Glucose-6-phosphate dehydrogenase modulates vascular endothelial growth factor-mediated angiogenesis.

Glucose-6-phosphate dehydrogenase (G6PD), the first enzyme of the pentose phosphate pathway, is the principal intracellular source of NADPH. NADPH is utilized as a cofactor by vascular endothelial cell nitric-oxide synthase (eNOS) to generate nitric oxide (NO*). To determine whether G6PD modulates NO*-mediated angiogenesis, we decreased G6PD expression in bovine aortic endothelial cells using an antisense oligodeoxynucleotide to G6PD or increased G6PD expression by adenoviral gene transfer, and we examined vascular endothelial growth factor (VEGF)-stimulated endothelial cell proliferation, migration, and capillary-like tube formation. Deficient G6PD activity was associated with a significant decrease in endothelial cell proliferation, migration, and tube formation, whereas increased G6PD activity promoted these processes. VEGF-stimulated eNOS activity and NO* production were decreased significantly in endothelial cells with deficient G6PD activity and enhanced in G6PD-overexpressing cells. In addition, G6PD-deficient cells demonstrated decreased tyrosine phosphorylation of the VEGF receptor Flk-1/KDR, Akt, and eNOS compared with cells with normal G6PD activity, whereas overexpression of G6PD enhanced phosphorylation of Flk-1/KDR, Akt, and eNOS. In the Pretsch mouse, a murine model of G6PD deficiency, vessel outgrowth from thoracic aorta segments was impaired compared with C3H wild-type mice. In an in vivo Matrigel angiogenesis assay, cell migration into the plugs was inhibited significantly in G6PD-deficient mice compared with wild-type mice, and gene transfer of G6PD restored the wild-type phenotype in G6PD-deficient mice. These findings demonstrate that G6PD modulates angiogenesis and may represent a novel angiogenic regulator.

Angiogenesis, the formation of new blood vessels in response to tissue ischemia or injury, is dependent upon a coordinated sequence of events involving vascular endothelial cell migration, proliferation, and tube formation (1,2). Initially, vascular endothelial cells must acquire an angiogenic phenotype to migrate toward an angiogenic stimulus, proliferate behind the front of migration, and differentiate to form endothelial tubes and capillary-like structures.
Nitric oxide (NO ⅐ ) 1 has been shown to modulate angiogenesis by mediating growth factor-stimulated endothelial cell migration and proliferation. Nitric oxide is permissive for endothelial cell migration and enhances directional migration by inducing a switch from a stationary to a mobile phenotype (3,4). In addition NO ⅐ promotes endothelial cell proliferation, and proliferating endothelial cells demonstrate increased expression of the endothelial isoform of nitric-oxide synthase (eNOS) compared with quiescent cells (5). In vivo studies utilizing eNOS Ϫ / Ϫ mice have demonstrated the absolute requirement for endothelium-derived NO ⅐ for effective angiogenesis. In this murine model, compared with mice with normal eNOS activity, vascular endothelial growth factor (VEGF)-stimulated angiogenesis and vascular permeability are significantly attenuated (6).
Glucose-6-phosphate dehydrogenase (G6PD), the first and rate-limiting enzyme in the pentose phosphate pathway, is the principal intracellular source of NADPH. NADPH, in turn, is utilized directly as a cofactor for eNOS and, indirectly, to maintain levels of another important cofactor, tetrahydrobiopterin, via de novo synthesis and the dihydrofolate reductase salvage pathway. In this manner, G6PD regulates eNOS activity and NO ⅐ levels. In this study, we demonstrate that G6PD activity modulates endothelial cell migration, proliferation, and tube formation by mediating NO ⅐ levels. G6PD may, therefore, serve as a novel regulatory determinant of the angiogenic phenotype.
for 30 min at 37°C, and neutralized with 10 N HCl. Thymidine incorporation was measured by liquid scintillation counting.
