Mitochondrial H2O2 Regulates the Angiogenic Phenotype via PTEN Oxidation*

Recent studies have demonstrated that the tumor suppressor PTEN (phosphatase and tensin homolog deleted from chromosome 10), the antagonist of the phosphosphoinositol-3-kinase (PI3K) signaling cascade, is susceptible to H2O2-dependent oxidative inactivation. This study describes the use of redox-engineered cell lines to identify PTEN as sensitive to oxidative inactivation by mitochondrial H2O2. Increases in the steady state production of mitochondrial derived H2O2, as a result of manganese superoxide dismutase (Sod2) overexpression, led to PTEN oxidation that was reversed by the coexpression of the H2O2-detoxifying enzyme catalase. The accumulation of an oxidized inactive fraction of PTEN favored the formation of phosphatidylinositol 3,4,5-triphosphate at the plasma membrane, resulting in increased activation of Akt and modulation of its downstream targets. PTEN oxidation in response to mitochondrial H2O2 enhanced PI3K signaling, leading to increased expression of the key regulator of angiogenesis, vascular endothelial growth factor. Overexpression of PTEN prevented the H2O2-dependent increase in vascular endothelial growth factor promoter activity and immunoreactive protein, whereas a mutant PTEN (G129R), lacking phosphatase activity, did not. Furthermore, mitochondrial generation of H2O2 by Sod2 promoted endothelial cell sprouting in a three-dimensional in vitro angiogenesis assay that was attenuated by catalase coexpression or the PI3K inhibitor LY2949002. Moreover, Sod2 overexpression resulted in increased in vivo blood vessel formation that was H2O2-dependent as assessed by the chicken chorioallantoic membrane assay. Our findings provide the first evidence for the involvement of mitochondrial H2O2 in regulating PTEN function and the angiogenic switch, indicating that Sod2 can serve as an alternative physiological source of the potent signaling molecule, H2O2.

Reactive oxygen species (ROS) 1 have long been established to play an important role in many disease pathologies and have also emerged as efficient signaling molecules. The principal mediator of ROS-dependent signaling is the two electron reduction product of oxygen (O 2 ), hydrogen peroxide (H 2 O 2 ). H 2 O 2 is generated in response to receptor stimulation and is an efficient signal transducing molecule by its ability to reversibly oxidize active site cysteines (1,2). Many protein tyrosine phosphatases are particularly susceptible to H 2 O 2 -dependent inactivation because of the lowered pK a of the active site cysteine (3)(4)(5). The essential cysteine residue in the signature active site motif Cys-(X) 5 -Arg exists as a thiolate anion (Cys-S Ϫ ), which at neutral pH is susceptible to nucleophilic attack by H 2 O 2 . Oxidation of the active site cysteine generates a sulfenic derivative (Cys-SOH), leading to enzyme inactivation that can be reversed by cellular thiols (6).
Oxidative inactivation of PTEN likely occurs close to the site of H 2 O 2 production in response to receptor activation.
However, the exact source of H 2 O 2 in response to receptor engagement is an area of controversy and may involve activation of the noninflammatory NADPH oxidase family members and subsequent spontaneous or enzymatic dismutation of superoxide (O 2 . ) to H 2 O 2 (16). The mitochondria are also a source for the generation of intracellular H 2 O 2 under physiologic conditions (17) and in response to receptor stimulation (18,19). However, whether mitochondrial derived H 2 O 2 contributes to the oxidative inactivation of PTEN or other phosphatases has not been established.
In this study, we report that alterations in the steady state production of mitochondrial H 2 O 2 by antioxidant enzyme overexpression modulate the redox state of PTEN. This mechanism is responsible for modulating PI3K/Akt signaling, VEGF production, and angiogenesis. These findings define a novel role for mitochondrial H 2 O 2 and Sod2 in regulating the angiogenic switch via PTEN phosphatase inactivation.
Measurement of Oxidation Rate in Intact Cells-Measurement of ROS levels in redox-engineered cell lines was performed using the redox-sensitive dye Redox Sensor Red TM CC-1 (Molecular Probes). Cells were harvested with phosphate-buffered saline (PBS) solution containing 1 mM EDTA. Cells were washed with PBS and resuspended in Hanks' balanced salt solution containing 6.7 g/liter glucose to a final density of 5 ϫ 10 5 cells/ml. Redox Red CC-1 was added to a final concentration of 10 M at room temperature. Samples were analyzed in a flow cytometer (BD Biosciences) at a wavelength of ϳ540/600 nm.
