The endosomal RIN2/Rab5C machinery prevents VEGFR2 degradation to control gene expression and tip cell identity during angiogenesis

Sprouting angiogenesis is key to many pathophysiological conditions, and is strongly regulated by vascular endothelial growth factor (VEGF) signaling through VEGF receptor 2 (VEGFR2). Here we report that the early endosomal GTPase Rab5C and its activator RIN2 prevent lysosomal routing and degradation of VEGF-bound, internalized VEGFR2 in human endothelial cells. Stabilization of endosomal VEGFR2 levels by RIN2/Rab5C is crucial for VEGF signaling through the ERK and PI3-K pathways, the expression of immediate VEGF target genes, as well as specification of angiogenic ‘tip’ and ‘stalk’ cell phenotypes and cell sprouting. Using overexpression of Rab mutants, knockdown and CRISPR/Cas9-mediated gene editing, and live-cell imaging in zebrafish, we further show that endosomal stabilization of VEGFR2 levels is required for developmental angiogenesis in vivo. In contrast, the premature degradation of internalized VEGFR2 disrupts VEGF signaling, gene expression, and tip cell formation and migration. Thus, an endosomal feedforward mechanism maintains receptor signaling by preventing lysosomal degradation, which is directly linked to the induction of target genes and cell fate in collectively migrating cells during morphogenesis. Supplementary Information The online version of this article contains supplementary material available (10.1007/s10456-021-09788-4).

While these findings emphasize the importance of endocytosis for VEGF signaling, less is known about the mechanisms and factors that maintain endosomal VEGFR2 levels and prevent premature degradation. Studies in zebrafish have shown that Rab5C is particularly highly expressed in endothelium and important for endothelial Notch trafficking [35][36][37][38], but its role in VEGFR2 traffic is unknown. Moreover, it is unclear if and how endosomal VEGF signaling regulates the induction of VEGF target genes, or the acquisition of tip versus stalk cell properties during sprouting angiogenesis.
Here we show that the regulated maintenance of internalized VEGFR2 and its diversion from the degradation pathway is tightly linked to the induction of VEGF target genes, and is crucial for tip cell specification and endothelial cell migration. The endosomal VEGFR2 pool is protected from degradation by Rab5C, which is recruited to endosomes by the GEF RIN2. Finally, we show by a number of in vitro and in vivo approaches that manipulation of the RIN2/Rab5C machinery leads to premature VEGFR2 degradation, thus disturbing normal endothelial VEGF/Notch signaling and VEGF-dependent gene expression, generation of functional tip cells, and sprouting angiogenesis. In summary, an endosomal feedforward loop controlled by RIN2/Rab5C prevents VEGFR2 degradation, and maintains a VEGF signaling window required for VEGF-induced gene expression, tip/stalk cell specification, and vascular sprouting.

Rab5C protects VEGFR2 from VEGF-induced lysosomal degradation
We first depleted Rab5C from human umbilical vein endothelial cells (HUVECs) with shRNA pools (leading to ~ 90% reduction at the mRNA level; Fig S1A), and assessed the cell-surface levels of VEGFR2 by flow cytometry. Intriguingly, VEGFR2 surface levels in steadystate HUVECs (growing in the presence of VEGF) were consistently reduced two-fold in sh_Rab5C cells, compared to HUVECs expressing scrambled sequences as a control (sh_Ctrl) (Fig. 1a). The decrease in VEGFR2 was further confirmed by quantification of the total VEGFR2 protein levels by Western blotting (Fig. 1b). In contrast, VEGFR1 protein levels were not affected, suggesting that VEGFR1 is not regulated by Rab5C ( Fig. 1b; Fig S1B). We then investigated if and how Rab5C regulates VEGFR2 endocytic recycling using flow cytometry. To test whether the internal pool of VEGFR2 can be recruited normally to the plasma membrane, we starved HUVECs for 30 min, a timeframe that allows mobilization of the internal VEGFR2 pool to the plasma membrane but is too short to induce strong changes in de novo protein synthesis. Starvation significantly increased the cell-surface levels of VEGFR2 compared to those in steady-state in sh_Ctrl but not sh_Rab5C cells, indicating that the recruitment from intracellular compartments was impaired (Fig. 1c, d). VEGF induced VEGFR2 endocytosis in both sh_Ctrl and sh_Rab5C cells, although in the latter the internalization was slightly reduced (Fig. 1c, e). The endosomal pool of VEGFR2 is protected from degradation in the absence of ligand, while VEGF stimulation results in rapid degradation [16,22]. Therefore, we hypothesized that the reduced levels of VEGFR2 in Rab5Cdepleted cells were due to increased degradation. To assess VEGFR2 degradation, HUVECs were deprived of growth factors overnight and then stimulated with VEGF, which resulted in a rapid decrease of up to 40% of VEGFR2 levels in 30 min (Fig. 1f), consistent with previous reports [16,22]. Strikingly, VEGFR2 degradation was enhanced twofold in Rab5C-depleted cells, confirming our hypothesis that the reduction in VEGFR2 levels in these cells is caused by increased degradation (Fig. 1f). Moreover, inhibition of lysosomal proteases blocked VEGF-induced degradation, and nullified the differences in VEGFR2 degradation between sh_Ctrl and sh_Rab5C cells (Fig S1C). In HUVECs that were maintained in the absence of VEGF, VEGFR2 levels declined only modestly (20% over 120 min), and this was not significantly altered by depletion of Rab5C (Fig. 1g).
In summary, these results show that Rab5C protects the endosomal VEGFR2 pool from VEGF-induced lysosomal degradation.

