Endothelial-Specific Notch Blockade Inhibits Vascular Function and Tumor Growth through an eNOS-Dependent Mechanism Free

Nitric oxide (NO) is an important mediator of normal and pathologic vascular remodeling (1, 2). In endothelial cells, NO is mainly produced by endothelial NO synthase (eNOS; refs. 3 and 4). Genetic ablation of eNOS and pharmacologic approaches have shown that NO contributes to tumor progression by maintaining blood flow (5–7), inducing vascular hyperpermeability (7, 8), recruiting pericytes and promoting vessel morphogenesis (4), and reducing endothelial cell–leucocyte interactions (6). eNOS is activated by phosphorylation of the Ser1117 residue, which is regulated by multiple signaling pathways (1, 2). In this context, the prototypical pro-angiogenic agent VEGFA was shown to activate eNOS in various studies (1, 2). The VEGFA-NO signaling axis is also believed to play a role in tumor angiogenesis but the mechanisms involved are not fully characterized. Recently, we have shown that Notch activates eNOS during embryonic heart development (9).

The Notch family comprises 4 heterodimeric transmembrane receptors and 5 ligands (10). Ligand-mediated Notch activation culminates in the cleavage of the Notch receptor, resulting in untethering of the intracellular domain that then translocates to the nucleus to induce transcription of genes involved in various aspects of cellular function (10). During normal and tumor angiogenesis induced by VEGFA, Notch signaling regulates blood vessel patterning (11). VEGFA induces the expression of the Notch ligand Dll4 in endothelial tip cells, which in turn activates Notch in the stalk endothelial cell, and results in decreased expression of VEGFR2. This diminishes endothelial cell responsiveness to VEGFA and thus limits endothelial cell sprouting. Conversely, strategies that hinder Dll4 signaling produce vascular networks with increased ramifications but with reduced blood perfusion, consequently inhibiting tumor growth (12–14). In contrast, activation of Notch by its ligand Jagged1 has a distinct role on angiogenesis during neonatal retina development (15). Jagged1 promotes angiogenesis by blocking Dll4 interaction with the Notch1 receptor, restricting Notch activation in stalk cells, and increasing VEGFR3 expression in tip cells (15). The glycosamamnylltransferases of the Fringe family fine-tune this tip-stalk cell selection by increasing the affinity of the Notch receptor for Delta member Notch ligands while decreasing Jagged1-Notch affinity (15).

Despite several studies documenting that Notch inhibition results in the generation of nonproductive vasculature when Dll4 is blocked, the mechanism for this phenomenon has not been fully defined (12–14). Furthermore, the effect of endothelial-specific pan-Notch inactivation in endothelial cell during tumor neovascularisation remains unexplored as previous studies used strategies to block Dll4 to inhibit Notch in endothelial cell, which is expressed mainly in tips cells (12–14).

Here we used an inducible transgenic mouse model to study the role of endothelial-specific Notch inactivation during tumor neovascularization. In this model, all activated Notch members are blocked by Tet-inducible, endothelial-specific expression of dominant-negative (dn)MAML (9). Lewis lung carcinoma (LLC) cells engineered to express VEGFA showed poorly perfused vasculature and diminished tumor growth in vivo in dnMAML mice without inducing aberrant increased ramification of the microvasculature as reported for Dll4 blockade (12–14). This observation led us to hypothesize that the Notch-VEGFA signaling axis in endothelial cell may have an important role in tumor blood vessel function independent of regulating sprouting angiogenesis. Indeed, we found that specific endothelial Notch activity in the context of increased VEGF signaling is required for eNOS activation, and that the NO–soluble guanyl cyclase (sGC) ligand–receptor pairing and activation is critical for permitting tumor perfusion in larger blood vessels but not in capillaries. Improving vascular function by activating sGC independent of NO bypasses the Notch inhibitory effect on tumor perfusion. Our results suggest that endothelial cell Notch is important for tumor growth and blood vessel perfusion in response to VEGF by regulating NO production.

Animal studies were approved by the University of British Columbia Animal Care Committee. LLC control cells transduced with empty vector (LLC-neo) or a VEGFA₁₆₅ construct (LLC-VEGFA) were generated with the retroviral system MSCV-pac as previously described (16). B16F10 melanoma cells expressing VEGFA (B16F10-VEGFA) were generated as described for LLC-VEGFA (16). All parental cell lines were from American Type Culture Collection and not authenticated. All cells lines were regularly tested and found negative for mycoplasma.

