VEGF and EG-VEGF

Supplement 1 to Campochiaro PA, First ARVO/Pfizer Institute Working Group. Ocular versus Extraocular Neovascularization: Mirror Images or Vague Resemblances. Invest Ophthalmol Vis Sci. 2006;47:462-474.

Napoleone Ferrara

Genentech, San Francisco, CA

Vascular endothelial growth factor (VEGF) plays an essential role in . This includes developmental angiogenesis as the loss of a single allele is embryonic lethal due to lack of development of endothelial cells. During early development, endothelial cells depend on VEGF for survival and if it is withdrawn, massive results. This is not true in adult organisms; withdrawal of VEGF does not result in widespread apoptosis of endothelial cells, but angiogenesis still appears to be dependent upon VEGF. This includes cyclical angiogenesis in the female reproductive system, skeletal growth, and pathologic angiogenesis, such as tumor angiogenesis. There are three receptors for VEGF, VEGFR1, R2, and R3. VEGFR3 mediates lymphangiogenesis. VEGFR2 is the major mediator of mitogenesis of endothelial cells. It binds VEGF-A and proteolytic fragments of VEGF-C and –D. VEGFR1 is not as well understood. It binds VEGF-A and –B, and (PlGF). Depending on the setting, VEGFR1 may act as a decoy or as a signaling receptor. It doesn’t mediate mitogenesis, but seems to promote differentiated functions of endothelial cells such as release of growth factors. It also has a role in hematopoiesis and recruitment of monocytes/macrophages. Neuropilin 1 and 2 are also receptors for VEGF; they are isoform specific receptors/coreceptors. The Vegf gene has eight exons and alternative splicing of mRNA results in several isoforms. The different isoforms have different properties: VEGF189 and VEGF206 bind very tightly to , so they are sequestered in extracellular matrix. VEGF165 is the predominant isoform. It binds heparin less tightly and therefore has some solubility but also associates with matrix. VEGF121 lacks a heparin-binding domain and so it is very soluble. Other active moieties of VEGF are generated by proteolytic degradation. cleaves VEGF165 resulting in an N-terminal 110 amino acid peptide which lacks a heparin- binding domain. VEGF110 can associate with VEGF165 to form VEGF110-VEGF165 heterodimers, and then with more digestion, VEGF110 homodimers can result. Matrix -3 (MMP3) cleaves VEGF165 resulting in an N-terminal 113 amino acid peptide. Inflammation causes increased levels of these enzymes and therefore increased levels of the VEGF fragments. In tumor xenograft models, tumors are obtained when Vegf null cells are implanted, because VEGF derived from the host is sufficient. Tumor growth still occurs when up to 70% of VEGF activity is eliminated. It appears that blockade of all VEGF isoforms and fragments is necessary to achieve sufficient reduction of VEGF activity to prevent tumor growth. In ocular fluids obtained from patients with diabetic retinopathy, total VEGF activity is significantly higher than activity attributable to heparin binding VEGF. This means that VEGF165, which is included within the heparin binding fraction, is partially responsible for VEGF activities and the remaining, mediated by VEGF121 and/or by proteolytic products of heparin-binding VEGF, is likely to be responsible for a significant amount of ocular pathology.

Downloaded from iovs.arvojournals.org on 09/30/2021 2

The first ocular disease that has been clinically targeted with anti-VEGF treatment is neovascular age-related macular degeneration (NVAMD). Clinical trials are underway testing Ranibizumab (Lucentis), an affinity-matured Fab variant of (Avastin). The affinity of Lucentis is increased by ~100-fold above that of Avastin Fab. Like Avastin, Lucentis inhibits all isoforms and active fragments of VEGF. Its molecular weight is 48 kD. It has been demonstrated to suppress CNV at Bruch’s membrane rupture sites in monkeys. Endothelial gland-derived VEGF (EG-VEGF) is an angiogenic that is structurally unrelated to VEGF. It is a selective mitogen for adrenal endothelial cells and has no effect on human umbilical vein endothelial cells. It is expressed in steroidogenic tissues such as adrenal gland, ovary, testis, and placenta. Like VEGF it can induce fenestrae in endothelial cells. It is a mediator of tissue-specific angiogenesis; it induces angiogenesis in the ovary, but if it is expressed in the cornea, there is no effect. Transgenic mice in which the rhodopsin promoter drives expression of EG-VEGF in the retina develop sprouting of neovascularization from choroidal vessels, which are fenestrated, but not from retinal vessels. Prokineticin 2 (BV8) is highly homologous to EG-VEGF. It is selectively expressed in testis, bone marrow, and circulating leukocytes. It can induce hematopoiesis and it is a potent chemoattractant for monocytes.

Downloaded from iovs.arvojournals.org on 09/30/2021 3

VEGF Gradients

Supplement 2 to Campochiaro PA, First ARVO/Pfizer Institute Working Group. Ocular versus Extraocular Neovascularization: Mirror Images or Vague Resemblances. Invest Ophthalmol Vis Sci. 2006;47:462-474.

David Shima

Eyetech Pharmaceuticals, Woburn, MA

VEGF is central to normal angiogenesis and pathological angiogenesis. What goes wrong in pathological angiogenesis? One hypothesis is that VEGF isoforms contribute to the multifunctional nature of VEGF. The heparin-binding domain of VEGF has many types of activities. Evidence for this is derived from mice that express only VEGF120, which have no heparin-binding VEGF. They have normal organ development until late in gestation and then develop problems related to vascular insufficiency due to abnormal density of blood vessels. There is a deficiency in branching of blood vessels. Microvessels are about 60% larger in diameter than normal. The number of dividing vascular cells is the same as in wild type mice, but they are organized differently. Larger vessels have less complex branching than normal. There are stunted blood vessel sprouts that are bulbous and lack filopodia. There is proliferation at ends of vessels, which is normally not a characteristic of tip cells at the ends of vessels. This increase in proliferation may give rise to larger vessels. We have hypothesized that heparin- binding VEGFs are needed to set up VEGF gradients, which promote pathfinding for vessels and branching morphogenesis. In contrast to mice with only VEGF120, mice that express only VEGF188 have ectopic branching, but mice that express only VEGF164 are normal. Mice that express both VEGF120 and VEGF188, but not VEGF164 are also normal. Both soluble and matrix-binding VEGFs are needed, which are provided by VEGF164 alone or the combination of VEGF120 and VEGF188. There is more to the heparin-binding region story than just interaction with extracellular matrix. Heparin-binding domain also mediates neural cell migration via neuropilins. VEGF120/120 mice have another phenotype, abnormal assembly of the facial motor nucleus. There is a delay in assembly and change in morphology. A similar phenotype is present in neuropilin1 knockout mice. Implantation of a soaked bead also results in an abnormal facial motor nucleus. This suggests that VEGF164 is involved in motor neuron migration, which in fact is isoform-specific. In contrast, neuroprotection is not isoform-specific. In a model of ischemia-induced retinal cell death, retinal neurons can be rescued by either VEGF120 or VEGF164. What about the heparin-binding domain in the context of disease? In xenograft tumor models, tumor growth is best when VEGF164 or a combination of VEGF120 and VEGF188 are available. In some way the heparin-binding domain provides growth advantage. Mice depleted of leukocytes show less ischemia-induced retinal neovascularization than wild type mice. In ischemic retina, VEGF164 is preferentially upregulated and mice lacking VEGF164 do not develop ischemia- induced neovascularization that breaks through the internal limiting membrane. Mice lacking VEGF164 also develop less leukostasis.

Downloaded from iovs.arvojournals.org on 09/30/2021 4

Alanine scanning mutation analysis was performed in the heparin-binding domain of VEGF. There are three amino acids required for heparin binding and the same three amino acids are also required for leukostasis. VEGF mutated in these three amino acids bind to VEGFR2, but have diminished interaction with VEGFR1. Leukostasis is mediated through VEGFR1. So it appears that the heparin-binding domain helps to mediate inflammation and amplify the pathologic effects of VEGF164. Summary of functions of the heparin-binding domain of VEGF: 1) Critical for setting up gradients needed for branching morphogenesis of vessels during development. 2) Acts through neuropilin 1 for proper motor neuron nucleus assembly. 3) Amplifies inflammation in disease.

Downloaded from iovs.arvojournals.org on 09/30/2021 5

Lymphangiogenesis

Supplement 3 to Campochiaro PA, First ARVO/Pfizer Institute Working Group. Ocular versus Extraocular Neovascularization: Mirror Images or Vague Resemblances. Invest Ophthalmol Vis Sci. 2006;47:462-474.

