Endothelial insulin receptors promote VEGF-A signaling via ERK1/2 and sprouting angiogenesis

Endothelial insulin receptors (Insr) promote sprouting angiogenesis, although the underpinning cellular and molecular mechanisms are unknown. Comparing mice with whole body insulin receptor haploinsufficiency (Insr +/- ) against littermate controls, we found impaired limb perfusion and muscle capillary density after inducing hind-limb ischemia; this was in spite of increased expression of the pro-angiogenic growth factor Vegfa . Insr +/-neonatal retinas exhibited reduced tip cell number and branching complexity during developmental angiogenesis, which was also found in separate studies of mice with endothelium-restricted Insr haploinsufficiency. Functional responses to VEGF-A, including in vitro angiogenesis, were also impaired in aortic rings and pulmonary endothelial cells from Insr +/- mice. Human umbilical vein endothelial cells (HUVEC) with shRNA-mediated knockdown of Insr also demonstrated impaired functional angiogenic responses to VEGF-A. VEGF-A signaling to Akt and eNOS was intact, but downstream signaling to ERK1/2 was impaired, as was VEGF receptor-2 (VEGFR-2) internalization, which is required specifically for signaling to ERK1/2. Hence, endothelial insulin receptors facilitate the functional response to VEGF-A during angiogenic sprouting and are required for appropriate signal transduction from VEGFR-2 to ERK1/2. Microscopy was performed using a Zeiss LSM880 upright confocal microscope with 10x/0.3NA, 20x/0.8NA and 40x/1.4NA objectives and Zen software UK). Tile scanning was used to image entire retinal segments with the 20x objective and maximum intensity projection of 5 consecutive 1 Airy unit thickness z-slices was used with the 40x to define tip cells and filopodia. Image analysis used ImageJ (NIH, Bethesda, MD). Radial outgrowth was defined as the distance from the optic disc periphery to the emerging vascular front measured at 12 points in each retina. Vascular area was defined by binary thresholding of the Isolectin B4 signal and expressed as a percentage of the region of interest, bounded either by the peripheral or central half of the vascularized area. Vascular branching was quantified in multiple 200x200μm regions of interest placed between arteries and veins, in the peripheral or central vascular plexus. Tip cell abundance was normalized to the perimeter of the contiguous vascular front in each image and filopodia were normalized to tip cell number. Capillary regression was defined as Collagen IV staining without colocalized IB4 staining, and expressed as total length per mm 2 in complete retinal segments, as per the method of Franco et al .(21) Endothelial proliferation, defined by EdU + nuclei co-staining with IB4, was quantified in multiple 200x200μm regions of interest placed between arteries and veins, in the peripheral vascular plexus. Image with the number of FITC staining sprouts per ring and the mean length of these sprouts; mean data were then produced for each experimental animal from at least 4 rings.

M a n u s c r i p t

INTRODUCTION
Insulin is a primary regulator of systemic carbohydrate and lipid metabolism,(1) but also has an important role in vascular function, for example promoting vasodilation and tissue perfusion. (2) Indeed, loss of endothelial insulin receptors, or perturbation of their signaling function, induces endothelial dysfunction, hypertension and atherosclerosis. (3)(4)(5) Sprouting angiogenesis, the phenomenon of new capillary formation, is another fundamental element of vascular biology that is intrinsically linked to metabolism. (6) In this highly orchestrated and conserved process, endothelial 'tip cells' emerge from existing vessels, followed by proliferating stalk cells that extend the sprout and form a lumenized vessel; these neovessels then anastomose into an immature network that remodels to meet local demands for oxygen and metabolite transport.(7) Insulin has been reported to promote angiogenesis in vitro and in vivo; (8)(9)(10)(11)(12) these studies found pro-angiogenic effects in nanomolar concentrations in vitro, but did not explore more physiological picomolar concentrations. In vivo, insulin receptor expression is known to be enriched in human tumor endothelial tip cells. (13) and loss of endothelial insulin receptors has been shown to impair angiogenesis in murine retinopathy. (14) However, it remains unclear how endothelial insulin receptors influence the cellular and molecular processes of angiogenesis and so we set out to define this.

Animal models
All experimental mice were kept in a conventional animal facility with a 12-hour light/dark cycle and received a standard chow diet. Genotyping was performed using PCR of ear notch A c c e p t e d M a n u s c r i p t 6 Insulin receptor halpoinsufficient (Insr +/-) mice As we have previously described, (15) Insr +/mice (also known as IRKO) were obtained from the Medical Research Council Mammalian Genetics Unit (Harwell, Oxfordshire, U.K.), and were maintained as heterozygotes on a C57BL/6J background. Insr +/were compared with age-matched wild-type (WT) littermates.

