Endothelial Notch signaling directly regulates the small GTPase RND1 to facilitate Notch suppression of endothelial migration

Abstract

To control sprouting angiogenesis, endothelial Notch signaling suppresses tip cell formation, migration, and proliferation while promoting barrier formation. Each of these responses may be regulated by distinct Notch-regulated effectors. Notch activity is highly dynamic in sprouting endothelial cells, while constitutive Notch signaling drives homeostatic endothelial polarization, indicating the need for both rapid and constitutive Notch targets. In contrast to previous screens that focus on genes regulated by constitutively active Notch, we characterized the dynamic response to Notch. We examined transcriptional changes from 1.5 to 6 h after Notch signal activation via ligand-specific or EGTA induction in cultured primary human endothelial cells and neonatal mouse brain. In each combination of endothelial type and Notch manipulation, transcriptomic analysis identified distinct but overlapping sets of rapidly regulated genes and revealed many novel Notch target genes. Among the novel Notch-regulated signaling pathways identified were effectors in GPCR signaling, notably, the constitutively active GTPase RND1. In endothelial cells, RND1 was shown to be a novel direct Notch transcriptional target and required for Notch control of sprouting angiogenesis, endothelial migration, and Ras activity. We conclude that RND1 is directly regulated by endothelial Notch signaling in a rapid fashion in order to suppress endothelial migration.

Figure 1

Ligand Notch activation rapidly induces known and novel targets including RND1 in endothelial cells. (A) Time course of induction after seeding HUVEC onto DLL4-Fc coated plates as compared to Fc coated control plates (TLA assay). Robust activation of expression of known direct Notch targets was achieved as early as 6 h (red arrowhead). Single, double, and triple asterisks indicate p values of < 0.05 (*), < 0.01 (**), and < 0.001 (***), respectively, in all panels. (B–C) The 20 genes with the highest average fold induction in HUVEC (B) or HREC (C) by DLL4 TLA after 6 h. All heatmaps indicate z-score (standard deviations from the mean). RND1 is among the most significantly regulated Notch targets in both EC cell types. Red rectangle indicates RND1 in all heatmaps. (D) Induction time course of known Notch targets in HREC after 15 min of EGTA treatment or treatment with both EGTA and CpE. Robust activation of known direct Notch targets occurs between 30 and 90 min (0.5–1.5 h). (E) Heatmap of the 20 genes most highly upregulated by EGTA induction in HREC within 1.5 h shows that RND1 is rapidly significantly upregulated within the same timeframe as known direct Notch targets. (F) GO pathways significantly enriched in the 340 genes significantly upregulated by DLL4 TLA induction in both HUVEC and HREC (top) or the 1,105 genes significantly upregulated by EGTA (bottom). Inset indicates number of significantly regulated genes contributing to each pathway. The Notch signaling pathway was significantly regulated in both contexts. (G) Overlap between genes significantly upregulated by DLL4 TLA induction in both HUVEC and HREC, genes induced by EGTA, and genes repressed by CpE. (H–J) Volcano plots of gene expression changes in DLL4-stimulated HUVEC, DLL4-stimulated HREC, and CpE-inhibited EGTA-stimulated HREC (left to right, respectively). Uncolored circles indicate p_adj > 0.05, colored circles indicate p_adj < 0.05, red circles = fold change (FC) <|1.2|, green circles = FC >|1.2|.

Figure 2

RND1 is the most highly regulated of the novel Notch targets in multiple contexts. (A) 12 genes were significantly regulated under all in vitro and in vivo RNA-seq screening conditions. Genes in purple are established Notch targets. (B) qPCR confirms that 7 of the 8 novel genes are regulated by DLL4 TLA in HUVEC and HREC, with RND1 showing the highest fold induction and the strongest inhibition by CpE. Black stars indicate significant induction by DLL4, red stars indicate significant inhibition of induction by CpE. (C) All novel targets are induced by EGTA; the most strongly induced gene is RND1 (blue line). (D) EGTA induction (blue or black lines) and CpE inhibition of EGTA (red lines) of the 8 novel putative endothelial Notch targets in HREC. Black stars indicate significant induction, red stars indicate significant inhibition of induction. The y axes are scaled for each gene to visualize induction of different magnitudes. (E) RND1, HIC1, and RGS3 remain induced after 48 h of DLL4 overexpression, while the other novel targets have transient expression that has returned to baseline after 48 h. (F) JAG1 TLA induces expression of most novel endothelial Notch targets at lower levels, suggesting that RND1 is a target of both DLL4 and JAG1 signaling. (G) qPCR of endothelial transcripts immunoprecipitated from the P8 mouse brain using an endothelial-specific RiboTag allele. Input = bulk brain homogenate, IP = immunoprecipitated endothelial mRNA.

