PTEN and Phosphatidylinositol 3′-Kinase Inhibitors Up-Regulate p53 and Block Tumor-induced Angiogenesis

INTRODUCTION

The PTEN, tumor suppressor gene, which is located on the long arm of chromosome 10 (10q23), is mutated in 40–50% of high-grade gliomas as well as prostate, endometrial, breast, lung, and other tumor types (1, 2, 3) . PTEN is a 55-kDa dual specificity protein and lipid phosphatase that is composed of a NH₂-terminal catalytic domain identified as a segment with homology to the cytoskeletal protein tensin and containing the sequence HC(X)₅R, which is the signature motif of members of the protein tyrosine phosphatase family. It contains a COOH-terminal C2 domain with lipid-binding and membrane-targeting functions (4) . The COOH-terminal regulatory tail is involved in posttranslational control of PTEN activity through phosphorylation, complex formation, and potentially degradation (5) . Importantly, PTEN possesses both protein and lipid phosphatase activities but preferentially dephosphorylates PIP₃ 3 at its D3 position. It is one of three enzymes known to dephosphorylate PIP₃, suggesting that PTEN may function as a direct antagonist of PI3K and PIP₃-dependent signaling (6, 7, 8, 9) . Crystallographic data suggest that the active site cleft of PTEN is wider and deeper than is observed with other PTPases and dual specificity phosphatases (4) . The basic amino acid residues (K125, K128, and R130) in the P-loop active site cleft, which is defined by the sequence 123-HCKAGKGR, is 100% conserved throughout different species and likely is important in the coordination of the negatively charged 3′, 4′, and 5′ inositol phosphates present in the phosphoinositol ring. These amino acid residues may provide PTEN with its capacity to accept PIP₃ as a preferred substrate. Reconstitution of PTEN in tumor cells that carry a mutation in the PTEN gene within the P-loop, which ablates lipid PIP₃ phosphatase activity (G129E), has established that this phosphatase regulates the PI3K-dependent activation of AKT, a major player in cell survival (10 , 11) . Other data suggest that PTEN regulates the activity of the nonreceptor protein tyrosine kinase, focal adhesion kinase by the possible direct dephosphorylation of tyrosine residues (12 , 13) . Because PTEN controls TSP-1 expression (14) and TSP-1 expression is regulated by p53 (15 , 16) , we reasoned that the mechanism for PTEN suppression of angiogenesis could be through control of p53 transcription. Another report has demonstrated an effect of PTEN on p21^waf-1 (17) . Recent evidence from our group has established a mechanistic link between PTEN and p53 function through the control of MDM2 phosphorylation state, which modulates the nuclear localization of MDM2 and the ubiquitination of p53 (18 , 19) . MDM2 is a RING finger ubiquitin ligase known to negatively regulate p53 via the capacity to bind to and mediate its proteosomal degradation (20, 21, 22, 23) . We reasoned that the derangement of the PTEN-AKT-MDM2-p53 signaling axis may promote cancer progression but would also be an attractive target for therapy.

An article by Sabbatini et al. (24) is consistent with a connection between the PI3K cascade and the regulation of p53 signaling. Activation of PI3K/AKT signaling pathways delays cellular apoptosis dependent upon p53. We recently reported that PTEN controls tumor-induced angiogenesis (14) . Others have reported effects of PTEN on cell growth and apoptosis (11) . These combined observations suggest a potential mechanism for the coordination of signals coming from growth factor receptors through PI3K cascades, which would jointly regulate apoptosis, proliferation, and recruitment of a new blood supply (neovascularization/angiogenesis). This process would be highly regulated in normal tissues and is likely deregulated during tumorigenesis, malignant transformation, and tumor progression. We hypothesized that PTEN may play a role in coordinating these signaling events within the cell. Loss of PTEN would lead to deregulation and tumor progression. If this hypothesis were correct, the PI3K inhibitor, LY294002, could reestablish this feedback system in favor of regulated growth and suppress angiogenesis in malignant tumors. To test the hypothesis that PTEN is connected to p53 transcription, we performed experiments in a U87MG glioma cell line, which is wild type for p53 and deficient in PTEN (25 , 26) . We conditionally expressed in these U87MG cells, wild-type PTEN, or catalytically defective mutants of PTEN to determine whether PTEN regulates p53-transactivating activity (using the mdm2-luciferase P2 promoter) in the tumor cells. We then performed experiments with the PI3K inhibitor, LY294002, using the parental U87MG cells to determine whether the control of PIP₃ metabolism would prevent tumor growth and block angiogenesis in vivo. Finally, we examined the effect of PTEN and LY294002 on p53 function in human brain endothelial cells (HBEC). The results suggest that PI3K inhibitors may potentially target both tumor and endothelial cells to control tumor-induced angiogenesis and progression.

