Ginsenoside-Rg1 Induces Vascular Endothelial Growth Factor Expression through the Glucocorticoid Receptor-related Phosphatidylinositol 3-Kinase/Akt and β-Catenin/T-cell Factor-dependent Pathway in Human Endothelial Cells*

Ginsenoside-Rg1, the most prevalent active constituent of ginseng, is a potent proangiogenic factor of vascular endothelial cells. This suggests that Rg1 may be a new modality for angiotherapy. Rg1 can activate the glucocorticoid receptor (GR). However, the regulatory steps downstream from GR that promote Rg1-induced angiogenesis have not been elucidated. Here we showed for the first time that Rg1 was a potent stimulator of vascular endothelial growth factor (VEGF) expression in human umbilical vein endothelial cells, and importantly this induction was mediated through a phosphatidylinositol 3-kinase (PI3K)/Akt and β-catenin/T-cell factor-dependent pathway via the GR. Rg1 stimulation resulted in an increase in the level of β-catenin, culminating its nuclear accumulation, and subsequent activation of VEGF expression. Transfection of a stable form of β-catenin (S37A) or the use of a glycogen synthase kinase 3β inhibitor to stabilize β-catenin induced VEGF synthesis, whereas small interfering RNA-mediated down-regulation of β-catenin did not, confirming that the effect was β-catenin-specific. Using a luciferase reporter gene assay, we observed that Rg1 increased T-cell factor/lymphoid enhancer factor transcriptional activity. These events were mediated via a PI3K-dependent phosphorylation of the inhibitory Ser9 residue of glycogen synthase kinase 3β. In addition, the GR antagonist RU486 was able to inhibit Rg1-induced PI3K/Akt and β-catenin activation. These findings provide new insights into the mechanism responsible for Rg1 functions.

Ginseng, the root of Panax ginseng C.A. Meyer, has been a key component in Chinese traditional medicine for more than 1000 years. It is now one of the most extensively used alternative medicines throughout the world and appears in the pharmacopoeias of several countries, including the United States and Europe and employed for cancer, diabetes, and cardiovascular concerns. In the United States, it was the second largest selling herbal supplement in 2001, with gross retail sales of U.S. $62 million (1). The molecular components responsible for ginseng actions are ginsenosides, which are triterpene saponins that have a rigid steroidal skeleton with sugar moieties (2). Within more than 30 different ginsenosides, Rg1 is among the most abundant and active ingredients of P. ginseng (3). Rg1 has been reported to trigger transcriptional activation of glucocorticoid-responsive element-containing reporter gene, suggesting that Rg1 can activate the glucocorticoid receptor (GR) 2 (4,5). In prior work, Rg1 has been demonstrated to promote functional neovascularization into a polymer scaffold in vivo and the proliferation, chemoinvasion, and tubulogenesis of endothelial cell in vitro (6,7). Although these data indicate that Rg1 can be used as a novel therapeutic modality for inducing angiogenesis, for example in wound healing and tissue regeneration, the downstream targets that transmit these proangiogenic effects are not clearly understood.
Angiogenesis is the formation of capillaries as outgrowths from preexisting vasculature; it is a tightly regulated event integral to many physiological and pathological situations, including development, wound healing, and tumor growth (8). Vascular endothelial growth factor (VEGF) is probably the most important angiogenesis inducer because of its potency, selectivity for endothelial cells, and ability to regulate key steps in angiogenesis, including proliferation and migration of endothelial cells (9). Overexpression of VEGF and its receptors promote blood vessel formation, and VEGF inhibition blocks angiogenesis (8). Moreover, a number of other angiogenic cytokines and growth factors act, at least in part, by up-regulating VEGF expression (10).