Kari Alitalo

Biomedicum Helsinki, Helsinki, Finland

Lymphatic capillaries differ from blood vessel capillaries in that they lack smooth muscle cells and pericytes and have very poorly developed basement membrane. Many types of leukocytes, particularly dendritic cells, migrate through lymphatics. Lymphangiogenesis is controlled by VEGF-C and VEGF-D binding to VEGFR3. VEGF-C and -D must be activated by proteolytic cleavage. The first cleavage activates for binding to VEGFR3, the second cleavage in amino terminal domain activates for binding to VEGFR2. So under conditions of extensive proteolytic activity, VEGF-C and -D are angiogenic. VEGFR3 is initially expressed in all blood vessels. Deletion of the Vegfr3 gene is embryonic lethal, because there is failure to reorganize the primary capillary plexus into a hierarchy of large and small vessels. This is an important blood vascular function in the early embryo, not a lymphatic function. The heterozygote embryos survive and the lymphatics develop by sprouting from embryonic veins. The first lymphatic endothelial cells are marked by the homeobox transcription factor PROX1. During this time, the VEGFR3 levels in blood vessels decrease as they increase in lymphatic vessels. After this time, VEGFR3 provides good marker for lymphatic endothelial cells. When lymphatic endothelial cells are put in culture, they maintain expression of PROX1 and VEGFR3, which allows them to be separated from blood vascular endothelial cells. With pure populations of the two types of endothelial cells, it is possible to make comparisons of gene expression. There is about 1-1.5% difference in gene expression between them. Blood vascular endothelial cells express more proteolytic enzymes (urokinase, MMPs) and STAT6. Lymphatic endothelial cells express more TIMP3, protease inhibitors, lymphatic chemokine, PROX1, and Net transcription factor, and they fail to express STAT6. Vegfc null embryos start to swell at mid-gestation. Without VEGF-C, there is no sprouting of lymphatics. Heterozygotes have a phenotype in which blood vessels are normal, but lymphatics are delayed in development and remain severely hypoplastic in the skin. So there is haploinsufficiency just as is the case for VEGF-A. The FoxC2 transcription factor is mutated in some patients with lymphedema and swelling of extremities. FoxC2 is controlled by the VEGFR3 signal transduction pathway, and it blocks basement membrane production and expression of PDGF-B by lymphatic endothelial cells. The absence of PDGF-B expression prevents recruitment of smooth muscle cells and pericytes around lymphatic endothelial cells. The absence of perivascular cells and basement membrane is important to the normal functioning of lymphatics, because their presence impedes fluid from entering lymphatics. In FoxC2 null mice, the lymphatics produce basement membrane and release PDGF- BB, which causes smooth muscles cells to surround the lymphatics; these alterations impede fluid entry into the lymphatics and results in lymphedema. In veins, the regions of valves show expression of FoxC2, which prevents smooth muscle cell recruitment in the areas of valves. Patients with

Downloaded from iovs.arvojournals.org on 09/30/2021 6

FoxC2 mutations have smooth muscle cells in the regions of blood vessel valves, which disrupt functioning of the valves and lead to varicose veins. If adenoviral vector-mediated gene transfer is used to overexpress VEGF-C, there is excessive sprouting of lymphatics resulting in many new lymphatics. This could be a treatment for lymphedema, and a phase I trial has been started to test this possibility. Expression of 1 also causes lymphatic sprouting. can stimulate both Tie1 and Tie2 in both blood and lymphatic endothelial cells. What happens when you shut off this pathway after lymphatics have already developed? This is achieved with VEGF-C/D trap, an Fab fragment that binds VEGF-C and VEGF-D. If the keratin 14 promoter is used to drive expression of the VEGF-C/D trap, it begins to appear in skin at embryonic day (E) 15. This results in regression of lymphatics due to apoptosis of lymphatic endothelial cells, and blood vessels are normal. VEGFR3 is upregulated in some pathologies, such as Kaposi sarcoma (spindle cells are high in VEGFR3) and breast carcinoma, resulting in necklace vessels. There is not much lymphangiogenesis in most tumors, but when VEGF-C is expressed, sprouting and dilation of lymphatics occurs. Mice in which the insulin promoter drives expression of VEGF-C get lymphatics in the pancreas. If these mice are crossed with RipTag mice, the double transgenics show extensive metastasis, whereas RipTag mice do not develop metastasis. The double transgenic mice can be used to test treatments for metastasis, and two things that suppress metastasis are VEGF-C/D trap and blocking monoclonal to VEGF-C. Cell lines that are selected for metastasis have upregulation of VEGF-C. Xenografts with these cell lines get extensive lymphatic vessels around tumors and metastasis. If they are treated with AdVEGFC/D-trap early, the development of excessive lymphatics and the metastasis are blocked. But if treatment is started after day 25, metastasis is not prevented. When tumors grow and enlarge, they sometimes begin to produce VEGF-C or -D, and that causes sprouting and hyperplasia of lymphatics from tumor vessels and promotes metastasis.

Questions

1. Is Ang1 a ligand for Tie1? Ang1 does not bind to Tie1 extracellular domain in solution. But if you overexpress Tie1 on cell surface and stimulate with Ang1 (recombinant form called CompAng1), you can stimulate tyrosine kinase activity of Tie1. So there is a missing component in the complex. Tie2 increases the activation of Tie1, so they form a complex. By using kinase negative mutants of the receptors, you can see that Tie2 participates in Tie1 signaling. In normal endos, Ang1 (or Ang4) stimulates Tie1 and in transfected cells that do not have Tie2. Ang2 does not give activity for Tie1 phosphorylation, but rather inhibits the phosphorylation stimulated by Ang1.

2. Does lymphangiogenesis come from lymphatics or veins? It comes from lymphatics- you don’t see it coming from veins and don’t see contribution by bone marrow cells. If you block VEGF-C you block about 2/3 of metastasis, but there is always a little left and this could be due to active migration of tumor cells into lymphatics.

3. Are there lymphatic vessels in the eye? You can provoke lymphangiogenesis in the cornea by induction of VEGF-D, so it is secondary.

Downloaded from iovs.arvojournals.org on 09/30/2021 7

Neuropilins and Semaphorins

Supplement 4 to Campochiaro PA, First ARVO/Pfizer Institute Working Group. Ocular versus Extraocular Neovascularization: Mirror Images or Vague Resemblances. Invest Ophthalmol Vis Sci. 2006;47:462-474.

Gera Neufeld

Israel Institute of Technology, Haifa, Israel

Our investigations started with the question, why are there so many VEGF isoforms? One possibility that was considered was that some isoforms may interact with receptors other than VEGFR1 and VEGFR2, the only VEGF receptors known at that time. Cross-linking studies using labeled VEGF165 and human umbilical vein endothelial cells (HUVEC) revealed two cross- linking products that were missing in cross-linking experiments in which labeled VEGF121 was used. The band unique to VEGF165 was isolated and was found to be the product of the neuropilin-1 (npn1) gene. Npn1 was already known to be a receptor for class-3 semaphorins (semas) such as sema- 3A, which function as axon guidance factors during the development of the central nervous system. Subsequent studies showed that another family member, npn2, also functions as a VEGF receptor that binds VEGF165 and VEGF145, but not VEGF121. Npn2 also binds placental growth factor (PlGF2) and VEGF-C. In early development, npn1 is expressed in arteries and npn2 is expressed in veins. The expression pattern of npn1 is similar to that of ephrinB2, but starts earlier. Class 3 semas interact differentially with different npns. Sema-3A binds npn1, but not npn2, and sema-3F binds to npn2 with high affinity and to npn1 with low affinity. Npns do not have tyrosine kinase domains, but when they bind semas, they form a complex with plexins, which undergo phosphorylation on tyrosine residues located in their intracellular domains and initiate intracellular signaling. Sema stimulation usually leads to repulsion of axonal growth cones expressing the appropriate semaphorin receptors. Stimulation of HUVEC with sema-3F results in a small contractile response. However, sema-3F inhibits VEGF-induced proliferation of HUVEC. This effect is not due to competition for shared receptors and appears to be an active signal that counteracts the effects of VEGF, because VEGF and sema-3F do not compete for binding to npn2, and because sema-3F also inhibits FGF2-induced proliferation of HUVEC. Sema-3F also inhibited ERK activation by both VEGF and FGF. The inhibitory effect of sema-3F is blocked by siRNA directed against npn2, but not siRNA directed against npn1. These observations suggested that sema-3F may function as an inhibitor of angiogenesis. Indeed, sema-3F was able to inhibit VEGF and FGF2-induced angiogenesis in-vivo in alginate beads and matrigel plug assays. The efficiency of the inhibition depended on the ratio between the angiogenesis inducer and sema-3F, since in the presence of high VEGF concentrations the sema-3F induced inhibition was ineffective. Tumor formation from HEK-293 tumor cells over- expressing sema-3F was strongly inhibited in xenograft tumor formation assays performed in immune-deficient mice as compared to tumor formation from empty vector transfected HEK-293 cells. Furthermore, tumors that did develop from the sema-3F expressing cells contained low

Downloaded from iovs.arvojournals.org on 09/30/2021 8

concentrations of blood vessels and developed much more slowly. However, the efficiency varies between different kinds of tumor cells, and s3f inhibited tumor development from PC3 prostate cancer cells less efficiently, possibly because these cells express higher amounts of pro- angiogenic factors. There are also some stimulatory semaphorins. Sema-4D is membrane bound and released by proteolysis. It binds to its receptor, plexinB1, which then forms complexes with c- met, receptor for HGF. Sema-6D binds to plexinA1, which can form complexes with VEGFR2 and stimulate it.

