The presence of hypoxic areas in glioblastoma is an important determinant in tumor response to therapy and, in particular, to radiotherapy. Here we have explored the involvement of integrins, up to now known as regulators of angiogenesis and invasion, in the regulation of tumor hypoxia driven from the tumor cell. We first show that hypoxia induces the recruitment of αvβ3 and αvβ5 integrins to the cellular membrane of U87 and SF763 glioblastoma cells, thereby activating the focal adhesion kinase (FAK). We then show that inhibiting αvβ3 or αvβ5 integrins in hypoxic cells with a specific inhibitor or with siRNA decreases the hypoxia-inducible factor 1α (HIF-1α) intracellular level. This integrin-dependent regulation of HIF-1α is mediated through the regulation of FAK, which in turn activates the small GTPase RhoB, leading to the inhibition of GSK3-β. Furthermore, silencing this pathway in glioma cells of established xenografts dramatically reduces glioma hypoxia, associated with a significant decrease in vessel density. Our present results unravel a new mechanism of hypoxia regulation by establishing the existence of an αvβ3vβ5 integrin–dependent loop of hypoxia autoregulation in glioma. Targeting this hypoxia loop may be crucial to optimizing radiotherapy efficiency. [Cancer Res 2009;69(8):3308–16]

Tumor hypoxia has extensively been described as a frequent cause of failure in anticancer therapies and, in particular, radiotherapy of solid tumors. Recent interest has focused on effectively inhibiting the pathways of growth factors and their receptors, which influence tumor oxygenation. Targeting Ras, which is a crucial downstream effector of growth factor pathways, with a farnesyltransferase inhibitor increases radiosensitivity and oxygenation of Ras activated-tumor xenografts (1, 2). In the same way, targeting the small GTPase RhoB, which we have shown to regulate radioresistance mechanisms (3, 4), with another farnesyltransferase inhibitor in wild-type Ras-expressing glioma xenografts produces tumor oxygenation (5). All these recent data enforce the idea that increased attention should be given to the combination of radiotherapy with drugs targeting proteins regulating both radioresistance and hypoxia mechanisms.

Cells respond to hypoxia by regulating the expression of genes involved in angiogenesis, glycolysis, proliferation and pH regulation through an essential regulator of oxygen homeostasis, the transcriptional hypoxia-inducible factor 1 (HIF-1). HIF-1 is a heterodimer composed of HIF-1α and HIF-1β subunits. Our previous works have shown that the small GTPase RhoB, which mediates the basic fibroblast growth factor 2 (4) and epidermal growth factor (6) pathways, is a central regulator of hypoxia and radioresistance pathways. Indeed, inhibiting RhoB in vitro leads to an increased sensitivity of cervix carcinoma (4) and glioma (3) to ionizing radiation. Targeting this protein in vivo leads to an oxygenation and normalization of the xenografts vascularization (5, 7). This reduction of tumor hypoxia is in part a consequence of the RhoB-dependent proteasome degradation of HIF-1α through GSK-3β and Akt under hypoxic conditions (8). Very recently we shown that RhoB is also a mediator of the radioprotective signal activated by another class of membrane receptors, the integrins (9). Integrins are cell surface heterodimers composed of nonconvalently associated different α and β chains (for review, see ref. 10). Ligand binding to integrins leads to integrin clustering, and their association with proteins, such as integrin-linked kinase (ILK) or focal adhesion kinase (FAK), leads to focal adhesion clusters and the recruitment of actin filaments regulated by the Rho proteins (for review, see ref. 11). We showed that αvβ3 and αvβ5 integrins, known to be highly expressed in glioblastoma (12), activate a signaling pathway from the membrane that leads to the inhibition of the radiation-induced mitotic cell death through ILK and RhoB in glioblastoma cell lines (9). A few reports have described a regulation of integrin expression by hypoxia in various cancer cells. For example, the α6β4 integrin, but not α3β1 integrin, expression is induced by hypoxia in breast cancer cells (13) and regulates hypoxia-induced cell death in pancreatic cancer cells (14), whereas the αvβ3 integrin surface expression is also increased by hypoxia in melanoma cells (15). These data and our results led us to hypothesize that the RhoB-dependent regulation of HIF-1α in hypoxic glioma could be controlled by an upstream signaling pathway dependent on integrins, which, in turn, may regulate intrinsic radioresistance and hypoxia mechanisms in tumor cells.

The aim of this work was to investigate the role of two integrins, αvβ3 and αvβ5, in the control of hypoxia in glioblastoma cell lines expressing either one of these integrins (9). Our present work clearly identifies a new biological pathway of hypoxia regulation in glioblastoma cells controlled from the membrane by αvβ3 and αvβ5 integrins through FAK, RhoB, and GSK-3β activated under hypoxic conditions, which suggests the existence of a self-perpetuating loop of glioma hypoxia partially regulated by integrins. Furthermore, we provide evidence that inhibiting the αvβ3vβ5 integrin-FAK pathway in glioblastoma cells dramatically reduces hypoxia in glioblastoma xenografts, associated with a significant decrease in vessel density.

Cell culture. Human U87 and SF763 glioblastoma cells were routinely maintained in DMEM supplemented with 10% calf serum at 37°C in 5% CO2–humidified incubators and were subcultured weekly. Hypoxic conditions (1% oxygen) were obtained by incubating U87 and SF763 cells in a sealed “Bug-Box” anaerobic work station (Ruskinn Technologies/Jouan). To test the effect of EMD121974 (kindly provided by Merck; for review, see ref. 16), a RGD peptidomimetic inhibitor of αvβ3 and αvβ5 integrins, on HIF-1α stabilization, glioblastoma cell lines were treated with different concentrations of EMD121974 or vehicle 2 h before the end of the hypoxic incubation.

siRNA and DNA transfection. Cells were transfected with siscramble (siscr) or different siRNAs against human ILK, β3 subunit integrin, β5 subunit integrin, or RhoB as previously described (9) or with two siRNAs targeted to human FAK messenger, FAK1 sense (5′-GCAUGUGGCCUGCUAUGGA-dTdT-3′; ref. 17) and FAK2 sense (5′-GAAUGACCAUCUGGUUGAAA-dTdT-3′; ref. 18). Exponentially growing U87 cells were transfected with 40 nmol/L of FAK1 and FAK2, and exponentially growing SF763 cells were transfected with 40 nmol/L of FAK1 and FAK2 siRNA combined with Oligofectamine reagent in the conditions recommended by the manufacturer (Invitrogen).

