A mechanosensitive transcriptional mechanism that controls angiogenesis

p190RhoGAP and TFII-I regulate VEGFR2

To explore whether p190RhoGAP modulates vascular development by altering gene transcription, we knocked down p190RhoGAP in human microvascular endothelial (HMVE) cells using siRNA. TFII-I protein levels were approximately 1.5 times higher in the nuclear fraction of knockdown cells (Supplementary Fig. S2a), and this was confirmed by fluorescence microscopy (Supplementary Fig. S2b). As first shown in fibroblasts¹⁷, we found that p190RhoGAP co-precipitates with TFII-I (and vice versa) (Supplementary Fig. S2c), suggesting that p190RhoGAP may also bind TFII-I and sequester it in the cytoplasm of human capillary cells.

TFII-I upregulates VEGFR2 protein expression in human aortic endothelial cells by binding to the VEGFR2 promoter¹⁸. But TFII-I knockdown in our HMVE cells increased (rather than decreased) VEGFR2 mRNA and protein levels by 2- to 3-fold relative to control cells, and overexpressing TFII-I (delta isoform²¹) using lentiviral transduction produced the opposite effect (Fig. 1a, Supplementary Fig. S3a). This may be due to the fact that macrovascular and microvascular endothelial cells undergo distinct morphogenetic programs (produce large tubes versus branching capillaries).

Fig. 1

TFII-I and GATA2 control VEGFR2 expression via p190RhoGAP

a)VEGFR2 and TFII-I mRNA levels in cells treated with TFII-I siRNA or lentiviral vectors (virus) relative to control cells (*, p < 0.05; unpaired Student's t-test is used throughout). b) Immunoblots showing VEGFR2, TFII-I, p190RhoGAP, and GAPDH protein levels in cells treated with TFII-I or p190RhoGAP siRNAs, or both. c) Immunofluorescence micrographs showing GATA2 distribution and nuclear DAPI staining in control and p190RhoGAP knockdown cells cultured with 0.3% serum (bar, 5μm). d)VEGFR2 mRNA level in cells transfected with p190RhoGAP siRNA alone or with GATA2 siRNA (*, p<0.01). VEGFR2 promoter activities (e) and mRNA levels (f) in cells transfected with GATA2 or TFII-I siRNA alone or in combination (*, p<0.01; **, p<0.05). g) ChIP analysis showing VEGFR2 promoter co-immunoprecipitating with GATA2 or TFII-I antibodies in cells transfected with TFII-I or GATA2 siRNA (Ct, control; n=3; error bars = s.e.m. throughout).

TFII-I binds to the Inr region of the VEGFR2 gene promoter, and smaller portions of this promoter (human VEGFR2: -225 to +268, -570 to +268, and -780 to +268) have similar or greater activity compared to the full length promoter (4kb)¹⁸22. When these various VEGFR2 promoters were characterized using a luciferase assay in human umbilical vein endothelial (HUVE) cells, TFII-I knockdown increased, and overexpression of TFII-I using DNA transfection decreased, VEGFR2 promoter activity by 1.5 to 2-fold and one-fifth normal levels, respectively (Supplementary Fig. S3b). The promoter-less pGL3 basic reporter showed no promoter activity (not shown). Moreover, mutagenesis of the VEGFR2 Inr (-780+268; 780MUT) decreased promoter activity in both control and TFII-I knockdown cells (Supplementary Fig. S3b), confirming that TFII-I decreases VEGFR2 promoter activity through the Inr region. Importantly, the specificity of these effects of TFII-I knockdown were confirmed by demonstrating that reconstitution of TFII-I can reverse these effects (Supplementary Fig. S3c).