Questions

1. What is likely to happen if you block Npn1 to try and treat tumor angiogenesis? It is expected that there should be inhibition.

2. How do the Inhibitory effects of semas work? Dependent upon Npn1 and plexins – it is not known if they form complexes with VEGF receptors.

Downloaded from iovs.arvojournals.org on 09/30/2021 9

Tissue-Specific Expression or Knockdown of VEGF

Supplement 5 to Campochiaro PA, First ARVO/Pfizer Institute Working Group. Ocular versus Extraocular Neovascularization: Mirror Images or Vague Resemblances. Invest Ophthalmol Vis Sci. 2006;47:462-474.

Eli Keshet

Hadassah Hebrew University, Jerusalem, Israel

It has been suggested by Shahin Raffi that injured tissue releases VEGF, which recruits progenitor cells from bone marrow that differentiate into endothelial cells and other cells that are not incorporated into new vessels but may stimulate their growth. Transgenic mice with inducible promoters coupled to the gene for Vegf can be used to test this hypothesis. The tetracycline-inducible promoter (tet-off) system combined with tissue-specific promoters allows for expression of VEGF in a particular tissue by omitting doxycycline (dox) from the drinking water of the mice and for its reversal at will. Expression of a VEGF trap (the extracellular domain of VEGF receptor 1 coupled to an Fc fragment) instead of VEGF allows for tissue- specific knockout of VEGF. Using the tet-off system and a heart-specific promoter, VEGF expression was initiated in the hearts of adult mice by omitting dox. This caused an influx of cells into the heart when dox was added. When the production of VEGF was halted, the cells dissipated. These cells were demonstrated to be hematopoietic cells, because they stained for the pan-hematopoietic marker CD45. They surrounded blood vessels within the heart tissue. In some experiments, mice were given whole body irradiation to eliminate all endogenous bone marrow cells, and then the bone marrow was reconstituted by transplanting with cells labeled with LacZ or GFP. Mice were then treated with dox to turn on VEGF in heart, and this caused an accumulation of labeled cells around blood vessels in the heart. The cells were not incorporated into the endothelium of blood vessels, but rather remained in a perivascular location. What is the nature of these cells? Using cell-type specific stains, they were found to be monocytes. The CD45 positive cells recruited to the organ are also positive for CXCR4 and can respond to stromal derived factor-1 (SDF-1). When VEGF is turned on, SDF-1 expression is induced around the blood vessels, which explains the accumulation of monocytes around blood vessels. If AMD3100, a CXCR4 inhibitor, is given by osmotic minipump, the perivascular cells disappear. These cells are necessary for the angiogenic response. When the cell recruitment is blocked with AMD3100, then the angiogenic response is markedly attenuated. This is due in part to production of proangiogenic agents by the monocytes, because they can be isolated and grown in culture, and their conditioned medium stimulates angiogenesis in the aortic ring assay. So it appears that when expression of VEGF is increased in a tissue, it causes altered expression of genes in local tissues. For instance, SDF-1 expression is induced around blood vessels, which promotes recruitment of cells in a perivascular location, which enables their paracrine mode of action. The recruitment of cells results in importation of pro-angiogenic activities that contribute to proliferation of the endothelial cells.

Downloaded from iovs.arvojournals.org on 09/30/2021 10

Conditional knockdown of VEGF was achieved in the heart using the tet-off system with a heart specific promoter to provide inducible expression of VEGF trap in the heart. When the VEGF trap was activated soon after birth, they developed “myocardial hibernation,” a sparing response to reduced perfusion in which the cardiac myocytes cease contracting and thereby save energy. In the neonatal period, the heart is still growing, and as it grows, vascular density increases. The expression of VEGF trap during this period of growth prevents the increase in vascular density, and therefore the heart becomes ischemic and this leads to reduced contractility. In the hibernation state, the heart remains viable and can be rescued by increasing perfusion. This is achieved by re-administration of dox and switching off VEGF trap expression, which results in recovery of normal microvascular density. This shows that VEGF is responsible for adjusting the microvascular density to hypoxia.

Downloaded from iovs.arvojournals.org on 09/30/2021 11

Platelet-Derived Growth Factors (PDGFs) and Perivascular Cells

Supplement 6 to Campochiaro PA, First ARVO/Pfizer Institute Working Group. Ocular versus Extraocular Neovascularization: Mirror Images or Vague Resemblances. Invest Ophthalmol Vis Sci. 2006;47:462-474.

Mats Hellstrom

AngioGenetics Sweden AB, Gotenborg, Sweden

The characterization of pdgfB knockout mice demonstrated that PDGF-BB plays a critical role in recruitment of perivascular cells by endothelial cells of blood vessels. The knockout mice die during embryonic development with severe vascular defects and almost complete lack of pericytes surrounding vessels. Blood vessels are dilated, variable in size, and there are many microaneurysms. Undifferentiated mesenchymal cells are induced to differentiate into pericytes during early embryonic life, and then PDGF-BB produced by vascular endothelial cells promotes proliferation and migration of pericytes and smooth muscle cells (SMCs) so that they surround blood vessels. In order to assess the role of PDGF-BB in adult animals, conditional pdgfB knockouts were generated by engineering loxP sites around the pdgfB gene and using a Tie1 promoter to express cre recombinase in endothelial cells. The efficiency of recombination was assessed by quantitative RT-PCR for pdgfB mRNA on isolated capillary fragments. The recombination rate was found to vary, and in some mice was fairly low, while in others was as high as 90%. Even mice with very low levels of pdgfB mRNA in capillary fragments survived into adulthood. Mice with the conditional pdgfB allele were crossed with a LacZ reporter strain (XlacZ4) in which the pericytes are labeled. In control mice, there was heavy staining in arteries and veins, while in mice with low pdfgB mRNA in capillary fragments, there was marked reduction in coverage of arteries and veins with SMC. Some of these mice developed retinopathy in which some part of the retina were normal, some areas showed dilated vessels and capillary dropout, and some regions showed retinal neovascularization breaking through the internal limiting membrane into the vitreous cavity. Areas of retina showing proliferative changes in the inner retina often showed evidence of traction and rosette formation in photoreceptors and NV growing into the vitreous. This phenotype is similar to that seen in patients with proliferative diabetic retinopathy. The severity of retinopathy correlated with the amount of pericyte loss. The threshold for development of retinopathy appeared to be reduction of pericytes coverage by about 50% or less. Angiogenic sprouts contain specialized endothelial cells called tip cells. They are migratory, nonproliferative, and have numerous long filopodia and a distinct pattern of gene expression. Adjacent to tip cells are stalk cells, which proliferate and form the lumen of the vessel. VEGF is the cue that guides tip cells to migrate into the tissue. The γ-secretase protease complex is best known in relation to APP in Alzheimer’s disease, but it is also essential for notch signaling. When postnatal day (P3)-P5 mice were treated with a γ-secretase inhibitor (DAPT), there was a distinct increase in microvascular density in the peripheral retina. Mice treated at P2-P6 showed numerous filopodia extensions and sprouting from the vascular network giving the appearance of branching and fusion of

Downloaded from iovs.arvojournals.org on 09/30/2021 12

vessels. There was a doubling of the number of tip cells per unit area. This suggests that γ- secretase inhibits the development of tip cells. To test whether the effect of DAPT was due to a direct effect on endothelial cells, its effect was tested on HUVECs cultured in collagen gels. When HUVECs are placed in collagen gels, they start to sprout within 24h and the sprouts can be quantified. The number of sprouts are increased by addition of VEGF or DAPT, and the effects of VEGF and DAPT appeared additive. In DAPT-treated neonatal mice, pericytes recruitment appeared normal, although in regions of very high vessel density, interactions between pericytes and endothelial cells appeared disturbed. The astrocyte networks in front of the developing vascular network didn’t seem to be affected. VEGF-A protein levels are not altered. In the mouse model of oxygen-induced ischemic retinopathy, new vessel tufts sprout from veins in regions bordering avascular zones. Treatment with DAPT resulted in a reduction in the size of avascular zones, doubling vascular density and reducing vascular tuft formation by one-third of that in controls. In summary, DAPT treatment promotes tip cell formation, endothelial cell proliferation, and branching. A hypothesis would be that a signal from the tip cell to the stalk cell, possibly a Delta4/notch signal, could be important for notch-mediated lateral inhibition of the stalk cells. The combination of the presented and published data supports this hypothesis. (1) γ-secretase is essential for notch signaling, (2) γ-secretase inhibitors can promote endothelial cell proliferation by notch inhibition, (3) the tip cells express Delta4, (4) cell autonomous notch signaling regulates endothelial cell branching during tubule formation, and the number of tip cells are increased when notch is inhibited. The VEGF gradient is important for maintaining a tip cell phenotype, and it is known that VEGF signaling can trigger Delta4 expression via notches 1 and 4. Notch signaling can suppress VEGFR2 signaling. In tip cells there is high Delta 4 and VEGFR2 expression, while in stalk cells there is no Delta4-like expression and low VEGFR2 expression. This suggests that lateral inhibition could take place. Similar cells are exposed to VEGF, but a future tip cell is closer to the VEGF source, which would cause it by the mechanism described above to suppress the tip cell phenotype in the neighboring cell so that it becomes a stalk cell. Both VEGF-A and Delta4-like knockouts are haploinsufficient; they are extremely dose sensitive. In Drosophila trachea development, FGF2 induced tip cell and notch signaling is important for differentiation of stalk cells. VEGFR2 signaling is important for chemotaxis. There are important repulsive signals mediated by netrin and semaphorins, and as suggested above, there may also be notch-mediated lateral inhibition from tip cells to stalk cells.