Recent studies have implicated the Wnt/␤-catenin signaling in vessel development in normal and pathological conditions (11). Stability of ␤-catenin is a critical point in Wnt signaling that is regulated by many cytoplasmic proteins, including glycogen synthase kinase 3␤ (GSK3␤), axin, and adenomatous polyposis coli. In the absence of Wnt, cytoplasmic ␤-catenin is constitutively degraded by the ubiquitin-proteasome pathway. * This work was supported by Research Grant Council, Hong Kong SAR Gov-Wnt signaling inhibits ␤-catenin degradation by inactivation of GSK3␤ as a result of phosphorylation at Ser 9 . As a result, ␤-catenin accumulates in the nucleus, where it interacts with T-cell factor (TCF)/lymphoid enhancer factor (LEF) family transcription factors and regulates Wnt target genes (12). Several studies have demonstrated the expression of Wnt ligands and frizzled (Fz) receptors in vascular cells (11). Mice lacking Wnt-2 or Fz-5 display severe vascular abnormalities, including defective placental vasculature, and are embryonic lethal (13,14). Mutations in human Fz-4 have been linked to familial exudative vitreoretinopathy, a hereditary disorder in which retinal angiogenesis is severely impaired (15). ␤-Catenin is also frequently observed in the cytoplasm and nucleus of vascular cells during angiogenesis and vessel remodeling in disease states (11). The endothelial cell-specific, conditional inactivation of the ␤-catenin gene in mice leads to a defective vascular pattern, ultimately resulting in embryonic lethality (16,17).
In the present study, we sought to characterize the molecular mechanism by which Rg1 mediates VEGF expression in human umbilical vein endothelial cells (HUVEC). We demonstrate a role for Rg1 in the regulation of VEGF expression, and this induction is through the activation of a PI3K/Akt 3 GSK3␤ 3 ␤-catenin-TCF pathway via the GR.

EXPERIMENTAL PROCEDURES
Experimental Reagents-Ginsenoside-Rg1 is a reference compound (purity Ͼ98%) purchased from the Division of Chinese Material Medica and Natural Products, National Institute for the Control of Pharmaceutical and Biological Products, Ministry of Public Health, China. A stock solution of Rg1 (50 mM) was prepared in sterile double distilled H 2 O. Medium 199, endothelial cell growth supplement, heparin, lithium chloride (LiCl), staurosporine, and the polyclonal anti-␤-actin were purchased from Sigma. The PI3K inhibitor, LY294002, and the Akt inhibitor, SH-6, were purchased from Calbiochem. Anti-␤catenin and anti-VEGF were obtained from Transduction Laboratories (Lexington, KY) and Upstate Biotechnology, Inc. (Lake Placid, NY), respectively. Biotinylated phosphatidylinositol 4,5-bisphosphate (PIP 2 ) and the PhosphoSensor kit were purchased from PerkinElmer Life Sciences, whereas growth factor-reduced Matrigel was obtained from BD Biosciences. The polyclonal phospho-specific antibodies to Akt (Ser 473 ) and GSK3␤ (Ser 9 ) and polyclonal anti-Akt and GSK3␤ were purchased from Cell Signaling, Inc. (Austin, TX). Alexa Fluor 488conjugated rabbit IgG was from Molecular Probes, Inc. (Eugene, OR), and peroxidase-conjugated secondary antibodies were purchased from Zymed Laboratories Inc. (San Francisco, CA).
Human ␤-catenin was first amplified from a cDNA library by PCR, and site-directed mutagenesis was then used to generate the activated mutant form of ␤-catenin (S37A). The PCR product was confirmed by DNA sequencing and subcloned into KpnI and BamHI sites of pEGFP expression vector (Clontech, Palo Alto, CA). Small interfering RNA (siRNA) oligonucleotide targeting ␤-catenin (18) and a nonspecific RNA control were obtained from Dharmacon Research (Lafayette, CO). Dominant negative Akt, the TCF-binding site reporter plasmid (TOPFLASH), and a mutated control reporter (FOPFLASH) were purchased from Upstate Biotechnology. Constitutively active and dominant negative constructs of TCF were kindly provided by Dr. B. Gumbiner (University of Virginia, Charlottesville, VA).
Cell Culture and Treatments-HUVEC was obtained from Clonetics (San Diego, CA) and cultured in medium 199 supplemented with 20% fetal bovine serum, 20 g/ml endothelial cell growth supplement, 90 units/ml heparin, and 1% penicillin/ streptomycin/neomycin in a humidified incubator at 37°C with 95% air, 5% CO 2 . HUVEC between passages 2 and 8 were used in these studies to ensure the genetic stability of a culture. For inhibition assays, HUVEC was pretreated with inhibitors (10 M LY294002, 10 M SH-6, 10 mM LiCl, and 10 M RU486) for 30 min before the addition of Rg1. Cell transfections were performed using Lipofectin reagent (Invitrogen) according to the manufacturer's instructions. Cells transfected with siRNA duplex were incubated for 24 h prior to protein level determination or TCF reporter gene assay. To express cDNA construct, 1.5 g of plasmid DNA was used per 35-mm dish.