Cells were also transfected, or cotransfected, with 2 μg of plasmid containing an activated form of RhoB, RhoBV14 (previously described in ref. 19).

Western blot analysis. Cells, untransfected or transfected with siRNA, were incubated under normoxia or hypoxia. Western blots were done as previously described (8, 9) and were normalized using a monoclonal anti-actin (Chemicon; diluted 1:10,000).

Immunofluorescence experiments. U87 and SF763 cells, on glass coverslips, were incubated under normoxia or hypoxia for 16 h and then fixed with 4% paraformaldehyde in PBS for 15 min before incubation or not with a permeabilization buffer (PBS containing 0.1% Triton X-100) for 5 min, before the application of the following primary antibodies: a mouse monoclonal anti–HIF-1α (BD Transduction Laboratories; diluted 1:500), a rabbit polyclonal anti–CA-9 (Santa Cruz Technology; diluted 1:200), a mouse monoclonal anti-vinculin (Sigma; diluted 1:200), a mouse monoclonal anti–β3 integrin (Chemicon International; diluted 1:50), and a rabbit polyclonal anti–β5 integrin (Chemicon International; diluted 1:100) overnight. Cells were then incubated with secondary antibodies antirabbit FITC (Rockland) at 1/500 dilution for 1 h or with FITC mouse antibodies (Sigma) at 1/500 and viewed using a confocal microscope (Zeiss). For each condition, the average number of focal adhesions per cell was determined for 15 randomly selected cells. More precisely, vinculin and integrin fluorescent spots, representing the focal adhesions or integrin staining, respectively, were counted for 15 cells using Zeiss AIM software.

Tumor xenograft generation and immunohistochemistry experiments. Mice were housed in an Institut Claudius Regaud Animal Care–accredited facility and maintained in accordance with Institut Claudius Regaud review committee and national guidelines. Mice were housed aseptically and inoculated as previously described (5, 7). Once the U87 xenografts were generated, mice were treated with siRNA (scr, FAK1, and β31) delivered directly into the tumor 5 d before the sacrifice, and immunohistochemistry analysis was done on paraffin-embedded sections (5 μm).

For hypoxia detection, after completion of the 5-d siRNA administration, mice were injected with 10 mmol/L EF5 (kindly given by Dr. Cameron Koch) as previously described (5, 7). Stained sections were viewed on a Nikon microscope through immersion lens.

For HIF-1α and vessel detection, after a rinse, an endogenous peroxidase block was applied for 15 min. The sections were incubated for 90 min with a mouse anti–HIF-1α (Becton Dickinson; diluted 1:100) or a rat monoclonal anti-CD31 (platelet/endothelial cell adhesion molecule 1; diluted 1:100; Becton Dickinson). Slides were counterstained with hematoxylin and viewed on a Nikon microscope through immersion lens.

RhoB activation assay. The Rho binding domain of Rhotekin (TRBD) was expressed as a recombinant fusion with glutathione S-transferase (GST) in E. coli as previously described (20). On reaching ∼80% confluence, U87 and SF763 cells, untransfected or transfected with siRNA, were incubated under hypoxia and RhoB activation assays were done as previously described (8).

Statistical analysis. Student's test was done to compare the means of values from different experiments. Differences were considered statistically significant at P < 0.05.

Hypoxia stimulates focal adhesion formation and the αvβ3vβ5 integrin pathways in human glioblastoma cells. We first examined whether hypoxia might stimulate focal adhesion formation and whether this stimulation favors αvβ3 and αvβ5 integrin recruitment at the membrane in two human glioblastoma cell lines, U87 and SF763, expressing αvβ3 and αvβ5 integrins, respectively (9). To accomplish this, we performed immunocytochemistry analysis with antibodies directed against the hypoxia marker carbonic anhydrase IX (CA-IX), vinculin, and β3 or β5 integrins in U87 and SF763 cells under normoxic or hypoxic conditions. As shown in Fig. 1A, hypoxia significantly increases the number of focal adhesions in both cell lines (vinculin labeling: 39.67 ± 8.37 for hypoxic U87 cells versus 15.2 ± 6.3 for normoxic U87 cells, P < 0.02; 34.73 ± 9.46 for hypoxic SF763 cells versus 10.53 ± 5.22 for normoxic SF763 cells, P < 0.02) associated with a significant increase of β3 integrin labeling in U87 cells (β3 integrin labeling: 40.87 ± 10.97 in hypoxic U87 cells versus 10.4 ± 6.21 in normoxic conditions; P < 0.02) and of β5 integrin labeling in SF763 cells (β5 integrin labeling: 38.53 ± 10.43 in hypoxic SF763 cells versus 9.33 ± 5.89 in normoxic SF763 cells, P < 0.02). These results strongly suggest that hypoxia might activate αvβ3 and αvβ5 integrins by enforcing their membrane recruitment.

Figure 1.

Hypoxia stimulates glioma cell focal adhesion by activating FAK. A, immunocytochemistry analysis of HIF-1α, carbonic anhydrase IX (CA9), vinculin, and β3 or β5 integrin was done as described in Materials and Methods in normoxic and hypoxic U87 and SF7653 cells. Data are representative of at least three different experiments. B, U87 and SF763 cells were incubated under conditions of normoxia or hypoxia for 16 h. Total cell lysates were resolved on SDS-PAGE and immunoblotted with the anti–P-FAK and total anti-FAK antibody. Representative of at least three different experiments. Histograms represent the ratio of P-FAK levels to total FAK levels as described in Materials and Methods. Columns, mean of three different experiments; bars, SD. ☆, under hypoxic (filled columns) conditions, U87 and SF763 cells exhibited a higher amount of P-FAK compared with normoxic (open columns) conditions (P < 0.01).