VE-cadherin (VEC)-tTA transgenic (Tg; ref. 17), control (wild-type and single Tg VEC-tTA or TetOS-dnMAML-GFP Tg; ref. 9) and double Tg mice VEC-tTA-TetOS-dnMAML-GFP from the mating of C57Bl6 with CD1 background in which the different transgenic lines were generated as previously described (9). In some experiments, CD1 mice backcrossed into the C57Bl6 background for 8 generations to conduct isogenic tumor experiments were also used. Doxycycline (Dox; 100 μg/mL) was added to the drinking water from breeding, changed every other day, and withdrawn 7 days before tumor implantation to induce dnMAML-GFP expression. To generate a short induction of dnMAML-GFP expression, Dox was replaced by tetracycline (Tet; 100 μg/mL) 5 days before tumor implantation. Tet was changed every day and withdrawn 16 hours before tumor harvest as indicated. BAY41-2272 (10 mg/kg) was injected intraperitoneally at days 8, 10, and 12 posttumor implantation to activate soluble guanylyl cyclase (NO receptor) independent of NO. The vehicle dimethyl sulfoxide (DMSO) was injected as a control treatment. The eNOS inhibitor N5-(1-iminoethyl)-L-ornithine, dihydrochloride (L-NIO; Cayman) was injected intraperitoneally (20 mg/kg) every day from day 7 to 13. The vehicle saline was injected similarly as a control treatment. LLC-VEGFA or B16F10-VEGFA cells were implanted subcutaneously in the flank of 6- to 8-week-old mice.

Flow cytometry following staining with 4′,6-diamidino-2-phenylindole or propidium iodide, CD45-APC-Cy7, CD31-PerCPeFluor710, or PDGFRβ-APC (eBiosciences) was performed on an Influx cell sorter (BD BioSciences) as described (18). Tumor cryosections were fixed and stained with rat anti-mouse VE-cadherin, rat anti-mouse CD31 (BD Pharmingen), rabbit anti-human Desmin (Lab Vision), rabbit anti-phospho-Ser1117 eNOS (Cell Signalling), and anti-goat anti-GFP-FITC (Abcam). Functional blood vessels were quantified by tail vein injection of either DiOC₇ or DilC₁₈ (3) depending on the fluorophore used for the vascular markers. The NO-reactive dye diaminorhodamine 4M-AM (DAR4M, 25 μmol/kg; Santa-Cruz) was injected intraperitoneally 1 hour before mice were killed as described (9). A custom robotic microscope was used to image immunofluorescent staining across entire tumor sections at 0.75 μm/pixel resolution, as previously described (19). Images were quantified using customized NIH-ImageJ software (19). The quantification of NO production in the microvasculature and larger vessels was performed with the activated eNOS (phospho-Ser1117)-specific antibody and the NO-specific probe DAR4M. For microvasculature quantification of activated eNOS and DAR4M staining was performed across the entire cryosection. For larger vessels, images of cross-sections from mural cell-invested blood vessels, which have larger lumina were separately analyzed from the microvasculature. Identical thresholds for object pixel size and brightness were used for the quantification of microvasculature and larger vessels.

HMEC control (empty vector) or overexpressing a human Jagged1 (Jag1) construct (20) were cultured with MCDB (Sigma) supplemented with 20% FBS, penicillin–streptomycin, glutamine, ECGS, and heparin. A total of 10,000 HMEC parental cells along with 10,000 HMEC-control or HMEC-Jag1 cells were seeded in multiculture chamber slides (BD Biosciences) the day before treatment with 2 μmol/L of the specific VEGFR2 inhibitor Ki8751 for 16 hours (21). Cells were stained with DAR4M (10 μmol/L; ref. 9) for the last hour treatment with Ki8751, washed with PBS, fixed with PFA 4% for 10 minutes, rinsed, dried, and imaged by fluorescence microscopy.

FACS-sorted cells were washed once with PBS and snap frozen at −80°C in lysis buffer until RNA extraction with the RNeasy Mini Kit according to manufacturer instructions (Qiagen). The cDNA from total RNA samples of sorted or total cell fractions were generated using SuperScript II with random nonamers primers according to the manufacturer's instructions (Invitrogen). The final 50 μL cDNA mixture was diluted by adding 150 μL H₂O and 2.5 μL of this dilution was used with 5 μL of primers (1 μmol/L each) and 7.5 μL of SybrGreen (Roche) for qPCR reactions using an ABI 7900 Thermal cycler instrument (Applied Biosystems). The Gapdh gene was used to normalize expression.

