NADPH Oxidase 4 Mediates Insulin-Stimulated HIF-1α and VEGF Expression, and Angiogenesis In Vitro

Acute intensive insulin therapy causes a transient worsening of diabetic retinopathy in type 1 diabetes patients and is related to VEGF expression. Reactive oxygen species (ROS) have been shown to be involved in HIF-1α and VEGF expression induced by insulin, but the role of specific ROS sources has not been fully elucidated. In this study we examined the role of NADPH oxidase subunit 4 (Nox4) in insulin-stimulated HIF-1α and VEGF expression, and angiogenic responses in human microvascular endothelial cells (HMVECs). Here we demonstrate that knockdown of Nox4 by siRNA reduced insulin-stimulated ROS generation, the tyrosine phosphorylation of IR-β and IRS-1, but did not change the serine phosphorylation of IRS-1. Nox4 gene silencing had a much greater inhibitory effect on insulin-induced AKT activation than ERK1/2 activation, whereas it had little effect on the expression of the phosphatases such as MKP-1 and SHIP. Inhibition of Nox4 expression inhibited the transcriptional activity of VEGF through HIF-1. Overexpression of wild-type Nox4 was sufficient to increase VEGF transcriptional activity, and further enhanced insulin-stimulated the activation of VEGF. Downregulation of Nox4 expression decreased insulin-stimulated mRNA and protein expression of HIF-1α, but did not change the rate of HIF-1α degradation. Inhibition of Nox4 impaired insulin-stimulated VEGF expression, cell migration, cell proliferation, and tube formation in HMVECs. Our data indicate that Nox4-derived ROS are essential for HIF-1α-dependent VEGF expression, and angiogenesis in vitro induced by insulin. Nox4 may be an attractive therapeutic target for diabetic retinopathy caused by intensive insulin treatment.


Introduction
Acute intensive insulin therapy causes a transient worsening of diabetic retinopathy with neovascularization in type 1 diabetes patients and is related to VEGF expression [1]. Insulin-induced VEGF expression has been reported to be mediated through the activation of PI 3-kinase/AKT and p44/42 MAPK pathways, and hypoxia-inducible factor-1a (HIF-1a) [2]. Insulin has been shown to increase intracellular reactive oxygen species (ROS) production in many cell types such as HepG2 cells [3], 3T3L1 adipocytes [4], human fibroblasts [5], and retinal microvascular endothelial cells [6]. ROS produced by insulin are involved in HIF-1a and VEGF expression [3,7]. However, the specific ROS sources that mediate insulin-induced HIF-1a and VEGF expression are not completely defined.
ROS such as superoxide, hydrogen peroxide, and hydroxyl radical are involved in the signaling pathways that mediate many stress and growth responses, including angiogenesis [8][9][10]. Noxfamily NADPH oxidases have proven to be a major source of ROS production in various cell types and have crucial roles in various physiological and pathological processes [11][12][13][14]. We showed previously that NADPH oxidase subunit 4 (Nox4)-derived ROS generation in response to insulin-like growth factor-1 (IGF-I) was linked to cell proliferation and migration in vascular smooth muscle cells (VSMCs) [15][16]. There is evidence that the level of Nox4 expression in endothelial cells in vivo is high as compared with other Nox isoforms [17]. Nox4 is important for ROS generation in human endothelial cells [18]. Moreover, Nox4 mediates endothelial responses to angiotensin-II [19], transforming growth factor-b [20] and epidermal growth factor [21], which include migration and proliferation. These data suggest that Nox4 has a critical role in modulating endothelial cell physiology.
In this study we examined the role of Nox4 in insulin-stimulated HIF-1a and VEGF expression, and angiogenic responses in human microvascular endothelial cells (HMVECs). Here we demonstrated that Nox4 mediates insulin-stimulated ROS generation, and angiogenesis in vitro, Nox4-derived ROS are essential for HIF-1a-dependent VEGF expression induced by insulin.

Cell culture
Human microvascular endothelial cells (HMVECs) immortalized with the human telomerase catalytic protein (hTERT) [22] were a kind gift from Dr. Rong Shao (University of Massachusetts, Springfield, MA). The cells were cultured as described previously [22].