Identification of Reduced and Oxidized PTEN by Immunoblot Analysis-Cells were harvested in PBS/EDTA, washed once with PBS, and resuspended in 0.2 ml of 100 mM Tris-HCl (pH 6.8) containing 2% SDS and 40 mM N-ethylmaleimide; 30 g of protein per cell line was subjected to SDS-PAGE under nonreducing conditions as described previously (7). The separated proteins were then transferred to polyvinylidene fluoride membrane and immunoblotted with a rabbit anti-PTEN 1 o Ab (1:1000). Immune complexes were detected by a horseradish peroxidase-conjugated 2 o Ab (1:10,000) (Amersham Biosciences) and enhanced chemiluminescence reagents (Pierce).
Sod Zymography-Lysates from cell lines were harvested in 0.005 M potassium phosphate buffer at pH 7.8 containing 0.1 mM EDTA. Lysate supernatants were analyzed by electrophoresis in a discontinuous polyacrylamide gel, consisting of a 5% stacking gel (pH 6.8) and 10% running gel. The gel was stained with a solution containing 2.5 mM nitro blue tetrazolium, 0.008 mM riboflavin, 30 mM TEMED, and 0.005 M potassium phosphate buffer (pH 7.8) containing 0.1 mM EDTA, was incubated for 15 min in the dark at room temperature, and was washed twice in deionized water. The gels were then exposed to fluorescent light until clear zones of SOD activity were evident.
5Ј-Fluoresceinated Iodoacetamide (5-IAF) Labeling-Cells were lysed in 0.05 mM KP i buffer at pH 7.0 containing 10 mM iodoacetamide (Sigma). Samples were rocked at room temperature in the dark for 10 min, and free cysteines were blocked at this step. Lysates were then immunoprecipitated for PTEN (Sigma) by using protein G-Sepharose beads (Pierce). This step also removed any excess iodoacetamide. After immunoprecipitation, the bead-antibody-protein complex was resuspended in 200 l of 0.05 mM KP i buffer at pH 7.0 containing 10 mM dithiothreitol (Roche Applied Science). This denatured the proteins, breaking all disulfides, thereby freeing previously oxidized cysteines. The proteins were then acetone-precipitated in order to remove the dithiothreitol. Samples were then resuspended in buffer containing 5 mM 5-IAF and were then vortexed and kept at 4°C for 30 min. The 5-IAF (Molecular Probes) will now bind previously oxidized cysteines. Lysates were then run on an SDS-polyacrylamide gel and subjected to Western blot using anti-fluorescein isothiocyanate antibody (Molecular Probes).
Transient Transfections of hVEGF Promoter Construct and Dominant Negative Constructs-Cell lines were transfected with the various hVEGF (obtained from Dr Kenneth Anderson at Harvard Medical School and described by Tai et al. (21)). The VEGF construct and pCMV.SPORT␤-gal were cotransfected with Lipofectamine Plus reagent according to the manufacturer's instructions (Invitrogen). The cells were lysed 18 h post-transfection, and the luciferase reporter activity was determined using the Promega assay system. All of the results were reported after normalization for transfection efficiency by measuring ␤-galactosidase activity.
In Vitro Angiogenesis Assay-The in vitro angiogenesis assay was described previously by Nakatsu et al. (22). In brief, bovine lung microvessels (BLMV) were complexed with Cytodex 3 microcarrier (Amersham Biosciences) in 1 ml of EGM-2 (Clonetics) at a concentration of 400 BLMVs per bead. The bead and cell solutions were shaken every 20 min for 4 h at 37°C and 5% CO 2 to allow the cells to bind to the beads. Following the incubation, the cell-bead complex was grown overnight in a T-25-cm 2 tissue culture flask (Corning Glass) in 5 ml of EGM-2 at 37°C and 5% CO 2. After incubating, the cells were washed three times in EGM-2 and resuspended at a concentration of 200 cell-coated beads/ml in EGM-2 containing 2.5 mg/ml of fibrinogen (Sigma) with 0.15 units of aprotinin (ICN). 500 l of the above solution was added to 0.625 units of thrombin (Sigma) in a 24-well tissue culture plate. The solution was allowed to clot for 5 min at room temperature and then at 37°C and 5% CO 2 for 20 min. One milliliter of EGM-2 containing 0.15 units aprotinin was added per well and allowed to equilibrate with the clot for 30 min at 37°C and 5% CO 2 . EGM-2 was removed and replaced with 1 ml EGM-2 containing 0.15 units of aprotinin and 20,000 cells of the indicated redox-engineered cell lines (CMV, Sod2, Sod2mCAT, eGFP, and Sod2GFP). For a positive control, 2.5 ng/ml of VEGF was added to the CMV control cell line, and for a negative control, 30 M of the PI3K inhibitor LY294002 was added (Calbiochem). To test if the effects were VEGF-specific, 0.16 g/ml of a monoclonal mouse antihuman IgG 2B VEGF neutralizing antibody (R & D Systems) or IgG 2B isotype control antibody (Pharmingen) was added to the Sod2-overexpressing cell lines, and the effect on sprout formation was analyzed. Clots were grown for 24 h, and the number of sprouts/bead were analyzed. Images of beads were captured on a Nikon Diaphot microscope with a 4ϫ objective. The number of sprouts per bead were counted, and all experiments were repeated 10 times with similar results.