MATERIALS AND METHODS

Constructs and Reagents.

Wild-type or mutant PTEN (G129E, lipid phosphatase dead) or (G129R, loss of catalytic activity toward lipid and protein substrates) cDNAs were subcloned into a retroviral expression vector containing a muristirone-inducible promoter or into the pBABEpuro vector for expression in U373MG cells (14) . U87MG clones were engineered to contain VgEcR and retinoid X receptor nuclear receptors/transcription factors that confer muristirone-responsive expression of PTEN or PTEN mutants when cloned into a retroviral vector, which contains a muristirone responsive promoter (27 , 28) . Stable clones of U87MG cells, which would allow for muristirone induced expression of PTEN, were established by subcloning the PTEN cDNA into the muristirone-inducible retroviral vector, which contains a puromycin selectable marker as described previously (2 μg/ml; Refs. 10 , 29 ). Cells selected in puromycin express PTEN in a graded manner only under conditions of muristirone induction (28) . The human brain endothelial cell line was characterized for endothelial markers as described previously (30 , 31) . Constructs used in endothelial cell experiments contained wild-type PTEN or mutant PTEN (C124S) in pRK5 vector. Antibodies specific for PTEN (10) , AKT, and phospho-S473-AKT were from New England Biolabs. Antibodies against p21^waf-1 (05-345) and IGFBP3 (06-108) were from Upstate Biologicals. LY294002 was purchased from Boerringer Mannheim.

Tumor Implantation.

U87MG cells were cultured in fresh medium for 24 h and harvested, adjusting the cell concentration to 1 × 10⁶ in 10 μl of RPMI medium. Mice, under general anesthesia, were placed into the stereotactic device (model 963; Kopf, Tugunga, CA). Stereotactically controlled drill assembly was used to provide a hole 0.3-mm deep and of 0.8-mm diameter in the cranium at a position 0.5-mm anterior and 1.2-mm lateral to the bregmal anatomical landmark. Tumor cells (1 × 10⁶) were introduced slowly through a 10-μl Hamilton syringe at a depth of 2.5 mm at a rate of 2 μl/min. We then slowly removed the needle at a rate of 0.5 mm/min. After needle removal, the hole was sealed with bone wax, and the incision was closed with a wound clip. In the same mice, 5 × 10⁶ tumor cells were implanted s.c. in the right flank to perform biochemical and immunohistochemical analysis of tumor tissue and to monitor tumor growth using calipers.

Treatment of Mice with LY294002.

LY294002 (25 mg/kg/dose) was administered in a volume of 50 μl of 100% DMSO by i.p. injection in a mouse twice/day for 2 weeks beginning 2 days after tumor implantation or when tumors had grown to size of 80–90 mm³. Minimal untoward effects were noted in mice treated with LY294002 or DMSO. Control mice were injected with same volume of 100% DMSO. Measurement of tumor volume was performed in three coordinates using calipers.

Biochemical Analysis.

Immunoblots were performed on cell lysates obtained from U87 cells grown in tissue culture. A Bradford assay was performed to determine the protein concentration of each lysate. Equivalent amounts of protein were resolved by SDS-PAGE and transferred to nitrocellulose. Membranes were probed with antisera specific for PTEN, AKT, or phospho-S473-AKT. We used a well-characterized mdm2 P2 promoter linked to firefly luciferase (mdm2-luc inserted into the pGL2 vector) that contains specific p53 DNA binding elements (32) to study p53-specific transcription in U87 cells under muristirone-induced PTEN expression conditions. U87MG cells containing the muristirone-inducible PTEN or mutant PTEN were transfected with pCMVβgal and either the mdm2 P2 promoter (RE) or a mutant lacking the p53 binding element (DRE) luciferase reporter plasmid using Lipofectamine. After transfection, the expression of PTEN was induced with vehicle or 0.5 μm muristirone for 48 h. For HBEC experiments, PTEN and mdm2 luciferase promoter constructs were contransfected using Lipofectamine. The relative luciferase activity was derived from luciferase activity normalized to β-galactosidase activity 48 h after transfection. The Tropix-galacto-light kit and the Promega luciferase assay system were used to quantitate β-galactosidase and luciferase activities, respectively.