Rab5C regulates VEGF signaling, the VEGF-induced immediate transcriptome, and endothelial cell sprouting
We next investigated if and how Rab5C regulates VEGF signaling. For this purpose, sh_Ctrl and sh_Rab5C HUVECs were starved overnight and then stimulated with VEGF, whereafter AKT phosphorylation was assessed as a readout for PI3-K-mediated VEGF signaling. VEGF triggered robust and persistent AKT phosphorylation within 10 min in sh_Ctrl cells, which was almost completely abolished in Rab5C-depleted cells, indicating that Rab5C is required for VEGF-induced PI3-K signaling (Fig. 2a). We also investigated VEGF-induced ERK-1/2 phosphorylation. In sh_Ctrl cells, robust induction of ERK-1/2 phosphorylation was observed within 5 min after VEGF addition and persisted for up to 30 min, while this was significantly reduced in sh_Rab5C cells (Fig. 2b).
We then assessed if and how the reduced VEGFR2 signaling observed upon depletion of Rab5C impacts on VEGFregulated gene expression by mRNA sequencing analysis (RNA-seq). Differential expression was assessed by empirical Bayes analysis followed by correction of p values for multiple testing using the Benjamini-Hochberg false discovery rate (FDR), with a cut-off adjusted p value of 0.05. Using these criteria, the expression of 40 genes was significantly induced and 1 gene was repressed in sh_Ctrl cells after 1 h (t = 1) of VEGF stimulation, while after 4 h (t = 4) the expression of these genes was largely back to base-line levels (Table S1 and Fig. 2c). These kinetics are comparable with those found in previous reports [39][40][41]. The induced target genes include well-known immediate VEGF targets, 58% of which encoding transcriptional regulators such as Krüppel-like factors, nuclear receptors of the NR4A family, and members of the Fos/Jun and early growth response protein 1 3 families, also consistent with earlier findings (Table S1 and Fig. 2c) [39][40][41]. Strikingly, expression of a large number of VEGF target genes (34 out of 41) was altered in sh_Rab5C cells (Fig. 2c), and geneset enrichment analysis revealed that at t = 1, expression of the 'VEGF transcriptome' was significantly different with respect to that in sh_Ctrl cells (p = 3.12*10 -4 , FDR = 1.52*10 -2 ).
We then assessed the role of Rab5C in sprouting angiogenesis, using the fibrin bead assay [42,43]. Quantification of the extent of sprouting after 48 h revealed that depletion of Rab5C strongly impaired sprouting, by affecting both the numbers of formed sprouts and their length ( Fig. 2d-g). In addition, tip cells clearly formed multiple filopodia in the sh_Ctrl population, a process stimulated by VEGF signaling, whereas sh_Rab5C tip cells were on average more 'blunted' (Fig. 2d).
Together, these data show that Rab5C promotes endosomal VEGFR2 signaling and full induction of the immediate VEGF-induced transcriptome, and that Rab5C is required for efficient sprouting angiogenesis.