All data are shown as mean ± SEM. P values were calculated using the Student t test (GraphPad Prism).

To study the role of endothelial cell–specific Notch blockade on tumor angiogenesis, we used a tetracycline (Tet)-based inducible system expressing a dominant-negative form of the Notch coactivator MAML fused to GFP (dnMAML-GFP; ref. 9). dnMAML has been used to assess the effect of pan-inhibition of Notch signaling used in various vascular models, and although found to be specifically inhibitory to Notch signaling (22–26), one cannot entirely rule out the possibility that some unknown pathway is also impacted. We hypothesized that increasing VEGFA signaling would sensitize tumors to Notch inhibition, as VEGFA activates endothelial cell Notch signaling by increasing Dll4 expression in tip cells (12–14, 27). As expected, tumor growth of LLC overexpressing VEGFA (LLC-VEGFA) was significantly reduced in VEC-tTA × TetOS-dnMAML-GFP double Tg animals compared with littermate control mice (Fig. 1A), but LLC-vector cells, which express low levels of VEGFA (28), were not impacted by Notch inhibition (Fig. 1B). Flow cytometry of tumor cell suspensions (Fig. 1C) confirmed endothelial cell–specific expression of the dnMAML-GFP transgene. As expected, the expression of Notch target genes Hey1, Hes1, and Dll4 were reduced in the tumor endothelial cell (Fig. 1D; ref. 15). In addition, Hey1 expression was decreased in both Dll4⁻ and Dll4⁺ sorted endothelial cell of double Tg mice (Fig. 1E), confirming pan-endothelial cell Notch inhibition in double Tg mice in both tip (Dll4⁺) and stalk (Dll4⁻) cells.

Immunofluorescence analysis confirmed endothelial cell–specific expression of the dnMAML-GFP transgene (Fig. 2A). In contrast to Dll4 blockade, which leads to excessive sprouting of the tumor vasculature, microvessel density—as measured by the tumor area occupied by endothelial cell (CD31 or VE-cadherin)—was not affected by pan-Notch inhibition (Fig. 2B; refs. 12–14). Analysis of smaller LLC-VEGFA tumors at 6 days of growth when they reach a palpable size also showed that the abundance of blood vessels was not changed, suggesting that the absence of aberrant sprouting in endothelial cell-Notch–inhibited (double Tg) animals is not related to a late tumor effect (Fig. 2C). In addition, short induction of dnMAML-GFP expression for 16 hours (Tet pulse) did not affect the abundance of blood vessels (Fig. 2C). Together, these results suggest that tumor growth inhibition induced by specific endothelial Notch-inhibition in VE-cadherin-tTA × TeTOS-dnMAML-GFP is independent of the aberrant sprouting of the tumor vasculature.

To determine whether Notch inhibition affects vascular perfusion in this context, as described for Dll4 inhibition (12–14), we analyzed the effect of blocking endothelial cell Notch on blood vessel function. Concordant with studies using Dll4 blockade, injection of the fluorescent tracer DioC₇ intravenously showed reduced perfusion in tumors of double Tg mice (Fig. 3A and B and Supplementary Fig. S1). The decrease in functional vasculature correlated significantly with increased expression of dnMAML-GFP in tumor blood vessels (Fig. 3C).

Vascular smooth muscle cells and pericytes control blood flow and previous studies have demonstrated the importance of Notch signaling in regulating these cells in tumor and retinal angiogenesis (14, 15). However, staining of tumor sections did not show a difference in immunostaining for the pericyte marker Desmin between double Tg and control mice (Fig. 4A and B). Furthermore, pericyte distribution relative to endothelial cell was not different between control and double Tg mice (Fig. 4C), and Desmin-positive cells were closely opposed to dnMAML-GFP-expressing endothelial cell (Fig. 4A inset), suggesting that the recruitment of pericytes to blood vessels is not affected by endothelial cell Notch blockade. The differentiation of pericytes, which impacts on the function of blood vessels (29), was also similar in tumors of double Tg versus control Tg mice as determined by the expression of early (RGS5 and PDGFRβ) and late (NG2, Desmin, and α-SMA) markers of pericyte differentiation (Fig. 4D). In support of this, flow cytometry analysis showed similar abundance of PDGFRβ⁺ cells in LLC-VEGFA tumors of control Tg versus double Tg (Supplementary Fig. S2). Together, these results suggest that pericyte abundance, maturation, and recruitment to endothelial cells are not responsible for the functional vascular deficiency observed in response to blockade of endothelial Notch in tumor blood vessels.