Figure 3

The RND1 locus contains an active endothelial specific enhancer with Notch binding sites. (A) RND1 is induced by EGTA at similar rates and magnitude as well-characterized HES and HEY Notch targets. (B) ENCODE database view of the RND1 locus shows a DNAse hypersensitivity peak at the promoter region in all cell types (blue bar) and a second peak upstream that appears only in endothelial cell types (yellow bar). The endothelial-specific peak shows histone methylation patterns consistent with active enhancers (insert box). Gray bars mark CTCF-binding insulator regions. (C) Enhancer region sequence in the RND1 locus with two Notch consensus sequence (green highlight). Also highlighted is the primer sequence for ChIP PCR (orange highlight). (D) ChIP assay of the putative endothelial-specific RND1 enhancer with antibodies against the intracellular domain of Notch1 (N1ICD). N1ICD binds to the RND1 enhancer after EGTA induction (left) or DLL4 overexpression, but binding is blocked with CpE treatment, indicating that active Notch signaling is required. (E) HUVECs transfected with siCon or siRND1 or infected with Control or Notch1IC lentiviral constructs were serum-starved for 3 h, treated with 100 ng/ml EGF, and analyzed for Ras activity by G-LISA after 0, 5, and 10 min. (F) HUVECs as in (E) were treated with 50 nM thrombin and analyzed for RhoA activity by G-LISA after 0, 2, and 5 min.

Figure 4

RND1 knockdown does not rescue Notch-mediated suppression of endothelial proliferation. (A) Transfection with a DLL4-myc expression construct induces DLL4 protein expression in HUVECs 3 days after lentivirus induction. All images are from the same blot, with uncropped version provided in Fig. S4. (B) RND1 siRNA robustly knocks down RND1 in HUVEC with normal levels of Notch signaling (RFP) or Notch signaling activated by DLL4-myc overexpression (DLL4). HUVECs were lentivirally transduced with RFP or DLL4 and transfected the next day with scrambled (siCNT) or siRNA targeting RND1 (siRND1). 3 days after lentivirus infection, HUVECs were harvested and analyzed by qPCR for RND1. (C) Treatment with siRNA1 does not grossly affect DLL4 levels in HUVEC treated as in (B). (D) Treatment with siRND1 does not restore proliferation in HUVEC where Notch signaling has been activated by DLL4. 10⁴ HUVEC treated as in (B) were seeded onto 6-well plates cell and counted daily. Proliferation rate was normalized to counts at day 1.

Figure 5

RND1 is required for Notch-mediated suppression of sprout number and length. (A) Representative images of Fibrin bead angiogenesis (FiBA) assays using lentivirally-transduced HUVECs with full-length RND1 overexpression vectors (RND1OE) or empty vector (Ctrl). HUVEC were coated on cytodex beads, embedded in a fibrin gel with fibroblasts (D551) grown on the surface of the gel to provide growth factors, and allowed to develop for 5 days. (B) After 5 days, overexpression of RND1 reduces the number of sprouts, length of sprouts, and number of tip cells in a highly significant manner. (C) Lentivirally-transduced HUVECs with RFP or DLL4 expression constructs and were transfected with siCNT or siRND1 used for FiBA assays as above. Representative images of FiBA under each condition, scale bar indicates 320 µm. (D) After 5 days, Notch activation by DLL4 suppresses the number of sprouts from each bead, but knockdown of RND1 partially rescues sprout number (left). Notch activation by DLL4 suppresses the length of sprouts from each bead, but knockdown of RND1 partially rescues sprout length (right).

Figure 6

RND1 is required for Notch-mediated suppression of EC migration. (A) Lentivirally-transduced HUVECs with RFP or DLL4 expression constructs were transfected with siCNT or siRND1 and plated in modified Boyden chambers and incubated with 50 ng/ml of VEGF for 6 h. Cells that had migrated through the pores were counted and normalized to their respective controls. DLL4 overexpression activation of Notch signaling suppresses endothelial migration (third column), but knockdown of RND1 rescues the endothelial migration to nearly control levels. (B) Representative images of cells that had migrated through Boyden chamber pores under each condition in (A). (C) Control and siRND1 transfected HUVECs were plated on a Boyden chamber inserts coating with fibronectin (5 µg/ml) and either Fc (control) or 10 µg/ml DLL4-Fc (Notch activation) and incubated with 50 ng/ml of VEGF for 6 h. Physiologic activation of Notch signaling by tethered DLL4 ligand suppresses endothelial migration (third column), but knockdown of RND1 increases migration under endogenous Notch signaling conditions (second column) or ligand-induced Notch signaling (fourth column). (D) Representative images of cells that had migrated through Boyden chamber pores under each condition in (B). The scale bar indicates 210 μm.