Cell Migration-Migration was assessed by a cell wounding assay in BAEC grown to confluence in a P100 dish and synchronized in 1% serum for 24 h. A longitudinal incision was made in the midline of the plate with a sterile scalpel, and cells were scraped from one-half of the plate and stimulated with VEGF (100 ng/ml) for 12 h. After this time, cells that migrated across the midline were visualized using a Nikon TE 300 microscope, and 5 images per high powered field (hpf) per plate were captured digitally. Cells that crossed the midline were counted and averaged per plate. Data are presented as migrated cells/hpf. In addition, cell migration was assayed using a modified Boyden chamber (ChemoTx® plate, Neuro Probe, Inc., Gaithersburg, MD) according to the manufacturer's instructions.
Tube Formation-In vitro formation of capillary-like tube structures was examined using Matrigel. Matrigel (0.5 ml) was polymerized on dual-chamber microscope slides. Cells were then plated on Matrigel in full-growth media for 1 h. Once the cells were seeded, the media were replaced with media containing 1% serum or VEGF (100 ng/ml). Tube formation was visualized using an inverted microscope (Nikon TE 300) equipped with digital imaging. For each treatment, 10 high power field images were captured, and area of tubes/networks formed was quantified using Scion Corp. (NIH Image) area analysis with background subtraction and averaged. Data are presented as density units.
eNOS Activity-eNOS activity was measured in intact cells without the addition of exogenous cofactors as described previously (9).
Immunoprecipitation-Cells were incubated in lysis buffer for 1 h at 4°C and centrifuged to remove insoluble material. The lysates were incubated overnight at 4°C with protein-Sepharose A that had been incubated for 24 h with an antibody to eNOS (Transduction Laboratories) or G6PD (Sigma). After boiling, 100 g of cell protein was loaded per lane, and samples were resolved by SDS-PAGE and transferred to polyvinylidene difluoride membrane. Blots were blocked with 5% skim milk and developed with antibodies for eNOS (1:500) or G6PD (1:1000) and visualized using the ECL detection system as described previously (7).
Real Time NO ⅐ Measurements-BAEC were grown to confluence in P100 dishes, washed twice with Dulbecco's phosphate-buffered saline, and placed in a balanced salt solution (130 mmol/liter NaCl, 5 mmol/ liter KCl, 1 mmol/liter MgCl 2 , 1.5 mmol/liter CaCl 2 , 35 mmol/liter phosphoric acid, and 20 mmol/liter HEPES, pH 7.4) stimulated with VEGF (100 ng/ml). Real time NO ⅐ measurements were obtained using the inNO II Nitric Oxide Measuring System (Innovative Instruments) and the amino-FLAT probe. Data were recorded in real time; peak NO ⅐ levels were measured, and the area under the curve was integrated.
Murine Aortic Ring Model of Angiogenesis-Murine thoracic aortas were harvested and placed in ice-cold Dulbecco's phosphate-buffered saline (PBS). The aorta was flushed with ice-cold PBS until free of blood. The adventitia was dissected free, and the aorta was cut into 1-mm rings. The rings were embedded in type I collagen gel such that the lumen was oriented horizontally and placed in MCDB 131 media and maintained at 37°C. To assess the angiogenic response, the rings were visualized using a Nikon TE300 microscope, digital images captured, and image analysis performed using Scion Corp. (NIH Image) area analysis with background subtraction. Data are presented as density units (11).
G6PD-deficient Mouse Model-The Pretsch mouse, a murine model of G6PD deficiency on a C3H murine background (12), was bred at our institution from frozen embryos obtained from Medical Research Council (Harwell, UK). Hemizygous (X b Y) G6PD-deficient male mice (HEMI) and wild-type (XY) C3H control mice (WT) age 12-16 weeks were studied. Animals were genotyped as described previously (13). The animals were fed standard chow and handled following NIH guidelines. All procedures were approved by the Institutional Animal Care and Use Committee at Boston University Medical Center. The G6PD-deficient phenotype was confirmed by measuring hepatic G6PD activity and NADPH levels as described previously (7). Liver was harvested from mice at the time of sacrifice, snap-frozen, and stored at Ϫ80°C. G6PD and NADPH levels were measured in liver homogenates.