We next asked whether p190RhoGAP stimulates VEGFR2 expression by restricting nuclear translocation of TFII-I (Supplementary Fig. S2a,b). TFII-I knockdown increased, and p190RhoGAP knockdown decreased, VEGFR2 promoter activity (-780+268) relative to control cells (Supplementary Fig. S3d). The double knockdown of p190RhoGAP and TFII-I exhibited the same stimulation as observed in the TFII-I knockdown cells (Supplementary Fig. S3d). However, p190RhoGAP knockdown did not decrease VEGFR2 mRNA and protein levels; in fact, it increased VEGFR2 protein levels (Fig. 1b, Supplementary Fig. S3d). Moreover, p190RhoGAP and TFII-I double knockdown produced similar or slightly higher levels of VEGFR2 mRNA and protein expression (Fig. 1b, Supplementary Fig. S3d) compared to either p190RhoGAP or TFII-I knockdown alone. These results raised the intriguing possibility that an antagonist of TFII-I activity might exist that also contributes to p190RhoGAP-dependent control of VEGFR2 expression.

GATA2 upregulates VEGFR2

GATA2 is another transcription factor that binds the VEGFR2 promoter and regulates its activity²³24. Because many GATA family members antagonize the effects of other transcription factors at promoter sites²⁵28, we explored whether GATA2 mediates p190RhoGAP-dependent control of VEGFR2 expression by opposing TFII-I activity. When p190RhoGAP was knocked down, GATA2 levels in the nucleus increased dramatically (> 10-fold) relative to control cells even though total GATA2 levels remained the same, as shown by immunoblotting (Supplementary Fig. S3e) and fluorescence microscopy (Fig. 1c). p190RhoGAP also co-immunoprecipitated with GATA2 and vice versa (Supplementary Fig. S3f). Importantly, this increase in nuclear GATA2 was significantly higher than the 1.5-fold increase in nuclear TFII-I produced using p190RhoGAP siRNA (Supplementary Fig. S2a,b), suggesting that p190RhoGAP binds and sequesters GATA2 in the cytoplasm more efficiently than TFII-I in capillary cells. GATA2 knockdown using siRNA decreased VEGFR2 promoter activity, as well as levels of mRNA and protein (Supplementary Fig. S4a,b) as previously observed²³, while GATA2 overexpression produced the opposite effects. These effects of GATA2 siRNA on VEGFR2 promoter activity and protein expression were specific as they were reversed by Gata2 reconstitution (Supplementary Fig. S3c). Interestingly, p190RhoGAP knockdown increased the expression of VEGFR2 mRNA and protein, and double knockdown with GATA2 inhibited these effects (Fig. 1d, Supplementary Fig. S4c). Thus, GATA2 appears to upregulate VEGFR2 promoter activity and mediate the effects of p190RhoGAP on VEGFR2 gene expression in capillary cells. However, knockdown of p190RhoGAP, which releases more GATA2 than TFII-I and increases VEGFR2 expression, did not increase VEGFR2 promoter activity (Supplementary Fig. S3d). This may be because cellular TFII-I levels are 5 times higher than GATA2 levels (not shown). We also measured VEGFR2 promoter activity using only a portion of its promoter, and hence, TFII-I and GATA2 also may exert regulatory activities at other promoter sites; alternatively, p190RhoGAP might elicit signals that influence mRNA stability.

We next asked whether GATA2 and TFII-I directly antagonize each other. Simultaneous knockdown of TFII-I and GATA2 abrogated each other's effects on VEGFR2 promoter activity (Fig. 1e), mRNA production (Fig. 1f) and protein expression levels (Supplementary Fig. S4e), and simultaneous overexpression of TFII-I and GATA2 produced similar effects (Supplementary Fig. S4d,e). These effects were specific for VEGFR2 as knockdown of TFII-I or GATA2 did not alter expression of VEGFR1 or VEGFR3 in HMVE cells (Supplementary Fig. S5a). Importantly, although p190RhoGAP is a Rho inhibitor, altering Rho activity with constitutively active RhoA, membrane-permeable C3 exoenzyme or siRNA directed to another Rho-inhibiting RhoGAP (p73RhoGAP) ²⁹ did not change VEGFR2 mRNA or protein levels (Supplementary Fig. S5b,c). Thus, p190RhoGAP appears to control VEGFR2 expression solely by its ability to sequester these transcription factors.

