Dioxin Receptor Deficiency Impairs Angiogenesis by a Mechanism Involving VEGF-A Depletion in the Endothelium and Transforming Growth Factor-β Overexpression in the Stroma*

Angiogenesis has key roles in development and in the progression of human diseases such as cancer. Consequently, identifying the novel markers and regulators of angiogenesis is a critical task. The dioxin receptor (AhR) contributes to vascular homeostasis and to the endothelial response to toxins, although the mechanisms involved are largely uncharacterized. Here, we show that AhR-null mice (AhR−/−) have impaired angiogenesis in vivo that compromises tumor xenograft growth. Aortic rings emigration experiments and RNA interference indicated that AhR−/− endothelial cells failed to branch and to form tube-like structures. Such a phenotype was found to be vascular endothelial growth factor (VEGF)-dependent, as AhR−/− aortic endothelial cells (MAECs) secreted lower amounts of active VEGF-A and their treatment with VEGF-A rescued angiogenesis in culture and in vivo. Further, the addition of anti-VEGF antibody to AhR+/+ MAECs reduced angiogenesis. Treatment under hypoxic conditions with 2-methoxyestradiol suggested that HIF-1α modulates endothelial VEGF expression in an AhR-dependent manner. Importantly, AhR-null stromal myofibroblasts produced increased transforming growth factor-β (TGFβ) activity, which inhibited angiogenesis in human endothelial cells (HMECs) and AhR−/− mice, whereas the co-culture of HMECs with AhR−/− myofibroblasts or with their conditioned medium inhibited branching, which was restored by an anti-TGFβ antibody. Moreover, VEGF and TGFβ activities cooperated in modulating angiogenesis, as the addition of TGFβ to AhR−/− MAECs further reduced their low basal VEGF-A activity. Thus, AhR modulates angiogenesis through a mechanism requiring VEGF activation in the endothelium and TGFβ inactivation in the stroma. These data highlight the role of AhR in cardiovascular homeostasis and suggest that this receptor can be a novel regulator of angiogenesis during tumor development.

The formation of new blood vessels is an important task that takes place in healthy tissues but also during tumor development. Among the several mechanisms of neovascularization already known, angiogenesis is particularly relevant for tumor vasculature and consists in the formation of new blood vessels by sprouting from preexisting ones (1,2). Angiogenic sprouting and branching involve specific populations of endothelial cells that guide tube formation at the tip and respond to vascular endothelial growth factor (VEGF) 5 as well as stalk cells that proliferate to form the vascular structure (3,4). The control of endothelial cell branching is regulated by both positive and negative factors that ultimately determine angiogenesis (5,6). A large body of evidence strongly suggests that the endotheliumspecific factor VEGF has a prominent role in angiogenesis and is indispensable for vascular development (1,(7)(8)(9)(10). Thus, blocking VEGF-dependent signaling by inactivation of the Vegfr-3 gene severely impairs vascular network development in the mouse (4), whereas neutralizing antibodies against the VEGFR-3 receptor reduces vascular density, decreases sprouting and branching (4), and inhibits tumor growth by blocking angiogenesis (11). Importantly, antibody-based strategies have been recently transferred to the clinic in order to inhibit metastatic colorectal cancer angiogenesis by blocking VEGFdependent signaling through patient treatment with the humanized antibody bevacizumab (Avastin) (12).
Angiogenesis, on the other hand, not only depends on the characteristics of the endothelial cells but also on the interactions that they establish with other stromal cells, in particular with fibroblasts and pericytes (13). Despite the fact that fibroblasts have different properties depending on their tissue of origin, the analysis of ␣-smooth muscle actin-expressing myofibroblasts isolated from a panel of breast carcinomas revealed common patterns of gene expression (14), which supports their conserved role in the synthesis and maintenance of extracellular matrix (ECM) components. Transforming growth factor-␤ (TGF␤) is a cytokine secreted and activated in the ECM by * This work was supported by grants from the Spanish Ministry of Science and mesenchymal cells (e.g. fibroblasts) that exerts a relevant role in proliferation, differentiation, apoptosis, and migration (15). Interestingly, TGF␤ is also a major molecule in the regulation of endothelial cell behavior (16) and vascular development by functional interaction with VEGF (17). Indeed, low extracellular TGF␤ levels promote endothelial cell proliferation and migration and new blood vessel formation, whereas high TGF␤ concentrations induce differentiation of endothelial cells, inhibition of tube formation, and impaired invasion in gels (18 -21). In addition, previous work has shown that TGF␤ is as potent as VEGF in inducing angiogenesis and that VEGF could be a target for TGF␤ (22). Therefore, it is reasonable to assume that altered VEGF and TGF␤ activities could induce defects in angiogenesis.
The aryl hydrocarbon (dioxin) receptor (AhR) is a well known transcription factor with increasing importance in cellular physiology and tumor development. Remarkably, AhR has a relevant role in vascular development and homeostasis (23,24) and in TGF␤ activation (25). AhR-null mice fail to resolve the embryonic structure known as portosystemic shunting and thus exhibit a patent ductus venous in the adult liver (23,24). AhR expression also affects the cardiovascular system, as AhR Ϫ/Ϫ mice have a significant heart hypertrophy (26,27), which is probably associated with the hypertension and increased endothelin-1 expression reported in this animal model (28). Previous studies using AhR-null mice have shown that this receptor regulates TGF␤ activation in mouse embryonic primary fibroblasts (29) and TGF␤ levels in the liver (30). TGF␤ overactivation results in decreased proliferation (29,31) and migration (32,33) in AhR Ϫ/Ϫ fibroblasts and co-localizes with portal fibrosis in AhR Ϫ/Ϫ mouse liver (34). Thus, AhR could be a common signaling intermediate in the regulation of vascular homeostasis by growth factors and cytokines such as TGF␤.
In this study, we sought to analyze the contribution of AhR expression to endothelial cell function as related to vessel formation and angiogenesis. Using mouse and human endothelial cells, tumor xenografts, aortic ring explants, in vivo vessel recruitment to Matrigel, and co-cultures of endothelial and fibroblast cells, we found that lack of AhR expression significantly impaired formation of tubular structures from aortic explants and angiogenesis and tumor development in vivo. Such a phenotype was a consequence of reduced VEGF activity in AhR Ϫ/Ϫ endothelial cells and of increased TGF␤ activity in AhR-null mesenchymal myofibroblasts. We suggest that AhR is a regulator of angiogenesis and that its down-modulation could represent a potential novel strategy to inhibit tumor growth.
