Dynamin 2 mediates PDGFRα-SHP-2-promoted glioblastoma growth and invasion.

Dynamin 2 (Dyn2), a large GTPase, is involved in receptor tyrosine kinase (RTK)-promoted cell migration. However, the molecular mechanisms by which Dyn2 regulates RTK-induced cell migration have not been established. Recently, we reported that tyrosine-protein phosphatase non-receptor type 11 (SHP-2) and phosphatidylinositol 3-kinase (PI3K) mediate platelet-derived growth factor receptor-α (PDGFRα)-promoted glioma tumor growth and invasion. Here, we show that Dyn2 is an effector downstream of the PDGFRα-PI3K/SHP-2 signaling in glioma cells. Depletion of endogenous Dyn2 by short hairpin RNAs (shRNAs) inhibited PDGFRα-stimulated phosphorylation of Akt, Erk1/2, Rac1 and Cdc42 activities, glioma cell migration and survival in vitro and tumor growth and invasion in the brains of mice. Dyn2 binds to SHP-2 and PI3K and colocalizes with PDGFRα at the invasive fronts in PDGF-A-stimulated glioma cells. Inhibition of SHP-2 by siRNA knockdown abrogated Dyn2 association with activated PDGFRα and PDGFRα activation of Rac1 and Cdc42, and glioma cell migration, thereby establishing a link between SHP-2 interaction with Dyn2 and the PDGFRα signaling. Furthermore, a dominant-negative SHP-2 C459S mutant inhibited PDGF-A-stimulated glioma cell migration, phosphorylation of Dyn2 and concomitantly blocked PDGFRα-induced Src activation. Inhibition of Src by Src inhibitors attenuated PDGF-A-stimulated phosphorylation of Akt and Dyn2 and glioma cell migration. Additionally, mutations of binding sites to PI3K, SHP-2 or Src of PDGFRα impaired PDGFRα-stimulated phosphorylation of Akt and Dyn2, and Dyn2 association with activated PDGFRα. Taken together, this study identifies Dyn2 as an effector that mediates PDGFRα-SHP-2-induced glioma tumor growth and invasion, suggesting that targeting the PDGFRα-SHP-2-Dyn2 pathway may be beneficial to patients with malignant glioblastomas.

Introduction

Glioblastoma multiformes (GBMs) are the most common and malignant tumors in the central nervous system (CNS) (Jemal ). The inherent single cell invasive nature of GBMs contributes to resistance to therapy and high frequency of tumor recurrence of these tumors after maximal surgical resection, combined with radiotherapy and concurrent or adjuvant chemotherapies. The median survival time of patients with GBMs remains 14–16 months after diagnosis (Furnari ; Wen and Kesari, 2008). Thus, understanding the mechanisms underlying the intrinsic tumor infiltration is urgent to improve efficacy of treatments for highly invasive GBMs (Giese ). Recently, coordinated genomic analyses of a large cohort of clinical GBM specimens showed that the gene encoding platelet-derived growth factor receptor α (PDGFRα) was amplified in ~13% of the total GBMs analyzed, ranking PDGFRα at third among the top 11 amplified genes in GBMs (TCGA, 2008). In clinic, overexpression of PDGFRα is associated with a poor prognosis and shorter survival time for patients (Furnari ; Haberler ; Wen and Kesari, 2008). Although activation of PDGFRα-specific pathway involves tumor progression of malignant GBMs, molecular mechanisms by which PDGFRα promotes GBM growth and invasion remain largely unknown (Calzolari and Malatesta, 2010).

The superfamily of dynamin (Dyn) includes classical Dyns and Dyn-related proteins that play important roles in endocytosis (Damke ; van der Bliek ), cell migration and invasion (Kruchten and McNiven, 2006) and cancer cell proliferation (Joshi ). There are three different Dyn isoforms in mammalian: Dyn1 is restricted to express in neuronal cells; Dyn2 is ubiquitously expressed; and Dyn3 may be limited to brain, lung and testis (Kruchten and McNiven, 2006). Abundant evidence implicates Dyns as movers during cell migration and invasion (Kruchten and McNiven, 2006; McNiven ). For example, Dyn2 interacts with components of cytoskeleton, thereby influencing the movement of intracellular vesicles (Lee and De Camilli, 2002; Orth and McNiven, 2003), modulating cell shape (McNiven ), and affecting podosomal adhesions (Lee and De Camilli, 2002; Ochoa ), invadopodia (Baldassarre ), and phagocytosis (Gold ). Upon PDGF stimulation, Dyn becomes markedly associated with membrane ruffles and lamellipodia, leading to increased cell migration (Cao ; Krueger ; McNiven ). Moreover, a dominant-negative (DN) Dyn2 K44A mutant blocks PDGF-induced macropinocytosis (Schlunck ). However, although this evidence suggests that Dyn is involved in PDGF-stimulated cell migration and other studies show that Dyn interacts with PI3K and phospholipase Cγ-1 (PLCγ-1) that are common downstream modulators for receptor tyrosine kinases, including PDGFRα (Scaife ), how Dyn mediates PDGFRα-promoted cell migration is still elusive.