In Vivo Matrigel Plug Assay and Cell Recovery-Matrigel (0.5 ml) was injected subcutaneously in the ventral groin area. One side was injected with Matrigel alone and the other with Matrigel mixed with VEGF (100 ng/ml). In this manner, each animal serves as its own control. After 14 days, the mice were euthanized by CO 2 inhalation. Matrigel plugs were excised and fixed in 10% formalin, subjected to an ethanol dehydration series, and embedded in paraffin. Serial sections (10 m) were cut using a cryotome and applied to glass slides. Slides were deparaffinized and stained with hematoxylin and eosin. To recover cells, the excised Matrigel plugs were minced with a sterile scalpel, passed 10 times through a 14-gauge needle, treated with BD TM Cell Recovery Solution for 1 h at 4°C, centrifuged, and subjected to immunoprecipitation as described above to assay for proteins of interest.
Immunohistochemistry-Immunohistochemistry on deparaffinized slides was performed using a rabbit polyclonal anti-von Willebrand factor antibody (Santa Cruz Biotechnology, Inc.) at 1:50, as described previously (14). An angiogenic response was quantified by cell counts from 10 high power fields per section and image analysis using Scion Corp. (NIH Image) with background subtraction to determine the area occupied by endothelial cells.
Statistical Analysis-Continuous data were expressed as mean Ϯ S.E. Comparison between groups was performed by Student's paired two-tailed t test. Two-way analysis of variance was used to examine differences in response to treatments between groups, with post hoc analysis performed by the method of Student-Newman-Keuls. A p value of Ͻ0.05 was considered significant.

G6PD and Endothelial
Cell Proliferation-To examine the role of G6PD in VEGF-mediated endothelial cell proliferation, BAEC were transfected with an antisense oligodeoxynucleotide to G6PD mRNA with a 36% transfection efficiency to decrease G6PD expression and activity by 76% (AS-EC) or a scrambled control sequence (SS-EC), and cells were stimulated with VEGF (100 ng/ml) for 12 h. Compared with SS-EC, AS-EC demonstrated a significant decrease in basal and VEGF-stimulated [ 3 H]thymidine incorporation ( Fig. 2A). The observed decrease in cell proliferation was not the result of increased cell death as determined by lactate dehydrogenase activity in the media (data not shown).
To determine the contribution of NO ⅐ to VEGF-stimulated We next sought to determine whether overexpression of G6PD would augment cell proliferation. To perform these studies, BAEC were infected with AdG6PD (AdG6PD-EC) at a multiplicity of infection ϭ 10 plaque-forming units/cell with a 90 -95% infection efficiency to increase G6PD activity 5-fold.
G6PD and Endothelial Cell Migration-To examine the influence of G6PD on VEGF-stimulated endothelial cell migration, we performed a cell wounding assay. SS-EC and AS-EC were incubated with VEGF, and cell migration across the midline was observed. After 12 h, VEGF markedly increased cell migration across the midline in SS-EC (44 Ϯ 4 versus 142 Ϯ 16 cells/hpf, p Ͻ 0.001), an effect that was not observed in AS-EC (27 Ϯ 11 versus 35 Ϯ 5 cells/hpf, p ϭ not significant) (Fig. 3).
These findings were confirmed utilizing a modified Boyden chamber assay to determine whether G6PD activity influenced directed VEGF-mediated endothelial cell migration. SS-EC and AS-EC were stimulated with VEGF (100 ng/ml) for 12 h during which time cells could migrate across the membrane. In SS-EC treated with VEGF, there was a significant increase in fluorescence (13.3 Ϯ 2.6 versus 55.3 Ϯ 6.6 units, p Ͻ 0.001), indicating an increase in cell migration, that was abrogated in AS-EC with deficient G6PD activity (14.1 Ϯ 4.8 versus 20.2 Ϯ 9.9 units, p ϭ not significant).
