple growth factor signals converge on PI3K signaling to mediate cell motility in hNSCs. This suggests that PI3K functions as an important node for the intracellular transduction of multiple extracellular signals. PI3K is particularly important because of the subsequent divergence of signals that can occur to mediate its pleiotropic effects . In the context of HGF signaling, the Ras-PI3K branch directly regulates Rac1 and p21-activated kinase to control cytoskeletal rearrangement and cell motility , events that are recapitulated in our hNSCs. Regulation of the cytoskeleton is paramount for hNSC migration.
Cell migration in other systems is regulated by intracellular signal transduction that is mediated by PI3K, and one of the principal antagonists of PI3K signaling is the tumor suppressor phosphatase and tensin homologue deleted on chromosome 10 (PTEN) [36, 58, 59]. PTEN is commonly mutated in gliomas, and this mutation confers a significant advantage on the invasiveness of the tumor [60, –62]. The regulation of PTEN-PI3K appears to be critical for the activation state of the key effector GTP-binding proteins Rac1 and Cdc42 .
The shared signaling pathways between hNSCs and glioma invasion in the brain parenchyma highlights the common molecular mechanism of cell migration and suggests that normal stem cells and cancer stem cells use the same signaling pathways for transportation. The difference lies in the cells' use of shared cellular resources—regulated use in hNSC migration versus deregulated, constitutive use in tumor cells that enables glioma invasion. In summary, our data provide important insight into the signal transduction pathways that operate during the migration of NSCs to glioma, which is expected to aid in the development and optimization of NSC-based tumor-targeting therapeutic platforms. Finally, our data may also have implications for cell replacement therapies that use stem cells and rely on their migration to organs and tissues with various pathologies.
Disclosure of Potential Conflicts of Interest
The authors indicate no potential conflicts of interest.
We appreciate the City of Hope Microscopy Core Facility's helpful assistance with confocal microscopy. We are grateful to Dr. Steven Flanagan and the City of Hope Microarray Core for technical assistance with real-time PCR. We also thank Drs. Kristine Justus and Keely Walker for critical scientific evaluation and editing of the manuscript. This work was supported in part by grants to K.S.A. from the Stop Cancer Foundation, the Neidorf Family Foundation, the Marcled Foundation, and the H.L. Snyder Foundation and to C.A.G. from the NIH, National Cancer Institute (Grant CA107245).