Rab5C promotes tip cell formation
Because VEGF signaling and the expression of certain VEGF targets is important for the induction and maintenance of tip cells during sprouting angiogenesis, we next addressed whether the reduced sprouting observed in the absence of Rab5C is due to differences in VEGF-induced tip cell specification. For this purpose, we first investigated the expression of a panel of genes commonly associated with tip cell identity, including ANGPT2, APLN, DLL4, NID2, NRP2, PDGFB, TIE1, UNC5B, VEGFR2, and VEGFR3 [6,7]. Intriguingly, depletion of Rab5C significantly suppressed the transcription of ANGPT2, APLN, DLL4, NID2, UNC5B, and VEGFR3, while the expression of NRP2, TIE1, and VEGFR2 was mildly but not significantly decreased ( Fig. 3a). Of note, PDGFB expression appeared not to be regulated at all by Rab5C.
We then assessed tip cell formation in mosaic sprouting assays, by differentially labeling sh_Ctrl and sh_Rab5C cells using two distinct CellTracker dyes, mixing them in a 1:1 ratio, and determining the fraction of tip cells formed by each population after 48 h (Fig. 3b). The knockdown of Rab5C significantly reduced the numbers of tip cells in this assay (Fig. 3b). Similar results were obtained using a pool of mixed shRNAs (Fig. 3b), two different individual shRNAs (Fig S2A-C), and when the dyes were swapped between conditions (Fig S2B).
To address the role of Rab5C in an in vivo model system for sprouting angiogenesis, we next analyzed if Rab5C regulates intersegmental vessel (ISV) development in zebrafish. The zebrafish rab5c gene is highly homologous to its human counterpart, suggesting an important function in the vasculature [35,37,38]. We first generated a dominant-negative (DN) mutant Rab5C protein, carrying a single substitution (S35N) in the GTP-binding pocket ( Fig S3A). DN Rab mutants sequester GEFs but fail to bind GTP, and therefore cannot be activated [44]. The mutant was fused to mCherry and expressed in HUVECs to test its ability to localize on endosomes. Wild-type (WT) Rab5C localized predominantly on endosomes, some of which were positive for EEA-1, while occasional distribution to the Golgi network was also observed, using Trans-Golgi Network (TGN)46 as a marker (Fig S3B). In contrast, the mutant localized predominantly at the Golgi and hardly on endosomes ( Fig  S3B). We then cloned the mCherry-tagged WT or DN human RAB5C gene into a construct that integrates via Tol2-mediated transgenesis and is driven by the zebrafish fli-1a promoter, to achieve expression specifically in vessels ( Fig. 3c) [45]. The construct was injected into single-cell stage Tg(kdrl:GFP) s843 zebrafish embryos expressing GFP in all vascular endothelial cells, and mCherry-positive endothelial cells were examined in developing ISVs at 30-32 h postfertilization (hpf) (Fig. 3d, e) [46,47]. As these injections result in mosaic expression of mCherry-Rab5C in endothelial cells, mCherry-positive endothelial cells were scored for their position within the developing vessels (tip, stalk, or dorsal aorta; Fig S4A). The distribution of WT-Rab5C positive cells was similar to that of mCherry-expressing cells ( Fig. 3f and Fig S4B). However, cells expressing DN-Rab5C were less commonly found in the tip cell position (22% versus 56% in the mCherry control), and most DN-Rab5C positive cells localized to the stalk or the dorsal aorta ( Fig. 3f and Fig S4B).
Altogether, our findings indicate that Rab5C is required for tip cell formation, with Rab5C expressing cells having a competitive advantage for tip cell positioning over cells with impaired Rab5C function. shows the relative VEGFR2 levels normalized to total protein content. Levels in sh_Ctrl cells were set to 1. Results are means + SEM of 3-4 independent experiments. c sh_Ctrl and sh_Rab5C HUVECs were starved for 30 min and then either left untreated or stimulated for 30 min with 50 ng/ml VEGF, whereafter the cell-surface levels of VEGFR2 were analyzed by flow cytometry. d Recruitment of VEGFR2 upon starvation was calculated as the increase in surface levels with respect to those in steady-state. Results are means + SEM of 3 independent experiments. e VEGF-induced VEGFR2 internalization was calculated for sh_Ctrl and sh_Rab5C HUVECs. Mean fluorescence intensities were normalized to steady-state levels. Results are means + SEM of 3-4 independent experiments. f, g HUVECs were starved overnight and subsequently stimulated with 50 ng/ml VEGF (f) or maintained in starvation medium for the indicated times (g). Lysates were subjected to Western blot analysis for VEGFR2, with α-tubulin as a loading control. Blots are representative of 4-5 individual experiments. Quantification of Western blots shows the decline in VEGFR2 levels, expressed relative to the levels at t = 0.