We next sought to determine the contribution of NO ⅐ to endothelial cell migration. By utilizing the modified Boyden chamber assay, SS-EC or AS-EC were stimulated with VEGF for 12 h in the presence or absence of L-NAME (1 mmol/liter) or L-NMMA (100 mol/liter). In SS-EC, L-NAME significantly decreased cell migration in VEGF-stimulated cells (51.9 Ϯ 4.2 versus 23.6 Ϯ 7.0 units, p Ͻ 0.01) as did L-NMMA (51.9 Ϯ 4.2 versus 21.6 Ϯ 7.3 units, p Ͻ 0.003). Interestingly, in AS-EC, VEGF did not significantly increase endothelial cell migration, and L-NAME or L-NMMA did not alter this response.
To determine whether overexpression of G6PD would promote directed endothelial cell migration, we overexpressed G6PD in endothelial cells and examined migration in a modified Boyden chamber assay. Compared with Ad-EC, AdG6PD-EC demonstrated a significant increase in fluorescence under basal conditions ( G6PD and Endothelial Tube Formation-As endothelial cell proliferation and migration are processes integral to the formation of capillary-like tube structures, we examined the effect of G6PD activity on tube formation using an in vitro Matrigel assay. SS-EC and AS-EC were plated on Matrigel and stimulated with VEGF for 12 h. Under basal conditions, SS-EC plated on Matrigel formed tubes and networks, and the area occupied by SS-EC endothelial tubes was increased significantly following exposure to VEGF (9,911 Ϯ 640 versus 24,729 Ϯ 3,311 units, p Ͻ 0.001). Similarly, AS-EC formed endothelial tubes and the area occupied by these networks was not significantly different from that observed with SS-EC (9,911 Ϯ 640 versus 8,424 Ϯ 1,005 units, p ϭ not significant); however, when stimulated with VEGF, AS-EC tube formation was decreased significantly compared with SS-EC (10,554 Ϯ 895 versus 24,729 Ϯ 3,311 units, p Ͻ 0.002) (Fig. 4A).
To examine the role of NO ⅐ in endothelial tube formation, we performed the in vitro Matrigel assay in the presence of L-NAME or L-NMMA. In VEGF-stimulated SS-EC, the area occupied by endothelial tubes was decreased significantly in the presence of L-NAME (24,729 Ϯ 3,311 versus 16,734 Ϯ 2,396 units, p Ͻ 0.001) or L-NMMA (24,729 Ϯ 3,311 versus 17,789 Ϯ 1,923 units, p Ͻ 0.001). In contrast, there was no significant difference in the area occupied by endothelial tubes in VEGFstimulated AS-EC in the presence of L-NAME or L-NMMA.
As G6PD overexpression was associated with enhanced endothelial cell proliferation and migration, it is not surprising that compared with Ad-EC, tube formation was significantly increased in AdG6PD-EC under basal conditions (10,372 Ϯ 1,286 versus 19,549 Ϯ 1,502 units, p Ͻ 0.01). This response was augmented further following stimulation with VEGF (24,994 Ϯ 2,533 versus 43,212 Ϯ 4,208 units, p Ͻ 0.001) (Fig. 4B). Furthermore, NO ⅐ played a significant role in this effect as the addition of L-NAME to AdG6PD-EC decreased tube formation in unstimulated (19, G6PD, eNOS Activity, and NO ⅐ Levels-These observations suggest that one mechanism by which G6PD influences endothelial cell proliferation, migration, and tube formation is to regulate eNOS activity and, thereby, bioavailable NO ⅐ . To examine the effect of G6PD expression on eNOS activity, we measured eNOS activity in intact cells (without the addition of exogenous cofactors) (Fig. 5A). In AS-EC, compared with SS-EC, eNOS activity was significantly decreased (12,728 Ϯ 488  (Fig. 5D) (15).