Chicken Embryo Chorioallantoic Assay-The CAM assay was performed as described previously by Brooks et al. (23) with minor variations. Fertilized chicken eggs were incubated with constant rocking at 37°C under 60% humidity. On the 9th day, the CAM was pulled away from the shell by first drilling a small shallow hole at the end of the egg that contains the natural air sac; an additional hole was made on the broad side of the egg directly over the embryonic blood vessels, as determined by candling. Gentle suction was applied to the 1st hole, thereby displacing the air sac to the side of the egg shell where the 2nd hole was made. This allowed the CAM to be pulled away from the shell on the side of the egg. A window was opened above the air pocket in the shell and sealed with a piece of sterile tape. A sterile rubber O-ring was placed on the CAM, and 2 ϫ 10 5 cells in a total volume of 50 l of PBS were placed in the center of the ring. No cells were added to the control. 24 h later the allantoic vessels around the CAM were photographed using a digital camera, five fields per egg. Vessel density was then quantified by placing a grid over the photo in Adobe Photoshop 7.0 and counting the number of vessels per field.
Plasmid Constructs-The pEGFP-N1 vector was obtained from Clontech. Human Sod2 was PCR-amplified with 5Ј-AvaI and 3Ј-KpnI ends. AvaI was added 4 bp upstream of the Sod2 initiation codon and KpnI was added prior to the Sod2 termination codon and directionally inserted in-frame into the pEGFP-N1. The constitutively active myr-Akt and kinase dead (K170M) Akt DNA was kindly provided by Dr. Andrew Aplin, Albany Medical College. The wild type PTEN and phosphatase mutant (G129R) PTEN DNA were obtained from Dr. William Sellers, Harvard Medical School, and were described previously (24). The Akt-PH-YFP and PLC␦-PH-CFP constructs were obtained from Tobias Meyer, Stanford University. The mito-GFP construct was obtained from Yisang Yoon, University of Rochester.
Confocal Microscopy-Cell cultures were imaged on an inverted LSM 510 Meta laser-scanning confocal microscope (Zeiss, Thornwood, NY) by using a Plan-Apo 63 ϫ 1.4 NA oil immersion objective. Pinhole diameter was ϳ1 airy unit. (a) For cells labeled with the single fluorescence reporter YFP, the 514 laser line from an argon laser was used to excite YFP, and images were collected using an HFT 458/514 dichroic mirror and a bandpass filter of 535-590 nm at an 8-bit intensity resolution over 1024 ϫ 1024 pixels at a pixel dwell time of 6.4 s. (b) For cells labeled with GFP and mitoTracker Red®, the multitrack mode was used in which the 488 nm laser line from an argon laser was used to excite GFP; the 543 laser line from a HeNe laser was used to excite the mitoTracker Red®, and images were collected by using an HFT 488/ 543/633 dichroic mirror and a 500 -530 nm bandpass filter for GFP and LP 560 nm filter for the mitoTracker Red® at a 12-bit intensity resolution over 1024 ϫ 1024 pixels at a pixel dwell time of 3.2 s. (c) For cells labeled with the fluorescence reporter combinations of CFP/GFP or GFP/YFP, images were collected at a 12-bit intensity resolution over 1024 ϫ 1024 pixels at a pixel dwell time of 3.2 s by using spectral imaging followed by linear unmixing to separate the fluorescence contribution of each fluorescent protein on a pixel by pixel basis (25,26). A series of stack X-Y images was collected using the Zeiss META detector over a spectrum of wavelengths between 462 and 633 nm with bandwidths of 10.7 nm during excitation with the 458 nm laser line and a HFT458 dichroic mirror for the CFP/GFP pair. For the GFP/YFP pair, a spectrum of wavelengths between 494 and 633 nm with bandwidths of 10.7 nm was collected during excitation with the 488 nm laser line and an HFT 488 dichroic mirror. Reference stack images were collected and stored for each of the fluorescence reporters alone to provide spectral signatures for GFP, YFP, and CFP that were used in the linear unmixing procedure. For time lapse studies, images were captured at 512 ϫ 512 pixels, one image every 15 s for a total of 5 min. The Zeiss LSM Physiology software package and Microsoft Excel were used to analyze the time-lapse data.
Statistics-Analysis of variance with ␣ ϭ 0.05 was used for processing the data. A two-sample t test was used as post-test unless otherwise indicated.