Rab5C is required for maintenance of Vegfr2 levels, Vegf/Notch signaling, and sprouting angiogenesis in zebrafish
We showed by mosaic overexpression of DN-Rab5C that Rab5C activation is required for tip cell formation in vivo. While these experiments allowed us to identify the preference for tip or stalk cell position in a competitive situation, we also analyzed the effects of DN-Rab5C overexpressed simultaneously in all endothelial cells, by generating transgenic Tg(kdrl:GFP) s843 embryos with vascular-specific expression of human mCherry-WT-Rab5C (Tg(fli1a:mCherry-WT-hRAB5C) mu227 , or mCherry-DN-Rab5C (Tg(fli1a:mCherry-DN-hRAB5C) mu228 . Similar to our findings in HUVECs, we observed localization of WT Rab5C protein in rapidly moving vesicles in endothelial cells in vivo (Fig. S4C), while the DN mutant was mostly absent from these vesicles but was instead retained in a perinuclear compartment, most likely the Golgi (Fig. S4C). Indeed, quantification confirmed that the number of mCherrypositive vesicles in DN-Rab5C-expressing ISV cells was strongly decreased, compared to WT-Rab5C expressing cells (Fig. S4D).
We then scored ISV formation at 30-32 hpf, a timepoint at which normal ISV endothelial cells have migrated dorsally and interconnected, to form the dorsal longitudinal anastomotic vessel (DLAV) [46,47]. The survival of DN-Rab5C-expressing embryos was compromised at later stages but not before 30-32 hpf, thus allowing analysis of the developing ISVs. Without WT cells taking over the tip cell function, we observed an impairment of endothelial cell migration in the ISVs of embryos expressing DN-Rab5C (Fig. 4a). While in WT-Rab5C-expressing embryos ISV development was characterized by 83.3% reaching the DLAV and 16.7% migrating nearly to the top, DN-Rab5Coverexpressing endothelial cells migrated only to the level of the horizontal myoseptum ('half') in 31% of the embryos, with 47.3% reaching the ISV length, and only 21.7% reaching the DLAV (Fig. 4b).
To disrupt zebrafish Rab5c function in vivo via additional methods, we also used Morpholino (MO)-mediated block of rab5c translation (rab5c ATG MO) or splicing (rab5c e2i2 MO), as well as CRISPR/Cas9-mediated genome modification to induce mutations in the rab5c gene ( Fig. S5A-E). A heterozygous rab5c mutant line was generated by injections of rab5c specific guide RNAs (gRNAs) and Cas9 protein, growth to adulthood, identification of germline-transmitting adults, growth of the F1 generation, and identification of the mutation. Detailed genomic analysis revealed that the resulting rab5c mutant (designated rab5c mu229 ) contained a premature stop codon, thus terminating protein synthesis at 29 (instead of 221) amino acids (Fig. S5D,E). We then analyzed vascular development in MO-injected embryos and in the homozygous mutant offspring of rab5c +/mu229 parents. Similar to the DN-Rab5C-expressing embryos, the homozygous mutants died during later stages of development. We observed mispatterning of the ISVs and impaired dorsal migration in individual ISV sprouts in embryos injected with rab5c MOs (compared to Control MOs), as well as in rab5c mu229/mu229 embryos (compared to rab5c +/+ siblings) (Fig. 4c,d), thus confirming that Rab5c is required for tip cell formation and normal ISV migration in vivo. Additionally, abnormal sprouting angiogenesis was also revealed by live imaging of ISV development using timelapse microscopy (Movies S1,S2 and Fig. S6A).
We also investigated the levels of zebrafish Vegfr2 (Kdrl) protein in embryos injected with Control MO or rab5c MO by Western blotting (Fig. 5a). Similar to what we observed in HUVECs, a significant reduction in Kdrl protein levels was apparent in Rab5c-depleted embryos (Fig. 5a,b). Consistently, tip cells in these embryos formed shorter filopodia, as revealed by Rab5c knockdown in Tg(fli1a:lifeact-EGFP) mu240 embryos, which express LifeAct-eGFP in the vasculature to visualize the actin cytoskeleton (Movies S3,S4 and Fig. 5c) [48]. Vegfinduced tip cell specification in zebrafish requires initial activation of Notch signaling in tip cells, which subsequently directs them into developing arteries [49,50]. Since we found an impairment of tip cell morphology and behavior in Rab5c-depleted embryos, we also addressed whether Notch signaling was affected. For this purpose we used the Tg(TP1bglob:VenusPEST) s940 transgenic line to image the highly dynamic expression of a very shortlived Venus protein (Venus-PEST), that is driven by the TP1 promoter element and thus reports activation of Notch signaling [51]. Notch activation was visible in arteries, Fig. 2 Rab5C regulates endosomal VEGF signaling, the immediate VEGF transcriptome, and sprouting angiogenesis. a Western blot analysis (left) and quantification (right) of phosphorylation of (S473)AKT and total AKT levels at 0, 10, 20, and 30 min of VEGF stimulation (50 ng/ml). b Western blot analysis (left) and quantification (right) of (Y204)ERK-1/2 phosphorylation and total ERK-1/2 levels at 0, 5, 15, and 30 min of VEGF stimulation (50 ng/ ml). Representative blots are shown, α-tubulin served as a loading control. Quantifications are means + SEM from 3 independent experiments and expressed relative to t = 0. (c) HUVECs were starved overnight (t = 0), then stimulated with VEGF (50 ng/ml) for 1 h (t = 1) or 4 h (t = 4). VEGF-induced gene expression was then assessed by RNA-seq. Genes have been ordered based on the difference in mean log2 FC between sh_Ctrl and sh_Rab5C. Heatmap shows the corresponding centered and scaled mean expression values for VEGF-induced genes at the indicated time-points. Differences in mean log2 fold change (FC) between sh_Ctrl and sh_Rab5C after 1 h of VEGF stimulation are indicated on the right. Genes that have a higher mean expression in sh_Rab5C with respect to sh_Ctrl are indicated in red, those that have a lower mean expression in green. (d) Sprouting of sh_Ctrl (top) and sh_Rab5C (bottom) HUVECs from collagen-coated beads into fibrin gels. Cells were fixed at 48 h, stained for F-actin (magenta) and nuclei (cyan), and visualized by confocal microscopy. Representative images are shown (maximum projections from z-stacks). Scale bar, 75 μm. e Quantification of the average number of large sprouts/bead. Values represent the means + SEM of 3 independent experiments (20 beads per experiment). f Quantification of the total network length and g the total number of sprouts. A representative experiment is shown. Values represent the means + SEM of n = 16 beads per condition ◂ as well as in some tip cells and arterial ISVs (Fig. 5d), in addition to a number of non-vascular mesenchymal cells. As expected from our morphological data, a strong reduction of Venus-PEST expression in rab5c MO-injected embryos was observed, indicating that activation of Notch signaling was reduced and further supporting our hypothesis that Rab5c is required for efficient generation of functional tip cells (Fig. 5d).
Altogether, these data show that in line with our observations in HUVECs (Figs. 1-3), Rab5c is crucial for the maintenance of Vegfr2 levels in zebrafish, and thereby regulates tip cell formation and sprouting angiogenesis in vivo.