Figure 1.

Hypoxia stimulates glioma cell focal adhesion by activating FAK. A, immunocytochemistry analysis of HIF-1α, carbonic anhydrase IX (CA9), vinculin, and β3 or β5 integrin was done as described in Materials and Methods in normoxic and hypoxic U87 and SF7653 cells. Data are representative of at least three different experiments. B, U87 and SF763 cells were incubated under conditions of normoxia or hypoxia for 16 h. Total cell lysates were resolved on SDS-PAGE and immunoblotted with the anti–P-FAK and total anti-FAK antibody. Representative of at least three different experiments. Histograms represent the ratio of P-FAK levels to total FAK levels as described in Materials and Methods. Columns, mean of three different experiments; bars, SD. ☆, under hypoxic (filled columns) conditions, U87 and SF763 cells exhibited a higher amount of P-FAK compared with normoxic (open columns) conditions (P < 0.01).

Close modal

Because it is commonly known that FAK is activated by focal adhesion formation and is also a mediator of the integrin pathway, we then determined whether hypoxia might activate FAK through phosphorylation of its Tyr397 residue known to be activated by integrins (21). To accomplish this, we quantified the amount of activated FAK by studying the amount of the Tyr397-phosphorylated form of FAK in hypoxic cells. As shown in Fig. 1B, hypoxia statistically significantly increased the amount of FAK phosphorylated form (P < 0.01) without modifying the total amount of FAK (Fig. 1B). This strongly suggests that hypoxia stimulates the αvβ3 and αvβ5 integrin pathways through FAK and shows that hypoxia activates FAK in human glioblastoma cell lines.

αvβ3 and αvβ5 integrins regulate HIF-1α intracellular level under hypoxia via FAK but not ILK. To investigate the role of αvβ3 and αvβ5 integrins in the regulation of hypoxia mechanisms, we first treated glioblastoma cells with the specific αvβ3 and αvβ5 integrin inhibitor EMD121974 at various concentrations or vehicle for 16 hours at 37°C. As shown in Fig. 2A, inhibiting the activity of αvβ3 and αvβ5 integrins led to a significant (P < 0.01) and dose-dependent decrease of HIF-1α intracellular level under hypoxic conditions. We then analyzed the effect of integrin inhibition on U87 and SF763 cell response to hypoxia by specifically inhibiting either αvβ3 or αvβ5 integrin by using siRNA directed against the β3 (siβ3) or the β5 (siβ5) subunit. We have previously shown that transfecting U87 and SF763 cells with siβ3 or siβ5 dramatically reduced the amount of αvβ3 or αvβ5 integrins at the membrane (9). Silencing the αvβ5 integrin in U87 did not significantly modify the HIF1-α intracellular level. By contrast, transfecting U87 cells with two siRNAs directed against the β3 integrin subunit or transfecting SF763 cells with two siRNAs directed against the β5 integrin subunit significantly decreased the intracellular amount of HIF-1α [i.e., a 57% decrease for U87 siβ3-transfected cells (P < 0.001) and a 65% decrease for SF763 siβ5-transfected cells (P < 0.001); Fig. 2B]. These results show that inhibiting the αvβ3 or αvβ5 integrin pathway modulates the HIF1-α level under hypoxic conditions.

Figure 2.

Inhibiting the αvβ3 or αvβ5 integrins decreases HIF-1α accumulation in hypoxic glioblastoma cells. A, U87 cells, either treated with various concentrations of EMD121974 or vehicle, were incubated under conditions of hypoxia for 16 h. HIF-1α levels were analyzed by immunoblotting with anti–HIF-1α antibody as described in Materials and Methods. Data are representative of at least three different experiments. Histograms represent quantification of HIF-1α levels as described in Materials and Methods. Columns, mean of three different experiments; bars, SD. ☆, under hypoxic conditions, treated U87 cells (1.2 μmol/L) exhibited a lower amount of HIF-1α compared with untreated cells (P < 0.01). B, U87 and SF763 cells, either untransfected or transfected with siRNA directed against β3 or β5 integrin, were incubated under conditions of hypoxia for 16 h. HIF-1α levels were analyzed by immunoblotting with anti–HIF-1α antibody and quantified as described in Materials and Methods. The ratios of HIF-1α to actin, quantified as described in Materials and Methods, are represented at the bottom of the gel. Data represent the means ± SD of at least three different experiments.

Figure 2.

Inhibiting the αvβ3 or αvβ5 integrins decreases HIF-1α accumulation in hypoxic glioblastoma cells. A, U87 cells, either treated with various concentrations of EMD121974 or vehicle, were incubated under conditions of hypoxia for 16 h. HIF-1α levels were analyzed by immunoblotting with anti–HIF-1α antibody as described in Materials and Methods. Data are representative of at least three different experiments. Histograms represent quantification of HIF-1α levels as described in Materials and Methods. Columns, mean of three different experiments; bars, SD. ☆, under hypoxic conditions, treated U87 cells (1.2 μmol/L) exhibited a lower amount of HIF-1α compared with untreated cells (P < 0.01). B, U87 and SF763 cells, either untransfected or transfected with siRNA directed against β3 or β5 integrin, were incubated under conditions of hypoxia for 16 h. HIF-1α levels were analyzed by immunoblotting with anti–HIF-1α antibody and quantified as described in Materials and Methods. The ratios of HIF-1α to actin, quantified as described in Materials and Methods, are represented at the bottom of the gel. Data represent the means ± SD of at least three different experiments.