SHP-2 is a non-receptor protein tyrosine phosphatase (PTP) that has been implicated in the activation of the Ras-MAPK signaling by receptors for various growth factors and cytokines, including PDGFRα signaling (Matozaki ; Neel ). Moreover, activating SHP-2 mutations were observed in hematological malignancies and other types of cancers including two E69K and I292M mutations in GBMs and a T22A mutation in oligodendrogliomas (Navis ). Most recently, PTPN11 (encoding SHP-2 protein) was identified as a “Linker” gene for connections to various altered genes in GBMs by genomic analysis (Cerami ). Our recent report functionally validated this hypothesis demonstrating that SHP-2 mediates PDGFRα-promoted tumorigenesis and tumor invasion in mouse Ink4a/Arf null astrocytes and human glioma cells (Liu ). In this study, we investigated mechanisms of PDGFRα-SHP-2 stimulation of glioma cell growth and invasion. We show that Dyn2 is a downstream effector of the PDGFRα-PI3K/SHP-2 signaling. Mechanistically, Dyn2 regulates PDGFRα-stimulated glioma cell growth and invasion through interaction with activated PDGFRα, SHP-2 and PI3K as well as Src-dependent protein phosphorylation.

Results

Depletion of Dyn2 inhibits PDGFRα-stimulated glioma cell migration in vitro

To determine the role of Dyn2 in PDGFRα-regulated glioma cell growth and invasion, we first examined the expression of Dyn isoforms in various glioma cell lines, mouse fibroblasts (NIH3T3) and astrocytes (mAst). As shown in Figure 1A, Dyn1 is expressed at low levels in all cell lines examined. Dyn3 is highly expressed in LN428, U138, LN443, T98G, LN235 and NIH3T3 cells but at low levels in other cell lines. In contrast, Dyn2 was found at high levels in all cell lines analyzed. To determine the function of Dyn2, we chose SNB19 and LN444 cells with high levels of Dyn2. We first knocked down Dyn2 using lentivirus-encoded shRNAs in LN444 and SNB19 glioma cells that were stimulated with PDGF-A {PDGF-A only activates PDGFRα, (Tallquist and Kazlauskas, 2004)}. As shown in Figure 1B and Figure S1, knockdown of Dyn2 in these glioma cells by two separate shRNAs but not a control shRNA reduced PDGF-A-induced phosphorylation of Akt and Erk1/2 as well as GTP-loading of Rac1 and Cdc42, two critical small GTPases that mediate Dyn-modulated cell motility (Kruchten and McNiven, 2006) without affecting the expression and tyrosine phosphorylation (p-Y) of PDGFRα. While PDGF-A stimulated cell migration of LN444 and SNB19 cells (Figure 1C and Figure S2, Supplemental movies 1 and 2), knockdown of Dyn2 markedly inhibited PDGF-A-stimulated cell migration of both cell lines (Figure 1C and Figure S2). Interestingly, depletion of Dyn2 by shRNAs did not affect FBS-, EGF-, and HGF-stimulated glioma cell migration (Figure S3), suggesting that Dyn2 is important and specific for PDGFRα-promoted glioma cell migration.

Figure 1

Knockdown of Dyn2 inhibits PDGFRα-promoted glioma cell migration in vitro

A. IB assays of expression of Dyn2, Dyn3 and Dyn1 in various glioma cells, mouse astrocytes (mAst) and fibroblast (NIH3T3) cells. β-actin was used as loading controls.

B. IB assays. Knockdown of Dyn2 by shRNAs inhibits PDGFRα stimulation of protein phosphorylation of Akt and Erk1/2, activities of Cdc42 and Rac1 in PDGF-A-stimulated LN444 and SNB19 glioma cells. A GFP shRNA was used as a control. Akt, Erk1/2, Cdc42, Rac1, β-actin and PDGFRα were used as loading controls.

C. In vitro cell migration assays using cells generated from (B). Fifty ng/ml PDGF-A was included in lower wells of a Boyden Chamber to induce cell migration for 8–10 h. Data is presented as a percentage of migrated cells normalized to the stimulated control (100%) from six replicates per pair per cell line; Bars, SD. *, P < 0.05, one-way ANOVA followed by Newman-Keuls post hoc test.

Results in A to C represent three independent experiments with similar results.

Knockdown of Dyn2 suppresses PDGFRα-stimulated glioma tumor growth and invasion in vivo

To determine the function of Dyn2 in PDGFRα-promoted glioma tumorigenesis, we knocked down Dyn2 in LN444/PDGF-A cells by two different lentivirus-encoded shRNAs (#1 and #2) and a control shRNA. As shown in Figure 2A, ~70% of endogenous Dyn2 in puromycin-resistant cell populations was reduced in LN444/PDGFA/shRNA cells when compared to the control shRNA cells. In vitro, depletion of Dyn2 in LN444/PDGF-A cells had a minimal impact on cell proliferation when compared to that of LN444/GFP or control shRNA cells (Figure 2B). However, knockdown of Dyn2 markedly attenuated PDGF-A/PDGFRα-stimulated cell survival (Figure 2C) and migration (data not shown). Additionally, treatment of LN444/PDGF-A/shDyn2 cells with AG1296, an inhibitor for PDGFR has no effect on PDGF-A-stimulated cell proliferation (Figure 2B) and survival (Figure 2C) and migration (not shown), suggesting that Dyn2 is not at upstream of PDGFR in glioma cells. These data suggest that under these experimental conditions, inhibition of Dyn2 by shRNA knockdown only affects PDGF-A-promoted glioma cell survival but not cell growth in vitro.