The concentrations of VEGF-A used (100 ng/ml in aortic ring assays or 300 ng/ml in Matrigel plugs in vivo) were shown to efficiently modulate VEGF activity (35,36). The anti-VEGF neutralizing antibody was used at 200 ng/ml based on the LD 50 provided by the manufacturer in human umbilical vein endothelial cells (50 -150 ng/ml). Recombinant TGF␤ was employed at 10 ng/ml taking into account our previous studies on cytokine activation in the ECM (37) and keratinocyte migration (38). The neutralizing anti-TGF␤ antibody was used at 1 g/ml considering its LD 50 (in-house determined to equal 60 ng/ml) (29) and its effect on keratinocyte migration (38).
Mice-AhR-null mice were produced by homologous recombination in embryonic stem cells as described (39). Mice were used at 10 -12 weeks of age to obtain endothelial cells or aortic rings or at 30 weeks for in vivo Matrigel plug assays. The experiments involving animals were performed in compliance with the guidelines established by the Animal Care and Use Committee of the University of Extremadura.
Aortic Ring Assays and MAEC Isolation-Aortic Ring Assays and MAEC isolation were done as described previously (40) with some modifications. Briefly, the thoracic aorta was removed and sectioned in 1-mm aortic rings, which were embedded in Matrigel. After solidification, Dulbecco's modified Eagle's medium-F12 medium (Invitrogen) containing 20% fetal bovine serum, 200 units/ml penicillin, 200 g/ml streptomycin, and 250 ng/ml endothelial cell growth supplement was added. The length of the sprouting vessels was measured at different times using light microscopy or 4,6-diamidino-2-phenylindole staining. MAECs were isolated at day 12 by solubilizing Matrigel-embedded aortic rings outgrowths in Matrisperse (BD Biosciences) followed by seeding in the same medium. MAEC cultures were assessed for purity by quantifying CD102-positive cells by flow cytometry as described previously (40). In some experiments, VEGF-A was added to the aortic rings at 100 ng/ml. Induction of Hypoxia by Cobalt Chloride Treatment-To analyze the effects of hypoxia on VEGF expression, AhR ϩ/ϩ and AhR Ϫ/Ϫ MAECs were cultured in complete medium and then treated with solvent (DMSO) or with 100 M CoCl 2 for 16 or 24 h. RNA was purified and used to analyze VEGF mRNA levels as indicated below. This concentration of CoCl 2 has been shown to efficiently mimic hypoxia in cell culture (41). In these experiments, Vegf mRNA levels were normalized by ␤-actin because hypoxia can alter Gapdh expression (see primers used in Table 1). The contribution of hypoxia inducible factor-1␣ (HIF-1␣) in maintaining VEGF levels was determined by adding 5 M 2-Me E2 to CoCl 2 -treated MAECs or to AhR small interfering RNA (siRNA)-transfected HMEC-1 cells. This concentration of 2-Me E2 was shown to induce degradation of HIF-1␣ in cultured cells (42,43).
Tumor Xenografts-Aliquots of 5 ϫ 10 5 B16F10 cells (in 100 l of phosphate-buffered saline) were injected subcutaneously into both flanks of five AhR ϩ/ϩ and AhR Ϫ/Ϫ mice. Tumors were collected at 7 or 14 days after injection and their volume calculated as length ϫ width 2 ϫ 0.4. After fixation in 4% paraformaldehyde, tissues were sectioned at 5 m and processed for hematoxylin and eosin staining or immunofluorescence.
In Vivo Matrigel Plug Assays and Hemoglobin Content-Aliquots consisting of 500 l of ice-cold solution of 70% Matrigel, 30% Hanks' salt medium (Invitrogen), 30 units of heparin, and 300 ng of endothelial cell growth supplement were injected in duplicate into the flanks of at least four independent AhR ϩ/ϩ and AhR Ϫ/Ϫ mice. One flank of each mouse was left as the control, and the other flank was used for the corresponding experimental treatments. In some experiments, VEGF-A 164 (300 ng/ml), TGF␤ (10 ng/ml), a neutralizing anti-VEGF-A antibody (200 ng/ml), or a neutralizing anti-TGF␤ antibody (1 g/ml) were added to the Matrigel plugs. After 7 days, plugs were harvested, photographed, homogenized in 1 ml of ice-cold H 2 O, and centrifuged at 10,000 ϫ g for 6 min at 4°C. The amounts of hemoglobin released to the supernatants were determined using Drabkin's reagent as indicated by the manufacturer (Sigma-Aldrich). Hemoglobin content is represented as mg/ml in 100 mg of Matrigel.
Isolation of Mouse Retina and Whole Mount Immunofluorescence-Retinas from five 3-day-old AhR ϩ/ϩ and AhR Ϫ/Ϫ mice were isolated and processed as described previously (44). Briefly, eyes were collected and fixed overnight at 4°C in Trisbuffered saline containing 4% paraformaldehyde. Retinas were dissected and permeabilized overnight at 4°C in Tris-buffered saline containing 1% bovine serum albumin and 0.05% Triton X-100. After three washes in Tris-buffered saline, retinas were incubated overnight at 4°C with isolectin B4-FITC (Sigma-Aldrich) in Tris-buffered saline. Flat-mounted retinas were ana-lyzed by fluorescence microscopy. Measurements were taken near the optic nerve, at the center of the retina.
Hematoxylin/Eosin Staining and Tissue Immunofluorescence-Hematoxylin/eosin staining of tumor tissues was performed as indicated previously (33). For immunofluorescence, tissue sections were rehydrated, blocked, and incubated with CD31 antibody overnight at 4°C. After washing in phosphatebuffered saline, sections were incubated with a TRITC-labeled secondary antibody (Sigma-Aldrich), mounted in Mowiol, and photographed under fluorescence microscopy.
In Vitro Tube Formation and Co-culture Experiments-Matrigel (70-l aliquots) was gelled at 37°C for 1 h. Then, 10 5 MAECs were plated in OptiMEM (Invitrogen), and the formation of capillary-like structures was photographed 24 h later. For endothelial-mesenchymal co-culture experiments, AhR ϩ/ϩ and AhR Ϫ/Ϫ immortalized T-FGM fibroblasts (33) were treated with mitomycin (Sigma-Aldrich) to inhibit cell proliferation. A sterile plastic ring was placed in a 35-mm tissue culture plate, and its interior volume was filled with Matrigel. After gelling, a monolayer of T-FGM AhR ϩ/ϩ or T-FGM AhR Ϫ/Ϫ fibroblasts was seeded in the culture area outside the ring using OptiMEM. After 48 h, 10 5 HMEC-1 cells were plated on the Matrigel plug located inside the ring. Once the endothelial cells adhered to the Matrigel, the ring was removed, and 16 h later, the formation of capillary-like structures was photographed. Experiments were also performed using T-FGM conditioned media rather than the cells themselves. T-FGM AhR ϩ/ϩ and T-FGM AhR Ϫ/Ϫ fibroblasts growing at 30% confluence were cultured for 48 h in OptiMEM. Conditioned Opti-MEM was then collected, processed as indicated previously (29), and used to analyze the formation of capillary-like structures as described above.