Enzyme-linked Immunosorbent Assay-VEGF protein in cell lysates and conditioned media were quantified using a commercially available enzyme-linked immunosorbent assay kit (Assay Designs) according to the manufacturer's instructions. Absorbance was measured at 450 nm using a microplate reader (BMG Labtech, Offenburg, Germany). VEGF protein levels were determined by serial dilutions of recombinant VEGF 165 standards assayed at the same time. All experiments were carried out in triplicate.
Cell Fractionation and Western Blot Analysis-The purification of membrane, cytosolic, and nuclear fractions of ␤-catenin was performed using a commercially available kit (Calbiochem) following the manufacturer's instructions. Calpain I and histone H1 antibodies were used to verify the separation of cytosolic and nuclear fractions, respectively. Protein concentration was determined using protein assay reagent from Bio-Rad. An equal quantity of proteins was separated by SDS-PAGE and transferred onto nitrocellulose membrane (Amersham Biosciences). The membrane was then incubated with the indicated primary antibodies for 3 h, followed by horseradish peroxidase-conjugated secondary antibodies, and revealed by an ECL detection system (Bio-Rad). Where indicated, the membranes were stripped and reprobed with another antibody. The density of the bands was quantified by densitometric analysis using Metamorph software.
Reverse Transcription-PCR-Total RNAs from HUVEC were extracted with TRIzol LS reagent (Sigma) and reverse transcribed using Superscript III First-strand Synthesis SuperMix (Invitrogen). The primers for human VEGF were 5Ј-CGA AAC CAT GAA CTT TCT GC-3Ј (forward) and 5Ј-CCT CAG TGG GCA CAC ACT CC-3Ј (reverse), and primers for ␤-actin were 5Ј-GTG GGG CCG CTC TAG GCA C-3Ј (forward) and 5Ј-TTT GAT GTC ACG CAC GAT TT-3Ј (reverse), purchased from Invitrogen. The number of amplification cycles during which PCR product formation was limited by template concentration was determined in pilot experiments. Amplification of human VEGF and ␤-actin cDNA were carried out for 33 and 30 cycles, respectively, with PCR SuperMix (Invitrogen).
Confocal Microscopy-Cells were grown on coverslips and treated with Rg1 in the presence or absence of inhibitors as described above. Cells were then fixed in 4% paraformaldehyde for 15 min. Primary and secondary antibodies were diluted (1:100) in phosphate-buffered saline, 0.1% Triton X-100, and 2% normal goat serum. Coverslips were mounted on glass slides with fluorescence mounting medium (Vector Laboratories, Burlingame, CA) and viewed using a Zeiss LSM-510 multitracking laser-scanning confocal microscope (Frankfurt, Germany).
TOPFLASH Luciferase Reporter Assay-In 6-well plates, cells were transiently transfected with 1.5 g of the TOPFLASH or FOPFLASH reporter plasmid using Lipofectamine (Invitrogen). As a control for transfection efficiency, 0.5 g of the ␤-galactosidase construct was included in each transfection. Cells were harvested 24 h after transfection, and extracts were prepared in 200 l of reporter lysis buffer (Promega). Luciferase and ␤-galactosidase activity were assayed according to the manufacturer's protocol using the luciferase assay kit from Promega. Luciferase activity in each well was normalized to the ␤-galactosidase activity. All experiments were assayed in triplicates, and the assay was performed in three independent experiments.
Assay for PI3K Activity-PI3K activity was determined using the PhosphoSensor kit according to the manufacturer's instructions. This kit measures PI3K activity by quantifying the phosphorylation of PIP 2 . Cellular extracts of HUVEC untreated or treated with Rg1 (150 nM) was incubated for 1 h at room temperature with biotinylated PIP 2 (500 nM) and ATP (3 M). Detection of phosphorylated PIP 2 was performed by using fluorescent streptavidin conjugates and measuring the emission of fluorescence at 520 nm (iEMS analyzer; Labsystems). The addition of the generic protein kinase inhibitor staurosporine (10 M) was used to control for nonspecific background signal.