Analysis of this mechanism of functional cross-antagonism revealed that GATA2 and TFII-I associate with each other, as GATA2 co-immunoprecipitated with TFII-I, and vice versa (Supplementary Fig. S6a). Furthermore, GATA2 even binds to TFII-I in p190RhoGAP knockdown cells and to p190RhoGAP in TFII-I knockdown cells (Supplementary Fig. S6b). These results suggest that the various heterodimeric combinations of TFII-I, GATA2, and p190RhoGAP exist in separate pools, and that these heterodimers then associate to form a larger ternary complex (Supplementary Fig. S1).

Chromatin immunoprecipitation (ChIP) analysis revealed that TFII-I knockdown cells exhibited increased recruitment of GATA2 to the GATA binding site (-150-+150) compared to control cells and vice versa (Fig. 1g), whereas control IgG did not immunoprecipitate these DNAs (Fig. 1g). p190RhoGAP knockdown decreased recruitment of TFII-I, but not GATA2, to this promoter site, which resulted in a relative net increase in GATA2 recruitment to this site (Supplementary Fig. S6c). TFII-I and GATA2 therefore compete with each other for occupancy of a common region of the VEGFR2 promoter, which is controlled by p190RhoGAP.

Mechanical control of VEGFR2

Soluble growth factors, integrin binding to ECM and mechanical distortion of the cytoskeleton all regulate p190RhoGAP activity in cells³⁰. Cell binding to growth factors and adhesive contact formation with ECM also control VEGFR2 expression³¹32, and soluble mitogens (5% serum plus VEGF, bFGF and PDGF) increase nuclear translocation of TFII-I and GATA2 in HMVE cells (Supplementary Fig. S7a). We therefore next explored whether changes in mechanical interactions between cells and ECM regulate this pathway as well. When HMVE cells were cultured in the absence of mitogens on fibronectin-coated polyacrylamide gels with different elasticity (Young's moduli of 150 to 4000Pa), they appeared round on the soft gels, while they flattened on the stiffer gels, as previously observed³³34, which is based on differences in their ability to physically resist cell traction forces⁸3334. Nuclear GATA2 levels were significantly higher in cells on the stiffer gels, whereas nuclei exhibit similar high levels of TFII-I regardless of ECM stiffness (Fig. 2a), and similar results were obtained in the presence of multiple soluble factors or VEGF alone (Supplementary Fig. S7b). Moreover, VEGFR2 mRNA and protein levels were higher in cells on the stiffer (4000 Pa) gels (Fig. 2b, Supplementary Fig. S7c). Interestingly, this relatively stiff, but still compliant, ECM gel appeared to support optimal responsiveness of this signaling pathway as VEGFR2 mRNA and protein levels were significantly lower in cells cultured on rigid glass ECM substrates (Fig. 2b, Supplementary Fig. S7c). ECM elasticity may therefore regulate VEGFR2 expression preferentially via GATA2, particularly over the stiffness range analyzed in this study.

Fig. 2

Matrix elasticity controls VEGFR2 expression via TFII-I and GATA2

a) Immunofluorescence micrographs showing GATA2 and TFII-I distribution in HMVE cells cultured on the fibronectin-coated gels of different stiffness (Young's moduli of 150 and 4000 Pa) in EBM2 with 0.3% serum (bar, 5μm). b)VEGFR2 mRNA level in HMVE cells cultured on rigid fibronectin-coated glass or the gels of different elasticity (150, 1000 and 4000 Pa; normalized to that in cells on the softest gels; *, p<0.01, **, p<0.05). c)VEGFR2 mRNA levels in HMVE cells on soft gels (150 Pa) transfected with TFII-I or GATA2 siRNA alone, or in combination (normalized to cells on the stiffest gels; *, p<0.01). d)VEGFR2 mRNA levels in HMVE cells on soft gels transfected with p190RhoGAP or GATA2 siRNA alone, or in combination (normalized to cells on stiffest gels; *, p<0.01). Error bars represent s.e.m. of 3 replica experiments.