Endothelial cell-specific insulin receptor halpoinsufficient (ECInsr +/-) mice
ECInsr +/mice were generated by crossing mice that have loxP sites flanking exon 4 of the insulin receptor (Line 006955, The Jackson Laboratory, Bay Harbor, ME)(16) with mice possessing a Cre transgene driven by the Tie2 promoter/enhancer (Line 004128, The Jackson Laboratory, Bay Harbor, ME) (17) and were maintained on a C57BL/6J background.

Tissue collection and processing
Retinal angiogenesis was assessed in P5 pups by precisely following the protocol of Pitulescu et al. (18) In brief, all pups from at least 3 litters were included in each experiment, with analysis blinded to the results of genotyping data. Both eyes were processed identically with a mean value from these to represent that pup. Vascular endothelium was stained with Isolectin B4 conjugated with Alexa Fluor 488 (I21411; Thermo Fisher Scientific, Warrington, UK). Co-staining with a rabbit anti-mouse anti-Collagen IV antibody (19) followed by an Alexa Fluor-647 conjugated goat anti-rabbit antibody (20) was used to visualize the vascular basement membrane. To define cell proliferation, pups were injected with 125μg 5-ethynyl-A c c e p t e d M a n u s c r i p t 7 2´-deoxyuridine (EdU) two hours prior to tissue collection; this was stained with Alexa Fluor 647 azide using Click-iT technology (C10640; Thermo Fisher Scientific, Warrington, UK).

Confocal microscopy and image analysis
Microscopy was performed using a Zeiss LSM880 upright confocal microscope with 10x/0.3NA, 20x/0.8NA and 40x/1.4NA objectives and Zen software (Carl Zeiss microscopy Ltd, Cambridge, UK). Tile scanning was used to image entire retinal segments with the 20x objective and maximum intensity projection of 5 consecutive 1 Airy unit thickness z-slices was used with the 40x to define tip cells and filopodia. Image analysis used ImageJ (NIH, Bethesda, MD). Radial outgrowth was defined as the distance from the optic disc periphery to the emerging vascular front measured at 12 points in each retina. A c c e p t e d M a n u s c r i p t 8

Surgical procedure
Following the protocol we have published,(22) 9-13 week-old male Insr +/mice were anesthetized with isoflurane before dissecting the left femoral artery, ligating it proximally at the inguinal ligament and distally at the bifurcation to saphenous and popliteal vessels, and excising the intervening segment.

Laser Doppler perfusion imaging
Laser Doppler analysis (Moor LDI2-HR, Moor Systems, UK) of ischemic and non-ischemic limbs was performed post-operatively in a temperature-controlled environment, to confirm induction of ischemia, and repeated weekly until day 21. Images were analyzed (MoorLDI software, Version 5.3, Moor Systems, UK) to derive an ischemic to non-ischemic limb perfusion ratio, based upon flux below the level of the inguinal ligament.

Tissue collection and processing
Ischemic and contralateral gastrocnemius muscle was harvested and fixed in 4% paraformaldehyde for 48 hours, whilst adductor muscles were snap frozen with liquid nitrogen for RNA isolation. Fixed muscle specimens were embedded in OCT (Tissue-Tek OCT compound, Sakura, Netherlands) before snap freezing in liquid nitrogen and cryosectioning at 10 µm thickness. Vascular endothelium was stained with Isolectin B4 conjugated with Alexa Fluor 488 (I21411; Thermo Fisher Scientific, Warrington, UK) and slides were mounted with DAPI-Fluoromount-G (Southern Biotech, AL) to define nuclei.
A c c e p t e d M a n u s c r i p t 9

Confocal microscopy and image analysis
Microscopy was performed using a Zeiss LSM880 upright confocal microscope with 20x/0.8NA objective and Zen software (Carl Zeiss microscopy Ltd, Cambridge, UK). Image analysis used ImageJ (NIH, Bethesda, MD). Vascular area was defined by binary thresholding of the Isolectin B4 signal and expressed as a percentage of the image area.