Immunohistochemistry and Histopathology.

MVD was determined for each s.c. tumor by CD31 staining performed on coronal cryostat sections (7 μm), fixed in acetone, blocked in 1% goat serum, and stained with anti-CD31 antibody (no. 01951D; PharMingen). Antibody staining was visualized with peroxidase-conjugated antirat and counterstained with hematoxylin. A negative control was performed on each tumor tissue stained with mouse IgG. Two sections from each tumor were scanned under low-power magnification (×40) to identify areas of highest CD31-positive vessel density (33) , followed by digitization of five fields from this area. The digitized images representing one ×100 field were counted for number of CD31-positive vascular elements. Data were collected by two independent observers without knowledge of which tumors were viewed. The average number of microvessels/digitized ×100 field was determined for five tumors/experimental group and analyzed by Student’s t test. For analysis of brain tumor incidence, each brain was removed and bisected in the coronal plane at the point of the stereotactic injection site and fixed in 10% buffered formalin. Serial coronal tissue sections on both sides of implantation site were cut at 70-μm intervals, stained with H&E, and examined under microscope for presence of tumor. Tissue sections were extended through the entire frontal and parietal lobe of brain to detect and assess the size of the tumor within the brain.

Statistical Analysis.

The Students t test was used to evaluate observed differences between PTEN reconstituted and/or LY294002-treated mice and controls.

RESULTS

Our previous work demonstrated that muristirone-inducible expression of PTEN in U87 cells resulted in augmented TSP-1 expression (14) , a negative regulator of angiogenesis (34 , 35) . The use of a specific point mutation in PTEN (PTEN G129E), in which the lipid phosphatase activity is impaired but protein phosphatase activity remains essentially intact allowed us to determine the PTENs capacity to regulate angiogenesis, resides in its capacity to control PIP₃ (14) . This result suggested a novel mechanism by which PTEN may regulate angiogenesis through the induction of TSP-1 (14) . One transcription factor that up-regulates TSP-1 is the tumor suppressor protein p53 (16) . Therefore, we sought to determine whether there was a link between PTEN, TSP-1, and p53. First, we confirmed that wild-type PTEN suppressed the activation of phospho-AKT without affecting total AKT in U87MG cells (Fig. 1A) ⇓ . The induced expression of the catalytically dead mutant of PTEN, G129R, or PIP₃ phosphatase dead G129E mutants had no effect on the phosphorylation of AKT at position 473 (Fig. 1A) ⇓ . The induction of wild-type PTEN and not mutant PTEN (G129R, G129E) promoted p53-dependent transcriptional activity of the mdm2-luciferase promoter in U87MG cells (7.5-fold induction; Fig. 1B ⇓ ; Ref. 32 ). From a comparison of effect of G129E mutant and G129R mutation in PTEN to wild-type PTEN on p53-transcriptional activity (Fig. 1B) ⇓ , we conclude that it is PTEN capacity to control lipid PIP₃ levels that determines its capacity to regulate p53 function. Hence, in this system, the loss of protein tyrosine phosphatase activity is not required for cells to deregulate p53 transcriptional function in vivo. The mdm2 promoter is a well-characterized p53 responsive promoter construct established to study p53 transactivation after specific p53 binding. Muristirone at 0.5 μm induced expression of wild-type PTEN or G129R protein was equivalent, whereas G129E induction was slightly greater (inset, Fig. 1B ⇓ ). Experiments were performed with the mdm2-luciferase construct deleted in critical p53 binding sites (DRE; Ref. 32 ) to confirm the specificity for PTEN induction of p53-specific transcription (Fig. 1B) ⇓ . The DRE mutant of the mdm2 P2 promoter is not induced by wild-type PTEN expression in U87MG cells, confirming a requirement for the p53 binding component of this reporter for PTEN induction. Our data provides direct evidence that PTEN regulates the capacity of p53 to transactivate a known p53 promoter element (mdm2-luciferase; Ref. 32 ). Other experiments then demonstrated that PTEN reconstitution of U87MG cells induced p21^waf-1 (Fig. 1C) ⇓ and IGFBP3 protein expression in U87MG cells (Fig. 1D) ⇓ . The p21 protein is an inhibitor of the cyclin-dependent kinase, cdk4, involved in G₁-S-phase transition (36) , and IGFBP3 protein inhibits IGF-I-induced activation of PI3K pathways via extracellular blockade. These results confirm an effect of PTEN and LY294002 on specific p53 activities in tumor cells. Our results may explain the effect of PTEN reconstitution on angiogenesis through the induction of p53-transcriptional activity, a factor known to regulate tumor-induced angiogenesis (15 , 16) .

