Enhanced notch signaling modulates unproductive revascularization in response to nitric oxide‐angiopoietin signaling in a mouse model of peripheral ischemia

Abstract Introduction Arteriolargenesis can be induced by concomitant stimulation of nitric Oxide (NO)‐Angiopoietin receptor (Tie)‐Vascular Endothelial Growth Factor (VEGF) signaling in the rat mesentery angiogenesis assay. We hypothesized that the same combination of exogenously added growth factors would also have a positive impact on arteriolargenesis and, consequently, the recovery of blood flow in a model of unilateral hindlimb ischemia. Results and Methods NO‐Tie mice had faster blood flow recovery compared to control mice, as assessed by laser speckle imaging. There was no change in capillary density within the ischemic muscles, but arteriole density was higher in NO‐Tie mice. Given the previously documented beneficial effect of VEGF signaling, we tested whether NO‐Tie‐VEGF mice would show further improvement. Surprisingly, these mice recovered no differently from control, arteriole density was similar and capillary density was lower. Dll4 is a driver of arterial specification, so we hypothesized that Notch1 expression would be involved in arteriolargenesis. There was a significant upregulation of Notch1 transcripts in NO‐Tie‐VEGF compared with NO‐Tie mice. Using soluble Dll4 (sDll4), we stimulated Notch signaling in the ischemic muscles of mice. NO‐Tie‐sDll4 mice had significantly increased capillary and arteriole densities, but impaired blood flow recovery. Conclusion These results suggest that Dll4 activation early on in revascularization can lead to unproductive angiogenesis and arteriolargenesis, despite increased vascular densities. These results suggest spatial and temporal balance of growth factors needs to be perfected for ideal functional and anatomical revascularisation.


| INTRODUC TI ON
PAD has become a global problem in the 21st century and is a common morbidity factor associated with some chronic diseases, such as diabetes. 1 Although amputation-free survival has steadily increased in high income countries, retrospective long-term follow-up studies reveal that the severe metabolic damage caused by ischemia still leads to poor prognosis in critical limb ischemia patients. 2 TreatmentofPAD relies upon effective vascular remodeling resulting in a "vascular tree," comprising the coordinated induction of arteriole, venule, and capillary identities that are capable of controlling blood flow. Moreover, regulated perfusion requires the formation of muscularized conduit vessels to deliver sufficient blood flow to the capillary exchange vessels.
The early promise of preclinical trials using VEGF as a therapeutic did not give rise to the expected results. 3 Angiogenesis induced by VEGF over-expression is abnormal as vessels lack pericytes or VSMCs, appear thin walled and hypertrophic, and are prone to regression upon VEGF withdrawal. [4][5][6] To develop a mature, stable microvasculature, other proteins appear to be necessary. 7 As simply over-expressing individual growth factors has demonstrated that some, but not all aspects of therapeutic revascularization can be achieved, it seems likely that the spatial, temporal, and biochemical co-ordination of specific growth factors will be key to effective vascular remodeling. As such, studies combining delivery of different growth factors have started to report more encouraging results. For example, combining VEGF 165 and Ang1 improved the outcome in a diabetic rat model of myocardial infarction by reducing fibrosis and ventricular remodeling, along with increased angiogenesis. 8 It has also been shown, in a rat dorsal skin flap model that the mechanism through which adenoviral-mediated delivery of VEGF (through Ad.VEGF) increases tissue viability and blood flow is due to an increase in NO synthesis and release. 9 Using a model of physiological angiogenesis, the rat mesentery assay, we previously described the process of arteriolargenesis, the generation of arterioles from capillaries. [10][11][12] We have shown that, by over-expressing eNOS (endothelial nitric oxide synthase), we promote a NO-mediated increase in blood flow, which gives rise to angiogenesis dependent uponVEGF-VEGFR andAng1-Tie signaling, 12 which was more beneficial than the combination of Ang1+VEGF alone. 11 Furthermore, the combination of adenoviral-mediated delivery of eNOS andAng1 (NO-Tie) is sufficient to promote arteriolargenesis, 12 but to a reduced degree compared with adenovirally mediated delivery of eNOS, Ang1, and VEGF (NO-Tie-VEGF). 10 VEGF has been shown to induce activation of the Dll4 notch pathway during neovascularization. Dll4 activation results in less permeable blood vessels, and a more mature vasculature, with more arterioles and fewer branch points. We therefore tested the hypoth-esisthataddinginDll4activationtotheNO-Tie-VEGFcombination might generate an improved revascularization.
Ad.eGFP (enhancedgreenfluorescentprotein,usedascontrol) and Ad.Ang1* (Angiopoietin 1) were a gift from Regeneron Inc., Tarrytown,USA. 12 Ad.eNOS(endothelialnitricoxidesynthase)was a gift from Prof Keith Channon and gave rise to increased production of nitric oxide. 12 Ad.VEGFwaspreviouslycharacterizedasgivingrisetotheover-expressionofthehumanVEGF-A 165 a isoform in the mesentery. 6 Ad.sDll4 was custom designed and purchased from Vector Ad.sDll4 gives rise to the over-expression of the extracellular portion of Delta-like ligand 4, which is sufficient to activate Notch signaling, nuclear translocation of NICD, and Hes/Hey upregulation in ECs. 14 Viruses were titrated by end-point dilution and purified using the AdEasy Virus Purification Kit (2 × 100). were then seeded on top of the fibroblasts and allowed to settle overnight. The addition of fibroblasts allowed for a continuously renewing layer of fibronectin and collagen to be synthesized, which supported spontaneous organization of ECs into tubule-like structures. 15 Thecellculturemediawasthenreplacedwithhalfvolume of fresh media and half volume adenovirus-conditioned media and changed every 2 days for 2 weeks, thus enabling us to test the functionality of the proteins produced via adenoviral-mediated infection. After 2 weeks, the media was removed, and the cells were fixed with ethanol, stained with polyclonal rabbit anti-VE-Cadherin (5 μg/mL, ab33168, Abcam), and counterstained with DAPI. z-stack images of 5 sections containing tubes from each coverslip were acquired, and image analysis was undertaken blinded using ImageJ. 16 We counted number of endothelial tubes, measured tube segment length, and expressed sprout point and branch point number per unit area.