Ex vivo aortic ring angiogenesis
Aortae were harvested from 8-12 week old Insr +/mice under terminal isoflurane anesthesia and then processed according to the protocol of Baker et al. (23) In brief, after dissection of perivascular fat and overnight storage in serum free OptiMEM media (Thermo Fisher Scientific, Warrington, UK), aortae were cut in to 1mm thick rings that were then embedded in rat type I collagen. Rings were incubated for 5 days at 37°C in 5% CO 2 in Opti-MEM media containing 2.5% fetal calf serum (FCS), 50ng/ml VEGF-A 165 (R&D Systems, Abingdon, UK) and penicillin-streptomycin, with a media change on day 3. Rings were then fixed with 4% paraformaldehyde, stained with BS-1 lectin-FITC (Sigma Aldrich, Gillingham, UK) to define endothelium, and then imaged with an inverted confocal microscope (LSM700, A c c e p t e d M a n u s c r i p t 10 Carl Zeiss Microscopy Ltd., Cambridge, UK); tiled images were collected using a 10x/0.2NA objective and stitched using Zen software. Image analysis was performed with Image J (NIH, Bethesda, MD), defining the number of FITC staining sprouts per ring and the mean length of these sprouts; mean data were then produced for each experimental animal from at least 4 rings.

Isolation and functional analysis
Pulmonary endothelial cells (PEC) were isolated from both lungs of 8-12 week old Insr +/mice, precisely following the protocol of Sobczak et al. (24) This uses immuno-magnetic selection of CD31 + cells, which are then cultured in EGM2 media (Lonza, Slough, UK) for 10-14 days before a second round of immuno-magnetic selection from ICAM2 + endothelial cells which were cultured for a further 5-7 days in EGM2 prior to functional assays.

Matrigel sprouting assay
Twenty-four well plates were coated with growth factor reduced Matrigel (BD Biosciences, Wokingham, UK) prior to seeding each well with 2x10 5 PEC suspended in EBM2 media A c c e p t e d M a n u s c r i p t 11 Scratch wound assay PEC were grown to confluence in EGM2 media on 1% gelatin coated 96 well plates before forming a scratch wound using the WoundMaker TM tool (Essen Bioscience, Royston, UK) and imaging wound closure hourly in a live cell imaging system (Incucyte, Essen Bioscience) to define residual wound area.

Boyden chamber
Following our published protocol,(25) 5x10 4 PEC were seeded in 1% gelatin coated Boyden chamber apparatus to define migration toward 50ng/ml VEGF-A 165 . The number of migrating cells per microscopic field was counted using standard light microscopy, and presented as net migration by subtracting the number of cells migrating in paired control experiments without VEGF-A 165 gradient.

Cell proliferation
Sparsely seeded PEC on 1% gelatin coated plastic, cultured in EGM2 media, were exposed to 10µM EdU two hours prior to fixation with 4% paraformaldehyde and processing with the Click-iT® EdU cell proliferation assay (Thermo Fisher Scientific, Warrington, UK) to label nuclei containing actively forming DNA with Alexa Fluor 488 and a Hoechst nuclear counterstain. Confocal microscopy (LSM700, Carl Zeiss Microscopy Ltd., Cambridge, UK) was used to define the proportion of EdU + nuclei. and 5ng/ml human basic FGF (PeproTech, NJ, USA). After 48 hours of incubation at 37°C in 5% CO 2 , 25 beads per condition were imaged with phase contrast microscopy (Olympus CX41, Olympus Life Sciences, Southend-On-Sea, UK) and analyzed with Image J (NIH, Bethesda, MD), defining sprouts per bead and the mean length of these sprouts; mean data were then produced for each experimental condition.

Scratch wound assay
HUVEC were grown to confluence in EGM2 media on 1% gelatin coated 96 well plates before forming a scratch wound using the WoundMaker TM tool (Essen Bioscience, Royston, UK) and imaging wound closure 8 hours later to define percentage wound closure from baseline.
A c c e p t e d M a n u s c r i p t 13 Adhesion assay HUVEC were seeded on to 1% gelatin coated 24-well plates in EBM2 media with 1% FCS, with or without 50ng/ml VEGA-A 165 , at a density of 4x10 4 cells per well and left for one hour before washing three times with phosphate buffered saline and fixing with 4% paraformaldehyde. Cells were counterstained with Hoechst and Phalloidin Alexa Fluor 488 conjugate and imaged with confocal microscopy (LSM700, Carl Zeiss Microscopy Ltd., Cambridge, UK) to count adherent cells per mm 2 .