Measurement of ROS
DCF-DA was used to measure ROS as described [16]. Growtharrested HMVECs were stimulated with 100 nM insulin for 5 minutes. After stimulation, HMVECs were incubated in phosphate buffered saline (PBS) containing DCF-DA (5 mM) for 15 minutes in a light-protected humidified chamber at 37uC. Cells were then rinsed in PBS and fixed with 10% buffered formalin. Images were obtained with a LSM 510 laser scanning confocal microscope (Carl Zeiss Inc., Thornwood, NY). The excitation was at 488 nm with an emission at 540 nm. The fluorescence image was collected by a single rapid scan with identical parameters for all samples. Fluorescent levels were expressed as percent increase over the control. The assays were carried out in duplicates, and data were obtained from three independent experiments.

Small interfering RNA (siRNA) transfection
Non-silencing (control) siRNA (100 nM) or Nox4 siRNA (100 nM) was transfected into HMVECs using the RNAi MAX transfection reagent according to the manufacturer's instructions (Invitrogen). The non-silencing siRNA was used as a negative control. 48 hours after transfection, the cells were starved for 16 hours followed by treatment with or without insulin.

Transient transfection and luciferase assay
Transient transfection and luciferase assay were performed as described previously [23]. A 2.65 kb human VEGF gene promoter in pGL2-basic luciferase reporter and a 46 bp human VEGF gene promoter containing the HIF-1 binding site in pGL2-basic luciferase reporter were kindly provided by Dr. Jiang (Thomas Jefferson University, USA). The human wild-type Nox4 cDNA Nox4 Regulates Insulin-Induced HIF-1a and VEGF PLOS ONE | www.plosone.org was cloned into the pcDNA3.1 vector. HMVECs were seeded in 12-well plates and cultured to 60-70% confluence. The cells were transiently transfected with VEGF reporter or pGL2-basic luciferase reporter and pCMV-b-galactosidase plasmid using Lipofectamine 2000 from Invitrogen (Carlsbad, CA). The transfected cells were cultured for 18-20 hours, and then incubated in serum-free medium for 12 hours followed by the incubation with insulin (100 nM) for 20 hours. Cells were lysed with reporter lysis buffer. Luciferase (Luc) activities of the cell extracts were determined using the Luciferase Assay System from Promega (Madison, WI). b-Galactosidase (b-gal) activity was measured. Relative Luc activity was calculated as the ratio of Luc/ b-gal activity, and expressed as the ratio of VEGF/pGL2-basic. All luciferase assays reported here represent at least three independent experiments, each consisting of three wells per transfection.

Quantification of VEGF protein
Human VEGF-A was quantified in the cell culture supernatant using ELISA assay kit from R&D Systems, Inc. (Minneapolis, MN). In brief, HMVECs were transfected with 100 nM Nox4 siRNA or control siRNA for 24 hours, the cells were growth arrested and then treated with insulin (100 nM) for 24 hours. The supernatants were collected, and total cellular protein was measured by protein assay kit (Bio-Rad Laboratories, Hercules, CA). VEGF-A production was expressed as pg/mg protein. The assays were carried out in duplicates, represent three independent experiments.

Cell migration assay
Migration of HMVECs was determined by using transwell chambers (24-well plates, 8-mm pore size, Corning, Inc.) as described previously [24]. The top well was loaded with 1610 5 cells and insulin of 100 nM was added in the lower chamber to induce chemotaxis. After 10 hours of incubation, the cells were fixed and stained with 0.1% crystal violet. The number of cells migrated to the bottom side was counted under a light microscope. The assays were carried out in triplicates, represent three independent experiments.

Cell proliferation assay
Cell proliferation was measured using a CCK-8 assay kit (Dojindo, Japan) as described previously [25]. HMVECs were transfected with the control or Nox4 siRNA for 24 hours, and growth arrested for 16 hours. The cells were then seeded in 96well plate with insulin (100 nM) treatment for 24 hours. CCK-8 solution (150 ml) was added to each well followed by incubation for 2 hours. After incubation, the absorbance was measured at 450 nm using a microplate reader. The assays were measured in sextuplicates, represent three independent experiments.