The Rab5 GEF RIN2 maintains VEGFR2 levels in endothelial cells and promotes angiogenic sprouting
The previous sections have shown that activation of endothelial Rab5C is required for sprouting angiogenesis. Because Rabs are activated by GEFs, we next investigated which of the eight known GEFs for Rab5 GTPases are expressed in endothelial cells, using publicly available genome-wide mRNA expression profiles. This analysis suggested that HUVECs express at least seven different Rab5 GEFs, while RINL was not represented on the used arrays ( Fig. 6a and Table S2). To investigate which of these GEFs is involved in the Rab5C-dependent effects on VEGFR2 and angiogenic sprouting, we silenced the expression of each individual GEF, including RINL, and assessed VEGFR2 cell-surface levels by flow cytometry. Of all investigated GEFs, only the depletion of Ras and Rab interactor (RIN) 2 significantly reduced the surface levels of VEGFR2, to the same extent as the depletion of Rab5C (Fig. 6b). Furthermore, total protein levels of VEGFR2 were also reduced by RIN2 depletion, as assessed by Western blotting (Fig. 6c, d). As in the case of Rab5C depletion, the reduction of VEGFR2 levels was associated with an increase in VEGF-induced VEGFR2 degradation (Fig. 6e). Finally, we performed a sprouting assay using RIN2-depleted cells. As expected, both the number of formed sprouts as well as the total network length were suppressed by RIN2 depletion (Fig. 6f-i).
Collectively, our data show that knockdown of RIN2 expression recapitulates the effects of Rab5C depletion, and strongly suggest that RIN2-mediated Rab5C activation prevents VEGFR2 degradation to promote sprouting angiogenesis.

RIN2 regulates Rab5C recruitment and is required for sprouting angiogenesis in vivo
We next tested whether forced Rab5C activation can circumvent the requirement for RIN2 to drive sprouting. For this purpose, we generated a constitutively active (CA) mutant (Q80L) of human Rab5C, which blocks GTP hydrolysis, fused to mScarlet (Fig. S6B). Expression of this mutant into HUVECs caused enlargement of EEs, indicative of constitutive Rab5C activation (Fig. S6C) [52]. We then expressed mScarlet-CA-Rab5C into RIN2-depleted cells, which slightly lowered endogenous Rab5C expression, whereas RIN2 knockdown by itself did not affect endogenous Rab5C levels or vice versa ( Fig. 7a; Fig. S6D). Importantly, expression of CA-Rab5C in RIN2-depleted HUVECs rescued the sprouting defects caused by RIN2 deficiency, and thus bypassed the requirement for RIN2 (Fig. 7b). These data confirm that 1) Rab5C activation promotes sprouting angiogenesis, and 2) the primary role of RIN2 in sprouting angiogenesis is to activate Rab5C.
To assess the effects of RIN2 knockdown on Rab5C function in vivo, we generated a splice-blocking MO (e3i3) against the zebrafish rin2 gene (Fig. S7A,B), which was injected into Tg(f li1a:mCherr y-W T-hRAB5C) mu227 ;Tg(kdrl:GFP) s843 zebrafish embryos to examine Rab5C localization. Depletion of zebrafish Rin2 protein resulted in a partial failure of Rab5C to localize in vesicular compartments (Fig. 7c), indicating that the recruitment of Rab5C requires Rin2 and hence, that the Rin2-Rab5c interaction is functionally conserved in zebrafish.
We also addressed the functional role of zebrafish Rin2 in sprouting angiogenesis in Tg(kdrl:GFP) s843 embryos. Knockdown of Rin2 resulted in impaired endothelial cell migration in the ISVs, similar to that induced by deficiency Fig. 3 Rab5C is essential for tip cell formation. a sh_Ctrl and sh_ Rab5C HUVECs were subjected to qPCR analysis for the indicated tip cell markers. Data are means + SEM of 3-5 individual experiments and are expressed relative to the means in sh_Ctrl cells (indicated by dashed line). b sh_Ctrl and sh_Rab5C cells were differentially labeled using CellTracker dyes, whereafter they were mixed in a 1:1 ratio, adhered to collagen-coated beads, and subjected to VEGF-stimulated sprouting in fibrin gels. After 48 h, cells were fixed and processed for confocal microscopy. Quantification of the number of tip cells was performed by counting the tip cells from confocal z-stacks and normalizing to the total number of cells in that color. Results shown are the means + SEM of 3 pooled independent experiments. Between (Fig. 7d, e). The migration defect was again accompanied by reduced activation of Notch signaling, as analyzed by Tp1:Venus-Pest expression (Fig. S7C).
In summary, these results indicate that Rin2 is crucial for Rab5C activation and recruitment in endothelial cells, which regulates Vegf/Notch signaling, gene expression, tip/ stalk cell specification, and sprouting angiogenesis (Fig. 7f).