Close modal

We then determined whether the well-known integrin signaling mediators, FAK and ILK, were able to transduce the αvβ3vβ5 integrin–dependent regulation of hypoxia mechanisms. U87 and SF763 cells were transfected by two different specific siRNAs directed against FAK (siFAK) or ILK (siILK), inducing a 90% inhibition of FAK and a 90% inhibition of ILK in both cell lines (9). Inhibiting ILK in U87 and SF763 cells did not significantly modify the HIF-1α intracellular level under hypoxia (Fig. 3A). However, inhibition of FAK induced a significant decrease in the amount of HIF-1α in the two cell lines [i.e., a 63% decrease for U87 siFAK-transfected cells (P < 0.001) and a 60% decrease for SF763 siFAK-transfected cells (P < 0.01); Fig. 3A] showing that FAK, but not ILK, controls the glioma cell response to hypoxia.

Figure 3.

αvβ3 or αvβ5 integrins regulate HIF-1α intracellular level through FAK but not ILK. A, U87 and SF763 cells, either untransfected or transfected with two different siRNAs directed against FAK or ILK, were incubated under conditions of normoxia or hypoxia for 16 h. HIF-1α levels were analyzed by immunoblotting with anti–HIF-1α antibody as described in Materials and Methods. The ratios of HIF-1α to actin, quantified as described in Materials and Methods, are represented at the bottom of the gel. Data represent the means ± SD of at least three different experiments. B, U87 and SF763 cells, transfected or not with siRNA directed against β3 or β5 integrins, were incubated under conditions of normoxia or hypoxia for 16 h. Total cell lysates were resolved on SDS-PAGE and immunoblotted with anti–P-FAK and total anti-FAK antibody. The ratios of P-FAK to total FAK, quantified as described in Materials and Methods, are represented at the bottom of the gel. Data represent the means ± SD of at least three different experiments.

Figure 3.

αvβ3 or αvβ5 integrins regulate HIF-1α intracellular level through FAK but not ILK. A, U87 and SF763 cells, either untransfected or transfected with two different siRNAs directed against FAK or ILK, were incubated under conditions of normoxia or hypoxia for 16 h. HIF-1α levels were analyzed by immunoblotting with anti–HIF-1α antibody as described in Materials and Methods. The ratios of HIF-1α to actin, quantified as described in Materials and Methods, are represented at the bottom of the gel. Data represent the means ± SD of at least three different experiments. B, U87 and SF763 cells, transfected or not with siRNA directed against β3 or β5 integrins, were incubated under conditions of normoxia or hypoxia for 16 h. Total cell lysates were resolved on SDS-PAGE and immunoblotted with anti–P-FAK and total anti-FAK antibody. The ratios of P-FAK to total FAK, quantified as described in Materials and Methods, are represented at the bottom of the gel. Data represent the means ± SD of at least three different experiments.

Close modal

To check that αvβ3 or αvβ5 integrin hypoxia regulation was mediated by FAK, we then respectively transfected U87 and SF763 cells with siRNA directed against the β3 or β5 subunit and quantified the amount of phosphorylated FAK in these cells. As shown in Fig. 3B, inhibiting the αvβ3 integrin in hypoxic U87 cells or the αvβ5 integrin in hypoxic SF763 cells significantly reduced the amount of phosphorylated FAK [i.e., a 76% reduction of phospho-FAK (P-FAK) amount in siβ3-transfected U87 cells (P < 0.001) and a 90% decrease in siβ5-transfected SF763 cells (P < 0.01)] thereby showing that αvβ3vβ5 integrins regulate FAK activation in hypoxic glioma cells. Moreover, inhibiting this pathway by silencing integrins, FAK, or RhoB decreased HIF-1α transcriptionally regulated genes such as VEGF (data not shown) or CA-IX (8). These results show that αvβ3 or αvβ5 integrin regulates the HIF-1α intracellular level through FAK activation.

RhoB mediates the cell response of glioma cells to hypoxia controlled by αvβ3vβ5 integrins. Because we have previously shown that RhoB regulates HIF-1α proteasome–dependent degradation in U87 cells (8), we then studied the involvement of this small GTPase in the hypoxia-induced cell signaling regulated by αvβ3 or αvβ5 integrin through FAK. We compared the levels of RhoB-GTP in siβ3-transfected αvβ3 integrin–expressing U87 cells, siβ5-transfected αvβ5 integrin–expressing SF763 cells, and si-FAK-transfected cells versus siscr-transfected cells using the GST-Rhotekin RBD pulldown assay. As shown in Fig. 4A, RhoB activation was induced by hypoxia in U87 cells, as we previously reported (8), and in SF763 cells. This hypoxia-induced activation was inhibited after silencing αvβ3 integrin in U87 cells, αvβ5 integrin in SF763 cells, and FAK in the two cell lines (Fig. 4A; Supplementary Fig. S1). These results showed that αvβ3 and αvβ5 integrins control the hypoxia-induced RhoB activation in glioma cells expressing these integrins.

Figure 4.

The αvβ3vβ5 integrin-FAK pathway regulates HIF-1α degradation through RhoB activation and GSK3β in hypoxic glioma cells. A, U87 and SF763 cells transfected with siRNA directed against β3 or β5 integrin, ILK, or FAK as described in Materials and Methods were incubated under conditions of normoxia or hypoxia for 16 h. TRBD assay was done as described in Materials and Methods. The ratios of RhoB-GTP to total RhoB, quantified as described in Materials and Methods, are represented at the bottom of the gel. Data represent the means ± SD of at least three different experiments. B, U87 and SF763 cells transfected with siRNA directed against β3 or β5 integrin or against FAK or RhoB and then with cDNA encoding for active form of RhoB (RhoBV14), as described in Materials and Methods, were incubated under hypoxic conditions for 16 h. HIF-1α levels were analyzed by immunoblotting with anti–HIF-1α antibody as described in Materials and Methods. The ratios of HIF-1α to actin, quantified as described in Materials and Methods, are represented at the bottom of the gel. Data represent the means ± SD of at least three different experiments. C, U87 and SF763 cells transfected with siRNA directed against β3 or β5 integrin or against FAK or RhoB, as described in Materials and Methods, were incubated under conditions of normoxia or hypoxia for 16 h. Total cell lysates were resolved on SDS-PAGE and immunoblotted with anti–phospho-GSK3β (P-Ser9-GSK3β) and total anti-GSK3β antibodies. The ratios of P-Ser9-GSK3β to total GSK3β, quantified as described in Materials and Methods, are represented at the bottom of the gel. Data represent the means ± SD of at least three different experiments.