In vitro angiogenesis assay
In vitro endothelial cell tube formation assay in three dimensional Matrigel culture was used as a surrogate assay for angiogenic potential. 200 ml of growth factor-reduced Matrigel was preplaced in a 48-well plate. HMVECs were preincubated with serum-free medium for 16 hours. The cells (1.5610 4 ) treated with insulin (100 nM) were transferred onto 48-well Matrigel. Endothelial tube formation was evaluated after incubation for 12 hours. The capillary-like tube structures were examined using an inverted phase contrast microscope, and the tube formation index was determined by measuring the average length of tubes in

Statistical analysis
Data are expressed as mean6SE from three independent experiments. Differences among groups were determined using prism software with one-way analysis of variance followed by the Newman-Keuls test, differences between two groups were assayed by two-tailed student t-test. The 0.05 level of probability was used as the criterion of significance.

Insulin increases ROS production through a Nox4-based NADPH oxidase
We measured ROS generation using DCF-DA, a membranepermeable dye that can be oxidized by intracellular ROS to the fluorescent product 29,79-dichlorodihydrofluorescein (DCF). The exposure of HMVECs to insulin (100 nM) for 5 minutes increased intracellular ROS generation, the flavoenzyme inhibitor diphenylene iodonium (DPI) inhibited insulin-derived ROS generation ( Fig. 1A). Quantitative RT-PCR was performed to identify the Nox isoforms of NADPH oxidase in HMVECs. Nox4 was highly expressed in HMVECs, and the expression of Nox4 markedly exceeded that of Nox2 (Fig. 1B), which is consistent with the previous report [26]. Nox1 expression was detected at very low level only (ct value of 37). Nox3 and Nox5 were undetectable in HMVECs.
We next determined the role of Nox isoforms in insulinstimulated ROS production. None of Nox isoforms protein expressions were changed by exposure to insulin for 5 minutes or 2 days (data not shown). Considering the predominant expressions of Nox4 and Nox2 in HMVECs, we used specific siRNA-mediated gene silencing to modulate Nox4 or Nox2 expression. Transfection of Nox4 siRNA resulted in a 50% decrease of Nox4 protein expression in HMVECs without decreasing Nox2 expression (Fig. 1C). ROS generation by insulin was significantly inhibited by Nox4 siRNA, whereas Nox2 siRNA had no significant effect on insulin-induced ROS production (Fig. 1D). These results show that insulin increases ROS production via a Nox4-based NADPH oxidase in HMVECs.

Nox4 is involved in insulin-stimulated insulin receptor and IRS tyrosine phosphorylation, ERK1/2, and AKT activation
To explore the role of Nox4 in the insulin signaling pathway in HMVECs, we measured tyrosine phosphorylation of the insulin receptor and IRS-1. Insulin stimulated an increase in the tyrosine phosphorylation of insulin receptor and IRS-1 in cells transfected with the control siRNA. In cells transfected with the Nox4-specific siRNA, however, insulin-stimulated insulin receptor b tyrosine phosphorylation was decreased by 35% (P,0.01), and IRS-1 tyrosine phosphorylation was decreased by 32% (P,0.05) compared to the control response ( Fig. 2A and 2B). Importantly, downregulation of Nox4 expression did not decrease the serine phosphorylation of IRS-1 induced by insulin (Fig. 2B). We next assessed the role of Nox4 in insulin-stimulated ERK1/2 and AKT activation. As shown in Fig. 2C and 2D, insulin increased the phosphorylation of ERK1/2 and AKT in cells transfected with the control siRNA, whereas the activation of ERK1/2 and AKT by insulin was significantly inhibited by Nox4 siRNA (P,0.05). As the MAPK phosphatase-1 (MKP-1) mediates the dephosphorylation and inactivation of ERK1/2 [27] and SHIP-1 (Src homology 2 domain-containing inositol 59-phosphatase 1) negatively regulates the activation of AKT [28], we tested the effect of Nox4 silencing on MKP-1 and SHIP-1 expression. Importantly, downregulation of Nox4 had little effect on the expression of these phosphatases including MKP-1 and SHIP-1 (Fig. 2E).

Nox4 mediates insulin-induced VEGF transcriptional activation through HIF-1
We investigated the potential involvement of Nox4 in the transcriptional activation of VEGF induced by insulin in HMVECs. HMVECs were transfected with a VEGF promoter reporter containing a 2.65-kb human VEGF promoter. Inhibition of ROS production by DPI or Nox4 siRNA inhibited insulininduced VEGF transcriptional activation (P,0.01, Fig. 3A and  3B). To test whether the HIF-1 binding site at the VEGF promoter is important for Nox4-mediated VEGF transcriptional activation, a VEGF reporter containing a functional promoter fragment with the HIF-1 binding site was used. Inhibition of Nox4 expression also inhibited this VEGF reporter activity induced by insulin (P,0.01, Fig. 3C), indicating that Nox4 mediates insulin-induced VEGF transcription through HIF-1 activation. In addition, overexpression of wild-type Nox4 enhanced the VEGF reporter activity (Fig. 3D). Compared to cells transfected with the control vector, insulin-stimulated VEGF transcriptional activation was increased by 115% (P,0.01) in cells overexpressing wild-type Nox4 (Fig. 3D).