Measurement of Total Tube Length (TTL) and Calculation of Branching Points-TTL and branching points were measured in aortic ring explants and in Matrigel plugs by blinded analyses as described in Fig. 2A. Raw images were analyzed using ImageJ software to identify tubular structures. These images were subsequently employed to calculate TTL as the total length of the individual tubes present and branching as the number of bifurcations along the tubes ( Fig. 2A, arrows).
Cell Viability and Cell Adhesion-After seeding for 16 h, 10 5 HMEC-1 cells were washed, fixed, and stained with crystal violet. Following extensive washing with water, cultures were photographed and the number of viable cells counted. For adhesion experiments, aliquots of 10 5 HMEC-1 cells were plated on gelled Matrigel plugs. After incubation at 37°C for 1 h, cells were fixed, stained with crystal violet, and analyzed as indicated above.
Invasion in Matrigel Chambers and Wound Closure in Vitro-Invasion assays were done essentially as described (33). Briefly, 5 ϫ 10 4 HMEC-1 cells were plated in the upper chamber of a Matrigel invasion Transwell (BD Biosciences) in either Opti-MEM or fibroblast-conditioned medium. Dulbecco's modified Eagle's medium-F12 medium containing 10 ng/ml insulin-like growth factor-1 (Sigma-Aldrich) was added to the lower chamber as a chemoattractant. After 24 h, Transwells were washed with phosphate-buffered saline, fixed at 4°C in 70% ethanol, treated for 15 min with RNase (10 ng/ml), and stained with propidium iodide. Cell invasion into Matrigel was analyzed in 5-m steps using a Zeiss LSM 510 confocal microscope.
Wound closure experiments were performed using HMEC-1 in the absence or presence of AhR siRNA essentially as described (25,33). RNA Interference (RNAi) and Western Immunoblotting-siRNA for AhR or unspecific scramble RNA (Dharmacon) was transiently transfected in HMEC-1 cells by electroporation using a MicroPorator MP-100 (Digital-Bio). siRNA or scramble RNA were transfected at 100 nM using three pulses of 10 ms with the voltage set at 1400 V. HMEC-1 cells were grown for 60 h and then trypsinized and used for in vitro Matrigel tube formation. A fraction of the transfected cells was used to analyze AhR expression by Western immunoblotting as described (45).
Real-time PCR and ELISAs-Quantitative real-time RT-PCR and data analyses for the expression of angiogenesis-related genes were done as described earlier (45). Oligonucleotide sequences used are indicated in Table 1. The amounts of active VEGF-A, TGF␤-1, and TGF␤-2 secreted by MAECs or T-FGM fibroblasts were determined using ELISA kits from Bender MedSystems following the manufacturer's instructions. To activate latent TGF␤, conditioned medium was treated for 10 min at room temperature with 165 mM HCl and then neutralized with NaOH.
Image Processing and Statistical Analyses-Light and fluorescence microscopy measurements of aortic rings and branching of capillary-like structures in Matrigel plugs were quantified in at least five random fields using ImageJ software. All determinations were done in triplicate in at least two independent experiments. Data are shown as mean Ϯ S.D. Statistical comparison between experimental conditions was done using GraphPad Prism 4.0 software. A comparison between conditions was done using the unpaired Student's t test, and multicomparison was done using oneway analysis of variance followed by Dunn's post test.

Loss of AhR Expression Impairs Tumor Growth and Compromises
Angiogenesis in Vivo-In previous work studying the potential of AhR-null transformed fibroblasts to induce tumors in mice, we suggested that lack of AhR expression could compromise tumor development by reducing cell migration and/or angiogenesis (33). These results prompted us to analyze whether AhR modulates angiogenesis and how its expression in the endothelial and stromal compartments contributes to the process. We first performed xenografts of B16F10 mouse melanoma cells in the dorsal area of AhR ϩ/ϩ and AhR Ϫ/Ϫ mice (Fig. 1A). Tumors produced by B16F10 cells in AhR-null mice had a significantly reduced volume as compared with those isolated FIGURE 1. AhR ؊/؊ mice have impaired tumor growth and inefficient angiogenesis. A, B16F10 mouse melanoma cells were injected subcutaneously in the dorsal area of AhR ϩ/ϩ and AhR Ϫ/Ϫ mice, and tumors were collected at 7 and 14 days after grafting. Tumor volume was calculated as length ϫ width 2 ϫ 0.4. B, tumors were fixed, sectioned, and stained with hematoxylin and eosin or used to detect blood vessels by immunofluorescence with an antibody for the endothelial marker CD31. Vessels were counted at each time point, and the results corresponding to tumors at 14 days were represented. Note that hematoxylin and eosin images appear dark because of the high levels of melanin expressed by B16F10 melanoma cells. C, Matrigel was prepared and injected subcutaneously in the dorsal area of AhR ϩ/ϩ and AhR Ϫ/Ϫ mice. After 7 days, the Matrigel plugs formed were removed, washed, and photographed, and the angiogenesis recruited in each genotype was quantitated by measuring the content in hemoglobin. Hemoglobin content is represented as mg/ml in 100 mg of Matrigel. Light microscopy was done at room temperature on a Nikon E-400 microscope equipped with a Nikon L16 camera. A ϫ10 objective (0.25 numeric aperture) was used on Eukitt-mounted sections. Immunofluorescence was done at room temperature on a Nikon TE2000U microscope equipped with a Nikon DS-5 M digital camera. A ϫ10 objective (0.25 numeric aperture) was used on Mowiol-mounted sections. Data are shown as means Ϯ S.D. Tumors were induced in both sides of five AhR ϩ/ϩ and AhR Ϫ/Ϫ mice at each time point. Matrigel plugs were done in duplicate in at least four animals of each genotype.
from AhR ϩ/ϩ animals at either 7 or 14 days after injection. In agreement to our hypothesis, this lower ability of AhR Ϫ/Ϫ mice to support tumorigenesis could be related to impaired angiogenesis because quantification of vessel area by both hematoxylin and eosin staining and CD31 labeling of endothelial cells revealed that AhR Ϫ/Ϫ tumors had significantly lower vessel numbers than wild-type tumors (Fig. 1B). Please note that hematoxylin and eosin images appear dark because of the high expression of melanin by B16F10 melanoma cells. To further support these observations, we next used an additional angiogenesis model based on the recruitment of blood vessels to Matrigel plugs inserted under the dorsal skin of AhR ϩ/ϩ and AhR Ϫ/Ϫ mice. Consistently, the vascular system of AhR Ϫ/Ϫ mice had a lower efficiency to produce new vessels able to colonize the Matrigel plug (Fig. 1C). Thus, loss of AhR expression appears to compromise angiogenesis in vivo.