Figure 4.

The αvβ3vβ5 integrin-FAK pathway regulates HIF-1α degradation through RhoB activation and GSK3β in hypoxic glioma cells. A, U87 and SF763 cells transfected with siRNA directed against β3 or β5 integrin, ILK, or FAK as described in Materials and Methods were incubated under conditions of normoxia or hypoxia for 16 h. TRBD assay was done as described in Materials and Methods. The ratios of RhoB-GTP to total RhoB, quantified as described in Materials and Methods, are represented at the bottom of the gel. Data represent the means ± SD of at least three different experiments. B, U87 and SF763 cells transfected with siRNA directed against β3 or β5 integrin or against FAK or RhoB and then with cDNA encoding for active form of RhoB (RhoBV14), as described in Materials and Methods, were incubated under hypoxic conditions for 16 h. HIF-1α levels were analyzed by immunoblotting with anti–HIF-1α antibody as described in Materials and Methods. The ratios of HIF-1α to actin, quantified as described in Materials and Methods, are represented at the bottom of the gel. Data represent the means ± SD of at least three different experiments. C, U87 and SF763 cells transfected with siRNA directed against β3 or β5 integrin or against FAK or RhoB, as described in Materials and Methods, were incubated under conditions of normoxia or hypoxia for 16 h. Total cell lysates were resolved on SDS-PAGE and immunoblotted with anti–phospho-GSK3β (P-Ser9-GSK3β) and total anti-GSK3β antibodies. The ratios of P-Ser9-GSK3β to total GSK3β, quantified as described in Materials and Methods, are represented at the bottom of the gel. Data represent the means ± SD of at least three different experiments.

Close modal

To fully show the involvement of RhoB in the integrin pathway controlling HIF-1α, we then inhibited RhoB by transfecting U87 and SF763 cells with a siRNA directed against the 3′ untranslated region of the RhoB gene (siRhoB3) or with siRNA directed against FAK or integrins, and transfected these silenced cells with cDNA encoding for constitutively activated RhoB, RhoBV14 (19). Transfection of cells with siRhoB3 respectively induced an 83% inhibition of RhoB in U87 cells and an 82% inhibition in SF763 cells, which were reversed when cells were transfected with the active form of RhoB (Supplementary Fig. S2). As previously shown, inhibiting the αvβ3 and αvβ5 integrins or FAK in U87 and SF763 cells led to a decrease of HIF-1α intracellular level under hypoxic conditions (Fig. 3). This inhibition was reversed when cells were also transfected with the active form of RhoB (Fig. 4B). Moreover, because we have previously shown that RhoB controls HIF-1α proteasome–dependent degradation under hypoxia through GSK-3β (8), we then checked that the control of HIF-1α intracellular level by the αvβ3 or αvβ5 integrin-FAK-RhoB pathway was mediated by GSK3. This enzyme is inactivated by phosphorylation of its Ser9 residue mediated by Akt. To determine whether inhibiting the αvβ3 or αvβ5 integrin-FAK-RhoB pathway would modify the activation of GSK3β, we first analyzed by Western blotting the level of GSK3β phosphorylated on Ser9 in hypoxic control cells or those transfected with siRNA directed against the β3 or β5 integrins, FAK, or RhoB (Fig. 4C). Under hypoxia, the amount of the phosphorylated Ser9 inactive form of GSK3β was significantly decreased when cells were transfected with siRhoB3 (P < 0.01) and siFAK (P < 0.01) in the two cell lines and with siβ3 (P < 0.01) in U87 cells or with siβ5 (P < 0.01) in SF763 cells. These results show that RhoB is a downstream effector of the αvβ3 or αvβ5 integrin-FAK–dependent HIF-1α regulation under hypoxic conditions. Considering our previous results showing that RhoB controlled the proteasome-dependent degradation of HIF-1α under hypoxia (8), we can conclude that the αvβ3 or αvβ5 integrin-FAK pathways are upstream regulators of RhoB-dependent HIF1α proteasome–dependent degradation. In conclusion, we showed that the αvβ3 or αvβ5 integrin signaling pathway through FAK, RhoB, and GSK-3 is a key regulator of the modulation of HIF-1α proteasome–dependent degradation in hypoxic human glioma cell lines.

In vivo inhibition of FAK or β3 integrins leads to oxygenation in U87 xenografts. We postulated that inhibiting this pathway in tumor cells in vivo would lead to an oxygenation of the glioblastoma xenografts. To investigate this hypothesis, we generated U87 xenografts in nude mice and treated them with siRNA directed against the human β3 integrin or FAK during 5 days. These siRNA did not silence mouse β3 integrin or FAK expression (data not shown). This treatment led to a significant inhibition of β3 integrin and FAK expression (74% and 58% inhibition, respectively; P < 0.01; Fig. 5A). We then studied the effects of the β3 integrin or FAK inhibition on the in vivo hypoxia regulation by analyzing the EF-5 and HIF-1α staining in the tumors (Fig. 5B). As expected, untreated xenografts contained large zones of EF-5 staining, showing the presence of large areas of viable hypoxic cells. Silencing either β3 integrin or FAK led to a dramatic decrease of EF-5 staining in the xenografts (P < 0.01; Fig. 5C), showing an oxygenation effect induced by β3 or FAK inhibition. We then checked whether the β3 integrin or FAK inhibition led to a reduction of HIF-1α expression in the treated tumors. Figure 5B shows, as expected, that hypoxic untreated control tumors presented HIF-1α staining whereas β3 integrin or FAK inhibition respectively led to a reduction of this staining in both siβ3- and siFAK-treated xenografts (P < 0.05; Fig. 5D). These results clearly showed that inhibiting αvβ3 integrin or FAK in human tumor cells reduced the hypoxia in human glioblastoma xenografts, just as we had previously shown after the in vivo inhibition of RhoB (5, 7).