Figure 2

Knockdown of Dyn2 inhibits PDGFRα-promoted glioma tumor growth and invasion in the brains of mice

A. IB assays. Knockdown of endogenous Dyn2 by shRNAs in puromycin-selected LN444/PDGF-A glioma cell populations. C, control GFP shRNA; #1 and #2, Dyn2 shRNAs that target different sequences of Dyn2 mRNA. PDGF-A, exogenously expressed PDGF-A. β-actin was used as loading control.

B. In vitro MTT assays. 4,000 cells of various cells with similar passages were seeded in one well of 96-well plates with DMEM plus 0.5% FBS containing AG1296 (AG, 2 μM) or vehicle at indicated times. Six replicates per cell line. Cell proliferation was determined by MTT assays. The data was normalized to the mean MTT values of the untreated cells at Day 0 (assigned as 1) for each type of cells. Bars, SD.

C. TUNEL assays. Various cells with similar passages were seeded in 8-well chamber slides with DMEM plus 0.5% FBS containing 2 μM AG1296 or vehicle. After 48 h of the treatment, cell apoptosis was determined by TUNEL assays. One thousand cells of each slide were randomly examined and numbers of TUNEL-positive cells were counted. Bars, SD. *, P < 0.05, one-way ANOVA followed by Newman-Keuls post hoc test.

D. Impact of Dyn2 knockdown on PDGFRα-promoted LN444 glioma growth, invasion, cell proliferation and apoptosis in vivo. Representative H&E and IHC images of brain sections of 3 to 5 mice per group from two independent experiments. Mice received various LN444 glioma cells and developed brain tumors in various sizes 7 to 8 weeks post injection. Panels a, d, g and j, LN444/PDGF-A tumors expressed a control shRNA; b, e, h and k, LN444/PDGF-A/Dyn2 shRNA #1; c, f, i and l, LN444/PDGF-A/Dyn2 shRNA #2. H&E staining, panels a to f. Panels d to f are enlarged areas in a to c marked with squares. Panels g to i, Ki-67 staining. Panels j to l, TUNEL staining; arrows in d to f, invasive tumor cells (d) or non-invasive tumor boarders (e and f), in g to i, Ki-67 positive cells and in j to l, TUNEL positive cells. Scale bars in a to c, 1 mm, in d to f, 200 μm, in g to i, 50 μm and in j to l, 100 μm.

E and F. Quantifications of tumor size and relative invasive fingers per tumor. E. Tumor volumes were determined by microscopically measuring tumor areas followed by estimation of the tumor volume as an oval sphere-shape. All mice received various LN444/PDGF-A/shControl or LN444/PDGF-A/shRNA-Dyn2 cells developed brain gliomas with volumes varied but within acceptable standard deviations. F. Relative invasive figures were estimated microscopically by counting protruded tumor tissue figures and disseminated areas as shown in panel a and d. Data in E and F are calculated from 3 to 5 tumors per group. Bars, SD. *, P < 0.05, one-way ANOVA followed by Newman-Keuls post hoc test.

G and H. Quantifications of Ki-67 or TUNEL staining. Data is calculated from 3 to 5 tumors per group from two independent experiments. Bars, SD. *, P < 0.05, one-way ANOVA followed by Newman-Keuls post hoc test.

Results in A to H represent two to three independent experiments with similar results.

Next, we examined determine the function of Dyn2 in PDGFRα-promoted glioma tumorigenesis in vivo. In the brains of mice, LN444/PDGF-A glioma cells formed highly tumorigenic and invasive gliomas whereas LN444/GFP cells only established small and non-invasive tumors (Liu ). Thus, we separately implanted LN444/PDGFA/control-shRNA (shControl) and LN444/PDGF-A/Dyn2-shRNA (shRNA#1 and shRNA#2) cells into the brains of mice. As shown in Figure 2D, compared with control LN444/PDGF-A gliomas (Figure 2D, panels a and d), knockdown of Dyn2 in LN444/PDGF-A tumors markedly suppressed PDGFRα-stimulated glioma growth and invasion in the brains of mice (Figure 2D, panels b, c, e, f, Figure 2E and 2F). When various brain sections with different tumors were examined by immunohistochemistry (IHC) staining, LN444/PDGF-A/shRNA-Dyn2 tumors showed a significantly decrease in cell proliferation compared to LN444/PDGF-A/shControl tumors (Figure 2D, compare panel g with h and i; Figure 2G). Similarly, when compared to LN444/PDGF-A/control shRNA tumors, LN444/PDGF-A/shRNA#1 and #2 tumors displayed a marked increase in cell apoptosis (Figure 2D, compare panel j with k and l, Figure 2H). Taken together, these results suggest that Dyn2 is important for PDGFRα-promoted glioma cell growth and invasion in vitro and in vivo.