To determine the effect of decreased or increased G6PD expression on VEGF-mediated NO ⅐ production, we measured real time NO ⅐ levels using a nitric oxide probe (Fig. 5B). Peak NO ⅐ levels were significantly lower in AS-EC compared with SS-EC To examine the mechanism(s) by which G6PD mediates NO ⅐ production, we first performed immunoprecipitation studies to determine whether G6PD co-localizes with eNOS. Under basal conditions, G6PD and eNOS co-localize, and following stimulation with VEGF, co-localization is enhanced (Fig. 5C). We next sought to determine whether G6PD influences activation of the VEGF receptor Flk-1/KDR to modulate Akt-eNOS activation. Interestingly, in AS-EC, tyrosine phosphorylation of a 230-kDa band, consistent with Flk-1/KDR, was decreased at both 5 and 15 min compared with what was observed in SS-EC resulting in decreased phosphorylation of Akt and eNOS in AS-EC compared with SS-EC (Fig. 5D). In contrast, in AdG6PD-EC, tyrosine phosphorylation of the 230-kDa band was enhanced at 5 and 15 min resulting in increased phosphorylation of Akt and eNOS compared with Ad-EC (Fig. 5D).
G6PD, VEGF, and ROS Accumulation-It has been demonstrated recently that VEGF promotes angiogenesis in human umbilical vein endothelial cells by stimulating NAD(P)H oxidase to increase ROS formation (16). Therefore, we measured ROS levels by 2Ј,7Ј-dichlorodihydrofluorescein diacetate fluorescence in endothelial cells with decreased or increased G6PD activity to determine whether our observations could be explained by changes in ROS levels. Interestingly, in AS-EC exposed to VEGF, ROS levels were significantly higher than in SS-EC (248 Ϯ 27 versus 180 Ϯ 22 units, p Ͻ 0.01), and conversely, in AdG6PD-EC, ROS levels were lower than in Ad-EC (156 Ϯ 14 versus 197 Ϯ 12 units, p Ͻ 0.01). These findings suggest that G6PD, which is recognized as an antioxidant enzyme in endothelial cells, may additionally modulate ROS accumulation by influencing NO ⅐ production to mediate the redox milieu to a level favorable for endothelial cell proliferation, migration, and tube formation.
Ex Vivo Aorta Implants and Vessel Outgrowth-We next examined the effect of G6PD on VEGF-mediated angiogenesis using an in vivo model of G6PD deficiency, the Pretsch mouse. Compared with C3H background mice (WT), hemizygous male G6PD mice (HEMI) demonstrate only 31% of WT G6PD activity with a concomitant reduction in NADPH levels (0.6 Ϯ 0.03 versus 0.3 Ϯ 0.01 mmol/mg protein, p Ͻ 0.001).
To examine specifically the effect of G6PD on aortic vessel outgrowth, we explanted thoracic aortas from WT and HEMI mice, sectioned them into rings, and embedded the rings in a collagen matrix. The rings were then observed over 6 days for vessel outgrowth. By day 3, there was significant outgrowth from the WT rings compared with the rings from the HEMI mice, and by day 6, there was a marked difference between outgrowth observed from WT and HEMI mice rings (Fig. 6). To quantify these observations, the area occupied by vessel outgrowth was determined using area image analysis with background subtraction. At day 3, the area occupied by vessel outgrowth was significantly greater in aortic rings from WT compared with HEMI mice (6,037 Ϯ 975 versus 2,767 Ϯ 391 units, p Ͻ 0.04), and this effect was more pronounced by day 6 (10,057 Ϯ 713 versus 3,475 Ϯ 295 units, p Ͻ 0.001).