| Hindlimb ischemia, gene transfer, and color speckle imaging
In accordance with the Animal (Scientific Procedures) Act (UK) of 1986 prepared by the Institute of Laboratory Animal Resources, C57Bl6 mice were subjected to surgical induction of unilateral hindlimb ischemia, as previously described (Couffinhal et al., 1998).
Anesthesia was induced with 5% isoflurane vaporized in 100% oxygen at a rate of 1 L/min and maintained at 2% isoflurane. Body temperature was controlled throughout with a homeothermic blanketandrectalprobe(HarvardApparatus).Theskinwasepilatedand sterilized using a chlorhexidine-based solution. All mice were administered pre-operative analgesic (buprenorphine at 0.05 mg/kg) and saline solution (0.9%NaCl at 40 mL/kg). An incision was cut in the groin region, and connective tissue was teased apart to expose the main artery supplying blood flow to the hindlimb. The nerve bundle was teased apart and the femoral artery was ligated in two sites above the epigastric branch and electrocoagulated in between, to induce ischemia to the left hindpaw. Immediately after ischemic induction, in all animals, 10 9 pfu of each adenovirus was injected intramuscularly in the ischemic adductor. The superficial blood flow to both ischemic and non-ischemic feet was measured using a color laserspeckleimagingsystem(MoorFLIP2;MoorInstruments)inanesthetized animals with 1% isoflurane at days 0, 3, 7, 14, and 21 after induction of limb ischemia. 17 The ratio of blood flow between the ischemic and contralateral foot was calculated to use as an index of percentage blood flow recovery.

| Digital droplet PCR
At day 3 post surgery, tissue was recovered from some animals and lysed using Trizol with mechanical homogenization to extract RNA.

| Immunofluorescence
At the experimental end-point, mice were euthanized and the ischemic adductor muscles were dissected and embedded in O.C.T. compound in liquid nitrogen-cooled isopentane. These muscles were sectioned and stained with IB4 and α-SMA-Cy3.Sectionswere imaged with a confocal microscope (Leica SPE), and counts of 30 random microscopic fields per mouse were averaged. Blood vessel profiles measuring around 5 μm in diameter stained only with IB4 were counted as capillaries, and those displaying a second cellular layerstainedwithSMA,surroundingtheinnerone,werecountedas arterioles(forexamples,seeFigure1C).Resultswereexpressedas density of capillaries and arterioles per mm 2 .