Dyn2 mediates PDGFRα-stimulated glioma cell migration through PI3K and SHP-2

Dynamin was previously shown to interact with PI3K in PC12 cells stimulated by PDGF (Scaife ). Since PI3K and SHP-2 mediate PDGFRα-stimulated glioma tumor growth and invasion in the brains of mice (Liu ) and Dyn interacts with PI3K (Scaife ), we hypothesized that Dyn2 mediates PDGFRα-SHP-2-stimulated glioma cell migration. To test this hypothesis, we first transiently transfected wild-type (WT) HA-Dyn2, dominant-negative (DN) HA-Dyn2-K44A mutant or a vector control into SNB19 and LN444 glioma cells, respectively. After validating the expression of exogenous Dyn2 by immunoblotting (IB, Figure S4A), we determined the impact on glioma cell migration by in vitro cell migration assays. As shown in Figure S4B, compared with the control SNB19 or LN444 cells, overexpression of WT Dyn2 did not significantly affect PDGFRα-stimulated glioma cell migration. However, expression of a DN Dyn2-K44A mutant significantly attenuated PDGF-A-stimulated cell migration of both cell lines. Thus, these data suggest that Dyn2 is important for PDGFRα-stimulated glioma cell migration.

Next, we determined whether Dyn2 interacts with SHP-2 and PI3K in SNB19 and LN444 cells stimulated by PDGF-A. As expected, PI3K was associated with Dyn2 in PDGF-A-treated glioma cells. Interestingly, Dyn2 also binds to SHP-2 in PDGF-A-stimulated glioma cells (Figure 3A), suggesting an involvement of Dyn2 in PDGFRα-SHP-2-stimulated glioma cell migration. To examine this possibility, we treated SNB19/WT HA-Dyn2 and SNB19/Control cells with inhibitors of PI3K (LY294002) or SHP-2 (PHPS-1 or NSC87877) with or without PDGF-A stimulation. As shown in Figure 3B, inhibition of PI3K or SHP-2 by their inhibitors effectively abrogated PDGF-A-stimulated glioma cell migration whereas overexpression of a WT HA-Dyn2 by SNB19 glioma cells did not fully but partially rescued the inhibitory effects by these inhibitors on PDGF-A-stimulated cell migration, suggesting Dyn2 as an effector downstream of the PDGFRα-PI3K/SHP-2 signaling. To further study this signaling, we examined the subcellular distribution of PDGF-A-stimulated Dyn2, cortactin (a Dyn2-binding protein that involves in PDGF-stimulated actin-remodeling) (Krueger ) and PDGFRα in serum-starved SNB19 cells that were transiently transfected with a WT Dyn2-GFP and pre-treated with or without LY294002, followed by PDGF-A stimulation. As shown in Figure 3C, consistently with previous reports (Krueger ; McNiven ), Dyn2 and cortactin distributed along the cell cortex and internal membrane compartments in non-PDGF-stimulated cells (Figure 3C, panel a and Figure S5, panels a to c). After PDGF-A stimulation, both proteins were redistributed and co-localized at the invasion fronts of cells (Figure 3C, panel b and Figure S5, panels d to f). Similarly, in non-stimulated cells, PDGFRα was distributed on internal membrane compartments (Figure 3C, panel d and Figure S5, panels j to l); PDGF-A stimulation induced PDGFRα redistribution and co-localization with Dyn2 at the invasion fronts of cells (Figure 3C, panel e and Figure S5, panels m to o). However, PI3K inhibitor LY294002 blocked their redistribution and co-localization at the invasion fronts of cells (Figure 3C, panels c and f and Figure S5, panels g to I and p to r). Taken together, these data show that activation of PDGFRα induces the association of Dyn2 with PI3K and SHP-2, and co-localization of PDGFRα, Dyn2 and cortactin, thereby augmenting PDGFRα-stimulated glioma cell migration.

Figure 3

Dyn2 mediates PDGF-A-stimulated glioma cell migration through PI3K and SHP-2

A. IP and IB assays of endogenous Dyn2 binding with PI3K and SHP-2. Dyn2 was used as a control.

B. Overexpression of HA-Dyn2 partially rescued the inhibitory effect of PI3K and SHP-2 inhibitors on PDGFRα-promoted glioma cell migration. Cells generated from (A) were serum-starved for 24 h, and then pre-treated with or without LY294002 (10 μM), PHPS-1 (100 μM), NSC87877 (100 μM) for 1 h on ice and analyzed by in vitro cell migration assays for 8 to 10 h as described in Figure 1C. Data is presented as a percentage of migrated cells normalized to the stimulated control (100%) from six replicates per pair per cell line. Bars, SD. *, P < 0.05, one-way ANOVA followed by Newman-Keuls post hoc test.

C. PDGF-A induced and a PI3K inhibitor inhibited co-localization of Dyn2 and cortactin, and Dyn2 and PDGFRα at the invasion fronts of SNB19 cells. Representative confocal microscopic images of immunofluorescent staining are shown. Individual images are shown in Figure S5. Serum-starved SNB19 cells transiently transfected with a WT Dyn2-GFP were pre-treated with 10 μM LY294002 for 1 h followed by 5 min stimulation with 50 ng/ml PDGF-A. Cells were then separately stained with anti-cortactin or anti-PDGFRα antibodies followed by secondary antibodies conjugated with different fluorophores (scale bar, 15 μm). N, nuclear; red arrows, co-localization of Dyn2 and cortactin or Dyn2 and PDGFRα at the invasion fronts; white arrows, Dyn2-GFP localized to the pericantriolar material and centrioles.

Results in A to C represent two to three independent experiments with similar results.

