The endothelial transcription factor ERG mediates Angiopoietin-1-dependent control of Notch signalling and vascular stability

Abstract

Notch and Angiopoietin-1 (Ang1)/Tie2 pathways are crucial for vascular maturation and stability. Here we identify the transcription factor ERG as a key regulator of endothelial Notch signalling. We show that ERG controls the balance between Notch ligands by driving Delta-like ligand 4 (Dll4) while repressing Jagged1 (Jag1) expression. In vivo, this regulation occurs selectively in the maturing plexus of the mouse developing retina, where Ang1/Tie2 signalling is active. We find that ERG mediates Ang1-dependent regulation of Notch ligands and is required for the stabilizing effects of Ang1 in vivo. We show that Ang1 induces ERG phosphorylation in a phosphoinositide 3-kinase (PI3K)/Akt-dependent manner, resulting in ERG enrichment at Dll4 promoter and multiple enhancers. Finally, we demonstrate that ERG directly interacts with Notch intracellular domain (NICD) and β-catenin and is required for Ang1-dependent β-catenin recruitment at the Dll4 locus. We propose that ERG coordinates Ang1, β-catenin and Notch signalling to promote vascular stability.

Figure 1. ERG regulates endothelial Notch signalling.

(a) Western blot (WB) analysis of Notch intracellular domain (NICD) expression in control (siCtrl) and ERG-deficient (siERG) HUVEC (n=4). (b) RBP-J TP-1 Notch reporter activity in control and ERG-deficient HUVEC (n=4). (c) qPCR of Notch target gene expression in siCtrl and siERG-treated HUVEC: Hes1, Hey1, Nrarp and p21 (n=4). (d) Microarray and PCR screen analysis of differential gene expression in HUVEC was performed at 24 and 48 h after ERG inhibition, with fold change of selected genes represented as high (red) and low (blue) expression compared to the median (grey). (e) qPCR analysis of Hes1 and Hey1 Notch target gene expression in siCtrl and siERG-transfected HUVEC stimulated with Dll4 or control BSA (n=4). (f) RBP-J TP-1 Notch reporter activity in control and ERG-deficient HUVEC plated on Dll4 or BSA (n=4). (g) WB analysis and quantification of NICD expression in siCtrl and siERG-transfected HUVEC stimulated with Dll4 or BSA (n=4). (h) qPCR analysis of Notch1 and Notch4 mRNA expression in siCtrl and siERG-transfected HUVEC stimulated with Dll4 or BSA (n=4). All graphical data are mean±s.e.m., *P<0.05, **P<0.01, ***P<0.001, Student’s t-test.

Figure 2. ERG transcriptionally activates Dll4 and represses Jag1 expression.

(a) mRNA levels of Dll4 in HUVEC treated with siCtrl or siERG for 24 and 48 h (n=6). (b) Representative WB and quantification of Dll4 protein expression in siCtrl and siERG-treated HUVEC for 24 and 48 h (n=3). (c) mRNA levels of Jag1 in siCtrl and siERG-treated HUVEC for 24 and 48 h (n=6). (d) Representative WB and quantification of Jag1 expression in siCtrl and siERG-treated HUVEC for 24 and 48 h (n=6). (e) Putative ERG binding sites (grey bars) are located within the Dll4 promoter downstream of the transcription start site (TSS) (arrow); ENCODE sequence conservation between 100 vertebrates is shown across this region. ENCODE ChIP-seq data profiles for H3K4me3, H3K27Ac and RNA polymerase II (RNA pol) in HUVEC indicate open chromatin and active transcription. Location of qPCR amplicon covering region R1 is indicated. (f) ChIP-qPCR using primers to region R1 on ERG-bound chromatin from siCtrl or siERG-treated HUVEC. Primers for a region within exon11 of the Dll4 gene were used as negative control. Data are shown as fold change over IgG (n=3). (g) Dll4 promoter luciferase reporter assay. ERG cDNA expression plasmid (pcDNA-ERG) or empty expression plasmid (pcDNA) were co-transfected with a Dll4 promoter-luciferase construct (pGl4-Dll4, covering region R1) in HUVEC, and luciferase activity was measured. Values represent the fold change in relative luciferase activity over the empty pGL4 vector alone (n=4). (h) Putative ERG binding sites (grey bars) located within the Jag1 genomic locus. TSS is indicated (arrow); ENCODE sequence conservation between 100 vertebrates and ChIP-seq data profiles for H3K4me3, H3K27Ac and RNA polymerase II in HUVEC are shown across this region. Location of qPCR amplicons covering R1, R2 and R3 are indicated. (i) ChIP-qPCR using primers covering Jag1 promoter regions R1, R2, R3 and Ctrl region on ERG-bound chromatin from siCtrl or siERG HUVEC (n=4). (j) Control or Jag1 promoter luciferase construct (pGl3-Jag1, covering regions R1 and R2) activity after siERG treatment. Results are expressed as luciferase activity relative to siCtrl-treated cells (n=3). All graphical data are mean±s.e.m., *P<0.05, **P<0.01, ***P<0.001, Student’s t-test.