Surface biotinylation and immunoprecipitation of VEGFR2
Surface VEGFR2 biotinylation, immunoprecipitation, and VEGFR2 Western blotting were performed according to our previously published protocol.(39) Briefly, HUVEC were incubated for 1 hour at 4°C with 0.5 mg/mL biotin sulfo-NHS (Sigma-Aldrich, Gillingham, UK), before being stimulated with 50ng/ml VEGF at 37C for either 5 or 15mins. At the end of the exposure period, the cells were washed 3x with PBS with calcium and magnesium before either: immediate lysis with cell extraction buffer to enable the measurement of total VEGFR2 in the sample; or, treated with 0.5ml of 0.05% trypsin/EDTA to cleave and remove any remaining biotin-labelled, cell surface VEGFR2, meaning detected biotin-labelled VEFGR2 would define only internalized protein. The trypsinized cell pellet was lysed using cell extraction buffer as before. Immunoprecipitation of VEGFR2 was carried out using protein A dynabeads (Thermo Fisher Scientific, Warrington UK) loaded with anti-VEGFR2 antibody (diluted 1:100)(35) for 30mins with rotation at room temperature. The beads were washed with RIPA buffer to remove any unbound antibody before incubating with cell lysate for 1hr with rotation at room temperature to allow the biotinylated antigen-antibody complexes to form. At the end of the pull down period, the beads were washed 5 x with RIPA buffer (50 mmol/L Tris-HCl (pH 8.0), 150 mmol/L NaCl, 0.5% (w/v) sodium deoxycholate, 0.1% (w/v) SDS, and 1% (v/v) Igepal) to remove any non-specific binding. The immunoprecipitated, biotinylated VEGFR2-complexes were mixed with dissociation buffer and boiled to release the complexes from the beads. The proteins were resolved by electrophoresis through 4-12% polyacrylamide gels and then transferred to nitrocellulose visualized using an enhanced chemiluminescence detection system (Amersham Biosciences). Quantification of immunoblots was performed using ImageJ software.

Statistics
All data are presented as mean [standard error of mean]. Comparison between groups was performed using Student's t-tests, or 2-way ANOVA for time series data. All tests were 2sided and statistical significance was defined as p<0.05.

Pathological angiogenesis is impaired in Insr +/mice and is associated with impaired responsiveness to VEGF
To study pathological angiogenesis, we induced hindlimb ischemia in Insr +/mice and quantified limb perfusion recovery every 7 days using Laser Doppler imaging. This revealed that, in spite of similar reductions in limb perfusion immediately post-operatively, Insr +/exhibited lower ischemic limb perfusion at all timepoints thereafter (Figure 1A), being 82% Vegfa is indicative of significant residual ischemia, and could also imply an inadequate functional response to this central regulator of angiogenesis. In order to address this possibility, we explanted aortae from a separate group of Insr +/mice to embed rings in a collagen matrix containing VEGF-A 165 which induces sprouting angiogenesis. This revealed fewer capillary sprouts emerging from Insr +/aortae in the presence of VEGF (30 [2.2] in WT vs 22 [2.4] in Insr +/-p<0.05 Figure 2A), and the length of sprouts was also reduced (1037µm [39] WT vs 847µm [38] Insr +/-p<0.05 Figure 2B). Next, we isolated PEC for functional studies; these exhibited appropriate reduction in Insr mRNA (85% [5] of 18S in WT vs 49% [4] in Insr +/-p<0.05 Figure 2C). We then conducted a Matrigel in vitro angiogenesis assay, which demonstrated reduced tubule formation in Insr +/-(28 [2] tubules per microscopic field in WT vs 14.5 [2.4] in Insr +/-p<0.05 Figure 2D) Figure 2G). Collectively, these data imply that Insr +/endothelial cells have selectively impaired migratory responses to VEGF-A.

Developmental angiogenesis is impaired in Insr +/and ECInsr +/mice
Next, we asked whether the abnormalities of pathological angiogenesis in Insr +/were recapitulated during developmental angiogenesis, which we assessed using whole-mounted retinas, at P5 when the vasculature is still developing. The radial outgrowth of the retinal vascular plexus was similar in Insr +/and WT (1367µm [50] WT vs 1354µm [56] Insr +/-, p=0.87; Figure 3A), although there was reduced vascular area in the peripheral vascular  Figure 3F). Next, we asked if EC proliferation was reduced in Insr +/-, but found similar numbers of EdU + EC in the peripheral retinal vasculature of both genotypes (781 [44] WT vs 814 [24] Insr +/-EdU + EC per mm 2 p=0.53; Figure 3G).
Overall, these data are compatible with a reduction in vascular sprouting in the emerging vasculature of Insr +/-, resulting in a less branched neo-vasculature.
To discern whether loss of insulin receptors expressed by EC contribute to the retinal vascular phenotype of Insr +/-, we then studied mice with endothelial cell restricted insulin receptor haploinsufficiency (ECInsr +/-). Again, the radial outgrowth of their retinal vascular plexus was similar to controls (1377µm [32] WT vs 1375µm [40] ECInsr +/-, p=0.96; Figure   4A Figure 4E). Collectively, these data suggest that endothelial cell insulin receptor expression is important in the generation of vascular sprouts, and the branching structure of the nascent vasculature.