In Vivo Matrigel Assay-To examine the effect of G6PD activity on angiogenesis in vivo, we performed an in vivo Matrigel migration assay in WT and HEMI mice. Each mouse was injected on one side with either Matrigel alone or Matrigel supplemented with VEGF, and the plugs were examined after 14 days. In WT mice, there is noticeable cell migration into the Matrigel plug, and this response is increased in Matrigel supplemented with VEGF (12 Ϯ 1 versus 52 Ϯ 2 cells/hpf, p Ͻ 0.001). In contrast, in HEMI mice, there is a marked decrease in cell migration into the Matrigel plug compared with WT mice, and exposure to VEGF only modestly improved this response (6 Ϯ 1 versus 9 Ϯ 1 cells/hpf, p Ͻ 0.03) (Fig. 7A). These findings confirm that G6PD modulates angiogenesis in vivo.
To determine that the cells that migrated into the Matrigel were endothelial cells, the Matrigel sections were immunostained for von Willebrand factor and counted. In Matrigel sections from WT mice, 91% of cells stained positive for von Willebrand factor, whereas in HEMI mice 89% of cells stained positive for von Willebrand factor.  6. Ex vivo aortic rings and vessel outgrowth. Thoracic aortas were harvested from C3H wild-type (WT) and G6PD-deficient X b Y hemizygous (HEMI) mice. Aortas were dissected free of adventitia, sectioned into 1-mm rings, and embedded in a collagen matrix. Rings were examined at high power magnification (ϫ40) after 3 days (A and B) and 6 days (C and D) for vessel outgrowth. These findings confirm that the cells that migrated into the Matrigel plugs are endothelial cells.
To determine whether we could "rescue" the G6PD-deficient phenotype, we implanted Matrigel plugs in HEMI mice that were supplemented with AdG6PD or Ad in the absence or presence of VEGF, and we examined them after 14 days (Fig.  7B). In Ad-treated plugs, VEGF stimulated cell migration into the plug (5 Ϯ 1 versus 9 Ϯ 1 cells/hpf, p Ͻ 0.001). In contrast, in plugs supplemented with AdG6PD, there was a marked increase in cell migration into the plug, which was enhanced in plugs treated with VEGF (15 Ϯ 1 versus 48 Ϯ 4 cells/hpf, p Ͻ 0.001). These findings demonstrate that the WT phenotype may be restored successfully in G6PD-deficient mice by local gene transfer of G6PD.
We next recovered cells from Ad-and AdG6PD-treated Matrigel plugs that had been implanted for 7 or 14 days, and we examined these cells for G6PD expression. Interestingly, G6PD expression was increased in cells from AdG6PD-treated plugs at 7 days compared with Ad-treated plugs, and this effect remained present at 14 days, although somewhat diminished (Fig. 7C). To confirm that our in vitro observations were operative in vivo, we next examined recovered cells for phosphorylation of Flk-1/KDR, Akt, and eNOS. In cells recovered from AdG6PD-treated plugs, there was enhanced phosphorylation of a 230-kDa protein consistent with Flk-1/KDR at day 7 resulting in increased phosphorylation of Akt and eNOS compared with cells recovered from Ad-treated plugs. This effect was also present, although to a lesser extent, at day 14 (Fig. 7C). DISCUSSION In these studies, we found that G6PD significantly influenced VEGF-mediated vascular endothelial cell proliferation, migration, and tube formation in an in vitro cell culture model and an in vivo murine model. Decreased G6PD expression and activity were associated with an impaired response to VEGF-stimulated cell proliferation, migration, and formation of tubes and networks in a Matrigel matrix. In contrast, increased G6PD expression and activity enhanced these processes. G6PD modulated these events, in part, by influencing basal and VEGF-stimulated eNOS activity and NO ⅐ levels. We have demonstrated that this may occur as a result of G6PD co-localizing with eNOS. In addition, we also demonstrate that G6PD influences tyrosine phosphorylation of the VEGF receptor Flk-1/KDR and, in turn, phosphorylation of Akt and eNOS. In endothelial cells with deficient G6PD activity, we have demonstrated previously that eNOS may "uncouple" to generate superoxide in preference to NO ⅐ (7). To determine that a decrease in eNOS-mediated NO ⅐ production, and not an increase in superoxide generation by eNOS, accounted for the diminished response to VEGF in G6PD-deficient cells, it was therefore critical to utilize the eNOS inhibitors L-NAME, which inhibits superoxide and NO ⅐ generation by eNOS, and L-NMMA, which inhibits only NO ⅐ production by eNOS (17). In studies performed in the presence of these inhibitors, we demonstrated that eNOS-mediated NO ⅐ generation was the critical determinant of basal and VEGF-stimulated endothelial cell proliferation, migration, and tube formation. These findings were confirmed by direct measurement of eNOS activity and NO ⅐ levels.