Further analysis revealed that TFII-I knockdown restored VEGFR2 expression in cells on soft gels to levels similar to those in cells on stiff gels, and double knockdown with GATA2 (which decreases VEGFR2 expression) abrogated TFII-I's effects (Fig. 2c, Supplementary Fig. S7d). p190RhoGAP knockdown also increased expression of VEGFR2 mRNA and protein, and double knockdown with GATA2 inhibited these effects on soft gels (Fig. 2d, Supplementary Fig. S7d). These data indicate that p190RhoGAP and the mutually antagonistic TFII-I and GATA2 transcription factors mediate the effects of ECM elasticity on VEGFR2 expression in these cells, which is preferentially shifted to GATA2 on stiffer gels.

Transcriptional control of angiogenesis

We next explored the functional relevance of this antagonism between GATA2 and TFII-I by analyzing capillary cell migration and differentiation (tube formation) in vitro. When HMVE cells were transfected with TFII-I siRNA and analyzed using the Transwell migration assay, VEGF-stimulated cell motility increased by 2-fold, whereas this induction was prevented by knocking down GATA2 simultaneously with TFII-I (Fig. 3a). The VEGFR2 kinase inhibitor SU5416 totally inhibited the migration of these cells (Fig. 3a), confirming that these effects are mediated by VEGFR2 signaling. Moreover, GATA2 overexpression increased VEGF-induced cell migration by 3-fold, and simultaneous overexpression of GATA2 and TFII-I abolished these effects (Fig. 3a). These effects were specific in that reconstitution of TFII-I or Gata2 reversed the effects of knocking down TFII-I or GATA2, respectively (Supplementary Fig. S8a).

Fig. 3

Antagonism between GATA2 and TFII-I controls capillary cell migration and tube formation in vitro

a) The motility of HMVE cells transfected with human siRNAs or transduced with lentiviral vectors encoding GATA2 or TFII-I, alone or in combination, was quantitated using the Transwell migration assay (*, p< 0.01). Where indicated, VEGF (10 ng/ml) was added to the lower chamber and SU5416 was added in both chambers. Error bars represent s.e.m. of 3 replica experiments. b) Micrographs showing in vitro tube formation induced by VEGF (10 ng/ml) in HMVE cells transfected with siRNAs or transduced with lentiviral vectors encoding GATA2 or TFII-I, alone or in combination.

We then studied capillary tube formation by HMVE cells cultured within the ECM gel, Matrigel, which supports angiogenesis in part because of its flexibility. Tube formation was stimulated by VEGF in a dose-dependent manner, whereas bFGF and PDGF were ineffective, and the effects of VEGF were inhibited by SU5416 (Supplementary Fig. S8b). GATA2 knockdown and TFII-I overexpression suppressed VEGF-stimulated capillary development, and once again either knocking down or overexpressing both transcription factors simultaneously negated these effects (Fig. 3b, Supplementary Fig. S8c). TFII-I knockdown or GATA2 overexpression alone did not produce significant effects on tube formation, apparently because it was already optimally stimulated under these conditions (Fig. 3b); knockdown or overexpression of p190RhoGAP also did not alter tube formation (Supplementary Fig. S8d). Thus, functional antagonism between GATA2 and TFII-I at the level of VEGFR2 transcription translates into biologically relevant changes in capillary cell behavior that are required for formation of 3D capillary networks.