| RE SULTS
The induction of unilateral hindlimb ischemia in 12-week-old male We have previously shown that activation of Notch signaling by soluble Dll4 (sDll4) decreases vascular permeability, mainly due to an increased expression of VE-Cadherin at intercellular junctions, 14 and thus we used an adenovirus (Ad.sDll4) to promote early vessel maturation in the hindlimb ischemia mouse model, in order to assess itseffectonrecoveryofbloodflow.Firstly,toconfirmthatsDll4was able to induce angiogenesis as well, we used an in vitro angiogenesis assay, where co-cultured fibroblasts and ECs were treated with adenovirus-conditionedmedia,stained,andimaged(Figure3).Intheunstimulatedinvitroangiogenesisassay(eGFP),weobservedsingular tubes that were not connected to one another, whereas the tubes in F I G U R E 1 VEGFinhibitsNO-Tiemediated revascularization in C57Bl6 mice. (A) Laser speckle images from representative animals over the course ofrecovery.(B)Two-wayrepeated measures ANOVA of the ischemic/ contralateral ratio reveals that virus had a significant effect in the variation of recovery profiles (5.89% of total variance). Thiswasparticularlyimportantat3and 7daysaftersurgery,whenNO-Tiemice (injectedwithAd.eNOS+Ad.Ang1)had faster blood flow recovery compared with controlmice(injectedwithAd.eGFP), while there was no difference between controlandNO-Tie-VEGFmice(injected withAd.eNOS+Ad.Ang1+Ad.VEGF). n=12mice/group.Two-wayrepeated measures analysis of variance followed byTukey'smultiplecomparisontest: *P < 0.05, **P < 0.01. (C) Confocal images from representative muscle sections of ischemic adductors stained with isolectin B4 for endothelial cells (green) and αsmooth muscle actin for vascular smooth musclecells(red).Scalebar=25μm. The role of Notch signaling in driving assembly of adherens junctions and regulating endothelial barrier integrity has recently been described. 14,18 Thesearemolecularpathwaysinvolvedinvessel maturation and, thus, precursors to arteriolargenesis. So, we assessed the effect of sDll4 in NO-Tie-induced recovery of blood flow after ischemia (Figure 4). We found that adenoviral-mediated sDll4 over-expression in the ischemic adductor had no effect in recovery of blood flow after ischemia when compared to control mice (Figure4A).However,sDll4significantlyimpairedthepositiveeffect inbloodfluxofNO-Tierecoveryafterischemia,specificallyatday14 aftersurgery (Figure4B).Surprisingly, additionof Ad.sDll4further worsened the outcome of NO-Tie-VEGF mice as well ( Figure 4C).
Moreover, sDll4 addition significantly increased both capillary and arterioledensitiesinallanimalgroups(Figure4D-I),whichpointsto a mismatch between metabolic demands and blood vessel densities, leading to lack of blood flow control at the vascular network level.