SHP-2 mediates interaction of Dyn2 with activated PDGFRα

Next, we assessed whether inhibition of SHP-2 affects the association between Dyn2 and PDGF-A-activated PDGFRα by knockdown of endogenous SHP-2 with a SHP-2 siRNA. As shown in Figure 4A, siRNA depletion of SHP-2 significantly decreased association of Dyn2 with activated PDGFRα compared with the control in PDGF-A-stimulated SNB19 and LN444 cells. Furthermore, impairment of association of Dyn2 with PDGFRα by knockdown of SHP-2 also markedly inhibited PDGF-A-stimulated Rac1 and Cdc42 activities (Figure 4A, lower panels) and glioma cell migration (Figure 4B). Thus, these data suggest that SHP-2 mediates Dyn2 binding to activate PDGFRα, thereby promoting Rac1 and Cdc42 activities and glioma cell migration.

Figure 4

SHP-2 mediates PDGF-A-stimulated Dyn2 association with PDGFRα, cell migration

A. Knockdown of SHP-2 inhibited PDGF-A-induced Dyn2 association with PDGFRα and PDGF-A activation of Rac1 and Cdc42. SNB19 and LN444 glioma cells were transiently transfected with a SHP-2 siRNA, or a control siRNA. Forth-eight h after transfection, cells were stimulated with 50 ng/ml PDGF-A for 5 min and analyzed by IP-IB for Dyn2 association with PDGFRα or Rac1 and Cdc42 activities. b-actin, SHP-2, Rac1 and Cdc42 were used as loading controls.

B. Knockdown of SHP-2 inhibited PDGF-A-stimulated cell migration. In vitro cell migration assays were performed for 8 to 10 h as described in Figure 1C using cells generated in (A). Data is presented as a percentage of migrated cells normalized to the stimulated control (100%) from six replicates per pair per cell line. Bars, SD. *, P < 0.05, one-way ANOVA followed by Newman-Keuls post hoc test.

Results in A and B represent two to three independent experiments with similar results.

PRD domain of Dyn2 is necessary for its association with SHP-2

It has been demonstrated that Dyn2 interacts with various Src homology (SH)-domain-containing endocytic adaptor proteins through its proline-rich domain (PRD) (Gout ; Kruchten and McNiven, 2006). To determine whether the PRD domain of Dyn2 is required for its association with SHP-2, we transiently transfected SNB19 cells with WT HA-Dyn2, a Dyn2 mutant lacking its PRD domain (HA-ΔPRD) or a vector control. As shown in Figure 6A, PDGF-A stimulation of SNB19 cells induced the association of WT Dyn2 with SHP-2 and GTP-loading activities of Rac1 and Cdc42 whereas the Dyn2 ΔPRD mutant failed to bind to SHP-2 and also inhibited PDGF-A-stimulated Rac1 and Cdc42 activities in SNB19 cells (Figure 5A). Next, we assessed the impact of loss of the association between Dyn2 and SHP-2 on PDGFRa-stimulated cell migration. As shown in Figure 5B, expression of WT Dyn2 did not significantly affect PDGF-A-induced cell migration whereas introduction of the ΔPRD mutant markedly impaired the PDGF-A-stimulated cell migration. Taken together, these data indicate that Dyn2 induces PDGFRα-SHP-2-promoted glioma cell migration through interaction with SHP-2 via its PRD domain.

Figure 6

SHP-2 regulates Src-dependent tyrosine phosphorylation of Dyn2 in PDGF-A-stimulated glioma cells

A. IB and IP assays. Expression of a DN C459S SHP-2 (catalytically-inactive) mutant inhibited PDGF-A-induced p-Akt but did not impair SHP-2 association with activated PDGFRα in SNB19 glioma cells. Akt and Myc were used as loading controls. Arrow, exogenously expressed a Myc-His tagged DN C459S SHP-2; dash arrow, endogenous SHP-2.

B. In vitro cell migration assays of the impact of expression of the DN C459S SHP-2 mutant on PDGFRα-induced glioma cell migration. Data is presented as a percentage of migrated cells normalized to the stimulated control (100%) from six replicates per pair per cell line.

C and D. IP and IB assays. Expression of the DN SHP-2 C459S mutant inhibited PDGF-A-induced p-Y of Dyn2 and p-SrcY418 in SNB19 glioma cells. Dyn2 and Src were used as loading controls.

E. IP and IB assays. Src family kinase (SFK) inhibitors PP2 and SU6656 attenuated PDGF-A-induced p-Y of Dyn2 in SNB19 glioma cells. Serum-starved SNB19 cells were pretreated with or without 2 μM PP2 or 2 μM SU6656 for 2 h and stimulated with or without 50 ng/ml PDGF-A for 5 min followed by IP-IB assays. Dyn2 and β-actin were used as loading controls.

F. In vitro cell migration assays of the impact of SFK inhibitors on PDGF-A-stimulated glioma cell migration. Data is presented as a percentage of migrated cells normalized to the stimulated control (100%) from six replicates per pair per cell line.

G. IB and IP assays. Effect of expression of PDGFRα WT or mutants on p-Y of Dyn2, Dyn2 association with PDGFRα, and p-Akt in mouse Ink4a/Arf null arstrocytes. Control, vector only; WT, PDGFRα wild type; F720, PDGFRα with a mutation at the SHP-2 binding site; F572/74, PDGFRα with mutations at the Src binding sites. Total Akt and PDGFRα were used as loading controls.

Bars in panels B and F, SD. *, P < 0.05, one-way ANOVA followed by Newman-Keuls post hoc test.

Results in A to G represent two to three independent experiments with similar results.