Discussion
In this study, we identify an endosomal regulatory mechanism dependent on Rab5C and RIN2 that is crucial for receptor signaling, transcriptional output, and sprouting angiogenesis. Our data indicate that RIN2 recruits Rab5C to EEs, where it prevents the degradation of internalized VEGFR2. This is required to maintain VEGFR2 levels and to sustain endosomal VEGF/VEGFR2 signaling toward PI3-K and ERK pathways. Furthermore, it is necessary for the normal expression of VEGF target genes, including various tip cell markers, as well as for the generation of functional tip cells and their migration (summarized in Fig. 7f). Our data emphasize the crucial role of endosomal trafficking for cell fate decisions, and have important implications.
First, the amplitude and duration of VEGF signaling toward PI3-K and to a lesser extent ERK is dependent on endosomal Rab5C. This observation implies that VEGFinduced PI3-K signaling occurs primarily from EEs and much less from the plasma membrane, which fits with the well-documented role of PI3-K as a direct effector of Rab5  GTPases on EEs [19]. Although it is possible that Rab5C depletion indirectly affects ERK signaling from cell-surface VEGFR2 due to the decrease of VEGFR2 levels, it seems likely that also VEGF-induced ERK signaling occurs mainly from EEs, which is consistent with the findings of others and the existence of an ERK scaffold complex on endosomes [2,34,53]. While VEGFR2 signaling in the absence of internalization can also occur from the plasma membrane [28,32], our data suggest that once VEGFR2 is endocytosed, Rab5C and the Rab5 GEF RIN2 protect the EE pool of VEGFR2 from VEGF-induced degradation, which sustains VEGF signaling and is required for the expression of immediate VEGF targets and tip cell genes. Together, these events regulate tip cell specification, endothelial cell migration, and sprouting angiogenesis components of the endosomal machinery, such as Rab5C, are required for normal VEGF signaling.
Second, our findings show that endosomal maintenance of VEGFR2 levels is required for transcriptional responses that determine cell fate. The time-frame in which most of the VEGFR2 degradation occurs in the absence of Rab5C, is also the time-frame that is necessary for VEGF signaling and the induction of immediate VEGF-responsive genes. Because Rab5C depletion leads to significant and quantitative differences in the expression of most VEGF-induced genes, our observations imply that accelerated VEGFR2 degradation results in disrupted VEGF-induced gene expression. Thus, the magnitude of VEGF-induced gene expression is closely related to VEGFR2 levels, and stabilization of endosomal VEGFR2 levels by Rab5C ensures the build-up of a sufficiently robust VEGF signal needed to surpass the threshold for transcription. Similar mechanisms likely regulate the duration of signaling of other receptors in different contexts, which is particularly interesting given the frequent overexpression of several Rab GTPases in cancer and many other diseases [54].
Third, the reduction in VEGFR2 levels induced by depletion of Rab5C is functionally and physiologically crucial, since we observe functional defects in tip cell behavior and angiogenic sprouting, both in vitro and in vivo. Tip cells are characterized by high VEGFR2 levels and thus high VEGF signaling, leading to the expression of tip cell markers [5][6][7]. Because VEGFR2 is downregulated in Rab5C-depleted cells whereas VEGFR1 levels are not affected, the VEGFR1/VEGFR2 balance is disturbed which confers a competitive disadvantage to develop tip cell properties [11][12][13]. Indeed, the expression of various tip cell genes was decreased by Rab5C depletion, and tip cell formation was impaired in mosaic sprouting assays in vitro. While it is known that balanced VEGFR2 signaling regulates tip cells, we show that this crucially depends on endosomal regulation of VEGFR2 in a Rab5C-dependent manner, and that Rab5C deficiency is sufficient to affect all aspects of VEGF-driven tip cell responses, namely tip cell specification, function, and characteristics. Interestingly, out of eight different Rab5 GEFs, only the depletion of RIN2 specifically recapitulates our main findings in Rab5C-depleted cells. Although RIN2 may activate several Rabs and Rab5C might use multiple GEFs for all of its functions, our data indicate that RIN2 is the most important Rab5 GEF for the effects of RabC on VEGFR2 traffic and sprouting angiogenesis. These results nicely complement previous work on RIN2 in endothelial cells, where RIN2 was found to be important for several processes in endothelial cells including tube formation [55]. In line with previous reports of a high molecular and functional conservation of zebrafish and mammalian angiogenesis [46,47], we could show full conservation of the Rin2/Rab5c-dependent Vegfr2signaling axis in zebrafish, allowing us to analyze tip cell behavior and sprouting angiogenesis in response to Rin2/ Rab5c-driven endosomal Vegfr2 signaling in vivo. In zebrafish embryos, depletion of Rab5c or Rin2 attenuated Vegf-induced activation of Notch signaling in the tip cells, filopodia formation, and endothelial cell migration, thus affecting key aspects of sprouting angiogenesis. Together, our data suggest that RIN2/Rab5C control a feedforward mechanism whereby inhibition of VEGFR2 degradation permits sprouting angiogenesis, by sustaining VEGF signaling and the expression of VEGF target genes. The endolysosomal machinery is therefore an important driver of the dynamic regulation of tip and stalk cells, and likely contributes strongly to their rapid and frequent interconversion observed during sprouting [56].
To date, few factors are known that protect VEGFR2 from degradation. Depletion of Numb or Rab GTPasebinding effector protein-2 (RABEP-2) also enhances VEGFR2 degradation, resulting in lower VEGFR2 levels and reduced VEGF signaling [27,33]. Interestingly, genetic deletion of RABEP-2 in mice disturbs arteriogenesis but not angiogenesis, while the deletion of Numb induces angiogenic defects [27,57]. It remains to be determined how different proteins along trafficking pathways can mediate such very diverse outcomes of VEGFR2 signaling. Future work should further establish how the fate of internalized VEGFR2 is linked to the transcriptional output of VEGF signaling and the dynamic regulation of endothelial cell phenotypes, as well as map the underlying molecular machinery along trafficking routes.