Endothelial Cells Lacking AhR Have Altered Tube Formation and Branching-Two important factors that modulate efficient angiogenesis are the ability of the endothelial cells to branch and to form capillary tubes (4,9,46). We investigated whether deficient angiogenesis in AhR Ϫ/Ϫ mice is related to altered branching and/or to the formation of tubular structures. Quantification of the total tube length (47) and of the number of branch points was performed in aortic rings and Matrigel plugs as indicated under "Experimental Procedures" and in the scheme shown in Fig. 2A. Culture of aortic rings (ar) from AhR ϩ/ϩ and AhR Ϫ/Ϫ mice on Matrigel revealed that AhR-null MAECs were significantly impaired in their ability to form tubes as determined by measuring sprouting from the aorta (Fig. 2B). However, reduced TTL in AhR Ϫ/Ϫ aortas was not due to a lower migration potential of AhR Ϫ/Ϫ MAECs, as they invaded Matrigel to a similar extend as AhR ϩ/ϩ MAECs (Fig. 2B). Staining the Matrigel plug with 4,6diamidino-2-phenylindole showed that whereas AhR ϩ/ϩ MAECs could orientate and align to form tubes, AhR Ϫ/Ϫ MAECs emigrated from the aortic ring but remained mostly dispersed and formed fewer tubes (Fig. 2C). We then decided to analyze branching in AhR Ϫ/Ϫ MAECs. MAECs were isolated from expanding aortic rings and plated in Matrigel (Fig. 3A). AhR-null cells not only formed a smaller amount of tubes but also exhibited a significantly reduced efficiency to branch as compared with AhR ϩ/ϩ MAECs. To confirm these results, vascular branching was also measured in retinas isolated from 3-day-old newborn mice. To avoid variations due to heterogeneity, measurements were taken near to the optic nerve, at the center of the retina. Fig. 3B FIGURE 2. AhR ؊/؊ mouse endothelial cells have a reduced potential for tubulogenesis and branching in aortic ring explants. A, schematic representation of the method used to calculate TTL and branching points in aortic ring explants and Matrigel plugs. Branching points are indicated by arrows. B, rings were prepared from the aortas of AhR ϩ/ϩ and AhR Ϫ/Ϫ mice and placed in Matrigel plugs. After 7 days in culture, tube formation was determined by measuring TTL for the major sprouting vessels as described (47). Maximum outgrowth was also determined as the total distance emigrated by the mouse endothelial cells (MAEC) from the aortic ring (ar). C, tube formation was also analyzed by staining the aortic rings with 4,6-diamidino-2-phenylindole. Tube formation was calculated as the ratio between tube-oriented/dispersed endothelial cells. Light microscopy was done at room temperature on a Nikon E-400 microscope equipped with a Nikon L16 camera. Objectives shows that isolectin B4-FITC-labeled AhR Ϫ/Ϫ retinas had fewer branching points per field than retinas from AhR ϩ/ϩ mice. Altogether, these experiments suggest that AhR expression is relevant to promote and to maintain angiogenesis, probably through the modulation of tube formation and branching of endothelial cells. The causal role of AhR in tube formation and branching was demonstrated by down-modulating AhR expression with RNAi. Human endothelial cells (HMEC-1) constitutively expressed AhR, and transfection of siRNA against this receptor significantly decreased its protein levels with respect to unspecific scramble-transfected cells (Fig. 3C). AhR was shown to be responsible for endothelial tube formation and branching, because siRNA-transfected HMEC-1 cells mimicked the phenotype observed in AhR Ϫ/Ϫ MAECs (Fig.  3D). Reduced branching and impaired tube formation in the absence of AhR could be due to a decrease in cell viability and/or migration potential. However, down-modulation of AhR expression by siRNA did not significantly reduced viability or proliferation of HMEC-1 cells within a 72-h period (Fig. 3E). Further, wound healing in culture revealed that AhR siRNA did not affect the migration potential of HMEC-1 (Fig. 3F).
VEGF-A Expression Is Down-modulated in AhR Ϫ/Ϫ MAECs: Role of Hypoxia and HIF-1␣-VEGF is a major growth factor involved in angiogenesis and in vascular homeostasis (7,9,10). Because loss of AhR expression markedly altered angiogenesis in vivo and tubulogenesis and branching both in vivo and in culture, we decided to analyze whether the AhR Ϫ/Ϫ phenotype was related to altered VEGF expression and/or activity. Realtime RT-PCR was used to measure mRNA expression of the angiogenesis-related genes Hif-1␣, Vegf-A, Vegf-B, Plgf, Vegfr-1, B, retinas were isolated from AhR ϩ/ϩ and AhR Ϫ/Ϫ 3-day-old newborn mice and stained with isolectin B4-FITC to label blood vessels. Branching points were determined and plotted for each genotype. Measurements were taken near to the optic nerve, at the center of the retina. C, HMEC-1 cells were transfected by nucleoporation with an AhR siRNA or unspecific scramble RNA, and AhR protein levels were determined by Western immunoblotting 72 h later. The expression of ␤-actin was used as negative control for RNAi and as loading control for Western blotting. D, siRNA-or scramble RNA-transfected HMEC-1 cells were plated in Matrigel. Tube formation and branching were analyzed after 24 h as described under "Experimental Procedures." E, cell viability was determined in HMEC-1 cells at 48 or 72 h after transfection with either scramble or AhR siRNA. Cells were stained with crystal violet and counted. F, HMEC-1 cells were transfected with scramble or AhR siRNA and grown to confluence. Serum-free culture medium was added, and wound healing performed as indicated under "Experimental Procedures." Data represent the extent of wound closure under both experimental conditions. Light microscopy was done at room temperature on a Nikon E-400 microscope equipped with a Nikon L16 camera. Objectives used were: ϫ10 (0.25 numeric aperture) (A and D) and ϫ20 (0.40 numeric aperture) (B). Data are shown as means Ϯ S.D. Branching was analyzed in at least five primary MAEC cultures obtained from aortic rings isolated from different mice. Retinas were isolated from eyes of three AhR ϩ/ϩ and AhR Ϫ/Ϫ mice. HMEC-1 cells were transfected in triplicate with AhR siRNA or unspecific scramble RNA. Calibration bar corresponds to 100 m.