Tube Formation Assay-HUVEC (1 ϫ 10 5 cells/well) was seeded in a growth factor-reduced Matrigel-coated 24-well plate. Cells were untreated or treated with 150 nM Rg1 or with a combination of Rg1 and various dominant negative constructs or siRNAs for 16 h at 37°C. Images were captured under phasecontrast microscopy (ϫ10) using a CCD camera. Twelve microscopic fields were randomly selected for each well, and the number of tubelike structures per field was counted.
Statistical Analysis-Each experiment was repeated at least three times, with each experiment yielding essentially identical results. Data were expressed as means Ϯ S.D. Statistical comparisons were carried out by one-way analysis of variance, with Tukey's least significant difference t test for post hoc analysis (GraphPad software, San Diego). p Ͻ 0.05 was considered statistically significant.

RESULTS
Ginsenoside Rg1 Induces VEGF Production in HUVEC-Both in vivo and in vitro studies have shown that Rg1 increases neovascularization of endothelial cells (6,7). Because VEGF is known to be a key activator of angiogenesis, we examined whether VEGF is up-regulated by Rg1. We found that VEGF production was significantly elevated in response to Rg1 stimulation as determined by an enzyme-linked immunosorbent assay (Fig. 1). Moreover, VEGF was mainly secreted outside of the cell, since the concentration of VEGF in tissue culture medium was much higher than that in the cell lysate ( Fig. 1). Treatment with increasing concentrations of Rg1 (50, 150, and 300 nM) increased the level of VEGF protein secretion by 38.0-, 41.2-, and 20.7-fold, respectively (Fig. 1A). In addition, the stimulatory effect of Rg1 on VEGF production in HUVEC was timedependent. Elevated levels of VEGF were noted at 6 h, became more evident at 12 h, and peaked at 24 h following 150 nM Rg1 stimulation (Fig. 1B). To confirm these results, Western blot analysis and semiquantitative reverse transcription-PCR for VEGF expression were also carried out using the antibodies and primers described under "Experimental Procedures." The results confirmed the up-regulation of VEGF protein and mRNA after cells were exposed to 150 nM Rg1 for 24 h (Fig. 1C).
Rg1-mediated Up-regulation of VEGF Involves Accumulation of ␤-Catenin-To uncover the molecular mechanisms for increased VEGF expression, we focused on signaling events that converge on ␤-catenin, which has an essential role in the regulation of angiogenesis (11). First, we examined the subcellular distribution of ␤-catenin in Rg1-treated HUVEC by Western blot analysis on fractionated membrane, cytosolic, and nuclear extracts. Although changes in membrane or cytosolic pools of ␤-catenin were not detected by this method, a significant amount of ␤-catenin appeared in the nucleus of HUVEC FIGURE 1. Rg1 induces VEGF production in HUVEC. The conditioned medium and cell lysate were harvested to assay for the VEGF concentrations. A, HUVEC were treated with Rg1 at 50, 150, and 300 nM for 24 h. B, cells were treated with 150 nM Rg1 for 6, 12, and 24 h. C (left), cell lysates and total RNA were prepared at 24 h after exposure to 150 nM Rg1. Cell lysates were subjected to immunoblotting with antibody against VEGF. cDNA was isolated, and reverse transcription (RT)-PCR was performed using VEGF sequence-specific primers. ␤-Actin was used as the internal control. C (right), signal intensities were determined by densitometry. Data points shown represent the mean, with error bars representing S.D. (n ϭ 3). *, p Ͻ 0.05 for Rg1-treated cells versus untreated control. after 1 h of incubation with Rg1 ( Fig. 2A). Equal loading of the membrane, cytosolic, and nuclear fractions were confirmed by the presence of VE-cadherin, calpain I, and histone H1, respectively ( Fig. 2A). To verify this highly intriguing result, we used immunofluorescent microscopy to directly visualize the localization of ␤-catenin. As shown in Fig. 2B, strong staining of ␤-catenin in the nucleus was detected in Ͼ90% of the cells following treatment with Rg1, confirming the data from the fractionation experiment ( Fig. 2A). Together, these data suggest that Rg1 induced rapid nuclear accumulation of ␤-catenin in HUVEC.