Figure 5

PRD domain of Dyn2 mediates PDGF-A-induced Dyn2 association with SHP-2, and Rac1 and Cdc42 activation

A. IP and IB assays. SNB19 glioma cells were transiently transfected with WT HA-Dyn2, HA-Dyn2-ΔPRD mutant or a vector control, respectively. Forty-eight h after transfection, cells were stimulated with 50 ng/ml PDGF-A for 5 min. Impact of Dyn2 ΔPRD mutant on Dyn2 association with SHP-2 (upper panels) and Rac1 and Cdc42 activation (lower panels) were determined by IP-IB analyses. β-actin, SHP-2, HA-Dyn2 WT, HA-Dyn2-ΔPRD, Rac1 and Cdc42 were used as loading controls. Arrow, HA-Dyn2 WT.

B. In vitro cell migration assays using cells generated from (A). Data is presented as a percentage of migrated cells normalized to the stimulated control (100%) from six replicates per pair per cell line. Bars, SD. *, P < 0.05, one-way ANOVA followed by Newman-Keuls post hoc test.

Results in A and B represent two to three independent experiments with similar results.

PDGFRα-SHP-2 induces glioma cell migration through Src-dependent tyrosine phosphorylation of Dyn2

SHP-2 is a ubiquitously expressed non-transmembrane PTP and essential in RTK signaling pathways (Chan ; Matozaki ). Thus far, we demonstrated that the interaction of SHP-2 and Dyn2 is required for PDGFRα-stimulated glioma cell migration. To assess whether activity of SHP-2 is required for Dyn2 function, we expressed a Myc-His tagged DN C459S SHP-2 (catalytically-inactive) mutant in SNB19 cells. As expected, the C459S SHP-2 mutant decreased PDGFRα-induced p-Akt (Figure 6A) and PDGFRa-promoted cell migration (Figure 6B). However, SHP-2 C459S mutant retained its ability to associate with PDGFRα induced by PDGF-A (Figure 6A). Next, since tyrosine phosphorylation (p-Y) of Dyn2 is critical for its function (Cao ; Weller ), we examined the impact of loss of SHP-2 activity on the induced p-Y of Dyn2. As shown in Figure 6C, expression of a DN C459S SHP-2 mutant, but not a WT SHP-2 markedly inhibited PDGF-A-induced p-Y of Dyn2. Furthermore, expression of C459S SHP-2, but not WT SHP-2, inhibited PDGF-A-stimulated p-SrcY418 (Figure 6D), consistent with a previous observation that SHP-2 regulates Src activity by affecting p-Y of SrcY418 (Zhang ). Additionally, since the DN C459S SHP-2 mutant inhibited p-Y of Dyn2 and p-SrcY418, it is possible that PDGF-A activated Src-induced p-Y of Dyn2, leading to increased cell migration. To test this hypothesis, we treated PDGF-A-stimulated SNB19 cells with two separate Src family kinase (SFK) inhibitors, PP2 and SU6656, and found that the inhibition of SFK activity significantly reduced PDGF-A stimulation of p-Y of Dyn2 (Figure 6E) and cell migration in vitro (Figure 6F).

Lastly, we investigated the roles of SHP-2 in recruiting Dyn2 to activated PDGFRα and Src in induction of p-Y of Dyn2 using mouse Ink4a/Arf null astrocytes that separately express PDGFRα WT and its various mutants (Liu ). As shown in Figure 6G and Figure S6, expression of WT PDGFRα enhanced p-Y of Dyn2, its association with PDGFRα and phosphorylation of Akt (p-AktS473) upon PDGF-A stimulation. In contrast, consistent with our recent report (Liu ), mutations of PDGFRα at the SHP-2 (F720), the Src (F572/74) or the PI3K (F731/42) binding sites did not significantly affect the overall p-Y of PDGFRα but reduced PDGF-A-induced p-AktS473. Furthermore, loss of capacities of interacting with these signal modules decreased PDGF-A-induced p-Y of Dyn2 and association with PDGFRα in these cells (Figure 6G and Figure S6). The residual PDGF-A-induced p-AktS473, p-Y of Dyn2 and Dyn2 association to activated PDGFRα in these mouse astrocytes were due to PDGF-A stimulation of endogenous PDGFRα in the astrocytes since these cells are not PDGFRα-deficient (Liu ). Taken together, these data suggest that SHP-2 activity is required for PDGFRα stimulation of p-Akt, p-Y of Dyn2, the association between Dyn2 and PDGFRα, and glioma cell migration, and that Src-dependent p-Y of Dyn2 mediates PDGFRα-induced glioma cell migration.

Discussion

In this report, as illustrated in a working model (Figure 7), we demonstrated a mechanism by which Dyn2 regulates PDGFRα-SHP-2/PI3K-stimulated glioma tumor growth and invasion. By using an established in vivo glioma model that activation of the PDGFRα-SHP-2/PI3K signaling promotes glioma growth and invasion in the brain (Liu ), we identified Dyn2 as a downstream effector for PDGFRα-induced glioma cell migration and survival in vitro, and tumor growth and invasion in vivo. Dyn2 mediates PDGFRα stimulation of glioma cell migration through association with SHP-2 and PI3K via its PRD domain, association and co-localization with the activated PDGFRα and activation of Rac1 and Cdc42. Inhibition of SHP-2 by siRNA knockdown, a DN SHP-2 mutant, SHP-2 inhibitors, deletion of the PRD domain of Dyn2 or a mutation of the SHP-2 binding site of PDGFRα suppressed PDGFRα-Dyn2-stimulated phosphorylation of Akt and Erk1/2, Rac1 and Cdc42 activities, and glioma cell growth and migration, thereby establishing a link of SHP-2 and Dyn2 with PDGFRα activation. Furthermore, we show that SHP-2 activity is required for Src-dependent p-Y of Dyn2, which is critical for PDGFRα-stimulated glioma cell migration.