and Vegfr-2. AhR Ϫ/Ϫ MAECs had a significant decrease in Vegf-A and, to a lesser extent, Vegf-B expression. Interestingly, the expression of the VEGF-A-target gene Hif-1␣ was also reduced in AhR Ϫ/Ϫ MAECs, suggesting that down-modulation of Vegf-A was functionally relevant (Fig. 4A). VEGF-A has several splice variants: VEGF 188 , which is slightly active and tightly bound to the extracellular matrix, and VEGF 120 and VEGF 164 , which are diffusible forms able to induce angiogenesis in vivo (48,49). Remarkably, AhR Ϫ/Ϫ MAECs had a significant downmodulation of Vegf 164 and Vegf 120 mRNAs as compared with AhR ϩ/ϩ MAECs, with Vegf 164 being the predominant isoform in these primary cells (Fig. 4B). Furthermore, the level of VEGF-A activity secreted to the medium (e.g. VEGF 164 and VEGF 120 ) was markedly reduced in AhR Ϫ/Ϫ MAECs as compared with wild-type MAECs (Fig. 4C), suggesting that lower levels of VEGF-A could be relevant to the defective angiogenesis observed in AhR-null mice. In addition, the high level of Vegf 188 produced by AhR Ϫ/Ϫ MAECs did not seem to compensate for the impairment in angiogenesis, further suggesting that Vegf 188 is not functionally relevant and that the reduction in Vegf 164 and Vegf 120 activities was mainly responsible for the lower angiogenic potential of MAECs.
The reduced expression of Hif-1␣ in AhR Ϫ/Ϫ MAECs suggested that their deficiency in producing active VEGF and inducing angiogenesis could be due, at least in part, to an impaired response to hypoxia. To address this possibility, we treated AhR ϩ/ϩ and AhR Ϫ/Ϫ MAECs with CoCl 2 to mimic hypoxia and then determined Vegf mRNA expression. Hypoxia readily increased Vegf expression with time in AhR ϩ/ϩ MAECs, whereas no significant induction was observed in AhR Ϫ/Ϫ MAECs (Fig. 4D). Moreover, HIF-1␣ was involved in this mechanism, because treatment with 2-Me E2, which induces HIF-1␣ protein degradation, significantly blocked Vegf induction by CoCl 2 in AhR ϩ/ϩ but not in AhRϪ/Ϫ MAECs (Fig.  4D). The causal role of AhR in HIF-1␣-mediated control of Vegf expression was further analyzed using HMEC-1 cells, which were found to respond to AhR down-modulation by RNAi (see Fig. 3). Transfection of HMEC-1 cells with AhR siRNA reduced the mRNA expression of both Hif-1␣ and Vegf-A (Fig. 5A). Accordingly, treatment with 2-Me E2 decreased Vegf levels in mock-transfected (scramble) but not in AhR siRNA-transfected HMEC-1 cells (Fig. 5B). Thus, reduced angiogenesis in AhR Ϫ/Ϫ endothelial cells seems to be the result of lower VEGF-A production due to decreased HIF-1␣ expression.
Based on these results, we next used recombinant VEGF protein (purified VEGF 164 , R&D Systems) and an inhibitory anti-VEGF antibody (neutralizes VEGF 164 and VEGF 120 ; Sigma-Aldrich) in an attempt to modulate angiogenesis in culture and in vivo. The addition of anti-VEGF antibody to AhR ϩ/ϩ aortic rings decreased the TTL to values similar to those found in AhR Ϫ/Ϫ aortic rings, whereas treatment of AhR Ϫ/Ϫ rings with recombinant VEGF protein did the opposite and rescued TTL levels to AhR ϩ/ϩ values (Fig. 6A). In vivo, the addition of anti-VEGF antibody to Matrigel plugs implanted in AhR ϩ/ϩ mice markedly reduced angiogenesis, whereas recombinant VEGF protein induced a modest, although very reproducible increase in blood vessels recruitment in plugs injected in AhR Ϫ/Ϫ mice (Fig. 6B).
Increased Secretion of TGF␤ Activity by Stromal Fibroblasts Also Contributes to Defective Angiogenesis in AhR Ϫ/Ϫ Mice-TGF␤ is a cytokine with a relevant role in angiogenesis that  Table 1. Determinations were done in triplicate in at least three MAEC cultures. B, mRNA expression for the Vegf 120 , Vegf 164 , and Vegf 188 isoforms was determined by real-time RT-PCR. Differences in expression between AhR Ϫ/Ϫ and AhR ϩ/ϩ MAECs are indicated as -fold change. The oligonucleotide sequences used to amplify each gene are indicated in Table 1. C, the amount of VEGF-A activity secreted by AhR ϩ/ϩ and AhR Ϫ/Ϫ MAEC and fibroblast cells was determined by ELISA in conditioned medium from each cell type and genotype. D, hypoxia was mimicked in AhR ϩ/ϩ and AhR Ϫ/Ϫ MAEC cultures by treatment with 100 M CoCl 2 for 16 or 24 h, and Vegf expression was determined at the mRNA level as indicated under "Experimental Procedures." Data were normalized and expressed as -fold change with respect to solvent (DMSO)-treated cultures. In some experiments, MAECs were co-treated with 100 M CoCl 2 and 5 M 2-Me E2 for 24 h to induce HIF-1␣ degradation under hypoxic conditions. Vegf expression was normalized using ␤-actin instead of Gapdh to avoid potential effects of CoCl 2 on the expression of the latter gene. The experiments were done in triplicate in three MAEC cultures. Data are shown as means Ϯ S.D.
cooperates functionally with VEGF in endothelial cell function (16,17,22). Because we had found previously that stromal fibroblasts from AhR Ϫ/Ϫ mice secrete increased levels of active TGF␤ (25,29), here we established co-cultures of Matrigelembedded HMEC-1 cells and fibroblasts isolated from AhR ϩ/ϩ or AhR Ϫ/Ϫ mice (Fig. 7A). HMEC-1 cells co-cultured with AhR Ϫ/Ϫ fibroblasts lost their ability to form tubes as compared with the same endothelial cells co-cultured with AhR ϩ/ϩ fibroblasts (Fig. 7B). Moreover, such an effect on the endothelial cells was produced by a soluble and secreted molecule, because conditioned medium (CM) from AhR Ϫ/Ϫ fibroblasts mimicked the effects induced by the fibroblasts themselves (Fig. 7C).