We have recently shown that Notch is a regulator of eNOS during cardiac cushion differentiation (9). We thus hypothesized that the decrease of functional vasculature in double Tg animals results from a defect of NO production. Initial quantitative analysis from whole tumor sections showed no significant changes of activated eNOS (phospho-Ser1117) and binding of the NO-specific probe DAR4M in LLC-VEGFA tumors of double versus control Tg animals (Supplementary Fig. S3A and S3B). We hypothesized that inhibition of Notch may have distinct effects in the production of NO in endothelial cell of larger vessels compared with the microvasculature, as NO would be expected to have a greater impact on vasodilation of larger vessels because of the thicker perivascular wall comprising vascular smooth muscle cells. To verify this hypothesis, cross-sections of larger tumor vessels were quantified for activated eNOS and DAR4M (NO) retention. As expected, activation of eNOS and NO production was reduced in tumors of larger vessels of double Tg mice (Fig. 5A and B).

To further validate the role of NO in the vascular function defect of double Tg mice, we performed a rescue experiment with BAY41-2272, a NO-independent activator of soluble guanylyl cyclase. Treatment with BAY41-2272 restored vascular function in double Tg mice (Fig. 5C), supporting the thesis that defective eNOS activation and consequent inhibition of NO production is responsible for limiting vasodilation and tumor perfusion in mice that have endothelial cell–specific Notch blockade. Treatment with BAY41-2272 also tended to rescue the tumor growth in double Tg, although the difference failed to reach statistical significance (P = 0.08; Supplementary Fig. S4). In addition, short induction of dnMAML-GFP expression for 16 hours (Tet removal) before tumor harvest also decreased phospho-eNOS in larger vessels (Fig. 5D), indicating that inhibition of eNOS activation is an early effect of Notch inactivation in endothelial cell of larger vessels rather than a secondary effect of tumor growth inhibition in double Tg animals. Accordingly, treatment with the L-NIO, a potent eNOS inhibitor (30), reduced LLC-VEGFA tumor growth (Fig. 5E). Similar observations were obtained using a B16F10-VEGFA tumor model (Supplementary Fig. S5). Taken together, our findings indicate that Notch inactivation-mediated inhibition of eNOS activity results in vascular dysfunction in larger vessels of the tumor vasculature affecting tumor growth. Our data support other studies showing that eNOS inhibition using genetic or pharmacologic strategies reduces tumor blood flow and growth (5–8, 31).

VEGFR2 is a major effector of VEGFA signaling (32, 33). To assess the role of VEGFR2 signaling in Notch-mediated production of NO in endothelial cell, we quantified DAR4M staining in cocultures of HMEC overexpressing the Notch ligand Jagged1 with HMEC parental cells (9) in the presence of VEGFR2-specific inhibitor Ki8751 (21). As expected, in this system NO production increased in HMEC-Jag1 cocultures compared with HMEC-control vector cocultures (Fig. 6A and B), and treatment with the VEGFR2 inhibitor Ki8751 reduced Jag1-mediated NO stimulation (Fig. 6A and B). This finding suggests that Notch is required for VEGFR2-induced NO production in endothelial cell.

A recent study shows that Notch in endothelial cell is important to control the expression of genes involved in vasodilation/vasoconstriction (34). We thus determined whether the expression of these genes is affected by endothelial cell–specific Notch inhibition in tumors. Interestingly, the expression of the vasodilation genes Nts and Adm was decreased in endothelial cell from LLC-VEGFA tumors of double Tg (Supplementary Fig. S6), compatible with reduced perfusion in the tumors of mice with endothelial cell–specific Notch inhibition. In contrast, expression of other genes with vasoactive function suggested to be affected by Notch were not detectable in tumor endothelial cell, although present in the bulk tumor population (Supplementary Table S1). These findings suggest that the gene expression program regulating vasodilation is diminished in response to endothelial cell–specific Notch inhibition.