Since we observed that Rg1-induced accumulation of ␤-catenin correlated with VEGF expression, we asked whether a reduction of the amount of ␤-catenin by siRNA would affect VEGF expression. Endogenous ␤-catenin protein levels were efficiently and specifically reduced in the presence of a specific siRNA for ␤-catenin (Fig. 3A). No inhibition was observed for nonspecific siRNA (Fig. 3A). Importantly, expression of ␤-catenin siRNA, but not nonspecific siRNA, reverted the effects of Rg1 on VEGF protein expression (Fig. 3B), indicating that ␤-catenin was required for Rg1 stimulation of VEGF. Similar results were obtained by measuring VEGF mRNA levels (Fig. 3C).

␤-Catenin Stimulates VEGF Expression in a TCF-dependent
Manner-In the nucleus, accumulation of ␤-catenin allows interaction with the TCF/LEF transcription factors and up-regulation of target genes. Hence we studied the effect of Rg1 on the transcriptional activity of ␤-catenin-TCF. The luciferase reporters, which have a minimal thymidine kinase promoter and either wild type (TOPFLASH) or mutated (FOPFLASH) binding sites for the ␤-catenin-TCF complex, have been widely used to characterize ␤-catenin-TCF-dependent gene expression (19) These reporter constructs were transfected into HUVEC, and luciferase activity was determined. The adenomatous polyposis coli mutant colorectal cancer cell line SW480 served as a positive control (data not shown). As shown in Fig.  4A, HUVEC had low basal TCF transcriptional activities, and Rg1 treatment increased TOPFLASH activation by ϳ2-fold.
To further test if ␤-catenin-TCF complex formation is necessary for mediating Rg1 effects, we utilized a dominant nega-  Equal loading was confirmed by the presence of ␤-actin. B, cells were transfected with ␤-catenin-siRNA or nonspecific siRNA and post-treated with Rg1 for 24 h. Immunoblot analysis was performed using VEGF antibodies. Equal loading was confirmed by the presence of ␤-actin. C, reverse transcription-PCR was performed using VEGF sequence-specific primers. ␤-Actin was used as the internal control. Right, signal intensities were determined by densitometry. Data points shown represent the mean, with error bars representing S.D. (n ϭ 3). *, significant differences from control, with p Ͻ 0.05. tive mutant of TCF (DN-TCF) that effectively blocks ␤-catenin-mediated TCF-dependent transcription (20). As shown in Fig. 4B, expression of DN-TCF repressed the stimulatory ability of Rg1 upon VEGF expression. On the other hand, expression of an activated mutant of TCF (VP16-TCF) (20) produced similar activation on VEGF expression as Rg1 (Fig. 4B). We also confirmed the effectiveness of these constructs to modulate TCF activity in HUVEC by measuring TOPFLASH reporter activation (Fig. 4A). These findings suggest that the Rg1 effects are attributable to changes in ␤-catenin-TCF nuclear signaling.
Rg1 Treatment Induces Phosphorylation of GSK3␤ in a PI3K/ Akt-dependent Manner-The results of Fig. 2 indicated that ␤-catenin accumulation in the nucleus was rapid (ϳ1 h) in response to Rg1 treatment, suggesting that the Rg1-induced ␤-catenin nuclear translocation in HUVEC is likely to be due to ␤-catenin stabilization rather than transcription induction. GSK3␤ is known to phosphorylate ␤-catenin on N-terminal serine residues, targeting ␤-catenin for rapid ubiquination and proteasome-mediated degradation (21). Conversely, inactivation of GSK3␤ by phosphorylation promotes ␤-catenin protein stability and transactivation activity. We therefore investigated whether Rg1-mediated ␤-catenin up-regulation involves changes in GSK3␤ phosphorylation. As shown in Fig. 5A, the level of Ser 9 phosphorylation of GSK3␤ was increased significantly after Rg1 treatment. Notably, Rg1 stimulation caused a parallel increase in the cellular level of ␤-catenin and subsequent VEGF expression (Fig. 5A). Treatment with LiCl, a pharmacological agent well known for its inhibitory effect on GSK3␤ activity (22,23), enhanced the cellular level of ␤-catenin and VEGF expression (Fig. 5A). Next, we asked whether expression of a constitutively active mutant of ␤-catenin (S37A) that is refractory to proteasomal degradation (24) could increase VEGF. Fig. 5B shows that the expression of S37A ␤-catenin increased ␤-catenin levels (2.7-fold), TOPFLASH activity (1.8fold; data not shown), and VEGF expression (2.3-fold) to similar levels as Rg1 treatment, lending additional support for GSK3␤dependent action of Rg1.