Importantly, by utilizing a murine model of G6PD deficiency, we demonstrated that G6PD regulates endothelial cell proliferation, migration, and tube formation in vivo. In aortas harvested from G6PD-deficient mice, there is decreased vessel outgrowth in a collagen matrix preparation compared with aortas from animals with normal G6PD activity. Similarly, in an in vivo Matrigel assay, there was a diminished endothelial cell response under basal conditions and following stimulation with VEGF in G6PD-deficient mice. Interestingly, the G6PDdeficient phenotype could be "rescued" by adding adenovirus encoding G6PD to the Matrigel plug, presumably resulting in local gene transfer of G6PD, to restore endothelial cell migration and proliferation to the degree observed in wild-type mice with normal levels of G6PD activity. The validity of this approach has been confirmed recently in a murine model using a different cDNA (18). By analyzing cells recovered from the plugs, we demonstrate further that there is increased expres-FIG. 7. In vivo Matrigel plug assay. A, Matrigel (0.5 ml) or Matrigel supplemented with VEGF (100 ng/ml) was implanted subcutaneously in the ventral groin of C3H wild-type (WT) and G6PD-deficient X b Y hemizygous (HEMI) mice. After 14 days, the mice were sacrificed, and the plugs were excised with surrounding tissue for orientation. The plugs were sectioned (10 m) using a cryotome and stained with hematoxylin and eosin and examined at ϫ20 magnification. Matrigel stains pink and cell nuclei stain blue. B, AdG6PD rescues the G6PD-deficient phenotype in an in vivo Matrigel plug assay. Matrigel was mixed with Ad, an empty adenoviral vector, Ad supplemented with VEGF, AdG6PD, an adenoviral vector encoding G6PD cDNA, or with AdG6PD and supplemented with VEGF and was injected subcutaneously in the ventral groin area of G6PD-deficient X b Y hemizygous mice. After 14 days, the mice were sacrificed, and the plugs were excised and processed as detailed in A. Data are presented as mean Ϯ S.E. #, p Ͻ 0.001 versus ϪVEGF; *, p Ͻ 0.001 versus Ad; **, p Ͻ 0.001 versus ϪVEGF. C, cells were recovered from Matrigel plugs after 7 or 14 days and examined for VEGF-mediated activation of eNOS by examining phosphorylation of a 230-kDa protein corresponding to the VEGF receptor Flk-1/ KDR, Akt, and eNOS by Western blotting as well as G6PD expression. Blots are representative of three experiments.

G6PD, VEGF, and Angiogenesis
sion of G6PD in these cells. Furthermore, this results in increased tyrosine phosphorylation of Flk-1/KDR, Akt, and eNOS in vivo to confirm our in vitro findings.
The association between G6PD activity and cell proliferation has been well described in several non-vascular cell types (19,20). Growth factors, such as platelet-derived growth factor and epidermal growth factor, have been shown to increase cell proliferation by increasing both basal G6PD activity and G6PD expression (19). Deficient G6PD activity is associated with decreased cell proliferation in both steroid receptor-positive and receptor-negative breast cancer cell lines (20), COS-7 cells, Swiss 3T3, Balb/c 3T3, and A31 fibroblasts (19). In contrast, increased G6PD expression enhances cell proliferation. Adenoviral gene transfer of G6PD to increase G6PD expression 2-3-fold in COS-7 cells resulted in a marked increase in [ 3 H]thymidine incorporation (19). Similarly, NIH 3T3 cells transfected with human G6PD cDNA exhibited contact-and anchorage-independent growth (21). Taken together, these findings demonstrate that G6PD activity is associated with cell proliferation in non-vascular cell lines. We now provide evidence that G6PD activity is associated with vascular endothelial cell proliferation.