Figure 5.

Silencing the αvβ3 integrin/FAK pathway induces in vivo oxygenation in U87 xenografts. U87 tumor–bearing mice were treated as described in Materials and Methods with scramble siRNA or siRNA directed against β3 integrin or FAK. A, Western blot analysis with anti–β3 integrin or anti-FAK was done on total tumor extracts as described in Materials and Methods. Data represent at least three independent experiments. B, animals were injected with EF5 and then sacrificed, and EF5 and HIF-1α stainings were studied by immunohistochemical analysis. Binding of EF5 appears in red. Exposure duration for each frame is 40 ms for 4′,6-diamidino-2-phenylindole (DAPI) and 1 s for EF5 staining. Immunohistochemistry with anti–HIF-1α was done as described in Materials and Methods. Magnification, ×20. Data represent at least three independent experiments. C, EF-5 staining of viable hypoxic tumor cells in control tumors (nontreated and siscr-transfected tumors) and in siRNA-treated tumors was determined by quantifying the fluorescence intensity using NIS Elements AR software (Nikon). Data represent the means ± SD of at least three different experiments. ☆, siβ3.1- or siFAK1-treated U87 xenografts exhibited a lower amount of EF-5 staining compared with control xenografts (P < 0.001). D, HIF-1α staining was quantified by counting the number of positively stained nuclei among 100 cells. Data represent the means ± SD of at least three different experiments. ☆, siβ3.1- or siFAK1-treated U87 xenografts exhibited a lower amount of HIF-1α staining compared with control xenografts (P < 0.05).

Figure 5.

Silencing the αvβ3 integrin/FAK pathway induces in vivo oxygenation in U87 xenografts. U87 tumor–bearing mice were treated as described in Materials and Methods with scramble siRNA or siRNA directed against β3 integrin or FAK. A, Western blot analysis with anti–β3 integrin or anti-FAK was done on total tumor extracts as described in Materials and Methods. Data represent at least three independent experiments. B, animals were injected with EF5 and then sacrificed, and EF5 and HIF-1α stainings were studied by immunohistochemical analysis. Binding of EF5 appears in red. Exposure duration for each frame is 40 ms for 4′,6-diamidino-2-phenylindole (DAPI) and 1 s for EF5 staining. Immunohistochemistry with anti–HIF-1α was done as described in Materials and Methods. Magnification, ×20. Data represent at least three independent experiments. C, EF-5 staining of viable hypoxic tumor cells in control tumors (nontreated and siscr-transfected tumors) and in siRNA-treated tumors was determined by quantifying the fluorescence intensity using NIS Elements AR software (Nikon). Data represent the means ± SD of at least three different experiments. ☆, siβ3.1- or siFAK1-treated U87 xenografts exhibited a lower amount of EF-5 staining compared with control xenografts (P < 0.001). D, HIF-1α staining was quantified by counting the number of positively stained nuclei among 100 cells. Data represent the means ± SD of at least three different experiments. ☆, siβ3.1- or siFAK1-treated U87 xenografts exhibited a lower amount of HIF-1α staining compared with control xenografts (P < 0.05).

Close modal

We then investigated whether this oxygenation effect was associated with a regulation of tumor vasculature. As shown in Fig. 6A and B, inhibition of either FAK or β3 integrin led to a significant reduction of CD31 staining (P < 0.05), the vessels being thinner and more regular, in the same way that we had previously observed when inhibiting RhoB (5, 7), suggesting a regulation of angiogenesis by this pathway.

Figure 6.

In vivo silencing of the αvβ3 integrin/FAK pathway modifies tumor vasculature. U87 tumor–bearing mice were treated as described in Materials and Methods by scramble siRNA or siRNA directed against β3 integrin or FAK. A, immunohistochemistry with anti–CD-31 was done as described in Materials and Methods. Exposure duration for each frame is 40 ms for DAPI staining. Magnification, ×20. Data represent at least three independent experiments. B, the percentages of CD-31–stained viable tumor cells in control xenografts (nontreated and siscr-transfected xenografts) and in siRNA-treated tumors were determined by quantifying the percentage of the tumor vessel area on at least eight random areas using NIS Elements AR software (Nikon). Columns, mean of at least three different experiments; bars, SD. ☆, siβ3.1- or siFAK1-treated U87 xenografts exhibited a lower amount of CD-31 staining compared with control xenografts (P < 0.05).

Figure 6.

In vivo silencing of the αvβ3 integrin/FAK pathway modifies tumor vasculature. U87 tumor–bearing mice were treated as described in Materials and Methods by scramble siRNA or siRNA directed against β3 integrin or FAK. A, immunohistochemistry with anti–CD-31 was done as described in Materials and Methods. Exposure duration for each frame is 40 ms for DAPI staining. Magnification, ×20. Data represent at least three independent experiments. B, the percentages of CD-31–stained viable tumor cells in control xenografts (nontreated and siscr-transfected xenografts) and in siRNA-treated tumors were determined by quantifying the percentage of the tumor vessel area on at least eight random areas using NIS Elements AR software (Nikon). Columns, mean of at least three different experiments; bars, SD. ☆, siβ3.1- or siFAK1-treated U87 xenografts exhibited a lower amount of CD-31 staining compared with control xenografts (P < 0.05).