Figure 3. ERG controls Dll4 and Jag1 expression in the plexus of the mouse retina.

Representative images and quantification of Dll4 and Jag1 (green) staining of postnatal day 6 retinal vessels in the (a and b) stable plexus and (c and d) angiogenic front from Erg^fl/fl and Erg^iEC-KO mice. Retinas are co-stained for isolectin B4 (IB4, red) and ERG (white, presented as a separate channel). Quantification represents the ratio between the sum of pixel intensity and isolectin B4 area (n=4 fields per mouse, n=4 mice per genotype). Scale bar, 70 μm. Arteries (A) and veins (V) are indicated. Arrowheads highlight tip cells at the angiogenic front. All graphical data are mean±s.e.m., *P<0.05, **P<0.01, ***P<0.001, Student's t-test.

Figure 4. ERG is required for Ang1 regulation of Dll4 and Jag1 expression and Notch signalling.

(a) Ang1-induced Notch transcriptional activity was determined by transfecting control and ERG-deficient cells with RBP-J TP1 reporter construct in the presence or absence of Ang1 (250 ng ml⁻¹ for 6 h) (n=3). (b) Ang1-induced transcription of downstream Notch target gene Hey1 in control and ERG-deficient HUVEC (n=4). (c) qPCR and (d) WB analysis of Dll4 expression in extracts of siCtrl and siERG HUVEC treated in the presence or absence of Ang1 (250 ng ml⁻¹ for 6 h) (n=5). (e) Jag1 mRNA and (f) protein expression in siCtrl and siERG HUVEC treated in the presence or absence of Ang1 (250 ng ml⁻¹ for 6 h) (n=5). (g) Model: ERG mediates Ang1-dependent reciprocal regulation of Dll4 and Jag1 in endothelial cells. All graphical data are mean±s.e.m., *P<0.05, **P<0.01, ***P<0.001, Student's t-test.

Figure 5. Ang1 requires ERG to promote vascular stability and drive Dll4 expression in vivo.

(a) Erg^fl/+ and Erg^cEC-het mice were subdermally injected with PBS, VEGF, Ang1, or a combination of reagents (Ang1+VEGF) (50 ng in 50 μl) for 1 h. Mice received intravenous injection of FITC-dextran 15 min before collecting skin samples. Representative images of whole-mount skin samples perfused with FITC-dextran (green). (b) Vessel permeability was quantified by measuring the intensity of FITC-dextran per field (n=5 fields per mouse, n=3 mice per genotype). The average of the mean intensity per mouse was converted to fold change compared to Erg^fl/+ mice injected with PBS. (c) qPCR analysis of ERG expression in extracts of skin samples from control Erg^fl/+ and ERG hemi-deficient (Erg^cEC-het) mice (n=3). (d) Representative images of corresponding skin samples showing extravasation of FITC-dextran (green) from blood vessels stained for isolectin B4 (IB4, purple). qPCR analysis of (e) Dll4 and (f) Jag1 expression in extracts of skin samples from control Erg^fl/+ and ERG hemi-deficient (Erg^cEC-het) mice treated with intradermal injection of PBS or Ang1 for 1 h (n=3). All graphical data are mean±s.e.m., *P<0.05, **P<0.01, ***P<0.001, Student’s t-test. $: Erg^cEC-het mice compared to Erg^fl/+ mice for each treatment, $ P<0.05, $$ P<0.01, $$$ P<0.001, Student’s t-test.

Figure 6. Ang1 induces ERG phosphorylation and binding to the Dll4 locus via PI3K/Akt.