As PTEN negatively regulates PI3K functions, experiments were conducted to determine whether PI3K exerts control over angiogenesis and/or the growth of glial tumors in vivo. An ectopic skin and orthotopic brain tumor model was developed in which tumor cells (U87MG) were stereotactically injected into the frontal cerebral cortex of nude mice. The same tumor cell line was also introduced by s.c. injection. In the orthotopic brain tumor model, 100% of mice implanted intracranially with the parental U87 cells displayed a highly angiogenic pattern of brain tumor growth. This resulted in 100% mortality within 25–27 days (14) . To assess the effect of LY294002 on angiogenesis, we treated mice with LY294002 (50 mg/kg/day × 2 weeks) or DMSO as negative control. Tumor volumes were recorded twice/week (Fig. 4, A and B) ⇓ . On day 7, we stained cryostat sections from s.c. tumors for CD31 (PECAM; Fig. 2, A and B ⇓ ). CD31 is an endothelial marker used to measure the MVD of these tumors. MVD was assessed from multiple digitized images of CD31-stained tumor tissue at ×100 magnification (five fields were evaluated/tumor) and counted blindly for the number of CD31-positive microvessels/unit surface area as described previously (33) . Quantitation of MVD in tumors treated with DMSO versus LY294002 is shown in Fig. 2C ⇓ . Compared with controls (Fig. 2, A–C) ⇓ , LY294002 markedly suppressed the tumor-induced angiogenic response in this model (MVD is 21.2 ± 6.5 in LY294002-treated tumors versus 49.6 ± 7.2 in the controls). It is likely that the effects of LY294002 are complex and that the size of tumor mass, hypoxia, and other factors may contribute at later time points to the induction of angiogenesis. Therefore, microvessel determinations were performed on day 7 after implantation (Fig. 2C) ⇓ to compare the angiogenic activity of tumors of similar size (Fig. 2, A–C) ⇓ . Our data demonstrate that LY294002 dramatically suppressed the angiogenic response of U87MG cells in vivo before the point of divergence in tumor size, suggesting a direct effect of LY294002 on tumor-induced angiogenesis similar to what was observed with PTEN reconstitution (14) .

To begin to address the possible effect of PTEN or LY294002 on the stromal compartment of the tumor, we examined the effect of PTEN expression and LY294002 treatment on human brain-derived endothelial cell p53-functional responses. Both PTEN transfection and LY294002 treatment induced transactivation of the MDM2 luciferase reporter, suggesting that the inhibition of PI3K activity in endothelial cells induces p53-transcriptional activity (Fig. 3, A and C) ⇓ . A catalytically dead mutant of PTEN (C124S) did not induce p53-dependent transcriptional activity. The p53 binding component of the mdm2-luciferase promoter was required for induction by PTEN and LY294002, confirming specificity for this known p53 binding element for transactivation. The effect of LY294002 was significantly greater than the transient transfection and overexpression p53 (Fig. 3C) ⇓ . Western blot analysis confirmed the overexpression of PTEN, PTEN mutant, and p53 proteins in HBECs (Fig. 3, B and D) ⇓ . To verify these results, we examined the p53 target gene, p21^waf-1, a cyclin-dependent kinase inhibitor (36) , in HBEC cells after exposure to LY294002 (Fig. 3E) ⇓ . The results demonstrate that LY294002 induced p21^waf-1 expression and increased total p53 levels in these cells (Fig. 3E) ⇓ . These data suggest that both PTEN and LY294002 have the capacity to activate p53 function in both tumor and stromal compartment, suggesting a dual mechanism for the antitumor effects of PTEN and LY294002 in vivo.