Close modal

In conclusion, our results clearly established that inhibiting the αvβ3 integrin-FAK-RhoB pathway in tumor cells regulated hypoxia and angiogenesis.

To date, no regulation of HIF-1, and thus hypoxia, by integrins has been reported in glioma. In fact, even if several works have shown that hypoxia induces adhesion molecules expression, including integrins, in cancer cells (1315), little is known about the reverse process, that is to say the control of HIF-1α regulation by integrin signaling. We show here that, in glioblastoma cells, the αvβ3 and αvβ5 integrins, when activated by hypoxia, regulate HIF-1α stabilization through FAK, RhoB, and GSK-3β in glioma cell lines. Moreover, we show that inhibition of this pathway in tumor cells leads to in vivo oxygenation and modification of tumor vasculature. This study clearly establishes not only that the αvβ3 and αvβ5 integrins are activated by hypoxia but also that they are key regulators of glioma response to hypoxic conditions by controlling HIF-1α degradation, thus suggesting the existence of a feedback loop of hypoxia maintenance mediated, at least in part, by integrins in these glioma cells.

To identify the downstream effectors of this biological pathway, we first investigated the role of the two well-known mediators of integrin pathways, ILK and FAK. Inhibiting ILK in glioma xenografts leads to a reduction of hypoxia due to a modification of tumor vasculature (22). Our present results establish that, in hypoxic human U87 glioblastoma cells, αvβ3 and αvβ5 integrins do not regulate HIF-1α and therefore the hypoxia pathways through ILK, but via FAK. A crucial role of the tyrosine kinase FAK in the promotion of glioblastoma cell proliferation in vitro and in vivo has largely been reported (for review, see ref. 23). However, except for a work reporting its implication in drug-induced HIF-1α expression in gastric cell carcinoma (24), the role of FAK in the regulation of hypoxia mechanisms via HIF-1α stabilization had been poorly described. We showed here that αvβ3 and αvβ5 integrins and FAK control HIF-1α stabilization through the small GTPase RhoB. We have previously shown that hypoxia activates RhoB, which in turn regulates HIF-1α proteasome degradation through GSK-3 (8). Potential links between FAK and different proteins from the Rho family have been widely described. For example, FAK has been involved in Rac1 translocation to focal adhesion, increasing the efficiency of mouse embryo fibroblast cell spreading (25). In keratinocytes, FAK is required for p190RhoGAP phosphorylation, suggesting that FAK phosphorylation of p190RhoGAP at focal adhesion might locally promote RhoGTP hydrolysis to suppress Rho-induced stress fiber formation and focal adhesion stabilization (26). However, to date, no regulation of RhoB by this kinase has been described. We showed here that silencing FAK in vitro led to the inhibition of hypoxia-induced RhoB activation, showing that FAK is a mediator of the hypoxia-induced RhoB activation controlled by integrins. In addition, we show hypoxia-induced FAK phosphorylation, which strongly suggests that FAK activation is necessary in this signaling transduction. The activation of Rho proteins, including RhoB, is mediated by specific guanine-nucleotide exchange factors (GEF), which catalyze the exchange of GDP for GTP. Because some of these GEF, such as Vav, can be phosphorylated on their tyrosine residues (27, 28), the tyrosine kinase activity of FAK may be important to RhoB activation, probably through RhoB GEF phosphorylation.

We then investigated whether inhibiting this pathway in tumor cells, using siRNA directed against human cells, might affect glioma hypoxia. We provided evidence that inhibiting β3 integrin or FAK in tumor cells led to an oxygenation of the U87 xenografts associated with a morphology normalization of the tumor vasculature. These results showed that the regulation of HIF-1α by this integrin-dependent pathway in glioma cells is crucial for the maintenance of hypoxia in the tumor. They are in accordance with our previous results that showed that inhibiting RhoB in U87 xenografts, either with a farnesyltransferase inhibitor (7) or with an inducible dominant negative expressed in tumor cells (5), led to an oxygenation of these tumors associated with a morphology normalization of the vessels. Considering the fact that these integrins also regulate glioma cell response to ionizing radiation (9), these results underline the interest of disrupting this pathway by using inhibitors of this integrin-dependent pathway not only to reduce hypoxia but also to reduce the resistance of glioma to radiotherapy. Our institute has developed a strategy whose aim is to inhibit a target controlling hypoxia, radioresistance, and angiogenesis. In fact, a phase I clinical trial combining a farnesyltransferase inhibitor targeting RhoB with radiotherapy has shown encouraging results in glioblastoma treatment (29, 30). Our present results broaden the spectrum of potential targets for specific inhibitors to improve the treatment of hypoxic, aggressive, and radioresistant tumors expressing these integrins, such as glioblastoma.

No potential conflicts of interest were disclosed.

Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).

Grant support: Ministère de la Recherche et de l'Enseignement Supérieur (N. Skuli, G. Favre, and E. Cohen-Jonathan Moyal), the Groupe de Recherche de l'Institut Claudius Regaud, and the Association pour la Recherche Contre le Cancer (E. Cohen-Jonathan Moyal and C. Toulas).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Dr. Cameron Koch (University of Pennsylvania, Philadelphia, PA) for giving us EF-5 and Dr. Simon Goodman (Merck, Germany) for kindly providing us EMD121974.