Downloaded from iovs.arvojournals.org on 09/30/2021 13

Macrophages, Wnts, and Programmed Vascular Regression

Supplement 7 to Campochiaro PA, First ARVO/Pfizer Institute Working Group. Ocular versus Extraocular Neovascularization: Mirror Images or Vague Resemblances. Invest Ophthalmol Vis Sci. 2006;47:462-474.

Richard Lang

Children Hospital Research Foundation, Cincinnati, OH

Programmed vascular regression occurs in three related vascular beds in the newborn rodent eye, the pupillary membrane, the tunica vasculosa lentis, and the hyaloid vessels. Understanding the signals involved may provide new insights as to how vascular regression occurs and could lead to new treatments that allow inducing regression of unwanted vessels. In mice, the hyaloid vessels regress between 3 and 9 days. Study of the hyaloid system is facilitated by the ability to visualize the vessels in vivo and the ability to analyze the entire vasculature en block in dissected flat mounts. There are a small number of resident macrophages in the pupillary membrane – 300 to 400 macrophages for a vessel network that contains 6000 to 7000 vascular endothelial cells. It is well-established that macrophages play a passive role in recognizing and engulfing apoptotic cells. They will simply engulf cells after there has been a macrophage-independent signal for cell death. But there is also evidence suggesting that they may also play an active role inducing apoptosis. Some of that evidence was obtained by observations in PU.1 mutant mice. PU.1 is a transcription factor that plays a role in hematopoiesis. One of the consequences of deleting the PU.1 gene is absence of mature tissue macrophages. The mutant mice have persistence of the hyaloid vessels, suggesting that macrophages may be important in hyaloid vessel regression. The canonical wnt pathway is also required for hyaloid vessel regression. The receptor complex for wnts consists of frizzled proteins and LRP5 or LRP6 co-receptors. After wnt ligands bind to the receptor complex, signal transduction events result in stabilization of β-catenin, which then complexes with transcription factors of the Lef-tcf class to regulate gene expression. Wnt signaling has important roles in development and is activated in tumorigenesis. It is also important for hyaloid vessel regression, because mice with a mutation in Lrp5, show no hyaloid vessel regression due to failure of cell death pathways. Mutations in Lef1 give a similar phenotype. These data suggest that the canonical wnt pathway is required for hyaloid regression. The Topgal mouse is a wnt pathway reporter mouse generated by Elaine Fuchs. The construct used to generate these mice consists of a minimal promoter with some Lef-tcf binding sites coupled to LacZ, so in these mice blue-staining cells are wnt-responsive. In Topgal mice, hyaloid vessel preparations show sporadically labeled endothelial cells; it is the endothelial cells and not the macrophages that are responding along this pathway. The canonical wnt pathway is very good at stimulating cell-cycle entry, and it might seem counter-intuitive that that might be required for cell death, but in some settings cell-cycle entry and cell death are coupled. In wnt pathway mutants, BrdU-labeled endothelial cells in hyaloid vessels are reduced, indicating that the wnt pathway is stimulating cell-cycle entry. In the pupillary membrane, endothelial cells die in a cell-cycle stage-dependent way that is also

Downloaded from iovs.arvojournals.org on 09/30/2021 14

macrophage-dependent. There is a particular point in G1 phase of the cell cycle, the restriction point, in which cells are monitoring their environment to determine if they have appropriate survival signals that can come from either growth factors or the extracellular matrix. If fibroblasts sense they lack survival signals, they will go into cell cycle arrest, but it is very difficult to generate cell cycle arrest in vascular endothelial cells, probably because they are programmed to die if those signals are missing. In hyaloid vessels, it was found that cell cycle entry and cell death were coupled. BrdU was injected at time 0 as a fate mapping marker and cells that enter S phase are identified; they are synchronous with respect to stages of the cell cycle. TUNEL staining showed a peak of BrdU-labeled, TUNEL-positive cells at about 17 hours. This indicates that cell death is dependent upon cell cycle stage. The staining is observed in whole mounts; you can identify essentially all cells that have divided and many are double- labeled. By definition they are in G1, so there is a strong correlation between death and G1. These findings could be explained if there is a signal that causes withdrawal of survival signals in G1 of cell cycle. A reasonable candidate is angiopoietin 2 (Ang2). Ang1 is an agonist for Tie2 and Ang2 can be context dependent, but in the context of the ocular vessels, it is antagonistic for Tie2. Some of the molecules that are downstream of Tie2 are PI3kinase and Akt and when Ang1 is active, these are also active. Akt is important in stimulating cell survival through the suppression of proapoptotic mediators such as Bad and FoxO transcription factor. Ang2 blocks Akt and therefore Bad and FoxO can stimulate apoptosis. Ang2 null mice have persistent hyaloid vessels. There is also a ras/raf MAP kinase component to Tie2 signaling. Is Akt sufficient for producing the Ang2 null phenotype? Using a Tet regulation system in which tet-off is used in combination with a VE-cadherin promoter, when tetracycline is withdrawn, expression is activated in endothelial cells. If this system is used to express a myristoylated form of Akt in neonatal mice, persistence of hyaloid vessels results, the same phenotype as that of Ang2 null mice. So, we can conclude that the function of the canonical wnt pathway is to stimulate cell-cycle entry and that once a cell is in G1, it is sensitive to the withdrawal of survival signals mediated by Ang2. The end result is cell death. Why are macrophages necessary? It could be that they produce Ang2, or a wnt ligand, or both. There is evidence against macrophage production of Ang2; it appears that the pericytes produce Ang2. The pericyte marker desmin has a similar distribution to LacZ in heterozygous mice in which the Ang2 promoter drives LacZ. Are macrophages producing wnt ligands? Laser capture microdissection was used to obtain macrophages from hyaloid vessel whole mounts. Using RT-PCR, it was found that these macrophages express mRNA for Wnt7B. In heterozygous Wnt7BLacZ knockin mice, hyaloid vessel preparations at P1-P5 when hyaloid vessel regression is starting show LacZ staining in macrophages. Lack of all Wnt7B activity as seen in homozygous Wnt7BLacZ knockin mice is embryonic lethal, but mice with the hypomorphic allele Wnt7BD1 have enough Wnt7B to survive long enough to assess hyaloid vessel regression. In these mice, there is persistence of the hyaloid vessels despite continued presence of macrophages, so macrophage production of wnt7B is necessary for regression. Injection of wild type macrophages into PU.1 mutant mice that lack mature macrophages rescues their phenotype; regression of the hyaloid vessels occurs. However, the phenotype is not rescued if macrophages from Wnt7BD1/D1 are injected. It appears that macrophages may be delivering Wnt7B locally and these ligands work over short range because they are palmitoylated. These data suggest the following model for regression of the hyaloid vessels. Macrophages deliver Wnt7B locally to vascular endothelial cells by cell-cell contact. Wnt7B

Downloaded from iovs.arvojournals.org on 09/30/2021 15

stimulates the endothelial cells to enter the cell cycle. Expression of Ang2 causes withdrawal of survival signals and without those, signals, endothelial cells that enter G1 undergo programmed cell death.

Questions

1. What tells the macrophages to release Wnt7B? It is not known. The macrophages are there from early on, so something must cause them to turn on wnt7B soon after birth.

2. Has blocking Fas/Fas ligand interaction been done to determine the effect on vascular regression? This hasn’t been tried in concerted fashion, but it would be interesting.

3. In situations in which there is poor development of retinal vessels, there is almost always persistence of hyaloid vessels. One possible mechanism is that retinal ischemia causes increased levels of VEGF, which promotes survival of hyaloid vessels and prevents their regression. Is there any evidence for this mechanism? There may be more than one mechanism that is operative. In PU.1 mutant mice, there is normal development of the retinal vessels, but persistence of hyaloid vessels. In other situations, lack of development of the retinal vessels may increase VEGF, and this could play a role in persistence. And it is possible that the two mechanisms may work together in some instances.

4. In many settings, macrophages stimulate angiogenesis. What is different here that causes the macrophages to promote apoptosis and regression? It may be that there is some global change in hyaloid vessels that causes macrophages to recognize cells destined to die and then upregulate Wnt signaling.

5. Heterozygous Bmp4 knockouts have lack of regression of hyaloid vessels and lack of infiltration of macrophages. BMP4 secreted from the lens may influence pupillary membrane regression. If you inhibit BMPs by injecting BMP inhibitors into the eye, you can prevent pupillary membrane regression.

6. Can VEGF turn off Wnt7B expression, and can IL-1 or TNF can promote it? Mice missing both TNF receptors have normal hyaloid vessel regression suggesting that if the TNF pathway has any role, it is limited. It is not known if VEGF can regulate Wnt7B expression.