Because GSK3␤ can be inactivated through phosphorylation at its Ser 9 residue via a PI3K-dependent activation of Akt, such events have the potential of increasing the amount of free ␤-catenin (25). Fig. 6A shows that Rg1 stimulated a time-dependent increase in PI3K activity, with a maximal response by 15 min. The Rg1-induced increase of PI3K activity was specific, because it was abrogated by the PI3K inhib-  . GSK3␤ plays a role in Rg1-induced ␤-catenin stabilization and VEGF expression. A, total cell lysates were prepared from HUVEC after treatment with Rg1 or LiCl. Western blots were probed with antibodies against phospho-GSK3␤ (p-GSK3␤) (Ser 9 ), GSK3␤, ␤-catenin, or VEGF. Equal loading was confirmed by the presence of ␤-actin. B, cells were treated or untreated with 150 nM Rg1 or transiently transfected with a constitutively active ␤-catenin S37A. The Western blots were carried out using antibodies against ␤-catenin or VEGF. Reverse transcription (RT)-PCR was performed using VEGF sequence-specific primers. ␤-Actin was the internal control. Right, signal intensities were determined by densitometry. Data points shown represent the mean, with error bars representing S.D. (n ϭ 3). *, significant differences from control, with p Ͻ 0.05. itor LY294002 (data not shown). By activating PI3K, Rg1 promoted phosphorylation of GSK3␤, accumulation of ␤-catenin, and VEGF expression, which were prevented by pharmacological inhibition of PI3K (LY294002) or Akt activities (SH-6) (Fig. 6B). In a different approach, we blocked the PI3K/Akt signaling pathway by expression of a dominant negative mutant of Akt (DN-Akt). Transfection of DN-Akt expression vector significantly inhibited Rg1-induced phosphorylation of GSK3␤ and subsequent ␤-catenin and VEGF expression (Fig. 6B). There was also a clear reduction of Rg1induced nuclear accumulation of ␤-catenin by inhibition of PI3K/Akt, as revealed by confocal microscopy (Fig. 6C). In contrast, the MEK1 inhibitor PD98059 did not have any inhibitory effect (Fig. 6). Together, these data suggest the potential involvement of PI3K/Akt and GSK3␤, but not ERK1/2, in Rg1-induced ␤-catenin activation and subsequent VEGF expression in HUVEC.

VEGF Is Involved in Rg1-induced
Angiogenic Actions in HUVEC-To examine whether Rg1-mediated induction of VEGF plays a role in the angiogenic activity of HUVEC, we used a tube formation assay. This is the most simple and classic angiogenesis assay in vitro and has been described as reminiscent of the multistep process of angiogenesis involving cell adhesion, migration, differentiation, and growth (26). As shown in Fig. 7, endothelial cells aligned and formed tube-like structures on Matrigel, and Rg1 treatment (150 nM) caused a significant increase in tube formation. The angiogenic effect of Rg1 was completely reversed by the addition of VEGF-neutralizing antibody (p Ͻ 0.05) but not control IgG, confirming that Rg1 actions were VEGF-dependent (Fig. 7). Importantly, inhibition of PI3K/Akt by LY294002 or DN-Akt or inhibition of ␤-catenin by ␤-catenin siRNA markedly attenuated Rg1-stimulated tube formation (Fig. 7). No inhibitory effect on tube formation was observed for nonspecific siRNA. These data indicate a role for VEGF in the angiogenic activity of Rg1 in HUVEC, and the effect was mediated via PI3K/ Akt and ␤-catenin.
Rg1 Effects on ␤-Catenin Require Glucocorticoid Receptor-Rg1 signals have been shown to be transduced via the GR (4,5). To investigate the possible role of GR in the course of Rg1-induced angiogenic effect, we used the specific GR antagonist RU486 (10 M) to block GR-related pathways. The addition of RU486 completely abolished Rg1-induced PI3K activity (Fig. 8A) as well as Akt and GSK3␤ phosphorylation (Fig. 8B), which is a critical step in Rg1-induced ␤-catenin activation and subsequent VEGF up-regulation. In support of this observation, RU486 also inhibited ␤-catenin nuclear accumulation, as revealed by Western blot analysis and immunofluorescent analysis (Fig. 8, C and D), and completely aborted the effect of Rg1 on VEGF expression (Fig. 8E). These studies suggest that the modulation of angiogenesis-related signaling by Rg1 is GR-mediated.