1
Cohen-Jonathan E, Muschel RJ, Gillies MW, et al. Farnesyltransferase inhibitors potentiate the antitumor effect of radiation on a human tumor xenograft expressing activated HRAS.
Radiat Res
2000
;
154
:
125
–32.
2
Cohen-Jonathan E, Evans SM, Koch CJ, et al. The farnesyltransferase inhibitor L744,832 reduces hypoxia in tumors expressing activated H-ras.
Cancer Res
2001
;
61
:
2289
–93.
3
Delmas C, Heliez C, Cohen-Jonathan E, et al. Farnesyltransferase inhibitor, R115777, reverses the resistance of human glioma cell lines to ionizing radiation.
Int J Cancer
2002
;
100
:
43
–8.
4
Ader I, Toulas C, Dalenc F, et al. RhoB controls the 24 kDa FGF-2-induced radioresistance in HeLa cells by preventing post-mitotic cell death.
Oncogene
2002
;
21
:
5998
–6006.
5
Ader I, Delmas C, Bonnet J, et al. Inhibition of Rho pathways induces radiosensitization and oxygenation in human glioblastoma xenografts.
Oncogene
2003
;
22
:
8861
–9.
6
Gampel A, Parker PJ, Mellor H. Regulation of epidermal growth factor receptor traffic by the small GTPase rhoB.
Curr Biol
1999
;
9
:
955
–8.
7
Delmas C, End D, Rochaix P, et al. The farnesyltransferase inhibitor R115777 reduces hypoxia and matrix metalloproteinase 2 expression in human glioma xenograft.
Clin Cancer Res
2003
;
9
:
6062
–8.
8
Skuli N, Monferran S, Delmas C, et al. Activation of RhoB by hypoxia controls hypoxia-inducible factor-1α stabilization through glycogen synthase kinase-3 in U87 glioblastoma cells.
Cancer Res
2006
;
66
:
482
–9.
9
Monferran S, Skuli N, Delmas C, et al. αvβ3 and αvβ5 integrins control glioma cell response to ionising radiation through ILK and RhoB.
Int J Cancer
2008
;
123
:
357
–64.
10
Hynes RO. Integrins: bidirectional, allosteric signaling machines.
Cell
2002
;
110
:
673
–87.
11
Hood JD, Cheresh DA. Role of integrins in cell invasion and migration.
Nat Rev Cancer
2002
;
2
:
91
–100.
12
Bello L, Francolini M, Marthyn P, et al. αvβ3 and αvβ5 integrin expression in glioma periphery.
Neurosurgery
2001
;
49
:
380
–9.
13
Yoon SO, Shin S, Mercurio AM. Ras stimulation of E2F activity and a consequent E2F regulation of integrin α6β4 promote the invasion of breast carcinoma cells.
Cancer Res
2006
;
66
:
6288
–95.
14
Baril P, Gangeswaran R, Mahon PC, et al. Periostin promotes invasiveness and resistance of pancreatic cancer cells to hypoxia-induced cell death: role of the β4 integrin and the PI3k pathway.
Oncogene
2007
;
26
:
2082
–94.
15
Cowden Dahl KD, Robertson SE, Weaver VM, Simon MC. Hypoxia-inducible factor regulates αvβ3 integrin cell surface expression.
Mol Biol Cell
2005
;
16
:
1901
–12.
16
Smith JW. Cilengitide Merck.
Curr Opin Investig Drugs
2003
;
4
:
741
–5.
17
Shiratsuchi H, Basson MD. Extracellular pressure stimulates macrophage phagocytosis by inhibiting a pathway involving FAK and ERK.
Am J Physiol Cell Physiol
2004
;
286
:
C1358
–66.
18
Tafolla E, Wang S, Wong B, Leong J, Kapila YL. JNK1 and JNK2 oppositely regulate p53 in signaling linked to apoptosis triggered by an altered fibronectin matrix: JNK links FAK and p53.
J Biol Chem
2005
;
280
:
19992
–9.
19
Prendergast GC, Khosravi-Far R, Solski PA, et al. Critical role of Rho in cell transformation by oncogenic Ras.
Oncogene
1995
;
10
:
2289
–96.
20
Ren XD, Schwartz MA. Determination of GTP loading on Rho.
Methods Enzymol
2000
;
325
:
264
–72.
21
Schaller MD, Hildebrand JD, Shannon JD, et al. Autophosphorylation of the focal adhesion kinase, pp125FAK, directs SH2-dependent binding of pp60src.
Mol Cell Biol
1994
;
14
:
1680
–8.
22
Edwards LA, Woo J, Huxham LA, et al. Suppression of VEGF secretion and changes in glioblastoma multiforme microenvironment by inhibition of integrin-linked kinase (ILK).
Mol Cancer Ther
2008
;
7
:
59
–70.
23
Natarajan M, Hecker TP, Gladson CL. FAK signaling in anaplastic astrocytoma and glioblastoma tumors.
Cancer J
2003
;
9
:
126
–33.
24
Liu L, Ning X, Sun L, et al. Involvement of MGr1-Ag/37LRP in the vincristine-induced HIF-1 expression in gastric cancer cells.
Mol Cell Biochem
2007
;
303
:
151
–60.
25
Chang F, Lemmon CA, Park D, Romer LH. FAK potentiates Rac1 activation and localization to matrix adhesion sites: a role for βPIX.
Mol Biol Cell
2007
;
18
:
253
–64.
26
Schober M, Raghavan S, Nikolova M, et al. Focal adhesion kinase modulates tension signaling to control actin and focal adhesion dynamics.
J Cell Biol
2007
;
176
:
667
–80.
27
Lopez-Lago M, Lee H, Cruz C, Movilla N, Bustelo XR. Tyrosine phosphorylation mediates both activation and downmodulation of the biological activity of Vav.
Mol Cell Biol
2000
;
20
:
1678
–91.
28
Bustelo XR. Regulatory and signaling properties of the Vav family.
Mol Cell Biol
2000
;
20
:
1461
–77.
29
Moyal EC, Laprie A, Delannes M, et al. Phase I trial of tipifarnib (R115777) concurrent with radiotherapy in patients with glioblastoma multiforme.
Int J Radiat Oncol Biol Phys
2007
;
68
:
1396
–401.
30
Laprie A, Catalaa I, Cassol E, et al. Proton magnetic resonance spectroscopic imaging in newly diagnosed glioblastoma: predictive value for the site of postradiotherapy relapse in a prospective longitudinal study.
Int J Radiat Oncol Biol Phys
2008
;
70
:
773
–81.

Supplementary data