Although a better understanding of Notch function in the microvasculature is being established, questions remain about the mechanistic role of endothelial Notch in the tumor vasculature. This study reveals that Notch signaling in endothelial cell of larger vessels facilitates blood flow by increasing production of the vasodilatory agent NO. These findings are compatible with previous studies showing that eNOS-targeted mice exhibit reduced LLC tumor growth and blood vessel permeability (8). Furthermore, our data suggest a mechanism in which Notch induces endothelial NO in a VEGFA–VEGFR2 axis-dependent manner. Elucidation of this mechanism in this study stems from multiple observations generated using various experimental strategies: (i) specific inhibition of Notch in endothelial cell inhibits blood vessel perfusion and tumor growth in experimental mouse tumor models in which the tumor cells produce VEGFA; (ii) NO production and activated eNOS activity is rapidly decreased in larger vessels in response to endothelial cell–specific Notch inhibition; (iii) activation of the NO receptors gastric cancer with the vasodilation agent BAY41-2272 rescues vascular function defects induced by endothelial cell–specific Notch inhibition; (iv) specific pharmacologic inhibition of VEGFR2 blocks Notch-mediated NO production in endothelial cell; and (v) endothelial cell–specific Notch inhibition reduces the expression of Nts and Adm, 2 genes that simulate vasodilation.

Our observations that inhibition of Notch in endothelial cell decreases tumor perfusion as well the expression of vasodilatory genes Nts and Adm is at odds with a previous study showing that Dll4-mediated Notch signaling induces the expression of vasoconstrictor genes, and diminishes vasodilator gene expression (34). The nature of this discrepancy is not clear, but may be because of the different models of vascular development, that is, tumor vasculature in this study versus retinal angiogenesis during development. In line with this supposition, tumor blood vessels are known to behave aberrantly and are notably more tortuous and leaky than vessels from normal vascular remodeling tissues such as the developing retina (32). Moreover, the genetic model used in our study to induce pan-Notch inhibition in endothelial cell versus Dll4 blockers used in the study of Lobov and colleagues (34) could also account for the difference observed in the 2 studies. Finally, we specifically sorted endothelial cells to examine expression, whereas the previous study examined the entire retinal tissue, which includes a variety of cell types.

The observation that endothelial-specific Notch inhibition does not induce sprouting angiogenesis defects in this study could be a consequence of Notch activation signals that have opposing effects during sprouting angiogenesisas seen with the distinct outcomes resulting from targeting Dll4 or Jagged1 (15). Interestingly, Jagged1 was shown to block Dll4-mediated Notch activation (15). Indeed, the observation that Jagged1 can also induce Notch target genes in the context of low Fringe expression (15) and during tumor neovascularization (35) raises the question whether Jagged1 (or Notch ligands other than Dll4) can also inhibit sprouting angiogenesis by inducing Notch activation signaling in a specific subset of endothelial cell of the nascent blood vessels. It is also possible that the strength of the Notch signal may activate different signaling pathways that have different functional effects during tumor angiogenesis. A recent study demonstrated this dose-dependence of Notch signaling for endocrine progenitors in the zebrafish intrapancreatic duct (36). Future studies in transgenic expression systems allowing dynamic assessment of Notch activity during tumor neovascularization will be needed to investigate these important questions.

A previous study has reported that NO production is important for recruitment of vascular smooth muscle cells and pericytes in tumors (4). In this study we did not find that the abundance of these perivascular cells was affected by Notch inhibition-mediated reduction of NO. The nature of this discrepancy could be explained by the increase of VEGFA signaling in LLC-VEGFA tumors. Indeed, VEGFA has been shown to reduce recruitment of pericytes (37) and thus high VEGFA-expressing tumors may be less sensitive to NO-dependent pericyte recruitment.

In conclusion, our study reveals that Notch activates the production of the vasorelaxing agent NO in endothelial cell of tumor vasculature, which is important for blood vessel function and consequently tumor growth.

No potential conflicts of interest were disclosed.

Conception and design: A. Patenaude, F. Wong, A. Karsan

Development of methodology: A. Patenaude, F. Wong, A.H. Kyle, E. Diaz, A.I. Minchinton, A. Karsan

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Patenaude, M. Fuller, L. Chang, F. Wong, G. Paliouras, R. Shaw, A.H. Kyle, J.H.E. Baker, J. Tong, A. Karsan

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Patenaude, F. Wong, G. Paliouras, A. Karsan

Writing, review, and/or revision of the manuscript: A. Patenaude, F. Wong, A. Karsan

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. Patenaude, M. Fuller, F. Wong, P. Umlandt, J. Tong, A. Karsan

Study supervision: A. Patenaude, F. Wong, A. Karsan

The authors are grateful to K. Lau for assistance with statistical analysis.

This research was supported by grants to A. Karsan from the Heart and Stroke Foundation of British Columbia and the Yukon and the Canadian Institutes for Health Research (MOP 64354). A. Patenaude and L. Chang were supported by research trainee awards from the Michael Smith Foundation for Health Research (MSFHR).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.