Analysis of Redox Engineered
Cells-To define molecular targets that are sensitive to alterations in the mitochondrial production of H 2 O 2 , we have developed a line of redox-engineered HT-1080 fibrosarcoma cell lines using antioxidant enzyme-based expression systems. We have focused our analyses on six redox-engineered cell lines that have been thoroughly characterized in terms of their antioxidant and oxidant profile (20,27). Enforced expression of mitochondrial manganese-containing superoxide dismutase (Sod2) led to a significant increase in intracellular H 2 O 2 , which was reversed by cytosolic or mitochondrial expression of catalase (Fig. 1A). Sod2 overexpression led to a 2-fold increase in the rate of oxidation of the H 2 O 2 -responsive fluorophore Redox-Red TM that was attenuated upon catalase coexpression (Fig. 1A). Sole catalase overexpression also decreased the basal rate of Redox-Red TM oxidation. These findings demonstrated that the enhanced rate of oxidation was attributed to the Sod2-dependent increases in the steady state production of H 2 O 2 that was subsequently reversed upon expression of the H 2 O 2 -detoxifying enzyme catalase.
Redox-dependent Oxidation of PTEN-PTEN is a dual specificity phosphatase that is reversibly sensitive to inactivation by H 2 O 2 (7) (Fig. 1B). The oxidation status of PTEN resulted in the conversion of the sulfhydryl groups to a disulfide, resulting in a more compact protein structure that can be visualized under nonreducing conditions (Fig. 1B) (7). The Sod2-dependent generation of H 2 O 2 showed accumulation of the oxidized form of PTEN (Fig. 1B, lane 4) that was prevented by coexpression of catalase (Fig. 1B, lanes 5 and 6). Addition of exogenous H 2 O 2 (0.5 mM) caused a prevalent accumulation of the oxidized form of PTEN that was blocked by mitochondrial targeted catalase (Fig. 1B, lane 6). PTEN inactivation resulting from mutational loss or oxidation increased the phosphorylation of the serine-threonine kinase Akt (4, 24, 28). Coordinate to the increase in PTEN oxidation upon Sod2 overexpression was a pronounced elevation in phosphorylation at serine 473 of Akt (Fig. 1B, lane 4), which was attenuated by the overexpression of catalase in either the cytosolic or mitochondrial compartment (Fig. 1B, lanes 5 and 6, respectively).
The substrates for Akt are numerous and are involved in regulating many diverse cellular functions, including proliferation, glucose utilization, angiogenesis, and cell survival (29). Glycogen synthase kinase 3␤ (GSK3␤), a negative regulator of cyclin D1, was inhibited by Akt-dependent phosphorylation. GSK3␤ phosphorylation was also reversibly sensitive to alterations in the steady state production of H 2 O 2 by antioxidant enzyme overexpression (Fig. 1B). The dominant negative isoform of Akt (K179M) can prevent phosphorylation of native Akt at serine 473 (30), and expression of this mutant in our Sod2 overexpressing cell line blocked the increase in Akt and GSK3␤ phosphorylation (Fig. 1B, lane 7). Enforced expression of a constitutively active myristoylated-Akt rescued the catalase-dependent inhibition of both Akt and GSK3␤ phosphorylation in the Sod2-overexpressing cell lines (Fig. 1B, lane 8). The PI3K complex was found in the plasma membrane, whereas PTEN is largely a cytoplasmic enzyme. It has been established previously that PTEN can be negatively regulated by phosphorylation of residues (Ser 380 , Thr 382 , and Thr 383 ) in its C-terminal tail (11)(12)(13). Given this, we sought to examine if PTEN phosphorylation was regulated in a redox-dependent manner. PTEN phosphorylation status did not change in any of our redox-engineered cell lines, indicating that its primary mode of redox regulation is through oxidation. These results suggested that reversible alterations in the steady state mitochondrial production of H 2 O 2 modulated the redox state of PTEN and the phosphorylation state of both Akt and the downstream target GSK3␤.