DISCUSSION
Previous study has revealed a proangiogenic role for Rg1 (6,7). Here we outline a plausible mechanism whereby Rg1-induced angiogenesis of vascular endothelial cells can be explained. We provide the first evidence for a role of Rg1 in the induction of VEGF, a critical mediator of angiogenesis, in HUVEC. Furthermore, our data suggest that PI3K/Akt and GSK3␤ are signaling molecules necessary for the Rg1-mediated up-regulation of ␤-catenin, its translocation into the nucleus, and changes in VEGF expression in these cells. In support of our data, the PI3K/Akt signaling pathway has been shown to be required for the angiogenic phenotype of Rg1 (7), indicating the functional significance of angiogenic signaling downstream of PI3K/Akt. The concentration of Rg1 (150 nM) used in this study is the optimal stimulatory dose of VEGF in HUVEC. A lower dose is also effective in VEGF induction (Fig. 1). Pharmacokinetics of Rg1 have not been determined in humans. However, the dosage (150 nM; 0.12 g/liter) utilized in this study is similar to what is observed in consuming 0.65-3.65 g of Rg1, based on the oral bioavailability of 3.29 -18.4% for Rg1 in animal studies (27,28).
Increased levels of free ␤-catenin result in an increased translocation of this protein to the nucleus, where its main task is to activate the TCF/LEF family of transcription factors (29). The present study suggests that ␤-catenin activity is an important signaling pathway contributing to VEGF induction by Rg1 in HUVEC. This conclusion is based on several lines of evidence: (a) Rg1 increased ␤-catenin nuclear accumulation and influenced TCF/LEF transcriptional activity; (b) transfection with a stable form of ␤-catenin (S37A) or treatment of cells with GSK3␤ inhibitor to stabilize ␤-catenin resulted in the production of augmented levels of VEGF mRNA and protein; and (c) a reduction in these levels was seen when the cells were transfected with ␤-catenin siRNA. These observations are consistent with the transcriptional data, which showed that VEGF promoter activity could be stimulated in response to stabilized ␤-catenin in several in vitro systems, including vascular endothelial cells (30 -32). Moreover, the presence of ␤-catenin-TCF binding motifs on the human VEGF promoter could further explain its activation by ␤-catenin nuclear signaling (30).
The observation that VEGF is upregulated by the ␤-catenin-TCF pathway in HUVEC provides a molecular basis for the angiogenesis observed after Rg1 treatment. VEGF is a pivotal mitogen that mediates endothelial cell proliferation and survival. The angiogenic potency of VEGF was confirmed in several in vitro and in vivo model systems, including wound healing, tumor xenografts, Matrigel plug assay, and tube-like structure formation. Moreover, mice lacking VEGF exhibit defective vascularization in most organs and are embryonic lethal (8). VEGF binds to a family of class III tyrosine kinase receptors, including Fms-like tyrosine kinase 1 (Flt-1 or VEGFR-1) and kinase insert domain-containing receptor (Flk-2 or VEGFR-2) (32). Recent works have demonstrated the co-expression of VEGF and its receptors in HUVEC, suggesting the presence of an autocrine mechanism of action to modulate angiogenesis (33). Therapeutic approaches that stimulate or inhibit VEGF have produced encouraging clinical results in treating disease resulting from insufficient (e.g. chronic wound) or excessive blood vessel formation (e.g. cancer) (8). Thus, it is rational to explore Rg1 as a source of a novel angiomodulator.
In this report, we also show that the transcriptional activator ␤-catenin is a novel target of Rg1 signaling. We showed that Rg1 increased ␤-catenin protein levels as well as ␤-catenin-dependent transcription. Moreover, this stimulation involves a signaling cascade consisting of PI3K/Akt/GSK3␤, suggesting crosstalk between the PI3K/Akt and ␤-catenin signaling pathways, and GSK3␤ is the convergent point of these two signaling path- Bottom, data are shown as mean Ϯ S.D. of the number of tube-like structure relative to the control, which was set to be 100%, counted in 12 microscopic fields of three independent experiments. Scale bar, 50 m. *, significant differences from untreated control, with p Ͻ 0.05.