AhR Ϫ/Ϫ fibroblasts expressed slightly increased levels of Tgf␤-1 and ϳ2-fold more Tgf␤-2 mRNA than wild-type cells as determined by real-time RT-PCR (Fig. 8A, left). Nevertheless, TGF␤ has to be secreted and activated in the extracellular matrix (15,50) to exert its effects, and because AhR-null fibro-blasts produce elevated levels of TGF␤ activity (25,29,37,51), we determined the amounts of active and total TGF␤-1 and TGF␤-2 released by AhR ϩ/ϩ and AhR Ϫ/Ϫ fibroblasts by ELISA. We found that the total amount of TGF␤-1 secreted to the medium was markedly higher than that of TGF␤-2 in both genotypes (Fig. 8A, center and right). Importantly, AhR Ϫ/Ϫ fibroblasts released higher levels of both active and total TGF␤-1 than AhR ϩ/ϩ cells, although no significant differences were observed with respect to active TGF␤-2 (Fig. 8A, center and  right). Thus, increased production of TGF␤-1/2 by estromal fibroblasts could contribute to the inhibition of angiogenesis. Consistent with this hypothesis, the addition of recombinant TGF␤ to medium conditioned by AhR ϩ/ϩ fibroblasts inhibited tube formation by HMEC-1 cells, whereas a neutralizing anti-TGF␤ antibody added to medium conditioned by AhR Ϫ/Ϫ fibroblasts increased tube formation (Fig. 8B). These detrimental effects on angiogenesis were not due to an inhibitory activity of the AhR Ϫ/Ϫ conditioned medium or the TGF␤ recombinant protein on HMEC-1 viability (Fig. 8C) or adhesion to Matrigel (Fig. 8D).
Because angiogenesis requires endothelial cell invasion of the surrounding tissue, we next used HMEC-1 cells as a model to analyze their invasion potential in the presence of medium conditioned by AhR Ϫ/Ϫ fibroblasts and to determine whether TGF␤ would contribute to such effect. HMEC-1 cells were plated in Matrigel-coated Transwells, and their invasive ability was determined by confocal microscopy. We found that the addition of medium conditioned by AhR Ϫ/Ϫ fibroblasts to the upper chamber significantly inhibited HMEC-1 invasion as compared with endothelial cells treated with medium from wild-type fibroblasts (Fig. 9A). Furthermore, the addition of recombinant TGF␤ to the medium conditioned by AhR ϩ/ϩ fibroblasts decreased HMEC-1 invasion, whereas a neutralizing anti-TGF␤ antibody did the opposite, increasing HMEC-1 invasion in the presence of medium conditioned by AhR Ϫ/Ϫ fibroblasts (Fig. 9A). Consistently, recombinant TGF␤ was able to decrease blood vessel recruitment to Matrigel plugs injected in AhR ϩ/ϩ mice, whereas a neutralizing anti-TGF␤ antibody enhanced vessel recruitment to Matrigel inserts in AhR Ϫ/Ϫ mice (Fig. 9B). Importantly, VEGF and TGF␤ acted together in AhR-dependent angiogenesis, because co-treatment of AhR ϩ/ϩ mice with anti-VEGF antibody plus TGF␤ decreased blood vessel recruitment to Matrigel plugs, whereas the addition of VEGF plus anti-TGF␤ antibody increased angiogenesis in Matrigel plugs implanted in AhR Ϫ/Ϫ mice (Fig. 9B). This possible interaction between VEGF and TGF␤ activities in endothelial cells is also supported by the observation that treatment of AhR Ϫ/Ϫ MAECs with exogenous TGF␤ further reduced their basal low level of VEGF-A secretion (Fig. 9C). Altogether, these data strongly suggest that TGF␤ activity secreted to the extracellular medium by stromal fibroblasts contributes significantly to angiogenesis and that overactivation of TGF␤ in the absence of AhR is a relevant factor in explaining their angiogenesis deficiency. In addition, the fact that an increase in VEGF cooperated with decreased TGF␤ activity in improving angiogenesis in AhR Ϫ/Ϫ mice (or the opposite in wild-type mice) further supports a functional interaction between endothelial and stromal cells in vessel recruitment and underlines AhR as a regulator of the process.  Table 1. Data were normalized, and fold change represents the difference between scramble-and AhR siRNA-transfected cultures. B, HMEC-1 cells were transfected as indicated in A, and some cultures were treated for 24 h with 5 M 2-Me E2. Total RNA was purified and analyzed for Vegf 165 mRNA expression as indicated in the legend for Fig. 4. The experiments were done in two different transfections using triplicate samples. Data are shown as means Ϯ S.D.

DISCUSSION
An increasing number of studies are revealing that adequate AhR expression and activity are required to maintain the homeostasis of important cell functions. In addition to the control of cell proliferation, differentiation, and migration (25,32,52), AhR also has a role in the cardiovascular system. The analysis of AhRnull mice has shown that, in the absence of xenobiotics, loss of AhR expression induces cardiac hypertrophy (26) with hypertension and elevated plasma levels of angiotensin II and endothelin-1 (53). In addition, AhR Ϫ/Ϫ mice have hepatic vascular defects characterized by the failure to resolve the embryonic portosystemic shunt (23,24). Moreover, we have suggested that the lower efficiency of transformed AhR Ϫ/Ϫ fibroblasts to generate tumor xenografts in vivo could be due to an impaired ability to induce angiogenesis (33). Because AhR seemed relevant to vascular homeostasis and angiogenesis, we decided to analyze further whether AhR expression compromises endothelial cell function and angiogenesis. Trying to accomplish that goal, experiments were performed using in vivo animal models, ex vivo tissue culture explants, and primary cultures of rodent and human endothelial cells. A major conclusion from our current work is that angiogenesis requires not only AhR-dependent expression of VEGF-A by endothelial cells but also AhR-dependent down-modulation of TGF␤ activity by stromal fibroblasts. The mechanisms by which AhR signaling modulates angiogenesis appear complex, as a recent report has revealed that AhR-null mice have increased response to ischemia-induced angiogenesis (54). That study suggests that because the aryl hydrocarbon receptor nuclear translocator (ARNT/HIF-1␤) is a common partner for both AhR and HIF-1␣ (55,56), and given that ischemia increases the expression of VEGF, ARNT, and HIF-1␣ to a greater extend in AhR Ϫ/Ϫ than in AhR ϩ/ϩ mice, the lack of competition between AhR and HIF-1␣ for common co-regulators in AhR-null mice would result in improved angiogenesis due to overactivation of the HIF-1␣ signaling pathway. Although this is a plausible hypothesis, and ARNT has an evident role in angiogenesis (57,58), other studies have shown that activation of either AhR-or HIF-1␣-dependent signaling does not inhibit the activity of the other pathway as a FIGURE 6. VEGF activity modulates angiogenesis in an AhR-dependent manner in vivo and in culture. A, aortic rings (ar) were obtained from AhR ϩ/ϩ and AhR Ϫ/Ϫ mice and plated in Matrigel plugs for 7 days. Sprouting of MAECs was analyzed by measuring the TTL as described (47). In some experiments, AhR ϩ/ϩ or AhR Ϫ/Ϫ aortic rings were treated with anti-VEGF-A antibody (200 ng/ml) or recombinant VEGF-A 164 protein (100 ng/ml), respectively. Light microscopy was done at room temperature on a Nikon E-400 microscope equipped with a Nikon L16 camera with a ϫ10 objective (0.25 numeric aperture). The experiments were performed in at least 10 aortic rings obtained from different AhR ϩ/ϩ and AhR Ϫ/Ϫ mice. Magnified images are shown below the main panels. B, Matrigel plugs were injected in the dorsal skin of AhR ϩ/ϩ or AhR Ϫ/Ϫ mice and left untreated or treated with anti-VEGF-A antibody (200 ng/ml) or recombinant VEGF-A 164 protein (300 ng/ml), respectively. Blood vessel recruitment to Matrigel plugs was quantitated by measuring their hemoglobin content. Hemoglobin content is represented as mg/ml in 100 mg of Matrigel. At least eight aortic rings from different mice were used for each genotype and experimental condition. Four mice of each genotype were used for the in vivo Matrigel experiments. One flank of each mouse was left untreated as a control, and the other flank was treated as indicated. Data are shown as means Ϯ S.D. result of competition for ARNT (59). Our data in fact indicate that genetic ablation of AhR in mouse endothelial cells or its down-modulation by RNAi in the human endothelium decreases HIF-1␣ levels, thus making the expression of HIF-1␣ AhR-dependent in these cell types. Therefore, the question remains as to how AhR participates in endothelial cell function under physiological and stress conditions and which are the regulatory pathways involved. This study focuses on the role of AhR in angiogenesis under normal cell conditions and identifies VEGF and TGF␤ as relevant growth factors that interact functionally in an AhR-dependent manner.