Lastly, we examined the effect of LY294002 on tumor growth and the formation of brain tumors in a nude mouse xenograft model. Our results show that s.c. tumor growth is suppressed by the systemic administration of LY294002 starting shortly after the time of implantation of tumor cells (Fig. 4) ⇓ . LY294002-treated tumors were 15 and 10% of control on days 22 and 26, respectively, after initiation of therapy. More importantly, the treatment of established xenografts with LY294002 induced a marked tumor regression in nude mice (Fig. 5) ⇓ . Tumor size was 99 mm³ in LY294002-treated mice as compared with 276 in control on day 28. This represents a 65% reduction in size of tumor by day 28 after initiation of therapy. The LY294002 treatment regimen was well tolerated. Initially, animals became somnolent for 5–10 min after i.p. injection of LY294002, which resolved over ensuing minutes without incident. This sedation effect did not persist beyond the first few injections. Our data provide the first direct evidence that the LY294002 has significant antiglioma activity in vivo. We also observed that LY294002 treatment, which began at the time of tumor implantation, markedly suppressed the intracranial growth of U87MG cells in nude mouse model. In control mice treated with DMSO, 11 of 12 (91.7%) had grossly visible and/or histologically confirmed brain tumors by H&E analysis of serial brain sections by day 25 after implantation, whereas 1 of 12 (8%) mice treated with LY294002 had grossly detectable and 4 of 12 (33.3%) had microscopic tumors only detected by extensive histological evaluation when examined on day 42 (Table 1 ⇓ ; n = 12, P < 0.01 comparing LY294002 treated to DMSO controls by gross examination; P < 0.05 for histological comparison of tumor incidence). The LY294002-treated mice remained neurologically normal for duration of study. From these data, we conclude that LY294002 prevented development of glioma growth in our orthotopic model. It is predicted that the effects of LY294002 would closely mimic the effects of PTEN reconstitution in PTEN-deficient glioma cells. To address this point, experiments performed using a PTEN-deficient glioma cell line, U373MG, reconstituted with wild-type PTEN, confirmed that both PTEN and LY294002 both suppress AKT activation. PTEN reconstitution dramatically induced p21^waf1 expression and suppressed tumor growth in vivo (Fig. 6, A and B) ⇓ . Unlike the U87MG cell line, the U373MG cells are p53 mutated (26) , hence, LY294002 does not induce p21^waf1 expression (Fig. 6B) ⇓ . In other experiments, we demonstrated that LY294002 treatment inhibits U373MG xenograft growth in nude mice (data not shown). These data provide additional evidence for a general effect of PTEN/PI3K signaling on glioma tumorigenesis and suggest that inhibitory effect of LY294002 may not require an intact p53-signaling axis.

DISCUSSION

Our results indicate that reconstitution of wild-type PTEN in U87 cells suppressed the malignant potential of these cells in an orthotopic animal model (14) . There was 90% survival at 40 days in animals implanted with the wild-type PTEN-reconstituted U87 cells compared with 100% mortality of mice implanted with the parental PTEN null cells at 27 days (14) . This observation is additionally supported by our more recent results obtained in another PTEN-deficient glioma cell line, U373MG, reconstituted with PTEN (Fig. 6) ⇓ . Herein, we also show that the PTEN tumor suppressor controls p53 transcription in U87 glioblastoma cells and HBECs. Importantly, the comparison of G129E and G129R mutants of PTEN provide evidence that PTEN′s control of p53 is exerted by the control of lipid second messenger PIP₃ and not control of protein phosphorylation. Holland et al. (37) reported that the introduction of activated AKT and Ras into glial cells of the mouse brain results in the development of glioblastoma multiforme. These results are consistent with our observations and suggest a potential therapeutic benefit for inhibitors of PI3K and its downstream targets such as AKT in treatment of malignant gliomas. LY294002 has shown efficacy against ovarian carcinoma where the tumor was grown as an ascitic tumor, and LY294002 was injected into a peritoneal cavity (38) . Thus, LY294002 may prove to be of general efficacy as a therapeutic agent in the treatment of malignant disease.