Nox4 mediates insulin-induced HIF-1a and VEGF mRNA and protein expression
To determine whether Nox4 is involved in HIF-1a and VEGF expression, the mRNA and protein levels of HIF-1a and VEGF were measured in HMVECs transfected with Nox4-specific or control siRNA. Insulin increased the mRNA and protein levels of HIF-1a, which were inhibited by Nox4 siRNA (Fig. 4A and 4B). To examine the molecular mechanism of HIF-1a protein inhibition by Nox4 siRNA, we next investigated whether silencing Nox4 alters the stability of HIF-1a. Cycloheximide (CHX), a general inhibitor of protein synthesis, was used to prevent HIF-1a de novo synthesis. HMVECs treated with CHX in combination with insulin displayed a time-dependent decrease of HIF-1a protein level (Fig. 4C). Knockdown of Nox4 expression did not significantly change the rate of HIF-1a degradation (Fig. 4C). Moreover, downregulation of Nox4 had little effect on HIF prolyl hydroxylase domin 2 (PHD2) protein expression (Fig. 4B). Similarly, Inhibition of Nox4 expression also inhibited insulininduced VEGF mRNA and protein expression (Fig. 5A and 5B).

Nox4 mediates insulin-induced HMVEC migration, proliferation and angiogenesis in vitro
To examine the role of Nox4 in insulin-induced HMVEC migration, proliferation and angiogenesis in vitro, we knocked down Nox4 Regulates Insulin-Induced HIF-1a and VEGF PLOS ONE | www.plosone.org Nox4 using siRNA in HMVECs. We observed that inhibition of Nox4 expression decreased insulin-induced cell migration (P,0.05, Fig. 6A) and cell proliferation (P,0.05, Fig. 6B). The functional effect of Nox4 on angiogenic potential was tested by three dimensional Matrigel tube forming assays. The results revealed that knockdown of Nox4 expression resulted in significant inhibition of insulin-induced tube formation (P,0.01, Fig. 6C and  6D). Of note, compromise in Matrigel tube formation of Nox4 knockdown could be significantly corrected by administration of exogenous H 2 O 2 to cells (P,0.05, Fig. 6C and 6D).