Figure 7

A working model for the PDGFRα-PI3K/SHP2-Dyn2 signaling in glioma cells. PDGF-A activation of PDGFRα induces Dyn2 association with PI3K and SHP-2 and activates SHP-2 and Src, leading to Src-dependent tyrosine phosphorylation of Dyn2. Activation of Dyn2 stimulates its downstream effectors including Akt, Erk1/2, Rac1 and Cdc42, resulting in increased glioma growth and invasion.

One of the important aspects of this work is the identification of Dyn2 as a downstream modulator for PDGFRα-SHP-2-stimulated glioma tumor growth and invasion. Dyn has been implicated in PDGF-stimulated cell migration through changing its own cytoplasmic distribution in fibroblasts (McNiven ). Disruption of Dyn2 function altered Rac localization and inhibited Rac-dependent cell spreading and lamellipodia formation in PDGF-induced macropinocytosis in fibroblasts even though Rac was activated (Schlunck ). Interestingly, Dyn is also recently implicated in the promotion of cancer cell proliferation (Joshi ). Small-molecule inhibitors of Dyn {myristyl trimethyl ammonium bromides (MiTMAB)} competitively impaired the ability of Dyn to bind phospholipids and abrogated receptor-mediated endocytosis. Cells treated with Dyn inhibitors (MiTMAB and octadecyltrimethyl ammonium bromide) prevented the growth of a wide range of human cancer cells but had much less effects on non-tumorigenic fibroblast cells (Joshi ). Our results are consistent with these studies. We show that Dyn2 is essential for PDGFRα-SHP-2-stimulated Akt, Erk1/2, Rac1 and Cdc42 activities, and glioma tumor growth and invasion. Depletion of Dyn2 by shRNAs inhibited these PDGFRα-stimulated signaling, glioma cell migration and survival in vitro, and tumor growth and invasion in the brains of mice. Overexpression of WT Dyn2 did not significantly affect PDGF-A-induced cell migration whereas a DN Dyn2-K44A mutant or a Dyn2 δPRD mutant incapable of associating with SHP-2 impaired the PDGFRα stimulation. Taken together, these results establish a link between Dyn2 and activation of the PDGFRα-SHP-2 signaling, leading to an enhanced glioma tumor growth and invasion.

The second important finding in this study is that our data reveals a collaborative function of Dyn2 and SHP-2 to mediate PDGFRα-stimulated glioma cell invasion. A previous study showed that poliovirus (PV) infection of cultured human brain microvascular endothelial cells depends on Dyn-dependent caveolar endocytosis and intracellular virus-triggered the PV receptor (PVR) signals through the association of PVR with SHP-2 (Coyne ). In breast cancer, prolactin (PRL) and insulin-like growth factor I (IGF-I) regulate the pathogenesis and progression together via increasing proliferation, survival, and invasion. Src family kinase activity is required for IGF-IR association with SHP-2, and knockdown of Dyn2 reduces SHP-2 association with IGF-IR (Carver ). Our results in this study suggest that Dyn2 mediates PDGFRα-SHP-2-stimulated glioma cell growth and invasion through interaction with SHP-2 upon PDGFRα stimulation. Moreover, inhibition of SHP-2 by siRNA depletion, SHP-2 inhibitors, disruption of SHP-2 binding to Dyn2 or mutation of SHP-2 binding site of PDGFRα (F720) abrogated Dyn2 binding to activated PDGFRα, and PDGF-A-stimulated cellular signaling and glioma cell migration, demonstrating that Dyn2 mediates the PDGFRα-SHP-2 signaling in promoting glioma cell growth and invasion in vitro and in vivo.

The last novel finding in this study is that SHP-2-mediated Dyn2 phosphorylation via Src is essential for PDGFRα-stimulated glioma cell migration. Src plays an important role in regulating the endocytosis of PDGFRβ-G protein coupled receptor complexes, in which Src induces p-Y of Grb-2 associated binder-1 (Gab1) and promotes accumulation of Dyn2 at the plasma membrane (Waters ). Src is recently found to activate Dyn2 by phosphorylation to induce Golgi fragmentation (Weller ). Moreover, Src-dependent p-Y of Dyn2 and its associated actin-binding protein, cortactin regulate clathrin-mediated endocytosis of transferrin in epithelial cells (Cao ). Additionally, SHP-2 activates Src kinase by inhibiting the recruitment of C-terminal Src kinase, which exerts its inhibitory role by phosphorylating of SrcY529 (Zhang ). Furthermore, during angiogenesis, VEGF-A-mediated VEGFR-2 activation is down-regulated by collagen I but not vitronectin. SHP-2 is recruited to the activated VEGFR-2, and then promotes p-Y of Dyn2 via Src to mediate VEGFR-2 internalization (Mitola ). Consistent with these reports, our data not only validates that Src-dependent p-Y of Dyn2 induced by PDGFRα activation is critical for glioma cell migration, but also demonstrates that SHP-2 activity is required for Src-dependent p-Y of Dyn2, underscoring the link between Dyn2 and SHP-2 with PDGFRα-Src activation in promoting glioma cell migration. However, although we show that Src induces p-Y of Dyn2 and Src inhibitors exert an opposite effect, we cannot rule out the involvement of other members of the SFK family or other non-receptor tyrosine kinases in this process. This possibility is illustrated by a study that cortactin, a partner of Dyn, can stimulate cell motility (Kowalski ), and is phosphorylated by SFKs as well as Abl and Arg (Boyle ; Wu and J.T., 1991). In addition, Dyn2 also binds to Grb2, Dock4 (a Rac1 guanine exchange factor) and PDGFRβ, forming a ternary complex to regulate PDGF-dependent cell migration on fibroblasts (Kawada ). These alternative possibilities of how Dyn2 is involved in the PDGFRα-SHP-2 signaling in promoting glioma cell invasion warrant further investigation.