qPCR
For analysis of relative mRNA expression levels in transduced HUVECs, total RNA was isolated using the RNeasy kit (Qiagen) according to the manufacturer's instructions. After reverse transcription to cDNA using the SuperScript III First-Strand Synthesis System (Thermo Fisher Scientific), qPCR was performed with the SensiFAST SYBR No-ROX kit (Bioline) and the indicated primers (Table S3), either on a LightCycler PCR system (Roche) or a StepOne-Plus system (Applied Biosystems). Duplicate reactions were performed for each gene and expression was normalized to that of β-actin.

Flow cytometry
HUVECs were treated as indicated, detached with accutase (Sigma-Aldrich) for 2 min, and washed in 2% FCS in PBS. Cells were stained with primary antibodies for 1 h on ice, washed twice, followed by incubation with secondary antibodies for 45 min on ice and washed twice. Cells were analyzed on a Canto-II flow cytometer (BD Immunocytometry Systems) equipped with FACSDiva software.

Bio-informatic analysis of mRNA expression
Expression of Rab5 GEFs in HUVECs (Table S2) was determined by analysis of Affymetrix U133 Plus 2.0 genomewide mRNA expression profiles in the public domain using the NCBI Gene Expression Omnibus (GEO: GSE7307) website [58,59]. GEO was first searched for studies on low-passage, non-recombinant, non-stimulated HUVECs with data normalization using the MAS5.0 algorithm (Affymetrix Inc., Santa Barbara, CA). This resulted in a total of 11 published studies comprising 29 separate arrays, 17 of which were 1 3 listed with present call analysis, that were queried for the expression of all known human Rab5 GEFs [20]. RINL was not represented on the Affymetrix arrays and was not further analyzed. Array data were analyzed as described using R2; an in-house developed Affymetrix analysis and visualization platform (http:// r2. amc. nl). Probes were ranked depending on high expression values and widespread expression (% of samples with significant expression for that gene) in the data collections tested.

Western blotting
Cells were washed in ice-cold PBS and lysed on ice in NP40 lysis buffer (50 mM Tris-HCl, pH 7.4, 100 mM NaCl, 10 mM MgCl 2 , 1% NP-40, 10% Glycerol), supplemented with protease inhibitor cocktail (Sigma-Aldrich). Cell lysates were cleared by centrifugation, heated at 95ºC in SDS sample buffer (50 mM Tris-HCl pH 6.8, 2% SDS, 10% glycerol, 1% β-mercaptoethanol, 12.5 mM EDTA, 0.02% bromophenol blue), and proteins were resolved by SDS-PAGE. Thereafter, proteins were transferred to nitrocellulose membranes (GE Healthcare Amersham) and aspecific binding was blocked using 5% (w/v) milk (Campina) in TBST (150 mM NaCl, 10 mM Tris, 0.1% Tween 20, pH 8.0) for 30 min. Immunoblotting was performed with the indicated antibodies by incubation with primary antibodies o/n at 4ºC and with secondary antibodies for 1 h at RT. Membranes were washed 3 × with TBST after each step. Proteins were visualized using ECL chemiluminescence reagent (Thermo Fisher Scientific), light sensitive films (Fuji Film) and a film processor (Konica Minolta, SRX-101A). Quantification of bands was performed by densitometry using ImageJ. Band intensity of the protein of interest was corrected to that of α-tubulin.

Degradation and signaling experiments
For VEGF signaling and VEGFR2 degradation assays, HUVECs in 6-well plates were grown to confluence. Then, cells were starved overnight and subsequently either left untreated or stimulated with 50 ng/ml VEGF for the indicated time-points, whereafter the cells were processed for Western blotting. Inhibitors were used as indicated.