Sod2-dependent Oxidation of PTEN and PI3K Signaling-Clonal isolation of transfected cells, as in the case of the redoxengineered cell lines, may give rise to variants whose altered signaling characteristics may be independent of the gene being studied. To further confirm the role of Sod2 in modulating PI3K/Akt via PTEN oxidation, we utilized a Sod2/GFP fusion construct to isolate populations of GFP-fluorescing cell lines with differing levels of Sod2 expression. Sod2 was targeted to the mitochondria via a mitochondrial sequence peptide located in its N terminus. Prior to fluorescent-activated cell sorting, mitochondrial localization of the Sod2GFP fusion construct in the HT-1080 cells was confirmed by fluorescent microscopy. Analysis of Sod2GFP-transfected HT-1080 cells showed distinct perinuclear GFP staining that colocalized with the mitochondrial specific dye, Mitotracker Red TM (Fig. 2B). By using fluorescent-activated cell sorting, distinct populations of GFPpositive cells were isolated into low, medium, and high fluorescing cell lines (Sod2 lo , Sod2 m1 , Sod2 m2 , and Sod2 hi ). A pure population of GFP-expressing cells was also isolated whose GFP fluorescence was similar to that of the Sod hi cell lines. Sod zymography of the distinct cell populations showed that Sod2 activity increased relative to the level of GFP intensity in the sorted cell populations and that Sod1 levels were unaffected by Sod2GFP transfection (Fig. 2C). Furthermore, GFP expression alone had no effect on Sod2 activity. The numerous bands in the region of Sod2 activity represented the various combinations of the native Sod2 monomer and the Sod2GFP fusion monomer that comprises the active tetrameric enzyme.
We next analyzed the redox state of PTEN in the sorted cell populations. Analysis of PTEN under nonreducing conditions showed Sod2-dependent increase in the levels of its oxidized form (Fig. 2C). Akt and GSK-3␤ phosphorylation were also dose-dependently sensitive to increases in Sod2 activity, which correlated with PTEN oxidation, whereas total Akt and GSK-3␤ levels remain unchanged (Fig. 2C). Cyclin D1 expression was quite sensitive to alterations in the mitochondrial redox environment as even a low level of Sod2 expression increased the levels of its immunoreactive protein.
To characterize further the PTEN oxidation state in these cell lines, we utilized a method for detecting free cysteines by using 5Ј-fluoresceinated iodoacetamide (5-IAF) labeling (5, 31) ( Fig. 2A). Cysteine oxidation of PTEN increased in response to Sod2 expression, indicating that the Sod2-dependent production of H 2 O 2 had a direct effect on PTEN oxidation (Fig. 2A). These findings are the first to demonstrate that changes in the steady state mitochondrial production of H 2 O 2 in tumor cells can modulate the redox state of PTEN and signaling pathways under its control.
Sod2-derived H 2 O 2 Regulates Phosphoinositide Distribution-PTEN plays an important role in restricting the generation and distribution of 3Ј-phosphoinositides following receptor tyrosine kinase engagement (32). Accumulation of an oxidized inactive fraction of PTEN would likely favor the distribution of PtIns(3,4,5)P 3 at the plasma membrane compartment (Fig.  3A). To test this hypothesis, the various control and redoxengineered cell lines were transfected with chimeras of green fluorescent protein and the pleckstrin homology domain of either phospholipase C-␦ or Akt, to monitor distribution of , medium 1 (m1), medium 2 (m2), and high (hi)) populations based on the GFP fluorescence. Purified populations were then evaluated for Sod2 activity using a SOD native PAGE (see "Experimental Procedures"). Upper bands represent areas of endogenous and GFP-fused Sod2, and lower bands show Sod1 activity. Lower panels, PTEN oxidation was determined using nonreducing PAGE and immunoblot analysis. In parallel, immunoblot analysis of p-Akt, p-GSK3-␤, and cyclin D1 was performed. As controls, immunoblot analysis for total Akt and GSK3-␤ was performed. Red, reduced; Ox, oxidized. PtIns(4,5)P 2 or PtIns(3,4,5)P 3 , respectively (Fig. 3A) (33). For simplicity, the localization of CFP-PLC␦-PH and Akt-PH-YFP hereafter will represent accumulation of PtIns(4,5)P 2 and PtIns(3,4,5)P 3, respectively. H 2 O 2 has been shown to activate PI3K leading to the production of PtIns(3,4,5)P 3 (34,35) and serves as an effective control to monitor PtIns(3,4,5)P 3 accumulation at the plasma membrane. Treatment of HT-1080 cells with 0.5 mM H 2 O 2 showed a rapid (30 s) linear increase in the localization of PtIns(3,4,5)P 3 to the plasmalamellar surface (Fig. 3B). In untreated cells, PtIns(3,4,5)P 3 binding construct remained cytosolic (Fig. 3, B and C, 1), and no time-dependent change in its localization was observed (Fig. 3B). The pattern of PtIns(3,4,5)P 3 distribution in the Sod2-overexpressing cell lines was distinct from that of the controls and showed prominent redistribution to the plasma membrane (Fig. 3C, 2), which was prevented by coexpression of mitochondrial targeted catalase (Fig. 3C, 3).