In summary, this study provides novel insights into the mechanisms by which Dyn2 mediates the PDGFRα-SHP-2/PI3K signaling through a SHP-2-induced and Src-dependent p-Y of Dyn2 in glioma cells. Our data identified Dyn2 as an effector downstream of the PDGFRα-SHP-2 signaling, critical in promoting glioma cell migration in vitro and tumor growth and invasion in vivo. Since Dyn2 is essential for various cell biological functions including cell motility and proliferation, our study suggests that targeting the PDGFRα-SHP-2-Dyn2 signaling may be beneficial for treatments of patients with glioblastomas and other types of human cancers that overexpress PDGFRα.

Materials and Methods

In vitro cell migration assays

In vitro cell migration assays were performed as we previously described (Imanishi ). Briefly, 5 × 105 cells/ml in 50 μl of DMEM plus 0.5% FBS were separately placed into the top wells of a Boyden chamber. The bottom wells contained 50 ng/ml PDGF-A or EGF, 40 ng/ml HGF, 10% FBS, or vehicle with or without indicated inhibitors. The cells were allowed to migrate through an 8-μm pore size membrane pre-coated with fibronectin (10 μg/ml) for 8–10 h at 37°C. Afterwards, the membrane was fixed and stained, non-migrating cells were removed and the remaining cells were counted. Six high-power fields (total magnification, X200) from each sample were randomly selected. Data were normalized to the stimulated control (100%) and analyzed using the GraphPad Software.

Living cell fluorescence real-time imaging

SNB19 cells expressing a nuclear-located H2A-GFP protein were seeded onto 35 mm No. 1.5 coverslip bottom dishes. After 24 h serum starvation, the medium was replaced with fresh medium with or without 50 ng/ml PDGF-A plus 0.5% FBS and incubated at 37°C, 5% CO2 coupled to a spinning disk UltraVIEW RS confocal imaging system (Perkin Elmer, Boston, MA). Cells were imaged at 10 min intervals, using a 300 ms exposure time during the 12 h time course. Images were compiled into a QuicktimeTM movie using the software Velocity (Improvision, Coventry, UK).

Cell proliferation and apoptosis

MTT assay

Various cells were seeded onto 96-well plates at a density of 4,000 per well, 6 replicates per cell line. After overnight incubation, the medium was replaced with fresh DMEM plus 0.5% FBS containing 2 μM AG1296 or vehicle. At indicated times, 50 μl MTT (2 mg/ml in PBS) was added to each well and incubated at 37°C for 4 h. MTT reduction by viable cells was measured colorimetrically at 570 nm using a Microplate Reader.

TUNEL assay

Various cells were plated on 8-well chamber slides. After overnight incubation, the medium was replaced with fresh DMEM plus 0.5% FBS with 2 μM AG1296 or vehicle. After 48 h, slides were fixed by 4% formaldehyde in ice-cold PBS and permeabilized with 0.2% Triton X-100 solution. After equilibration, slides were immersed in TUNEL reaction buffer at 37°C for 60 min inside the humidified chamber. Reaction was stopped by 2× saline sodium citrate (SSC) and slides were counterstained with Hoechst 33258 and mounted. One thousand cells of each slide were randomly examined and numbers of TUNEL-positive cells were counted under a microscope equipped with a SPOT digital camera.

Immunofluorescent staining and confocal microscopic analyses

SNB19 cells transiently transfected with Dyn2-GFP were plated on glass coverslips in media with 10% FBS. After overnight, the medium was replaced with fresh DMEM containing 0.5% FBS. The cells were cultured for an additional 24 h, pre-treated with 10 μM PI3K inhibitor, LY294002, for 1 h, and then stimulated with 50 ng/ml PDGF-A for 5 min. After the treatment, coverslips were washed and fixed by 4% formaldehyde in ice-cold PBS and permeabilized with 0.2% Triton X-100 solution. After equilibration, slides were stained with anti-cortactin (1:300) and anti-PDGFRα (1:50) antibodies in AquaBlock reagent in PBS at 4°C overnight, and then re-stained with secondary antibodies. Coverslips were mounted and examined under an Olympus Fluoview 1000 Confocal equipped with a digital camera.

Statistical analyses

One-way ANOVA or an unpaired, two-tailed Student's t test followed by Newman-Keuls post-test was performed using the GraphPad Prism software. A p value of 0.05 or less was considered statistically significant.