Discussion
Previous studies have shown that ROS are involved in insulin signaling pathways [3,7]. However, the role of Nox4 in insulinstimulated HIF-1a and VEGF expression, and angiogenic responses remains unknown. In this study, we provided evidence that Nox4-derived ROS have an important role in activating ERK1/2 and AKT, modulating HIF-1a-dependant VEGF expression, and angiogenic responses induced by insulin.
In this study, we showed that insulin-induced ROS production was reduced by the flavoenzyme inhibitor DPI or Nox4 siRNA, but it was not affected by Nox2 siRNA (Fig. 1). These data suggest that Nox4 is involved in the generation of ROS by insulin in HMVECs. Similar findings were reported previously for insulin in adipocytes [4,29]. Indeed, Nox4 is constitutively active and ROS formation by Nox4 is linked to its expression [30][31]. We found that a short (e.g. 5 minutes) exposure to insulin had no effect on the expression of Nox4 or other Nox isoforms (data not shown), but it enhanced ROS generation. It was known that Nox4/ p22phox complex formation is required for Nox4-derived ROS [30]. Recent study has demonstrated that IGF-I stimulated Nox4/ p22phox complex formation within 5 minutes [32]. The exact mechanism of acute insulin-induced ROS production still needs to be identified.
We found that knockdown of Nox4 expression reduced the tyrosine phosphorylation of IR-b and IRS-1, but did not change the serine phosphorylation of IRS-1 in HMVECs (Fig. 2), suggesting that insulin-stimulated tyrosine phosphorylation of Nox4 Regulates Insulin-Induced HIF-1a and VEGF PLOS ONE | www.plosone.org insulin receptor and its major receptor substrate is regulated by Nox4. These findings are consistent with the previous observation by Biswas and colleagues, which demonstrated that insulininduced Nox4-linked generation of ROS activated insulin receptor tyrosine kinase by inhibiting protein-tyrosine phosphatase 1B (PTP1B) in adipocytes [4]. Moreover, we found that inhibition of Nox4 had little effect on the expression of the phosphatases such as MKP-1 and SHIP. Our results are in contrast to that study in human preadipocytes, which demonstrated that downregulation of Nox4 attenuated MKP-1 expression, and promoted ERK1/2 activation and cell proliferation in response to insulin [33]. Differences between our findings and that study may be due to the differences in the cells used. Furthermore, we found that Nox4 gene silencing had a much greater inhibitory effect on insulininduced AKT activation than ERK1/2 activation (Fig. 2C and  2D). It has been shown that PI3K/AKT pathway is essential for retinal angiogenesis in the zebrafish [34]. Thus, our data suggest that Nox4 plays a more important role in the PI3K/AKT pathway than in the MAPK pathway for insulin signaling.
In this study, knockdown of Nox4 expression inhibited the transcriptional activity of VEGF reporter containing the HIF-1 binding site, indicating that Nox4 mediates insulin-induced VEGF transcription activation through HIF-1. Overexpression of Nox4 alone was sufficient to increase VEGF transcriptional activity, and Nox4 enhanced insulin-stimulated the activation of VEGF (Fig. 3). In addition, we also found that insulin-induced HIF-1a and VEGF expression were blocked by PI3-kinase inhibitor LY294002 and MEK1 inhibitor U0126 (data not shown). Thus, our findings suggest that Nox4-derived ROS regulate insulin-induced HIF-1a and VEGF expression through PI3K/AKT and ERK1/2 pathways.
Our findings demonstrated that insulin stimulated HIF-1 a mRNA and protein expression. In deed, consistent with our observation, it was recently demonstrated that insulin induced HIF-1 a expression in mRNA and protein level in adipocytes [35].
Here, we showed that inhibition of Nox4 impaired insulinstimulated mRNA and protein expression of HIF-1a (Fig. 4A an 4B), suggesting that Nox4-derived ROS are required by insulin in the stimulation of HIF-1a mRNA and protein expression. Moreover, we found that knockdown of Nox4 expression did not change the expression of PHD2 and the rate of HIF-1 a degradation ( Fig. 4B and 4C), indicating that HIF-1 a repression by silencing Nox4 is not through the decrease of stabilization of HIF-1 a protein. This is in line with the previous observation in 786-O renal carcinoma cells, which demonstrated that knockdown of Nox4 by siRNA decreased HIF-2 a mRNA and protein levels independent of pVHL, suggesting that Nox4-derived ROS regulate HIF-2 a by a translational mechanism [36]. Similarly, in the present study, Nox4 regulates the PI3K/AKT signaling, which is important for the translation of HIF-1a protein expression. Thus, it is likely that Nox4 mediates insulin-stimulated HIF-1 a expression through a transcriptional and translational mechanism in HMVECs. In the present study, we have demonstrated that Nox4-derived ROS are essential for HIF-1adependent VEGF expression induced by insulin. In contrast to our findings, a previous study showed that Nox3-derived ROS regulated insulin-induced VEGF through the activation of Sp1, but not by HIF-1a in HepG2 cells [37]. It suggests that insulinstimulated ROS modulate VEGF expression through multiple pathways, depending on different Nox isoform involved.
We found that inhibition of Nox4 impaired insulin-stimulated cell migration, cell proliferation, and tube formation in HMVECs, and the impaired tube formation by knockdown of Nox4 can be corrected by exogenous H2O2. These results support an involvement of Nox4-derived ROS in insulin-induced angiogenesis in vitro. In fact, Nox4 has been associated with angiogenesis, overexpression of Nox4 was sufficient to promote endothelial proliferation, migration, and tube formation [6]. A recent study has shown that Nox4 (2/2) mice exhibited attenuated angiogenesis, further confirming the important role of Nox4 in angiogenesis [38].
In summary, the data presented here demonstrated that Nox4 is a critical mediator of insulin-mediated ROS generation, HIF-1a and VEGF expression, endothelial cell migration, and angiogenesis in vitro. Nox4 may be an attractive therapeutic target for diabetic retinopathy caused by intensive insulin treatment.