(a) ChIP-qPCR assays of confluent HUVEC treated with Ang1 (250 ng ml⁻¹). ERG or IgG-immunoprecipitated DNA was analysed by qPCR with primers to −16 kb, −12 kb, promoter, intron 3 and +14 kb of Dll4 locus and negative control region exon 11. Results are expressed as fold change compared to IgG (n=3). *P<0.05, **P<0.01, ***P<0.001 indicates significant ERG enrichment compared to IgG and ^#P<0.05 indicates significant change in Ang1-induced ERG enrichment compared to control conditions. (b) Confluent HUVEC were pre-treated with LY294002 (20 μM) or Akt inhibitor IV (8 μM) and treated with Ang1 (250 ng ml⁻¹) or DMSO for 30 min. Proximity ligation assay analysis of ERG phosphorylation was performed using rabbit anti-ERG and mouse anti-phospho-serine antibodies (n>97 cells quantified per condition). Scale bar, 20 μm. (c) ChIP-qPCR analysis of HUVEC treated with Ang1 and LY294002 or Akt inhibitor IV. ERG or IgG-immunoprecipitated DNA was analysed by qPCR with primers to the Dll4 regulatory regions indicated and negative control region exon 11. Results are expressed as fold change compared to IgG (n=4). Graphical data are mean±s.e.m., *P<0.05, **P<0.01, ***P<0.001, Student's t-test.

Figure 7. Ang1 induction of β-catenin occupancy at Dll4 promoter and enhancers requires ERG.

ChIP-qPCR analysis of siCtrl and siERG-treated HUVEC, in the presence or absence of Ang1 (250 ng ml⁻¹). Chromatin was immunoprecipitated with (a) an anti-NICD antibody, (b) an anti-β-catenin antibody or control IgG. Immunoprecipitated DNA was analysed by qPCR with primers to −16 kb, −12 kb, promoter, intron 3 and +14 kb of Dll4 locus. Primers covering a negative control region within exon 11 were also used. Results are expressed as fold change enrichment compared to IgG (n=3). (c) HUVEC lysates were immunoprecipitated with an anti-ERG antibody. Immunoprecipitates were analysed by immunoblotting (IB) with anti-ERG, anti-NICD and anti-β-catenin antibodies. (d) Confluent HUVEC were pre-treated with LY294002 (20 μM) or Akt inhibitor IV (8 μM) and treated with Ang1 (250 ng ml⁻¹) or DMSO for 30 min. Proximity ligation assay analysis of localization of ERG-β-catenin interaction was performed using rabbit anti-ERG and mouse anti-β-catenin antibodies (n>99 cells quantified per condition). Scale bar, 20 μm. All graphical data are mean±s.e.m., *P<0.05, **P<0.01, ***P<0.001, Student's t-test.

Figure 8. Reciprocal regulation of Notch signalling and ERG expression.

(a) Dll4 mRNA expression in siCtrl and siERG-transfected HUVEC treated in the presence or absence of the γ-secretase inhibitor DAPT (n=4). (b) Dll4 mRNA expression in HUVEC transfected with ERG cDNA expression plasmid (pcDNA-ERG) or an empty expression plasmid (pcDNA) and treated in the presence or absence of the γ-secretase inhibitor DAPT (n=3). (c) ERG mRNA expression in siCtrl and siERG-transfected HUVEC treated in the presence or absence of DAPT (n=4). (d) mRNA expression of ERG and the Notch target gene Hey1 in HUVEC stimulated with control BSA or Dll4 (n=4). Representative images and quantification of ERG (red) staining of P6 retinal vessels in the (e) vascular plexus and (f) angiogenic front from control (Rbpj^fl/fl) and Rbpj^iΔEC mice. Retinas are co-stained for isolectin B4 (IB4, green). Quantification represents the ratio between the sum of pixel intensity and isolectin B4 area (n=4 fields per mouse, n=4 mice per genotype). Scale bar, 70 μm. Arteries (A) and veins (V) are indicated. All graphical data are mean±s.e.m., *P<0.05, **P<0.01, ***P<0.001, Student's t-test.

Figure 9. Model: ERG mediates Ang1 potentiation of Dll4/Notch signalling.

(a) Notch-ERG positive feedback loop. ERG drives expression of the Notch ligand Dll4 and is required for endothelial Notch signalling. Notch signalling itself upregulates ERG expression, suggesting that continued Dll4 expression and Notch signalling is maintained through this positive feedback loop. (b) ERG is required for Ang1 induction of Dll4. In confluent cells, Ang1/Tie2 signalling induces PI3K/Akt-dependent ERG phosphorylation (P). This increases ERG binding to the Dll4 gene locus and recruitment of β-catenin. The complex of ERG with NICD and β-catenin mediates Ang1-dependent Dll4/ Notch signalling in confluent EC.