Insulin receptor silencing impairs human endothelial cell functional responsiveness to VEGF
In order to explore the relevance of these data in human EC, we transduced HUVEC with lentivirus particles to deliver insulin receptor targeting shRNA (referred to as Insr shRNA), or control GFP targeting shRNA (referred to as control shRNA), reducing insulin receptor protein by 40% (Figure 5A). Transduced HUVEC were then coated on to Cytodex-3 carrier beads and embedded in a fibrin matrix to study VEGF-A 165 induced sprouting angiogenesis in vitro. Mirroring data from Insr +/aortic ring sprouting experiments (Figure 2A-C Figure 5E). Overall, these data suggest that the insulin receptor is also important for VEGF-A induced angiogenic sprouting and cell motility in human EC.

Endothelial insulin receptors are required for VEGFR2 internalization and subsequent ERK signaling
As we had demonstrated that insulin receptor expression influenced functional responses to VEGF-A in human and murine EC, we then asked if this was associated with altered VEGF-A signaling. VEGF-A promotes angiogenesis by binding to VEGF receptor 2 (VEGFR2), a cell membrane-bound receptor tyrosine kinase that initiates a complex intracellular signaling cascade. We therefore stimulated Insr shRNA and control shRNA HUVEC with 50ng/ml VEGF-A 165 and studied major VEGF-A signaling nodes 5-and 15-minutes later, along with unstimulated cells (Figure 6A). Insr shRNA HUVEC exhibited unaffected activation of VEGFR2 (measured by phosphorylation at Y1175), or the downstream nodes Akt (measured by phosphorylation at S473) and eNOS (measured by phosphorylation at S1177)(Data not shown). However, downstream activation of ERK1/2 (measured by phosphorylation at  Figure 6C). Importantly, the activation of Akt and ERK1/2 downstream of VEGFR2 follows highly distinct pathways, with internalization of VEGFR2 being essential for only the latter; (40) Figure   6D,E), which is known to selectively impede ERK1/2 activation. Overall, these data reveal a A c c e p t e d M a n u s c r i p t 20 selective deficit in VEGF-A signal transduction in Insr shRNA HUVEC, which is likely to result from impaired internalization of activated VEGFR2.

Major findings and implications
Our data reveal for the first time that endothelial insulin receptors are required for appropriate migration and angiogenic sprouting in response to VEGF-A, both in vitro and in vivo. At a molecular level, we found that insulin receptor expression promotes the internalization of VEGF-A activated VEGFR2, allowing signaling to ERK1/2. Our data suggest that the pro-angiogenic effects of insulin receptors relate to crosstalk with VEGF-A signaling, although the nature of this interaction, and whether insulin participates in the process, requires further study (Figure 7). This previously unappreciated crosstalk establishes a further link between systems regulating metabolism and angiogenesis.

Insulin and angiogenesis
A number of studies have shown insulin exerts pro-angiogenic effects, although they did not dissect the role of endothelial insulin receptors. (8)(9)(10)(11)(12) These mainly in vitro studies revealed pro-angiogenic effects of nanomolar range insulin, but did not explore more physiological picomolar concentrations. The extent to which picomolar insulin augments ERK signaling in endothelial cells is a source of disagreement in the literature, (11,(42)(43)(44) probably reflecting known heterogeneity between endothelial populations, including in their insulin receptor expression. Notably, insulin receptor expression is reported to be enriched in endothelial tip cells of human tumors,(13) yet tip cells generally lack a lumen,(45) so may be exposed to lower concentrations of insulin than other endothelial cells. Hence, one possible explanation for our data is a ligand-independent role of tip cell insulin receptors. Indeed, recent data from Ronald Kahn's group indicate a ligand-independent role of insulin receptors in the A c c e p t e d M a n u s c r i p t 21 membrane trafficking of brown pre-adipocytes, (46) potentially aligning with our findings.
However, whilst the concentrations of insulin experienced by sprouting endothelial cells in vivo are unknown, our in vitro data are likely to reflect low picomolar concentrations of insulin due to its presence in fetal calf serum. This may suggest a role for insulin in promoting VEGF-induced ERK signaling, as has been shown for epidermal growth factor signaling. (47) Another explanation may be that insulin regulates a common endocytic mechanism for its own receptor and VEGFR2, as discussed later.
The only existing data describing the role of endothelial insulin receptors in angiogenesis were published by Ronald Kahn's group in 2003. Using mice with complete deletion of endothelial insulin receptors, they found a 57% reduction in retinal neovascularization during oxygen-induced retinopathy. (14) However, they did not study the individual cellular processes contributing to angiogenesis, or examine VEGF signaling. Our in vivo data suggest that insulin receptors regulate endothelial tip cell emergence and migration, although have no major impact on endothelial cell proliferation. Notably, we found no reduction in non-ischemic muscle tissue vascularity of Insr +/mice, implying that impaired vascularization ultimately catches up; this is seen in many published examples of impaired angiogenesis, (48)(49)(50) presumably reflecting persistent activation of pro-angiogenic programmes.