Control of angiogenesis in vivo

We next examined whether ECM mechanics governs vessel formation in vivo using a modified Matrigel implant assay. Maximal levels of cell infiltration, capillary blood vessel formation, and VEGFR2 expression were observed in Matrigel with intermediate stiffness (800 Pa) compared to cells in more or less rigid gels (900 or 700 Pa, respectively)(Fig. 4a,c and Supplementary Fig. S9a,b). VEGFR2, GATA2, and TFII-I all localized within cells lining CD31-positive staining microvessels (Supplementary Fig. S9c). The finding that the optimal ECM stiffness required for angiogenesis in vivo (800 Pa) was different than that observed in vitro (4000 Pa) is likely due to different requirements by cells when cultured on 2D ECM versus within a 3D ECM gel; it also may relate to how the Matrigel is remodeled over time in vivo.

Fig. 4

Matrix elasticity controls vessel formation via TFII-I and GATA2 in vivo

Light (H&amp;E-stained) (a,b top), immunofluorescence micrographs (a,b bottom) and vessel densities per high power field (c,d) showing cell infiltration and VEGFR2 expression in Matrigel with different elasticity implanted in mice for 7 days without (a,c) or with TFII-I or Gata2 siRNAs (b,d) (bar, 25 μm). The number of VEGFR2-positive blood vessels was normalized to that in the 700Pa gels in (c) and to that in gels treated with control siRNA in (d) (n=6, mean ± S.E.M., *, p< 0.01).

We next performed siRNA-mediated gene knockdown in the in vivo Matrigel assay. Interestingly, TFII-I knockdown increased cell infiltration, capillary vessel formation and VEGFR2 expression by cells in the soft gels so that it mimicked the behavior of cells on the intermediate stiffness gel, whereas Gata2 knockdown produced the opposite effects (Fig. 4b,d and Supplementary Fig. S10a), and p190RhoGAP knockdown also increased the level of vessel formation in the soft gels (Supplementary Fig. S10b). These results indicate that TFII-I, GATA2, and p190RhoGAP mediate the signaling effects of ECM mechanics on vessel formation in vivo.

To unequivocally confirm the functional and clinical relevance of regulatory interactions between TFII-I, GATA2, and p190RhoGAP, we modulated their expression in the neonatal mouse retina because angiogenesis in this growing organ is tightly regulated by VEGF and its receptors¹⁴3536. TFII-I, GATA2, and VEGFR2 localized to the three layers of the retina where blood vessels are located in postnatal day 15 mice (Supplementary Fig. S11a,b). Consistent with in vitro data, knocking down TFII-I using intravitreal injection of siRNA to p14 mice resulted in increased Vegfr2 expression and the appearance of highly tortuous dilated vessels and a significant increase in vascular density in the retina, whereas Gata2 knockdown suppressed Vegfr2 expression, disrupted capillary network formation and decreased vascular density (Fig. 5a, Supplementary Fig. S12a,b). Knockdown of p190RhoGAP also increased Vegfr2 expression and vessel density, but it resulted in a slightly different capillary growth pattern, perhaps in part because of its known effects on vascular permeability³⁷ (Fig. 5a and Supplementary Fig. S12a,b). Similar effects were observed in earlier (P5) retina that contain rapidly growing microvessels (Supplementary Fig. S13). Furthermore, overexpression of TFII-I and GATA2 produced opposite effects on vascular density in P14 retina, and simultaneous overexpression of both factors abolished these effects, confirming our in vitro findings (Fig. 5b and Supplementary Fig. S12c). Knockdown of p190RhoGAP or TFII-I did not significantly change Vegf expression levels in the retina, whereas Gata2 knockdown increased (rather than decreased) its expression (Supplementary Fig. S12d). The decrease of Vegfr2 expression appears to be sufficient to abrogate this Vegf response in Gata2 knockdown retina, and thereby inhibit angiogenesis. Hence, p190RhoGAP controls Vegfr2 expression and vascular development by modulating the balance between TFII-I and GATA2 activities such that TFII-I activity dominates (and Vegfr2 expression is suppressed) in retina.