Sprouting assays
Sprouting assays were performed according to previously established protocols with minor modifications [42,43]. In brief, HUVECs were incubated with collagen-coated microcarrier beads (Sigma-Aldrich, C3275). The next day, the beads were detached by washing and transferred to 48-well plates containing fibrinogen in PBS (2 mg/ml) supplemented with 0.15 U/ml aprotinin and 6.25 U/ml thrombin. EGM-2 medium was added on top of the gels. Beads were imaged by time-lapse microscopy on a Widefield system using a 10 × dry lens objective (Carl Zeiss MicroImaging, Inc.). Images were recorded at 15-min intervals during 48 h. For each condition, 20 beads were analyzed per experiment. Sprouting was assessed as the average number of large (length of sprout > than the bead diameter) sprouts per bead. Alternatively, beads were embedded in fibrin gels in 'half-area' glass-bottom 96-well imaging plates (Corning, #4580). After 48 h, beads were fixed, stained with Hoechst and phalloidin, and visualized by generating z-stacks on a confocal microscope. Maximum projections of z-stacks were created using Fiji/ImageJ (version 1.52e), de-speckled once, and the contrast was enhanced with 0.3% and analyzed using the Angiogenesis plugin [60]. The phalloidin staining was used for correct segmentation of the sprouts. For the sprouts stained with CellTracker, the number of tip cells per condition was counted and normalized to the total number of cells in that condition.

RT-PCR and Western blotting in zebrafish
Total RNAs were prepared from 36 hpf embryos using TRIzol reagent (Invitrogen), and reverse-transcribed by random hexamer primers using Superscript III (Invitrogen) according to the manufacturer's instruction. PCR was performed using gene-specific primers as below. Efficiency of the knockdown was assessed by PCR using the primers fwd: AGA GCT GCG TAC ATC CAA CC and rev: TGC AAG CCA AAA TTC AAA GA.
After removing chorionic membrane and yolk sac, embryos injected with either control MO or rab5c MO were directly lysed in 1 × SDS sample buffer, and subjected to Western blot analysis with anti-Kdrl antibody (Kerafast, ES1003) and anti-alpha-Actin antibody (Santa Cruz Biotechnology, sc-47778) as described previously [67].

Confocal microscopy
HUVECs on glass coverslips were fixed with 4% paraformaldehyde (Merck) in PBS containing 1 mM CaCl 2 and 0.5 mM MgCl 2 (PBS + +) for 10 min, and permeabilized with 0.4% Triton X-100 (Sigma-Aldrich) in PBS + + for 5 min. Aspecific antibody binding was prevented by blocking with 2% BSA Fraction V (Sigma-Aldrich) in PBS + + for 15 min. Following incubation with the indicated primary antibodies, coverslips were washed with 0.5% BSA Fraction V in PBS + + and antibody binding was visualized using secondary antibodies. Coverslips were washed and subsequently mounted in 10% Mowiol, 2.5% Dabco, 25% glycerol, pH 8.5. Image acquisition was performed on a Leica SP8 confocal microscope using a 63 × oil immersion objective, after which images were processed using Leica Application Suite X and Fiji/ImageJ (version 1.52e) software.
Sprouting was visualized using a 25 × long-distance water objective on a Leica SP5 confocal set-up. z-stacks were created of 1.5 μm per slice and the different channels were sequentially scanned between stacks.
For time-lapse analysis in zebrafish, a temperature of 28.5 °C was maintained using a heating chamber. Images were collected every 6 min during 1 h (for imaging of filopodia) or every 15 min during 6 h (for imaging of ISV sprouting). Assembly of confocal stacks and time-lapse movies was conducted using Imaris 8/9 software (Bitplane).

mRNA sequencing and analysis
sh_Ctrl, sh_Rab5A or sh_Rab5C cells were seeded in 10 cm dishes and grown to confluence. Cells were starved overnight, whereafter they were either left untreated or stimulated with 50 ng/ml VEGF for 1 h or 4 h. Subsequently, cells were lysed using RLT buffer (Qiagen) with 1% β-mercaptoethanol and stored at −80 °C until analysis. RNA-seq libraries were prepared from samples of 2 independent experiments with the mRNA KAPA HyperPrep Kit (Illumina) using Truseq Illumina LT adapters. Sequencing was performed on an Illumina Hiseq4000 system (single-read 50 bp), 10 samples per lane.
Reads were subjected to quality control (FastQC, dupRadar, Picard Tools), trimmed using Trimmomatic v0.36, and aligned to the hg38 genome using HISAT2 v2.1.0. Counts were obtained with HTSeq v0.11.0. Statistical analyses were performed using the edgeR and limma/voom packages [69,70]. Genes with no counts in any of the samples were removed while genes with more than 2 reads in at least 3 of the samples were kept. Count data were transformed to log2-counts per million (logCPM), normalized by applying the trimmed mean of M-values method and precisionweighted using voom. Differential expression was assessed using an empirical Bayes' moderated t-test within limma's linear model framework including the precision weights.
Resulting p values were corrected for multiple testing using the Benjamini-Hochberg false discovery rate. Additional gene annotation was retrieved from Ensembl (release 94) using the biomaRt/Bioconductor package. Geneset enrichment was performed using CAMERA with a preset value of 0.01 for the inter-gene correlation using collections H, C1,