Downloaded from iovs.arvojournals.org on 09/30/2021 16

Norrin and Fz4: A Ligand-Receptor Pair That Controls Capillary Growth in the Developing Mammalian Retina

Supplement 8 to Campochiaro PA, First ARVO/Pfizer Institute Working Group. Ocular versus Extraocular Neovascularization: Mirror Images or Vague Resemblances. Invest Ophthalmol Vis Sci. 2006;47:462-474.

Jeremy Nathans

Johns Hopkins University, Baltimore, MD

I will start this talk by describing several apparently unrelated threads – the biology of Wnt proteins, the discovery of a family of orphan receptors called Frizzleds, and, later on, an orphan ligand called Norrin, several human retinal vascular diseases, and a puzzling knockout mouse phenotype – but by the end I will tie all of these threads together. Twenty years ago, the principal site of chromosomal integration of the mouse mammary tumor virus was found to reside next to a gene, Int1, which encodes a secreted protein of 45KD, now called Wnt1. When Wnt1 is expressed by cells, there is a localized effect, in that only responsive cells immediately adjacent to the expressing cells proliferate. This suggests that Wnt1 stimulates proliferation, but only very locally. The Wingless gene in Drosophila is involved in many aspects of development and was found to be a homologue of Int1. In the Drosophila embryo, Wingless interacts with Engrailed to produce the segmented cuticle polarity so that there are stripes of denticles and naked cuticle. In 1982, Gubb and Garcia-Bellido described mutations in the Drosophila “Frizzled” gene, in which the hairs and bristles are disorganized – a defect referred to as a planar cell polarity or PCP phenotype. Actually, it looks like the flies are having a bad bristle day. There are at least six other genes that have the same phenotype. Based on their analysis of PCP genes, Gubb and Garcia-Bellido suggested that there might be some type of Cartesian coordinate system in all animals that orients each local structure to the big game plan - the head is this way and the feet are that way - and that this system is somehow disrupted in Frizzled mutants. The first Frizzled receptor gene was cloned by Paul Adler. It has the structure of a G- protein receptor with 7 putative transmembrane domains and a single extracellular cysteine-rich domain. Based on this structure, Adler suggested that it might be a receptor of spatial information. In our laboratory, high throughput sequencing in the early 1990s identified homologues of Frizzled in a human retinal cDNA library. A LacZ knock-in of Frizzled-5 shows strong expression of in the retina; we now know that it functions in retinal development. In Drosophila, Frizzled-2, a second Frizzled that we identified in the mid-1990s, is expressed in stripes in the embryo. This observation suggested to us that Wingless might interact with Frizzled-2 to produce this striped pattern, and that, more generally, Frizzled family members were the receptors for Wnts. In a collaboration with the laboratory of Roel Nusse at Stanford, we showed that this hypothesis is correct: Wnts are the ligands for Frizzled receptors. There are 19 Wnts in mammals and extensive promiscuity in binding between Wnt ligands and Frizzled receptors. There is also emerging evidence from our laboratory and the laboratory of Kevin Struhl for genetic redundancy among different Frizzleds, and from Andy

Downloaded from iovs.arvojournals.org on 09/30/2021 17

McMahon and others, among Wnts. In Drosophila, mutation of either Frizzled or Frizzled-2 does not result in an embryonic phenotype, but double mutants have the same severe patterning defects as a Wingless mutant, and can be rescued by a transgene for either Frizzled receptor. Our recent work has focused largely on the role of Frizzled receptors in mammals. In mice, Frizzled-3 is critical for axonal development. It is expressed in the developing brain and spinal cord from very early on and its deletion results in severe defects: there are large ventricles and a thinner cortex, not because there are fewer cells, but rather because of the absence of many axons. The phenotype suggests that there is lack of directional cues, and if axons don’t know where to go, they don’t grow. The mouse Frizzled-4 gene is expressed in the brain, inner ear, and retina. Frizzled-4 knockouts have a phenotype that is quite specific, but with rather diverse effects, suggesting that Frizzled-4 does many different things. The knockouts have: (1) a progressive cerebellar degeneration, (2) an enlarged esophagus that is not surrounded by muscle, (3) progressive deafness due to degeneration of the stria vascularis, the capillaries that provide the fluid that fills the central chamber of the organ of Corti, and (4) bleeding in the retina. The growth of the major retinal vessels is slow and incomplete, but the most dramatic defect is a complete absence of intra-retinal capillaries. There are two human diseases in which the phenotype is similar to that seen in Frizzled-4 knockouts. One of these, Norrie disease, is an X-linked disease that presents with blindness in infancy due to incomplete retinal vascular development, severe neovascularization, and retinal detachment. The Norrie disease gene was identified over 12 years ago by positional cloning; it encodes a small secreted protein that has a signal sequence and a cysteine knot motif similar to that seen in TGF-β family members. A Norrin knockout mouse looks exactly like the Frizzled-4 knockout. The intra-retinal capillaries do not develop resulting in nonperfusion. Based on this series of observations, we hypothesized that even though Norrin has no primary sequence homology to Wnts, maybe it is a ligand for Frizzled-4 - and that turned out to be the case. Norrin binds directly to Frizzled-4 with nanomolar affinity and with high specificity in that it doesn’t bind other Frizzled receptors. Moreover, Norrin activates the “classical” Wnt signaling cascade when both Frizzled-4 and a coreceptor (Lrp5 or Lrp6) are present on target cells. Interestingly, Norrin acts locally because it has a high affinity for the extracellular matrix. Just like Wnts, it is a paracrine activator. There are many different mutations in Norrin that result in Norrie disease. We have expressed many of the mutant proteins and observed that they have a wide variation in bioactivity. It will be interesting to see in what way this correlates with clinical phenotype. A second disease with a phenotype like that seen in the Frizzled-4 knockout mouse is familial exudative vitreoretinopathy (FEVR). This is an inherited disease in which there is incomplete development of retinal vessels resulting in areas of retinal ischemia, which leads to retinal neovascularization, scarring and retinal detachments. Recently we and others identified families with FEVR who have inherited mutations in Frizzled-4. A second FEVR locus has been identified as the LRP5 co-receptor for Wnt signaling. In the families that we have studied, the Frizzled-4 mutations (both are single amino acid substitutions) lie very close together in the cysteine rich domain. They both have a rather specific functional defect: there is nothing wrong with Frizzled-4 protein biosynthesis, stability, cell surface localization, or Norrin binding, but the mutant Frizzled-4 receptors are defective in Norrin-induced signaling. In summary, we have identified Norrin and Frizzled-4 as a ligand-receptor system that is devoted to building the retinal vascular system. This begs the following questions. Do other

Downloaded from iovs.arvojournals.org on 09/30/2021 18

vascular beds have their own private signaling systems? Are there other private ligands for Frizzled receptors? The interaction of both Norrin and Wnts with Frizzled-4 is conceptually analogous to the interaction of both VEGF and semaphorins with neuropilins, and suggests that there may be other examples of structurally unrelated ligands interacting with the same receptor. Perhaps there are new players in receptor-ligand systems that were until now thought to be completely understood. A dramatic example of this idea is a recently discovered ligand for the insulin receptor: visfatin, a secreted protein made by visceral fat. Visfatin works in much the same way as insulin, but it binds to another site on the insulin receptor.

Questions

1. Does the Norrin/Frizzled system have any role in adult vascular biology? Not known.

2. Which cells produce Norrin and which produced Frizzled-4 in retina? Not known.

Downloaded from iovs.arvojournals.org on 09/30/2021 19

Integrins and Extracellular Matrix

Supplement 9 to Campochiaro PA, First ARVO/Pfizer Institute Working Group. Ocular versus Extraocular Neovascularization: Mirror Images or Vague Resemblances. Invest Ophthalmol Vis Sci. 2006;47:462-474.

Peter Brooks

New York University, New York, NY

Regulatory signals come from the cellular compartment (tumor cells, stromal cells, inflammatory cells, and vascular cells) and the acellular compartment. The acellular compartment is the extracellular matrix (ECM). There are cryptic sites in ECM that when exposed can participate in the regulation of angiogenesis. Three groups of molecules are involved: , matrix (MMPs), and ECM components such as collagens or laminin. Remodeling of the ECM by MMPs can expose cryptic sites that interact with integrins. To develop reagents that could recognize proteolyzed or denatured forms of collagen, but not the native triple helical form, subtractive immunization was used to generate monoclonal antibodies. One , HUIV26, recognizes a cryptic site in collagen IV and was used to stain normal skin and biopsies from malignant melanoma. Exposure of the HUIV26 cryptic epitope was observed around tumor blood vessels and associated with the invasive fronts of tumors as they penetrated epithelial basement membranes and little or no exposure as seen in normal skin. This was also the case in biopsies from several other tumor types. Systemic injections of HUIV26 inhibited FGF2-induced angiogenesis in chick chorioallantoic membranes (CAMs), and the antibody also blocked cellular adhesion and migration. How does this information get transferred to the cells? HUIV26 was capable of inhibiting purified αvβ3 binding to denatured collagen, but had no effect on α2β1 binding, suggesting that the cryptic site is recognized by αvβ3. This site is a non-RGD site. Does this site also play a role with regard to tumor cells? It was previously demonstrated that expression of αvβ3 by tumor cells gives them a growth advantage. M21 cells that express αvβ3 were compared to a variant of the cells that don’t express αvβ3, and it was found that the cells that express αvβ3 grew better in vivo, but not in vitro. This suggests that something in the in vivo environment is giving them a growth advantage. In a xenograft tumor model, the cells that lacked αvβ3 formed tumors with less vascular density than the tumors formed from cells that expressed αvβ3. Cs1 melanoma cells that lacked αvβ3 were transfected with β3 and put on chick CAM, and blood flow was measured. Tumors induced by cells expressing β3 had greater blood flow. Conditioned medium (CM) from cells that lack αvβ3 inhibited endothelial cell proliferation, inhibited angiogenesis in CAM assay, and inhibited tumor growth. Microarray analysis demonstrated many genes that were upregulated and many that were downregulated in αvβ3-negative cells. -1 (Tsp-1) was upregulated and the CM was found to have high levels of Tsp-1. Members of the insulin-like growth factor binding protein (IGF-BP) family were also upregulated. Knock down of αvβ3 with siRNA in cells that express it resulted in upregulation of Tsp-1.