VEGFR2 signaling
VEGF-A binding to VEGFR2 induces a complex intracellular signaling cascade,(51) a crucial element of which is internalization (endocytosis) of ligand-bound VEGFR2. This moves the phosphorylated receptor to a domain where it is less susceptible to the phosphatase PTP1B, hence sustaining signal transmission, which is particularly important for ERK1/2 signaling.(40) VEGFR2 internalization, and subsequent ERK1/2 signaling, is known to be A c c e p t e d M a n u s c r i p t 22 crucial to vascular biology, e.g. during arterial morphogenesis and lymphatic specification. (52,53) Whilst less well studied, ERK1/2 activation has recently emerged as regulating endothelial tip cell function and sprouting angiogenesis; (54,55) importantly, tip cells are exposed to the highest VEGF-A concentrations during angiogenesis, and tip cell ERK phosphorylation is prevented in vivo by VEGF inhibition.(7,55) There are many known interacting partners of VEGFR2, which can modify its signaling and internalization. (41,56) The insulin receptor signaling adaptor, insulin receptor substrate-1 (IRS-1), has been implicated in receptor endocytosis, (57) and is reported to interact with VEGFR2. (58) Interestingly, recent data show that insulin signaling to ERK1/2 (and Src homology phosphatase 2) feeds back via serine phosphorylation of IRS-1 to induce insulin receptor internalization, which augments ERK1/2 signaling.(59) Therefore, it is possible that VEGFR2 internalization is similarly impacted, and this putative insulin-dependent mechanism also warrants further exploration. Integrins are also well established to modulate the propagation of ERK1/2 signals downstream of many growth factor receptors, including during angiogenesis, so warrant future assessment. (60). Finally, the insulin receptor can regulate cytoskeletal actin remodeling, (61) another process that influences endocytosis, (62) warranting further exploration in future.

Limitations
Although we show that internalization of activated VEGFR2 is impaired, the underlying mechanism of this requires further investigation. It will also be interesting to explore signal transduction downstream of other receptor tyrosine kinases to assess the generalizability of this phenomenon. As alluded to earlier, it is also important to acknowledge that impaired VEGFR2 internalization may not be the only mechanism by which Insr silencing impairs VEGF-A responses; indeed, intracellular signaling networks are highly complex, as is their perturbation. Moreover, our work only sought to describe the fundamental links between insulin receptors and VEGF signaling during angiogenesis, and so we cannot comment on A c c e p t e d M a n u s c r i p t 23 disease relevance. However, obesity-induced insulin resistance is associated with reduced vascular IR expression and impaired angiogenesis,(63) so it would be interesting to explore endothelial VEGFR2 internalization and ERK signal transduction in this setting. Finally, it is important to note that our ECInsr +/control data comes from littermates expressing Tie2-Cre, and recent data show that Cre is not biologically inert;(64) however, the similar retinal vascular phenotype of Insr +/and ECInsr +/mice provides some reassurance that off-target Cre effects do not underpin our findings.

Conclusions
We show that endothelial insulin receptors are required for appropriate migration and angiogenic sprouting in response to VEGF-A, along with internalization of activated VEGFR2 and downstream signaling to ERK1/2. This novel link between major regulators of systemic metabolism and angiogenesis warrants further mechanistic exploration to understand its wider relevance.

Data availability
All datasets generated during and/or analyzed during the current study are not publicly available but are available from the corresponding author on reasonable request.