Others have suggested that LY294002 treatment and/or PTEN reconstitution may increase tumor responsiveness to DNA damage from chemotherapy and radiation and that PI3K/AKT pathways are linked to control of radiation sensitivity (18 , 39 , 40) and expression of cyclin-dependent kinase inhibitors p21^Waf1 and p27^Kip1 (41) . These combined observations suggest that the marked efficacy of LY294002 in vivo against malignant gliomas may relate to its capacity to control cell proliferation and angiogenesis in vivo. Our brain tumor model will be used to assess the efficacy of PI3K and AKT inhibitors for treatment of malignant brain tumors and the potential mechanisms involved in tumor suppression. Alternative targets for LY294002 and PTEN action would be known downstream targets for AKT, which would include the HIF-1α-vascular endothelial growth factor, mTOR-p70RS6K, or AKT-glycogen synthase kinase 3 signaling pathways in the potential control of angiogenesis (42 , 43) . Non-AKT targets would include signaling pathways impacted by PIP₃ separate from AKT to include PDK-1, certain nonreceptor protein tyrosine kinases, protein kinase C-ζ, and Rac GTPase activation (44) and many others (11) .

Our results provide evidence that LY294002 controls tumor-induced angiogenesis by a mechanism that could involve the regulation of p53 as measured by the transactivation of mdm2 promoter in both tumor and endothelial cells. Our data suggest that PI3K inhibitors may have efficacy in the treatment of pathologic conditions associated with augmented angiogenesis and demonstrate that PTEN and LY294002 exert control over p53 transcription. These data are consistent with our recent observation that the phosphorylation MDM2 by AKT on serine 166 and 186 within a conserved sequence motif (RxRxxS/T) regulates the movement of MDM2 into the nucleus and the induced degradation of p53 (18) . These data suggest that the PTEN and p53 tumor suppressor genes are localized on the same signaling axis for the coordinated control of growth and angiogenesis in tumor cells (Fig. 7) ⇓ . Hung et al. (45) have recently reported similar finding and have shown that AKT phosphorylates p21^waf-1 within the AKT phosphorylation motif (RxRxxS/T) resulting in cytoplasmic localization of this protein. This provides additional insight into how AKT may coordinately regulate cell cycle activity and p21 function. Moreover AKT is known to phosphorylate a significant number of other substrates, including glycogen synthase kinase-3β, BAD, caspase 9, FKHRL1, and IκB kinase α, which could contribute to effect of PTEN on angiogenesis (46, 47, 48, 49, 50) . The recent observation of Stambolic et al. (51) supports a model in which the PTEN promoter is regulated by p53, suggesting a positive feedback loop involving concerted and coordinated regulation of PTEN and p53 for the preservation of a nonproliferative, apoptosis-sensitive, antiangiogenic state. Although our results do not prove that p53 induction is required for PTENs suppression of angiogenesis, they are consistent with the possibility that PTEN and p53 are linked in a common pathway and therefore provides a logical mechanism by which PTEN may regulate the expression of TSP-1 through the regulation of p53 transcription. Others have reported that hypoxia is associated with reduced levels of TSP-1 expression in glioma cells and that this effect on TSP-1 was unrelated to p53 activity, suggesting a greater degree of complexity to the regulation of TSP-1 in glioma cells (52) . Additional experiments in PTEN reconstituted cells that are positive or negative for p53 under different conditions of growth factor stimulation ± hypoxia will be required to resolve this important point.

In other studies, it has been shown that PTEN/AKT pathway may control the expression of vascular endothelial growth factor, a potent angiogenic mediator, and other reports have shown a role for PTEN and p53 in the control of HIF-1α, suggesting a mechanism for regulation of hypoxia in tumor progression (43 , 53, 54, 55) . On the basis of our data and the literature, we propose a role for PTEN as an angioproliferative control point to provide rheostatic control over cell surface receptor-induced PIP₃ signals to coordinate cell growth and angiogenesis in vivo through the integration of upstream PIP₃ metabolism with nuclear p53 functions in the stromal and tumor compartments. Moreover, other components of PI3K and PTEN regulation will likely prove important in overall regulation of tumor angiogenesis. Importantly, a loss of this pathway through mutation in PTEN or upstream activation of PIP₃ pathways beyond the control set by endogenous PTEN leads to tumor progression. Our data support the contention that PI3K inhibitors have the capacity to reset this positive p53-PTEN feedback loop, resulting in the activation of endothelial p53 activity and the suppression of angiogenesis in vivo (Fig. 7) ⇓ . Finally, we suggest that the capacity to reset this feedback loop in vivo may have significant implications for the treatment of malignant disease. The results provide preclinical data that supports the additional development of PI3K and/or AKT inhibitors in the treatment of malignant gliomas.