In agreement with our prior study (33) and with reports showing that constitutive AhR activation enhances tumor growth (60,61), we have found that lack of AhR expression in mice severely impairs their capacity to support the growth of melanoma tumors, suggesting that an AhR-competent microenvironment is needed for such an effect. The fact that melanoma tumors from AhR Ϫ/Ϫ mice had decreased blood vessel content provided additional support for AhR in tumor angio-genesis. Indeed, blood vessels from mice lacking AhR expression were much less capable of invading angiogenesis-promoting Matrigel plugs. The sprouting of expanding and invading vessels markedly influences angiogenesis through branching of the endothelial cells located at the tip end and tube formation by trailing stalk cells (9). Ex vivo culture of aortic rings from AhR Ϫ/Ϫ mice and AhR siRNAtransfected HMEC-1 cells revealed that, although AhR-deficient endothelial cells (MAECs) migrated "in the same manner" as their wildtype counterparts, they had a reduced potential to branch and to align forming tubular structures. Branch deficiency was AhRdependent, as it was observed in retinas from AhR Ϫ/Ϫ mice under physiological conditions and also because interference of AhR expression by RNAi in human HMEC-1 cells inhibited branching angiogenesis. Thus, AhR expression in endothelial cells is required for efficient angiogenesis.
VEGF is a critical regulator of angiogenesis, and blockade of its signaling compromises endothelial cell growth, sprouting, and tubulogenesis (4,9,16). A previous study has shown that AhR activation by 2,3,7,8-tetrachlorodibenzo-pdioxin (TCDD) promotes vascularization of the mouse retina by a mechanism involving increased VEGF production (62). The implication of VEGF in AhR-dependent angiogenesis is sustained by our data considering that AhR Ϫ/Ϫ MAECs expressed reduced Vegf-A and Vegf-B mRNA levels and, consequently, secreted lower amounts of active VEGF-A to the extracellular medium. Furthermore, the analysis of specific VEGF-A isoforms revealed that AhR inactivation preferentially down-regulated mRNA expression of Vegf-A 120 and Vegf-A 164 , which are considered the active, secretable forms of this growth factor with angiogenesis-promoting activity (48,49). Interestingly, because the amount of VEGF activity produced by fibroblasts is lower than that of endothelial cells, regardless of the AhR genotype, our data suggest a relevant role for AhR-dependent VEGF secretion by endothelial cells in angiogenesis. We therefore hypothesized that VEGF could be an important intermediate molecule in AhR-dependent angiogenesis. This possibility proved to be correct, because modulating VEGF levels by a recombinant VEGF protein increased angiogenesis in Matrigel implants in vivo and in aortic rings of AhR Ϫ/Ϫ mice, whereas a neutralizing anti-VEGF antibody plugs were co-cultured with AhR ϩ/ϩ or AhR Ϫ/Ϫ fibroblasts, and their ability to form tubes was determined by measuring TTL as described (47). C, HMEC-1 cells were cultured in Matrigel plugs in the presence of medium conditioned (C.M.) by AhR ϩ/ϩ or AhR Ϫ/Ϫ fibroblasts, and TTL was determined as above. Light microscopy was done at room temperature on a Nikon E-400 microscope equipped with a Nikon L16 camera with a ϫ10 objective (0.25 numeric aperture). The experiments were performed in four HMEC-1 cultures for each experimental condition. Data are shown as means Ϯ S.D.  Table 1. Secretion of latent and active TGF␤-1 and TGF␤-2 proteins by AhR ϩ/ϩ and AhR Ϫ/Ϫ fibroblasts to the culture medium was determined by specific ELISA as indicated under "Experimental Procedures." To activate latent TGF␤, conditioned medium was treated for 10 min at room temperature with 165 mM HCl and then neutralized with NaOH. B, HMEC-1 were plated in Matrigel and treated with CM from AhR ϩ/ϩ or AhR Ϫ/Ϫ fibroblasts, with AhR ϩ/ϩ CM plus 10 ng/ml recombinant TGF␤, or with AhR Ϫ/Ϫ CM plus 1 g/ml neutralizing anti-TGF␤ antibody (Ab). TTL was determined as indicated (47). Light microscopy was done at room temperature on a Nikon E-400 microscope equipped with a Nikon L16 camera with a ϫ10 objective (0. reduced angiogenesis in AhR ϩ/ϩ mice under the same experimental conditions. The mechanism by which AhR modulates angiogenesis through VEGF is in large part unknown, although several sets of data suggest the contribution of hypoxia and HIF-1␣. Firstly, MAECs growing under hypoxia-mimicking conditions by CoCl 2 treatment were able to induce Vegf but only if they expressed AhR. Secondly, inducing HIF-1␣ degradation by 2-Me E2 blocked CoCl 2 -dependent Vegf induction in AhR ϩ/ϩ but not in AhR Ϫ/Ϫ endothelial cells. Thirdly, down-modulation of AhR by siRNA in HMEC-1 cells decreased Hif-1␣ and Vegf expression, although 2-Me E2 inhibited Vegf levels only in AhRexpressing cells. Finally, a previous study has indicated that AhR activation by dioxin in thymic epithelial cells induces Vegf expression as did CoCl 2 (63). Altogether, we propose that AhR expression in endothelial cells maintains the production of active VEGF isoforms, at least in part through HIF-1␣, and that alteration of such mechanism contributes to defective angiogenesis in AhR-null mice.