| D ISCUSS I ON
Similar to its effects on the rat mesentery assay, 12  ECs through endocytic vesicles regulates arterial morphogenesis 20 and requires the cytoplasmic domain of neuropilin 1. 20 Moreover, ablation of the scaffold protein synectin expression in ECs, but not VSMCs,resultedinimpairedarteriogenesisinadultmice 21 and this phenotype was rescued upon endothelial-specific knockout of phosphotyrosine phosphatase 1b. 22 Thus, it appears that VEGF signaling and VEGFR2 internalization in specific endosomes within ECs may be required for arteriolargenesis. However, when the source of VEGFcomesfromtheischemicmuscle,thesubsequentrecruitment ofVSMCsmaypreventfurthersproutingangiogenesisandleadto premature arterialization of the newly formed vasculature in the ischemic muscle. Therefore,wesoughtsomemolecularcuestohelpusunderstand this process and found that one of the genes involved in arterial specification,Notch1,wasupregulatedinNO-Tie-VEGFmice,when compared to NO-Tie mice ( Figure 2B). Jagged 1/Notch 1 signaling pathway has been implicated in endothelial-mesenchymal transition-it is activated in microvascular ECs in the bleomycin-induced model of pulmonary fibrosis in rats, where it may induce α-smooth muscle actin expression via a non-canonical pathway involving NFκB. 23 Activation of Notch 1 signaling has also been found in brain arteriovenous malformations in humans. 24 Furthermore, molecular alterations in Dll4-Hey2 signaling are associated with VSMC hypertrophy and hyperplasia in varicose veins. 25 Recent gene expression signatures associated with Notch ligands Dll4 and Jagged 1 in plaque material from PAD patients have been associated with disease progression. 26 Adult Dll4 +/− mice have increased angiogenesis, F I G U R E 3 Ad.sDll4 increased sprouting but reduced branching. (A) HDBECs were co-cultured on top of fibroblasts in an in vitro angiogenesisassay.TheyweretreatedwithconditionedmediaforeitherAd.eGFP,Ad.VEGF,Ad.sDll4,orAd.VEGF+Ad.sDll4everyother day for 2 weeks (n = 3 experimental replicates, experiment was repeated three times). Cells were then fixed and stained for the endothelial junctionalmarker,VE-Cadherin(green)andthenuclearmarker,DAPI(blue).(B)BothsDll4andVEGF+sDll4resultedinasignificantlyhigher tube density (sDll4: P=0.0004,sDll4+VEGF: P=0.0005),thanthecontroleGFP,yetshortermeantubelengths(C).TreatmentwithsDll4 resulted in tubes with the most sprouts (16.33 ± 0.677 P=0.0091);however,VEGFaloneandincombinationwithsDll4alsoincreasedthis (D).Additionally,allthreetreatmentsincreasedthedensityofbranches,withVEGFinducingthemostandsDll4aloneresultinginslightly less than the combination treatment. One-way ANOVA with a Bonferroni post-test, with a 95% confidence interval. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 decreased arteriolar density, more severe tissue ischemia, and poor bloodflowrecoveryafterfemoralarteryocclusion.Theyalsopresent increased plasma leakage, decreased flow-induced outward remodeling, increased arteriolar contractility, and impaired arteriolar responses to shear-stress and vasoactive molecules. 27 In the past, a different Ad.sDll4 has been used as a decoy, inhibiting Notch signaling, downregulating expression of Hes, and giving rise to a disorganized and non-perfused capillary network in a model of hindlimb ischemia. 28 However, we have found that sDll4 is sufficient to upregulate Hes expression and activate Notch signaling, tightening the endothelial barrier in vitro and in vivo. 14 Here, we find that sDll4 is sufficient to induce an angiogenic response. However, in an in vitro angiogenesis assay, the resulting tubule network has a different topology than the one elicited by VEGF. The EC tubes formed had more blind-endsandwerelessinterconnectedthanthoseformedbyVEGF alone.WhenVEGFandsDll4weresimultaneouslyapplied,theaspect of the network resembled that of sDll4-only stimulation (Figure 3).
Given the rising interest and often conflicting data on the role of Dll4/ Notch signaling in promoting vessel maturation [27][28][29][30] and our own experience researching the role of Dll4 in regulating vascular permeability and intercellular communication, 14 we also tested how Dll4/ Notch signaling affected recovery of blood flow after ischemia. We postulatedthatthedeleteriouseffectofVEGFcouldpossiblybeexplained by the activation of Notch in the ischemic muscle, and that we F I G U R E 4 Ad.sDll4impairsrecoveryofbloodfluxinNO-Tiemice,evenwhenparalleledbyincreasedangiogenesisandarteriolargenesis. (A) Ad.sDll4 had no effect on recovery of blood flux after ischemia when compared to control mice but (B) a significant effect in the variation ofrecoveryprofileswasobservedwhenthiswasaddedtoNO-Tiemice(3.22%oftotalvariance,***P = 0.0004). (C) Addition of Ad.sDll4 furtherworsenedtheoutcomeofNO-Tie-VEGFmice(1.80%oftotalvariance,**P=0.0070).Two-wayANOVAwithSidak'spost-test: *P<0.05,n=12mice/group.Inallanimalgroups,additionofAd.sDll4ledtoanincreaseinbothcapillary(D-F)andarterioledensities(G-I). n=5-6mice/group.Two-tailedMann-Whitneytestwith95%confidenceinterval:****P < 0.0001 could be able to recapitulate this phenotype using sDll4. We observed a phenotype of unproductive angiogenesis and arteriolargenesis elic- Together,theseresultssuggestthatthenegativeeffectofexog-enousVEGFinNO-Tiemicerecoveryfrombloodflowmaybeparalleled by, but is not synergistic with, Notch1 activation, and that the quality of the revascularization in terms of the balance between the temporal and spatial development of arterioles vs capillaries is critical for effective recovery from ischemia.