Downloaded from iovs.arvojournals.org on 09/30/2021 20

Cells were allowed to attach to vitronectin (αvβ3-mediated) or to triple helical collagen (β1-mediated), and when they were attached to vitronectin there was suppression of Tsp-1 expression. What happens if cells are plated on denatured collagen (exposes the cryptic site recognized by αvβ3, but also has β1 sites- so it is a mixed ligand) and then treated with LM609, which binds αvβ3? CM collected after 12 hours demonstrated upregulation of Tsp1 compared to controls. LM609 antibody was coated on a plate to drive a specific signal through αvβ3. Compared to cells plated on an anti-β1 antibody, there was suppression of Tsp1 levels in the CM of cells in which signaling was driven through αvβ3. So ligation or binding of αvβ3 results in suppression of a known , Tsp1. The results were similar for IGFBP4. Knockout of αvβ3 resulted in upregulation of IGFBP4. Treatment of cells plated on denatured collagen with LM609 resulted in upregulation of IGFBP4. M21 melanoma cells were implanted on CAM and the embryo was treated with LM609 or an isotype-matched control antibody. Immunohistochemistry showed upregulation of IGFBP4 in tumor cells from LM609-treated embryos. In another model, ng quantities of IGFBP4 inhibited FGF2-induced angiogenesis. When cells were plated on denatured collagen in the presence of HUIV26 (blocks cryptic site), there was upregulation of Tsp1 and IGFBP4. In endothelial cells, the results were similar. If you treat denatured collagen with HUIV26 and plate HUVECs, there is upregulation of Tsp1 and IGFBP4. We tested the effect of LM609 in a nude mouse model. M21 human melanoma cells were injected into mice and then mice were treated with LM609 or an isotype-matched control antibody. After 7 days, the tumors in mice treated with LM609 were almost avascular. This may be due to LM609 binding to αvβ3 on tumor cells and indirectly inhibiting angiogenesis by increased production of Tsp1 and IGFBP4. The working hypothesis is that αvβ3 is expressed on tumor cells (high in invasive malignant melanoma), and a number of MMPs are being produced by tumors, stroma, and inflammatory cells. The MMPs can degrade ECM and expose cryptic epitopes, and αvβ3 on tumor cells can then recognize and bind to these cryptic epitopes, drive a signal through FAK and ultimately through Ras (Ras activates PI3kinase, Rou DGS, Raf MAPkinase). Activation of the MAP kinase pathway can lead to activation of ERK which phosphorylates and activates cMyc, which acts as a repressor of Tsp1. Alternatively, the PI3kinase pathway is activated, which can activate cMyc and repress Tsp1. So one of these pathways may be involved in the ability of αvβ3 on tumor cells to suppress these endogenous inhibitors of angiogenesis, and if this is blocked by blocking αvβ3 or blocking the cellular access to the cryptic epitopes, the repression of Tsp1 is removed, and thereby Tsp1 is increased and suppresses angiogenesis.

Questions

1. Do you have any data on IGFBP3, because it is commonly considered a proapoptotic agent? IGFBP3 was upregulated at the mRNA level, but in the CM there was not an increase in IGFBP3 protein.

2. The increase in IGFBPs would have an effect on local IGF1 levels. Have you looked at that? Not yet, but it may be one part of how the system is working.

Downloaded from iovs.arvojournals.org on 09/30/2021 21

Abnormalities in Vessel Formation in a Mouse Model of Timp3 Deficiency

Supplement 10 to Campochiaro PA, First ARVO/Pfizer Institute Working Group. Ocular versus Extraocular Neovascularization: Mirror Images or Vague Resemblances. Invest Ophthalmol Vis Sci. 2006;47:462-474.

Bernhard H. F. Weber,1 Andreas Janssen,1 Heinrich Schrewe,2 Ernst Tamm,3 Chr. Albrecht May,4 Mathias Seeliger5

1Institute of Human Genetics, University of Regensburg, Germany; 2Max-Planck Institute for Molecular Genetics, Berlin, Germany; 3Department of Anatomy, University of Regensburg, Germany; 4Institute of Anatomy, University of Dresden, Germany; 5University Eye Clinic, University of Tuebingen, Germany

Sorsby fundus dystrophy (SFD) is an autosomal dominantly inherited degenerative disease of the retina and is characterized by rapid loss of central vision1. One of the early signs is the deposition of material in Bruch’s membrane, a five-layered extracellular matrix (ECM) between the retinal pigment epithelium (RPE) and the choriocapillaris. As a hallmark of this condition, there is a high rate of choroidal neovascularization (CNV), which closely resembles the exudative form of age-related macular degeneration (AMD), a prevalent blinding disorder of multifactorial etiology. SFD is caused by mutations in the tissue inhibitor of metalloproteinases- 3 (TIMP3).2 All TIMP3 mutations known so far are predicted to result in unpaired cysteine residues within the C-terminal portion of the protein which likely cause the formation of high- molecular weight complexes due to disulfate bridge formation. However, it is still unknown how this complex formation leads to retinal degeneration and CNV. TIMP3 is a member of a family of four secreted proteins (TIMP1 to TIMP4). While TIMP 1, 2, and 4 are soluble and diffusible, TIMP3 is unique as it is covalently bound to the ECM. Originally identified as inhibitors of matrix metalloproteinases (MMPs),3 the TIMPs are now recognized as proteins with multiple functions independent of their MMP inhibitory activities. Specifically, TIMP3 has recently been found to be a potent inhibitor of angiogenesis by competitively blocking the binding of VEGF to its receptor VEGFR2.4 In order to better understand the pathogenesis of SFD and hopefully provide new insight into mechanisms involved in CNV formation, our strategy is to generate mouse models carrying distinct types of Timp3 mutations. Using the Cre-loxP system, a knock-in mouse for a Ser156Cys mutation found to cause early signs of SFD in a large Austrian pedigree was established. On a C57BL/6 background there was a mild phenotype which was more pronounced on a CD1 albino background closely resembling early SFD manifestations. A second knock-in mouse model expressing a Timp3 Ser156Met mutation was found to be embryonic lethal as a homozygous trait, possibly emphasizing a role for Timp3 during embryonic development. Finally, Timp3 deficiency appears to cause multiple phenotypes affecting tissues that normally express Timp3. So far, reports include, for example, air space enlargement in the lung,5 dilated cardiomyopathy6 and chronic hepatic inflammation.7 In the eye, Timp3 knock-out mice on both a C57BL/6 and a CD1 albino background reveal a striking dilation of choroidal vessels, although without developing a retinal degeneration phenotype.

Downloaded from iovs.arvojournals.org on 09/30/2021 22

To analyze the underlying pathomechanism of the vessel phenotype in Timp3-deficient mice, we investigated the angiogenic response in an aortic ring angiogenesis assay. In comparison to wild type, the vascular response in Timp3-deficient tissue showed strongly enhanced vessel outgrowth under spontaneous conditions. Addition of recombinant murine Timp3 reduced this outgrowth to a level comparable to wild type. We further show that this response is mediated through the VEGF/VEGFR2 signaling cascade by pretreatment of Timp3- deficient aortic rings with an inhibitor of VEGFR2 (ZM323881) which dramatically reduced vessel outgrowth. Although to a much lesser degree than in knock-out animals, the excessive outgrowth of vessels also occurs in aortic rings from mutant Ser156Cys knock-in mice. This suggests that the Ser156Cys mutation may impair but not abolish proper binding of Timp3 to VEGFR2. Downstream signaling pathways of VEGFR2 such as ERK1/ERK2 and p38 were investigated. Both ERK1/ERK2 and p38 show higher levels of phosphorylated protein in knock- out and Ser156Cys knock-in mice compared to wild type. Stimulation with recombinant VEGF increased the amount of phosphorylated proteins regardless of mutation status. Interestingly, the p125FAK kinase showed increased protein levels in the knock-out but significantly reduced protein expression in the Ser156Cys knock-in. Nakatsu et al.8 reported a distinct function for VEGF in the regulation of blood vessel diameter. In a fibrin gel bead assay we coated mouse heart endothelial cells (MHEC) on beads which were then imbedded into fibrin gels. In this three-dimensional network, endothelial cells form capillary like structures. Treatment with increasing concentrations of recombinant VEGF resulted in increased diameters of newly formed capillaries. Addition of recombinant Timp3 strongly reduced this vessel dilation. This suggests a regulatory role of Timp3 in vessel dilation via mechanisms similar to vessel outgrowth. Together, Timp3 appears to be actively competing with VEGF for binding to the VEGFR2 receptor. Consequently, lack of Timp3 or dysfunctional mutated Timp3 results in increased VEGF/VEGFR2 signaling leading to enhanced MAPkinase activity and thus to endothelial cell proliferation and migration, eventually resulting in vessel dilation and increased vessel diameter.