The distinct distribution of the phosphoinositide-binding constructs in the Sod2GFP cells relative to the Sod2-overexpressing cells was likely due to the higher enzymatic activity of the Sod2GFP sorted population (45-versus 15-fold). Alternatively, the GFP protein of the Sod2GFP chimera when present in the mitochondrial compartment may alter the distribution of the phosphoinositide-binding constructs. To test this latter possibility, HT-1080 cell lines were transfected with a mitochondrial targeted GFP (mito-GFP), and phosphoinositide distribution was evaluated. Expression of the compartmentalized GFP in the mitochondria did not mediate mitochondrial accumulation of either phosphoinositide (Fig. 3D, 9 and 12). Furthermore, PtIns(4,5)P 2 and PtIns(3,4,5)P 3 distribution was similar to that observed in the control GFP-expressing cell lines residing in either the membrane or cytosolic compartments, respectively (Fig. 3D, 8 and 11). These findings imply that mitochondrial derived H 2 O 2 generated as a result of Sod2 overexpression and independent of receptor engagement contributed to a shift in the production and subsequent mobilization of both PtIns(4,5)P 2 and PtIns(3,4,5)P 3 .
Sod2-dependent Regulation of VEGF Expression-Because PTEN and reactive oxygen species have been postulated to play an important role in the angiogenic switch (37,38), we next examined the effects of alteration in the mitochondrial production of H 2 O 2 on expression of angiogenic markers and activity. First, the promoter activity and endogenous protein expression of a key stimulator of endothelial cell growth, vascular endothelial derived growth factor, were assessed. Overexpression of Sod2 led to a dramatic increase in the levels of endogenous VEGF immunoreactive protein as well as the activity of the transfected VEGF luciferase promoter construct (Fig. 4, A and  B). However, coexpression of catalase in either the cytosolic or mitochondrial compartment reversed the Sod2-dependent enhancement in VEGF promoter activity as well as its immunoreactivity (Fig. 4, A and B). Both the transcription factors Ets-1 and HIF-1␣ contribute to the activation of the angiogenic phenotype (39,40) and respond to changes in the mitochondrial redox environment (41,42). We have reported previously that Sod2 enhanced Ets-1-regulated gene expression in an H 2 O 2 -dependent fashion (43). Fig. 4B demonstrates that HIF-1␣ protein expression is also sensitive to alterations in the Sod2-dependent mitochondrial production of H 2 O 2 that is reversed by coexpression of both cytosolic and mitochondrial catalase. The increased HIF-1␣ expression was also associated with augmented VEGF promoter-driven luciferase activity that is reversed by mitochondrial catalase overexpression (Fig. 4A). These studies support a role for mitochondrial H 2 O 2 in the regulation of factors critical to the angiogenic switch.
PTEN plays an important role in the regulation of VEGF expression by blocking Akt-mediated VEGF gene transcription (37,44). To define whether the Sod2-dependent increase in VEGF expression may be attributed to the loss of PTEN activity and altered Akt signaling, Sod2-overexpressing cell lines were transfected with a wild type PTEN-GFP fusion or a dominant negative, kinase dead Akt (K179M) construct. Enforced expression of either PTEN or the Akt (K179M) prevented the Sod2-dependent increase in VEGF expression (Fig. 4B). Sod2GFP overexpression also led to an increase in VEGF expression that was not observed in the eGFP-overexpressing cell lines. This increase was also attenuated by the enforced expression of PTEN or the Akt (K179M). A myristoylated, constitutively active Akt (Myr-Akt) was used to confirm that VEGF expression in the eGFP HT-1080 cell line was sensitive to alterations in Akt signaling (Fig. 4B). In Fig. 4A, both basal and Sod2-dependent promoter activity were abolished by enforced expression of PTEN, whereas the PTEN (G129R) mutant lacking both lipid and protein phosphatase activity did not significantly alter VEGF promoter activity in any of the cell lines tested. These findings established that alterations in the steady state mitochondrial production of H 2 O 2 by antioxidant enzymes can modulate Akt-dependent signaling cascades involved in regulating VEGF production.