Nonetheless, angiogenesis is not only regulated through endothelial cells but also involves functional interactions with different components of the stroma (13). Among them, TGF␤ is considered an important growth factor able to inhibit angiogenesis (16,64). Considering our previous studies showing that AhR modulates TGF␤ activity in primary and immortalized stromal fibroblasts (25,29,34), we sought to analyze whether the increase in TGF␤ activity by AhR Ϫ/Ϫ fibroblasts could add to the deficient angiogenesis observed in AhR-null mice. Co-culture of human HMEC-1 cells with AhR Ϫ/Ϫ fibroblasts severely impaired their angiogenesis potential. Because the same effects were found by using conditioned medium from AhR Ϫ/Ϫ fibroblasts, we concluded that a secreted angiogenesis inhibitor should be involved. TGF␤ emerged as a potentially relevant candidate because AhR Ϫ/Ϫ fibroblasts expressed higher levels of Tgf␤-1 and Tgf␤-2 and, more importantly, because these cells secreted increased amounts of active and total cytokine, particularly TGF␤-1. This hypothesis was confirmed by experiments per-

cells, and modulation of VEGF and TGF␤ levels affects blood vessel recruitment in vivo.
A, HMEC-1 cells plated in Matrigel-coated Transwells were treated with CM from AhR ϩ/ϩ or AhR Ϫ/Ϫ fibroblasts, with AhR ϩ/ϩ CM plus 10 ng/ml recombinant TGF␤, or with AhR Ϫ/Ϫ CM plus 1 g/ml neutralizing anti-TGF␤ antibody (Ab). HMEC-1 invasion was quantitated by confocal microscopy taking measurements each 5 m. Confocal microscopy was done at room temperature on a Zeiss LSM 510 apparatus. At least three Transwells were used for each genotype and experimental condition. B, Matrigel plugs were injected into the dorsal skin of AhR ϩ/ϩ or AhR Ϫ/Ϫ mice and left untreated or treated with recombinant TGF␤ (10 ng/ml) (AhR ϩ/ϩ mice), neutralizing anti-TGF␤ antibody (1 g/ml) (AhR Ϫ/Ϫ mice), recombinant TGF␤ (10 ng/ml) plus anti-VEGF-A antibody (200 ng/ml) (AhR ϩ/ϩ mice), or neutralizing anti-TGF␤ antibody (1 g/ml) plus recombinant VEGF-A 164 protein (300 ng/ml) (AhR Ϫ/Ϫ mice). After 7 days, the plugs were recovered, photographed, and used to quantify vessel formation by measuring their hemoglobin content. Hemoglobin content is represented as mg/ml in 100 mg of Matrigel. The experiments were repeated in at least four mice of each genotype. One flank of each mouse was used as an untreated control, and the other flank was treated as indicated. C, AhR ϩ/ϩ and AhR Ϫ/Ϫ MAECs were left untreated or treated with neutralizing anti-TGF␤ antibody (1 g/ml) or recombinant TGF␤ (10 ng/ml), respectively. VEGF-A activity secreted by these cells was measured by ELISA as indicated in the legend for Fig. 4. Data are shown as means Ϯ S.D.
formed in the presence of recombinant TGF␤ protein or a neutralizing anti-TGF␤ antibody. Thus, the addition of TGF␤ inhibited tube formation in HMEC-1 cells and angiogenesis in Matrigel plugs implanted in AhR ϩ/ϩ mice, whereas a neutralizing anti-TGF␤ antibody did the opposite, promoting both tube formation in human endothelia and angiogenesis in Matrigel plugs injected in AhR Ϫ/Ϫ mice. In agreement, previous studies have also shown that TGF␤ or its neutralizing antibody either block or stimulate embryonic vasculogenesis (20). Thus, AhR expression in stromal fibroblast also contributes to angiogenesis by regulating the production of active TGF␤ in the extracellular matrix. Several mechanisms have been proposed to explain how AhR inhibits TGF␤ expression/activation in fibroblast cells. They include transcriptional repression of the Tgf␤-1 and Tgf␤-2 genes (this work and Ref. 65), down-regulation of the latent TGF␤-binding protein that binds TGF␤ to the extracellular matrix (29,51), and inhibition of extracellular proteases (e.g. plasminogen activators/plasmin and elastase) known to be responsible for TGF␤ activation (37). We suggest that alteration of these mechanisms in AhR lacking fibroblasts could result in TGF␤ overactivation, which, in turn, could cooperate with decreased VEGF activity to inhibit angiogenesis. It is also possible that the inhibitory effects on angiogenesis of stroma-derived TGF␤ could take place through reduction of the invasive properties of the endothelium, suggesting that whereas endothelia-derived VEGF modulates angiogenesis by regulating branching, stromal TGF␤ could exert its effects by affecting cell invasion.
In addition to their individual roles as angiogenesis regulators, our study provides experimental support for the proposed interaction between VEGF and TGF␤ in the control of vessel formation (9,17,20). Indeed, VEGF and TGF␤ had opposing effects on angiogenesis in such a way that their coordinated secretion by AhR-null endothelial and fibroblast cells probably determines the lower angiogenesis present in AhR Ϫ/Ϫ mice. Remarkably, the intracellular AhR receptor seems to be a common partner, able to regulate VEGF and TGF␤ activities in the extracellular medium and, ultimately, endothelial cell branching and tubulogenesis under basal and pathological conditions. In this context, the fact that natural AhR antagonists such as resveratrol can inhibit tumor angiogenesis (66) by interfering with VEGF expression and HIF-1␣ accumulation in human squamous carcinoma SCC-9 and hepatoma HepG2 cells (67) identifies AhR as novel potential therapeutic target to block angiogenesis in certain diseases such as cancer. From a mechanistic point of view, it is clear that AhR modulates TGF␤ activity by regulating its location and activation in the ECM through the transcriptional control of the latency protein LTBP-1 and the activity of ECM proteases (25,37,51) and by controlling the expression of TGF␤-related genes (68,69). Regarding the regulation of VEGF-A, our data suggest the involvement of HIF-1␣, a major regulatory protein for this growth factor (70,71) that is also down-modulated in AhR Ϫ/Ϫ MAECs. Alternatively, because the Vegf-A promoter contains xenobiotic response elements for AhR binding, AhR can also maintain constitutive VEGF expression as shown previously for the target genes Cyp1a2 (39), Vav3 (72), and p27 Kip1 (73).
In summary, this study provides additional support for the widely accepted implications for AhR in cell physiology and pathology. We propose here that the coordinated control by AhR of VEGF activity in endothelial cells and TGF␤ in fibroblasts represents a mechanism that can integrate these two growth factors in the progression of angiogenesis. It will be of interest to determine whether AhR, VEGF, and TGF␤ are coordinately regulated during tumor angiogenesis.