References 1. Sorsby A, Mason MEJ, Gardner N. A fundus dystrophy with unusual features. Br J Ophthalmol. 1949; 33:67-97 2. Weber BHF, Vogt G, Pruett RC, et al. Mutations in the tissue inhibitor of metalloproteinases-3 (TIMP3) in patients with Sorsby´s fundus dystrophy. Nat Genet. 1994;8:352-356. 3. Murphy G, Willenbrock F, Crabbe T, et al. Regulation of matrix metalloproteinase activity. Ann NY Acad Sci. 1994;732:31-41. 4. Qui JH, Ebrahem Q, Moore N, et al. A novel function for tissue inhibitor of metalloproteinases-3 (TIMP3): inhibition of angiogenesis by blockage of VEGF binding to VEGF receptor-2. Nat Med. 2003;9:407-415. 5. Leco KJ, Waterhouse P, Sanchez OH, et al. Spontaneous air space enlargement in the lungs of mice lacking tissue inhibitor of metalloproteinases-3(TIMP-3). J Clin Invest. 2001;108:817-829. Erratum in: J Clin Invest. 2001;108:1405. 6. Fedak PW, Smookler DS, Kassiri, et al. TIMP-3 deficiency leads to dilated cardiomyopathy. Circulation. 2004;110:2401-2409.

Downloaded from iovs.arvojournals.org on 09/30/2021 23

7. Mohammed FF, Smookler DS, Taylor SE, et al. Abnormal TNF activity in Timp3-/- mice leads to chronic hepatic inflammation and failure of liver regeneration. Nat Genet. 2004;36:969-977. 8. Nakatsu MN, Sainson RCA, Pérez-del-Pulgar S, et al. VEGF121 and VEGF165 regulate blood vessel diameter through VEGFR2 in an in vitro angiogenesis model. Lab Invest. 2003;83:1873 -1885.

Downloaded from iovs.arvojournals.org on 09/30/2021 24

Endogenous Protein Inhibitors of Angiogenesis

Supplement 11 to Campochiaro PA, First ARVO/Pfizer Institute Working Group. Ocular versus Extraocular Neovascularization: Mirror Images or Vague Resemblances. Invest Ophthalmol Vis Sci. 2006;47:462-474.

Olga Volpert

Northwestern University, Evanston, IL

Endogenous protein inhibitors may serve as built-in breaks for angiogenesis. Angiogenic stimuli cause changes in endothelial cells that lead to survival, proliferation, migration, and morphogenesis. Inhibitors cause different signaling responses that usually lead to death by apoptosis. Depending upon prevailing molecular cues, angiogenesis or regression occurs. Pigment Epithelium-Derived Factor (PEDF) PEDF is a secreted glycoprotein and a member of family, which actually lacks inhibitor activity and lacks an active serpin loop. It inhibits angiogenesis by promoting apoptosis of endothelial cells. It also induces neuronal differentiation. Its expression controlled by oxygen levels. The VEGF/PEDF ratio may be important for control of angiogenesis. (TSP1) Tsp1 is a large (180 kDa), secreted glycoprotein that has multiple functions, one of which is inhibiting angiogenesis by promoting apoptosis of endothelial cells by signaling through CD36 receptor. It is also a neurotrophic factor and can direct axon guidance. Its expression is regulated by glucose level. It also helps to maintain immune privilege by redirecting inflammation out of the eye and reducing inflammation in the eye. Blood vessels must remodel in order to be stopped. In animals treated with VEGF inhibitors, the resident, mature blood vessels are unaffected, so quiescent vessels are not destroyed. Apotosis is dependent upon cells in G1 phase of the cell cycle. What features of growing endothelium differentiate it from quiescent endothelium that target growing endothelium for destruction by inhibitors of angiogenesis? Angiogenic stimulators increase survival kinases and antiapoptotic factors that boost endothelial cell survival and promote angiogenesis. In parallel, the same molecules produce a totally different effect: they induce a death receptor, Fas, on endothelial cell surface. However, in the absence of Fas ligand this fails to induce cell death; for that you need inhibitors of angiogenesis, such as PEDF or Tsp1, which through different signaling pathways generate Fas ligand. Fas then binds Fas ligand resulting in caspase 8-dependent apoptosis. Other inhibitors of angiogenesis, angiostatin, , and canstatin, also seem to be acting at least in part by upregulating Fas ligand and inducing Fas- dependent apoptosis in activated endothelium. What are other points in the angiogenic cascade that can be attacked by angiogenesis inhibitors and cause endothelial cell death? Another critical molecule is nuclear factor of activated T cells or NFAT. It is involved in VEGF-induced angiogenesis. In quiescent cells, inactive NFAT is located in the cytosol. In the presence of VEGF, it translocates to the nucleus where it participates in transcriptional events. In cultured endothelial cells, NFAT is critical for VEGF-induced chemotaxis and tube formation; it is also needed for corneal angiogenesis in vivo.

Downloaded from iovs.arvojournals.org on 09/30/2021 25

How is NFAT regulated? Phosphorylation on 5 serine residues masks NFAT nuclear localization signal and maintains its cytoplasmic locale. It is bound to its regulatory phosphatase, calcineurin A, which is activated by calcium influx, which causes calmodulin activation, and in turn induction of calcineurin phosphatase. Dephosphorylation exposes NFAT nuclear localization signal so that it goes to the nucleus, binds to the consensus sites on the DNA and induces transcription. There are also opposing signals that regulate NFAT through several kinases that rephosphorylate NFAT in the cytoplasm and cause its retention, or in the nucleus, where rephosphorylation causes NFAT return to the cytoplasm through CrmA. When endothelial cells are treated with VEGF or FGF2 and cell lysates are immunoprecipitated with antiphosphoserine Ab and then immunoblotted for NFAT, they show dephosphorylation (activation) of NFAT. In control cells, NFAT is phosphorylated. NFAT activation is disrupted by PEDF and TSP1, which cause rephosphorylation of NFAT. When cells are treated with VEGF or FGF2, staining for NFAT highlights predominantly nuclei, but when treated with both VEGF and PEDF, there are few stained nuclei. What are the kinases regulating NFAT deactivation by PEDF? PEDF has no effect on the activation of ERK kinases. JNK kinases are activated by PEDF, but only in activated endothelial cells; VEGF or FGF2 cause mild activation of JNK kinases that is greatly augmented by PEDF. Jun, a substrate for JNK kinases, is highly phosphorylated in the presence of PEDF only in stimulated and not in quiescent endothelial cells. A JNK inhibitor blocks NFAT deactivation by PEDF (dephosporylation and nuclear localization are maintained). Is JNK binding NFAT? JNK2, but not JNK1, binds to cytosolic NFAT. VEGF disrupts the binding of JNK1 to NFAT, but PEDF restores the binding of JNK1 to NFAT in the presence of VEGF. In the nucleus, both JNK1 and JNK2 are able to bind NFAT upon PEDF treatment. We hypothesized that JNK1 performs NFAT retention in the cytoplasm, while both JNK1 and JNK2 are responsible for the shuttling of NFAT out of the nucleus. JNK inhibitors block PEDF- induced apoptosis of endothelial cells. PEDF blocks endothelial cell migration stimulated by VEGF: JNK inhibitor disrupts this blockade. PEDF blocks FGF2-induced corneal angiogenesis, but not in the presence of a JNK inhibitor. Electrophoretic mobility shift assay demonstrates that NFAT DNA binding is reduced in PEDF-treated endothelial cells. Of NFAT targets that might be important for apoptosis, cFLIP, an endogenous inhibitor of caspase 8, was reduced in PEDF- treated cells and this reduction was JNK-dependent both at the protein and mRNA level. Chromatin immunoprecipitation assays show that NFAT binds to the cFLIP promoter in VEGF- or FGF2-treated cells: this binding is disrupted by PEDF. In summary, VEGF or FGF2 can induce calcineurin phosphatase, which dephosphorylates NFAT, promotes its nuclear localization, and drives transcription of proangiogenic molecules such as cFLIP or Cox-2 resulting in neovascularization. However, when PEDF or TSP1 are present, there is induction of JNK kinases, which phosphorylate NFAT, causing it to relocalize to the cytoplasm thereby blocking proangiogenic transcriptional events. Other transcription factors that are changed in remodeling endothelial cells in reponse to PEDF include Egr-1, NFκB, cMyb and CREB. CREB, cMyb and Er-1, like NFAT, are decreased and may form a common transcription network. CREB is a co-activator for NFAT, as is Egr-1, which can co-operate with NFAT by forming heterodimers. CMyb is one of NFAT transcriptional targets, which can also be regulated by calcium levels. NFκB, which on the contrary is decreased by PEDF, competes with NFAT for binding sites. NFκB also may drive Fas ligand expression