Fig. 5

TFII-I, GATA2, and p190RhoGAP regulate retinal vessel formation in vivo

a) Confocal fluorescence micrographs showing projected images of vessels stained with Alexa 594-isolectin in the retina of a control mouse eye versus eyes transfected with TFII-I, Gata2, or p190RhoGAP siRNAs (bar = 0.1 mm). b) Confocal fluorescence micrographs showing projected images of vessels stained with Alexa 594-isolectin in the retina of a control mouse eye versus eyes overexpressed TFII-I or GATA2, alone or both in combination (bar = 0.1 mm).

Discussion

Transcription factors change their activities in a spatiotemporal manner during development, and thereby specify cell fate³⁸39. Here we show that p190RhoGAP, which has been shown to be regulated by growth factors, ECM binding and cytoskeletal distortion also controls VEGFR2 expression, as well as angiogenesis in vitro and in vivo, by altering the balance between two mutually antagonistic transcription factors: TFII-I and GATA2 (Supplementary Fig. S1). Moreover, we demonstrate that p190RhoGAP and this downstream transcriptional control mechanism are controlled by mechanical signals conveyed by variations in ECM elasticity. This mechanism is analogous to other developmental mechanisms used by hematopoietic cells and mammary epithelium³⁸40²⁸; however, this is the first demonstration that transcriptional cross-antagonism can govern histodifferentiation and tissue morphogenesis, and be sensitive to mechanical as well as chemical cues.

Our data suggest that an appropriate level of ECM stiffness may be required for optimal VEGFR2 expression and vascular development in vitro and in vivo. In fact, the fates of different cell types are exquisitely sensitive to distinct ECM elasticity values that often match those exhibited by their host tissues⁴¹. Abrupt local changes in ECM mechanics also accompany the switch between active growth and quiescent differentiation of functional capillary networks in living tissues⁴². Cell rounding suppressed p190RhoGAP activity within 1 to 2 hours by altering its binding to the cytoskeletal protein, filamin¹⁶. A similar cytoskeleton-based effect could mediate the effect of ECM elasticity on p190RhoGAP activity at later times, and thereby control the TFII-I:GATA2 balance and VEGFR2 transcription. Since VEGFR2 is expressed in neurons as well as in endothelial cells in retina, some of the effects of gene knockdown we observed may not be specific to capillary cells. But even these alterations might be relevant for control of vascular development because neuron-vessel interactions are indispensable for normal microvasculature function and patterning⁴³.

In summary, these data reveal a previously unknown mechanosensitive signaling pathway that controls VEGFR2 promoter activity and expression, and which represents a point of convergence for all three classes of microenvironmental signals that regulate capillary morphogenesis. Development of specific modifiers of this pathway could therefore lead to novel therapeutic approaches for various angiogenesis-dependent diseases, including proliferative retinopathy, arthritis, and cancer in the future.

Methods Summary

Expression of TFII-I, GATA2, and VEGFR2 were evaluated by qRT-PCR and immunoblotting. A luciferase reporter assay was used to measure VEGFR2 promoter activity. To test the effects of TFII-I and GATA2 on VEGFR2 expression and angiogenesis in vitro, siRNA-mediated knockdown or lentiviral transduction was performed in HMVE cells. In vitro analysis of angiogenesis was carried out using Transwell migration and Matrigel tube formation assays, and similar results were obtained using both native and growth factor-reduced forms of Matrigel. A subcutaneous Matrigel angiogenesis assay was used to analyze the effects of ECM mechanics on capillary formation in vivo. Retinal vessel formation was also studied in newborn C57BL/6 mice, and gene expression was manipulated in whole living retina by intravitreal injection of siRNA or DNA into the eye at P5 or P14.