ways. Studies in several cell types have shown that ␤-catenin is rapidly degraded by the proteasome in unstimulated cells via a GSK3␤-dependent mechanism (21). Our finding that the proteasome inhibitor MG132 (data not shown) and inhibitor of GSK3␤ LiCl (Fig. 4C) both increased cellular ␤-catenin levels in HUVEC argues that proteasome-mediated degradation keeps ␤-catenin levels low in unstimulated cells. In addition, Rg1 caused an increase in the levels of ␤-catenin in the nuclei within 1 h, which is more consistent with the kinetics of inhibition of ␤-catenin degradation as opposed to an increase in transcription. It is noteworthy that Rg1, similar to the Wnt signal (12), can activate the ␤-catenin-TCF pathway. Unlike Wnt, Rg1 stimulates PI3K and Akt, which are not downstream effectors of the Wnt signal. However, the involvement of PI3K/Akt in the regulation of the ␤-catenin signaling pathway has been described in other cell systems (25,34,35). Although the mechanism of ␤-catenin activation could also be mediated through a mitogen-activated protein kinase pathway (36), we ruled out this possibility in HUVEC, since treatment of cells with a specific inhibitor of MEK1 (PD98059) had no effect on TCF activity (data not shown). Nor did it alter the level of ␤-catenin (Fig. 6).
The mechanism by which Rg1 activates PI3K/Akt is not yet fully understood, but Rg1 has been shown to be a functional ligand of GR (4,5). Experiments using the steroid antagonists and siRNA provide further evidence that the action of Rg1 is mediated through GR but not estrogen receptor or progesterone receptor (5). Our findings that Rg1 rapidly induced activation of PI3K/Akt (within minutes) suggest a nongenomic rather than genomic action. Although the rapid nongenomic effects of GR have attracted increasing attention in recent years, the underlying mechanisms of action are not yet clear. One possibility is through a membrane-bound GR. However, the existence of functional GR associated with the plasma membrane has been debated. Membrane-bound GR have so far been detected only in amphibian brain and on immune cells, although specific nongenomic effects are widely observed in humans (37). Another intriguing hypothesis to explain the rapid effects is through the cytosolic GR, which not only causes classical genomic but also rapid nongenomic effects resulting in interaction with signaling processes, including PI3K/Akt (37,38). Our recent studies (5) showing that Rg1 can exert its genomic effects by binding to GR ligand binding domain and that Rg1 is also able to trigger rapid nongenomic activation of PI3K/Akt through a GR-dependent, transcription-independent mechanism are highly suggestive of this view. It is noteworthy that unlike many other members of the nuclear receptors that are predominantly localized in the nucleus, the majority of GR is located in the cytoplasm and stabilized by chaperones (37). Activation of PI3K/Akt is a key step for diverse biological effects, including cellular proliferation, cell cycle progression, viability, motility, invasion, and neovascularization. We show that by activating the PI3K/Akt pathway, Rg1 promotes the FIGURE 8. Rg1 signals via GR in HUVEC. Cells were pretreated with the GR antagonist RU486 (RU) (10 M) for 30 min before Rg1 treatment. A, PI3K activity was determined using a PhosphoSensor kit by quantifying the phosphorylation of PIP 2 as described under "Experimental Procedures." B, the Western blots were carried out using antibodies against phospho-Akt (p-Akt), Akt, phospho-GSK3␤ (p-GSK3␤), or GSK3␤. C, nuclear fractions from cells treated with Rg1 in the presence or absence of RU486 were blotted with the anti-␤-catenin antibody. Histone H1 served as loading control. D, ␤-catenin localization was detected by immunofluorescence with the anti-␤-catenin antibody. Micrographs were captured with a confocal microscope at ϫ63 magnification. N, nucleus. Scale bar, 20 m. E, Western blot was performed with lysates from above using antibodies against VEGF. Equal loading was confirmed by the presence of ␤-actin. Right, signal intensities determined by densitometry. Data points shown represent the mean, with error bars representing S.D. (n ϭ 3). *, significant differences from control, with p Ͻ 0.05. synthesis of VEGF and hence angiogenic activity. Whether Rg1 may play a role in other cellular functions via the P13k/Akt pathway warrants further investigation.
In summary, the findings reported here provide novel information about the mechanisms utilized by Rg1 to control ␤-catenin signaling pathways that culminate in the production of VEGF. The results shed light on the mechanism of action of Rg1 and also provide a further insight into the possible therapeutic use of Rg1 for inducing angiogenesis, such as in wound healing and tissue regeneration.