To date, there have been no studies to evaluate the influence of G6PD on cell migration. In fact, the current evidence linking G6PD to cell migration is indirect; G6PD activity has been associated with cell motility, as a surrogate for migration, in spermatogenic cells. In a rat model, treatment with agents that decreased G6PD activity resulted in a reduction in sperm motility (22). In our studies, we provide the first evidence that G6PD importantly modulates directed vascular endothelial cell migration.
The mechanism(s) and/or effectors by which G6PD enhances acquisition of an angiogenic phenotype remain incompletely characterized; however, it has been demonstrated that intracellular NADPH levels, and not the production of riboses by G6PD, are the critical determinant of cell growth (19). This suggests that one mechanism by which G6PD regulates endothelial cell migration, proliferation, and tube formation is by synthesizing NO ⅐ via eNOS, which has an absolute requirement for NADPH as a cofactor.
The relationship between G6PD and NOS activity has been demonstrated in several cell lines. For example, NO ⅐ production is impaired in G6PD-deficient granulocytes stimulated with lipopolysaccharide or phorbol 12-myristate 13-acetate. Interestingly, this response resulted from a decrease in activity, not expression, of the inducible form of NOS (23). Similarly, in peritoneal macrophages from p53 Ϫ / Ϫ mice treated with dehydroepiandrosterone (at concentrations that inhibit G6PD activity), NOS activity and indices of NO ⅐ production were significantly decreased (24).
In vascular endothelial cells, we have shown previously that G6PD influences eNOS activity by regulating substrate availability. G6PD-deficient endothelial cells produce decreased levels of bioavailable NO ⅐ in response to agonists because of eNOS "uncoupling" to generate reactive oxygen species in preference to NO ⅐ (7). This uncoupling of eNOS has been shown to occur when cofactors, such as NADPH and/or tetrahydrobiopterin, which is synthesized and salvaged in NADPH-dependent reactions, are depleted. Furthermore, gene transfer of G6PD to increase G6PD expression and activity in aortic endothelial cells resulted in enhanced NADPH levels, which in turn increased eNOS activity and NO ⅐ generation measured as cGMP, nitrate, and nitrite levels (10).
There are additional mechanisms by which G6PD may influence eNOS activity. Co-localization of G6PD with eNOS represents one mechanism by which G6PD may modulate NO ⅐ production. In fact, it has been shown previously that G6PD and nitric-oxide synthase co-localize in brush cells of rat stomach and pancreas (25). We now demonstrate in vascular endothelial cells that G6PD co-localizes with eNOS. In addition, VEGF-stimulated NO ⅐ production has been shown to occur via phosphorylation of the VEGF receptor Flk-1/KDR to phosphorylate and activate Akt and thereby eNOS (26,27). Activation of this signaling pathway has been implicated in angiogenesis both in vitro and in vivo (6,28). It is therefore not surprising that decreased phosphorylation of Flk-1/KDR, Akt, and eNOS associated with deficient G6PD activity results in diminished endothelial proliferation, migration, and tube formation both in vitro and in vivo and that increased G6PD activity enhances these responses.
These studies provide novel insight into the role of G6PD in vascular endothelial cell proliferation, migration, and tube formation. In vascular endothelial cells, G6PD activity correlates with the response to agonist-mediated cell proliferation and migration, which, in turn, determines tube formation. This effect occurs in part by the influence of G6PD on eNOS activity to generate NO ⅐ . Therefore, G6PD, by modulating eNOS activity to regulate vascular endothelial cell proliferation, migration, and tube formation, is a critical regulatory determinant of the angiogenic phenotype.