Methods

Materials

Anti-GATA2 polyclonal antibody was from Abcam (Cambridge, MA); anti-HA monoclonal antibody from Covance (Princeton, NJ); monoclonal antibodies against TFII-I, p190RhoGAP, PECAM (CD31), and paxillin from Transduction laboratory (Lexington, KY); anti-GAPDH antibody from Chemicon (Temecula, CA); anti-lamin monoclonal antibody from Upstate (Lake Placid, NY); anti-VEGFR2 antibody from Cell Signaling (Danvers, MA); and anti-myc antibody from Santa Cruz (Santa Cruz, CA). Protein G-sepharose was from Amersham-Pharmacia (Uppsala, Sweden) and SU5416 was from Calbiochem (SanDiego, CA). VEGF-A was from NIH; bFGF and PDGF were from Roche (Basel, Switzerland) and Biovision (Mountain View, CA) respectively. Cell permeable Rho inhibitor (C3 exoenzyme) was from Cytoskeleton (Denver, CO). HMVE and HUVE cells (Cambrex, Walkersville, MD) were cultured in EBM2 medium containing 5% FBS and growth factors (VEGF, bFGF, and PDGF) for all experiments¹⁶ except the nuclear translocation assays in which we used EBM2 with 0.3% serum. Cells were plated on plastic dishes for molecular biochemical assays, and on fibronectin-coated glass coverslips for cell staining, except for experiments using flexible substrates.

Plasmid construction and gene knockdown

pGL3-VEGFR2-225 (-225+268), -570 (-570+268), -780 (-780+268) were constructed using the reverse transcription (RT)-PCR with genomic DNA from HUVE cells, and subcloned into pGL3 vector (Promega) at the SacI/XhoI sites. For pGL3-VEGFR2-780Inr-MUT, the Inr (CACT to GTGC) was point mutated using the QuickChange mutagenesis kit (Stratagene, La Jolla, CA). For lentivirus construction, human myc-TFII-I (delta isoform), HA-GATA2, and HA-p190RhoGAP were constructed by PCR using template plasmids from Open Biosystems (Huntsville, AL) and H. Sabe (OBI, Osaka, Japan). For retrovirus construction, mouse myc-TFII-I (gamma isoform) and HA-Gata2 were constructed using template plasmids from Open Biosystems and T. Nakano (Osaka University, Osaka, Japan), respectively. To generate delta isoform of mouse TFII-I, 256-274 and 294-314aa were deleted from the gamma isoform. Construction of constitutively active RhoA and generation of viral vectors were previously described ³⁷. For gene knockdown, siRNA transfection was performed as described¹⁶. siRNA sequences are shown in Supplementary Table S1.

Biochemical Methods

For luciferase reporter assays, HUVE cells were transfected using Superfect (QIAGEN) and assayed using Dual-Luciferase reporter assay kit (Promega). Luciferase activity was measured in duplicate using a luminometer (TD20/20, Turner Designs). Cytoplasmic and nuclear cell extracts were prepared with a Nuclear Extraction Kit (Chemicon).

Molecular biological methods

Quantitative RT-PCR was performed with the Quantitect SYBR Green RT-PCR kit (QIAGEN) using ABI7300 real-time PCR system (Applied Biosystems, Foster City, CA); β2 microglobulin or cyclophilin controlled for cDNA content. The primers used are shown in Supplementary table S2. For ChIP assay, DNA from HMVE cells transfected with control or TFII-I siRNA was immunoprecipitated with the GATA2 antibody or control immunoglobulin (Jackson Immuno Research), or vice versa, according to the manufacturer instructions (Active Motif). GATA2- and TFII-I-binding region was amplified using primers, 5′-GTAAATGGGCTTGGGGAGCTG-3′ and 5′-GGCGGCTGCAGGGGCGTCT-3&amp;prime;.