Mitochondrial H 2 O 2 Regulates Angiogenic Activity-We next investigated whether mitochondrial redox-dependent alterations in VEGF production impact angiogenic activity. By using a recently developed three-dimensional in vitro angiogenesis assay, we were able to evaluate whether factors secreted from the fibrosarcoma cells were able to promote endothelial cell sprouting (22). BLMV were grown on Cytodex beads and embedded in fibrin gels followed by incubation with the various redox-engineered cell lines. In Fig. 5, A and B, only rudimentary sprout formation was observed in the vector-transfected control cell lines (CMV and eGFP) and Sod2mCAT coexpressors, whereas Sod2 expression led to a significant increase in the length and number of sprouts (Fig. 5, A and B). Treatment of the control cell lines with VEGF (2.5 ng/ml) led to an increase in sprout formation to levels observed in the Sod2 cell lines. Addition of SodGFP hi cells to the three-dimensional matrix also stimulated sprout formation that was not observed with the addition of eGFP-overexpressing cell lines. Treatment of either the Sod2 or Sod2GFP cell lines with the PI3K inhibitor, LY2949002, completely blocked sprout formation in both cell lines, indicating that PI3K signaling contributed to redox-dependent angiogenic activity. The above assays were also performed in the absence of a primary fibroblast cell line that was shown previously to be required for sustained sprout formation in the presence of VEGF due to their ability to secrete additional factors critical for sprout development (22). In the complete absence of a fibroblast population, no endothelial sprouting was observed (Fig. 5, A and B). To test if this response was specific to VEGF, 0.16 g/ml of a VEGF-neutralizing antibody was added to wells containing either Sod2 or Sod2GFP-expressing cell lines. Coincubation of the VEGF-neutralizing antibody with the Sod2-overexpressing cells resulted in an inhibition of angiogenic sprout formation. Furthermore, when the Sod2-overexpressing cells were incubated with an isotype control antibody, no hindrance in sprout formation was observed. The present study indicates that increases in the cellular production of mitochondrial H 2 O 2 leads to the secretion of necessary factors, including VEGF that can promote endothelial sprout formation, and that these factors are, at least in part, dependent on PI3K signaling.
In Vivo Determination of Angiogenic Activity-To determine the redox-dependent regulation on in vivo angiogenesis, we looked at capillary development on the CAM, a widely adopted in vivo method for studying angiogenesis. Sod2 expression resulted in an increase in the development of new embryonic blood vessels, many of which displayed a winding morphology as compared with that of the control CMV cell lines, a characteristic commonly associated with potent angiogenic activity (Fig. 6, A and B). CAMs incubated with the CMV control cell lines did not show a significant increase in blood vessel density compared with CAMs incubated without cells. Coexpression of catalase resulted in blood vessel density similar to that of the control. These studies indicate that Sod2 expression increases the angiogenic phenotype in vivo and that coexpression of the H 2 O 2 -detoxifying enzyme, catalase, reverses this phenotype.

DISCUSSION
In recent years it has become more evident that ROS are not merely toxic by products of metabolism but serve an important regulatory role in numerous signaling pathways (45). Our data suggest a mechanism by which alterations of mitochondrial antioxidant enzyme expression can modulate the levels of intracellular H 2 O 2 and subsequently alter the activity of the PI3K signaling cascade through oxidation of PTEN. Furthermore, we have defined an alternative redox-sensitive pathway utilizing H 2 O 2 of mitochondrial origin to control angiogenesis.
The reversible oxidative inactivation of the tumor suppressor PTEN by H 2 O 2 has emerged as a second mode of inhibitory regulation independent of PDZ phosphorylation (7). The membrane-bound NADPH oxidase family members are the primary candidates for the production of oxidants in the plasma membrane microenvironment, and their inhibition can attenuate Akt phosphorylation in activated macrophages (4). Our findings indicate that the H 2 O 2 -dependent enhancement of the PI3K signaling axis can be attenuated by enforced expression of PTEN or dominant negative isoforms of Akt, emphasizing the vital role of PTEN in the redox regulation of this signaling cascade. This is also the first evidence for the oxidative loss of PTEN function in the absence of mutation in a human tumor cell line by mitochondrial derived H 2 O 2 .
Several reports indicate that PtIns(3,4,5)P 3 accumulation occurs at the plasma membrane in response to ROS (4,46). The current findings using redox-engineered cell lines are consistent with the hypothesis that mitochondrial oxidants can impact the distribution of these same bioactive phosphoinositides (Fig.  3, B and C). Furthermore, mitochondrial localization of the lipid-binding chimeras was also observed in response to Sod2 overexpression (Fig. 3D). Akt has been shown to distribute to the mitochondria and phosphorylate the mitochondrial GSK-3␤ isoform (47). It is likely that phosphoinositides themselves recruit Akt as significant accumulation of inositol phosphates to the mitochondria has been observed (48). Thus, the enforced expression of Sod2 and the production of H 2 O 2 may recruit important signaling factors to the mitochondria that mediate PI3K signaling and augment vascular recruitment. H 2 O 2 itself has been shown to regulate proangiogenic responses in a variety of systems, including human retinal pigment epithelial cells (49), cultured keratinocytes (50), and bovine pulmonary artery endothelial cells (41). Monte et al. (51) has demonstrated that H 2 O 2 contributes to promoting the angiogenic activity of tumor-bearing lymphocytes and that this activity can be inhibited by coadministration of the H 2 O 2detoxifying enzyme catalase but not Sod. In vitro angiogenesis of bovine thoracic aorta can be induced with relatively low concentrations (1 M) of H 2 O 2 and can be blocked by catalase (41). The Nox1-dependent generation of H 2 O 2 is a potent trigger of angiogenesis, increasing the vascularity of tumors and inducing molecular markers of angiogenesis, including vascular endothelial growth factor (VEGF), VEGF receptors, and matrix metalloproteinase activity in cultured cells and in tumors. Coexpression of catalase blocks the increased activity of the angiogenic markers, indicating that hydrogen peroxide signals part of the switch to the angiogenic phenotype (38). Furthermore, metalloproteinases are important contributors to the angiogenic switch, and their expression is redox-responsive (27,43). Finally, Ets-1 is also an important regulator of angiogenesis (52) and is sensitive to H 2 O 2 production (41). Thus, the intracellular production of H 2 O 2 activates the angiogenic switch enhancing the activity of a variety of signaling pathways that lead to production of proangiogenic factors.