Cell analysis methods

Flexible polyacrylamide gel culture substrates were prepared as described³⁴44 and coated with fibronectin (1 μg/cm^{). Substrate flexibility was controlled by varying the acrylamide (2–4%) and the bis-acrylamide (0.1–0.5%) concentration; the Young's modulus (stiffness) was determined as described45}. HMVE cells were cultured for 6 hrs on the gels and immunostaining was performed and analyzed using confocal Leica SP2 microscope ¹⁵. For cell migration assay, Transwell membranes (Coster, NY) were coated with 0.5% gelatin, and cells were seeded (10 ^{cells/100 μl) with 0.3%FBS/EBM2. Cells were stained with Giemsa solution 16 h later, and counted in 10 random fields (×400). For the} in vitro angiogenesis assay, HMVE cells (10 ^{cells/150 μl of EBM-2) were plated on MatrigelTM (BD biosciences), and incubated for 12-16 hrs in the presence of VEGF (10 ng/ml); tube formation was assessed in 10 random fields (4×).}

In vivo Matrigel implantation assay

All animal studies were reviewed and approved by the animal care and use committee of Children's Hospital Boston. Matrigel plugs with different elasticity were cast in 4 × 4mm (ID×H) polydimethylsiloxane (PDMS) molds and incubated at 37°C overnight before implanting them subcutaneously on the backs of C57BL/6 mice. The stiffness of the Matrigel was modulated over a narrow range (i.e., without making it rigid) by altering ECM protein cross-linking using a microbial transglutaminase (2.5-20 U/g; Ajinomoto, Japan) ⁴⁶. The storage modulus (G′) of the gels was measured with an AR-G2 rheometer (TA Instruments) using a standard 20 mm aluminum parallel plate (1Hz, 1% strain, 37°C). The Young's modulus for an equivalent polyacrylamide gel was calculated by E = 2*G′(1+ν) using an average Poisson's Ratio (ν) of 0.5. After 7 days, the PDMS molds containing the gel plugs were harvested, fixed, and cryosectioned. H&amp;E staining and immunostaining were performed as described ¹⁵⁴⁷. Stacks of optical sections (20 μm thick) were compiled to form 3D images using Velocity4.4 (Improvision, PerkinElmer). Vessel formation was evaluated by counting the number of vessels that stained positive for fluorescein-conjugated ConA injected into the tail vein or VEGFR2 in 5 different areas (n=6), individual cell nuclei were identified by DAPI staining. Cell recovery solution (BD Biosciences) was used to collect cells from the recovered Matrigel plugs. In some studies, siRNA (7 μg) was mixed into the Matrigels and 10 μg of additional siRNA was injected into the implanted Matrigel after 3 days (n=6); the same amount of scrambled siRNA was used as a control. Gene knockdown was evaluated by counting the number of cells expressing each gene in the five different regions.

In vivo analysis for retinal vessel formation

For gene knockdown in living retina, siRNA (0.5μg) for each gene was injected intravitreally into one eye of C57BL/6 mouse at P5 or P14, and the same amount of control siRNA was injected to the other eye. To overexpress genes, the complex of DNA (0.5μg) for TFII-I and/or GATA2 and jetPEI transfection reagent (Polyplus transfection, CA) was injected to the eye at P14. Vascular network formation in the retina was assessed 2 days after injection using flat-mounted, fluorescein-conjugated isolectin-staining and immunohistochemical analysis (n=7). Retinal RNA was purified and gene expression was quantified using qRT-PCR (n=7). Quantification of vessel density was performed with Adobe photoshop.

Supplementary Material

Supple Fig S 1

Supple Fig S 10

Supple Fig S 11

Supple Fig S 12

Supple Fig S 13

Supple Fig S 2

Supple Fig S 3

Supple Fig S 4

Supple Fig S 5

Supple Fig S 6

Supple Fig S 7

Supple Fig S 8

Supple Fig S 9

SuppleTableS1

SuppleTableS2

Acknowledgments