Several reports have demonstrated that VEGF promoter expression is responsive to H 2 O 2 . Cho et al. (53) has determined that a region from Ϫ449 to Ϫ126 was required for VEGF promoter activity in macrophages in response to 1 mM H 2 O 2 . More detailed analysis of the VEGF promoter has identified a GC-rich region between Ϫ95 and Ϫ51 (54), containing an SP1binding site that is responsive to relatively low (50 M) concentrations of H 2 O 2 in human keratinocytes. Hocker and co-workers (55) have also shown that enhanced binding at the two SP1 and SP3 sites at Ϫ73/Ϫ66 and Ϫ58/Ϫ52 represent the core mechanism of oxidative stress-triggered VEGF transactivation. We have demonstrated previously (43) that the binding activity of SP1 is regulated in our redox-engineered cell lines. Thus, the SP1 sites at position Ϫ73 and Ϫ58 may also contribute to the regulation of VEGF expression in response to mitochondrial derived H 2 O 2 .
Active Akt leads to an increase in HIF-1␣ stabilization and subsequent transcriptional activation of VEGF and the angiogenic switch (56 -58). Loss of PTEN function leads to the increased production of VEGF (28). The present study provides evidence for the redox-sensitive link between these two parallel findings and strongly suggests that mitochondrial H 2 O 2 can also modulate the angiogenic switch through oxidative inactivation of the tumor suppressor PTEN. In the present study we have established that mitochondrial H 2 O 2 oxidizes PTEN leading to a concomitant increase in Akt signaling, VEGF expression, and angiogenic activity. The oxidant-dependent PTEN inactivation also enhances cyclin D1 expression in an Akt-dependent fashion and may regulate cell cycle progression. Sod2 overexpression protects from programmed cell death in response to various apoptotic stimuli (59 -61), and PTEN oxidation may contribute to this resistance. Future studies will be directed at defining whether a relationship between Sod2, PTEN, and apoptosis exists.
In recent years H 2 O 2 has come to the forefront as a key determinant of many redox-sensitive signaling pathways (2). The mitochondrion is the primary source for the intracellular production of H 2 O 2 , yet little emphasis has been placed on the role of this organelle in signaling processes. Receptor engagement leading to oxidant-dependent signaling has traditionally been attributed to the Rac-dependent activation of the phagocytic oxidase family members (62). Further support for this hypothesis was put forth with the discovery of the noninflammatory oxidases (63). However, a number of reports have emerged indicating a role for mitochondrial derived oxidants in receptor-dependent signaling process. Engagement of the ␣5␤1 integrin leads to alterations in cell shape via a process that leads to increased mitochondrial ROS production and subsequent expression of MMP-1 (18). Finkel and co-workers (64) have identified a feedback regulatory pathway whereby increased mitochondrial metabolic flow leads to enhanced oxidant production followed by JNK activation. JNK inhibits GSK3␤ and shifts glucose utilization toward increased glycogen synthesis by restricting the generation of diffusible H 2 O 2 . Thus, mitochondrial oxidants deliver a cytosolic signal that prevents their own production.
In the present study, JNK activation is also observed in response to Sod2 overexpression (data not shown) with a corresponding increase in GSK3␤ phosphorylation. However, GSK3␤ inhibition is attributed to Akt activation (Fig. 1B, lane  7). In addition, mitochondrial H 2 O 2 production is not feedbackinhibited because its production is driven by the enforced CMVdependent expression of Sod2.
Sod2 is positioned to efficiently dismutate O 2 . to H 2 O 2 at near diffusion limiting rates and is the only antioxidant enzyme that is induced in response to growth factors, cytokines, ionizing radiation, and redox cycling drugs (65). We propose that the mitochondria may play an important role in regulating the angiogenic switch by serving as a potent source of the small diffusible signaling molecule H 2 O 2 . Furthermore, the enhanced expression of Sod2 under varying pathophysiologic conditions may not solely be to prevent the deleterious actions of oxidants but also serves to modulate key regulatory pathways that control cellular function by its ability to generate H 2 O 2 .