Downloaded from iovs.arvojournals.org on 09/30/2021 26

The NFκB activation pathway is affected by PEDF and TSP1. IκB is an inhibitor of NFκB and its levels go down in VEGF-treated cells: this decrease is relieved by both inhibitors. IκB is regulated by phosphorylation/proteasomal degradation: phospho-IκB is increased by PEDF and by TSP1. In nuclear extracts, NFκB is increased by inducers of angiogenesis including VEGF and FGF2, but it is increased much more when PEDF or TSP1 are added. ChIP assay showed that NFκB binding to Fas ligand promoter is increased when activated endothelial cells are treated either PEDF or TSP1. VEGF or FGF2 induce NFAT activation and nuclear localization possibly in cooperation with cofactors Egr-1 and cMyb, which in turn increase levels of cFLIP, Cox2, and other targets. These signals are targeted by PEDF and TSP1 through JNK kinases. Inducers upregulate trafficking of CD95 or Fas to the endothelial cell surface, where it binds Fas ligand (also upregulated by PEDF or TSP1), causing endothelial cell apoptosis and blocking angiogenesis. PEDF may have a special role in prostate, because PEDF KO mice have hyperplasia and increased miciovascular density in the prostate. Testosterone suppresses PEDF protein levels, and PEDF is lower in metastatic prostate cancer than in non-metastatic one. Crystal structure of PEDF indicates two areas on the surface that are likely to be involved in receptor binding consisting (44 mer and a 34 mer N-terminal peptides). The 44 mer was found to have neurodifferentiation activity, and the 34 mer was anti-angiogenic. The 34 mer, but not 44 mer (1) blocks VEGF-induced migration of endothelial cells, (2) blocks FGF2-induced angiogenesis in the corneal pocket model, (3) induces JNK activation just like full-length PEDF, (4) blocks NFAT activation, and (5) reduces binding to the cFLIP promoter. The tet-inducible promoter system was used to express the 34 mer, the 44 mer, or a large N-terminal fragment in PC3 prostate cancer cells. The tranfected cells were implanted into mice and the mice were treated with Doxycycline. The large N-terminal fragment, 34 mer, and 44 mer all decreased tumor growth, but did so in a different fashion. The N-terminal fragment and the 34 mer decreased vascular density and increased endothelial apoptosis, while the 44 mer didn’t decrease vascular density or increase endothelial apoptosis, but induced differentiation and slowed tumor growth.

Downloaded from iovs.arvojournals.org on 09/30/2021 27

Hypoxia and HIF-1

Supplement 12 to Campochiaro PA, First ARVO/Pfizer Institute Working Group. Ocular versus Extraocular Neovascularization: Mirror Images or Vague Resemblances. Invest Ophthalmol Vis Sci. 2006;47:462-474.

Gregg L. Semenza

Johns Hopkins University, Baltimore, MD

Hypoxia-inducible factor-1 (HIF-1) regulates a large battery of genes that help cells survive oxygen deprivation or increase oxygen delivery to tissues. HIF-1 is a heterodimer composed of HIF-1α and HIF-1β. It binds to the hypoxia response element (HRE), short sequences that can be put into a heterologous gene, and then that gene will be induced by hypoxia. When HIF-1 binds to the HRE, it recruits the coactivators P300 and CBP, which both serve as a bridge to the transcription initiation complex and have histone acetyltransferase activity that is required for remodeling of chromatin. The expression of the HIF-1α subunit is tightly regulated by the cellular oxygen concentration. There is a modest increase in HIF-1α between 20% and 6% and then a sharp increase when oxygen concentration drops below 6%, which corresponds to 40 mm Hg (venous PO2). In vivo all cells are exposed to oxygen concentrations below 6%, and so any fluctuations occur along the steep part of the curve. This provides a mechanism for a graded response to hypoxia - the more severe the hypoxia, the greater the expression of HIF1α and the greater the transcription of downstream target genes. The molecular basis for this regulation of HIF-1α levels is oxygen-dependent hydroxylation of two proline residues in HIF-1α by a group of enzymes called prolyl hydroxylases (reviewed in Ref. 1). These enzymes use molecular oxygen as a substrate and have a high Km for oxygen so that oxygen is a rate limiting substrate under physiological conditions. Under non-hypoxic conditions these residues are hydroxylated, which is required for the binding of von Hippel-Lindau (VHL) tumor suppressor protein, the recognition component of an E3 protein ligase that ubiquitinates HIF-1α and targets it for degradation in the proteosome. Under hypoxic conditions, oxygen becomes limiting, the proline residues remain unhydroxylated, VHL does not bind, and HIF-1α accumulates within the cell. There is also an asparagine residue in the transactivation domain that is hydroxylated by a hydroxylase called FIH-1. This hydroxylation prevents the interaction of the transactivation domain with the co-activators. Under hypoxic conditions, the hydroxylation does not occur and the binding of the co-activators can occur. So both the half-life of HIF-1 and its specific activity as a transcription factor are regulated in an oxygen-dependent manner. In addition to regulation by oxygen concentration, a whole series of cytokines and growth factors can also increase HIF-1α levels, and rather than affecting degradation, they increase the synthesis of the protein. IGF -1 stimulates VEGF production through HIF-1α. Cultured cells exposed to IGF-1 show a time-dependent increase in HIF-1α.2 The induction of HIF-1α by IGF- 1 can be blocked by inhibitors of signal transduction pathways, including the MAP kinase and phosphatidylinositol 3 (PI3) kinase pathways and the downstream kinase mTOR. The levels of HIF-1α protein correlate with VEGF mRNA levels. The basis for the regulation of the HIF-1α

Downloaded from iovs.arvojournals.org on 09/30/2021 28

protein synthesis is through the stimulation of the PI3 kinase and MAP kinase pathways that ultimately regulate the activity eIF-4E, which is a critical regulator of cap-dependent mRNA translation. The signals coming through these pathways stimulate translation of a subset of mRNAs within the cell and included among these is HIF-1α mRNA.3 This pathway is relevant to diabetic retinopathy, because there is activation in the diabetic retina of the Akt protein kinase, increased levels of HIF-1α and VEGF. Pharmacological blockade of the IGF-1 receptor dramatically reduces the levels of activated Akt, HIF-1α, and VEGF. Thus, there are at least two general stimuli that regulate HIF-1α levels, hypoxia and growth factors.4 HIF-1α plays a role in the induction of VEGF in oxygen–induced ischemic retinopathy. There is basal expression of HIF-1α at P7 in normoxia and downregulation when mice are exposed to hyperoxia. When mice are taken out hyperoxia and the retina becomes hypoxic, HIF- 1α levels increase and remain elevated for several days, during which VEGF levels increase.5 HIF-1α is essential for normal embryonic development. HIF-1α knockouts arrest in development at day 8.5 and die by day 10.6 There is a dramatic effect on vascularization. There is initial establishment of the vasculature, but then it regresses. In embryonic stem cells from wild type mice exposed to 1% oxygen for 24 hours, there is strong induction of VEFG, Ang1, PlGF, and PDGF-B. The hypoxic induction of these mRNAs is dramatically reduced in HIF-1α- null cells. In primary cardiac fibroblasts, cardiac myocytes, arterial smooth muscle cells, and arterial endothelial cells, expression of VEGF, Ang1, PIGF, PDGF- B were compared after exposure to 1% or 20% oxygen for 24 hours. Each of these cell types responds in a different manner in terms of the regulation of these genes. VEFG is uniformly upregulated, but the others are not.7 Thus, the response to hypoxia is cell type-specific. A mutant form of HIF-1α that is constitutively expressed independent of hypoxia has been engineered by a series of point mutations and deletions that allow it to escape proteosomal degradation under non-hypoxic conditions. An adenoviral vector was used to transfect the above cells with constitutively active HIF-1α or LacZ. The pattern of expression was identical to the pattern of expression induced by hypoxia; the cell type specificity was maintained.7 The adenoviral vectors were injected into mouse eyes. After intravitreous injection of adenoviral vector expressing constitutively active HIF-1α (AdCA5), there was neovascularization along the surface of the retina and in the anterior chamber of the eye. There was no effect from injection of the adenoviral vector expressing LacZ (AdLacZ). VEGF is necessary, but not sufficient to get neovascular sprouts from superficial capillaries like that seen by expressing the constitutively active HIF-1α. Expression of mRNAs 24, 48, and 72 hours after intravitreous injection of AdCA5 was compared to that in eyes injected with AdLacZ. HIF-1α mRNA levels peaked at 48 hours after injection and there were dramatic increases in PIGF, VEGF, Ang1, Ang2, and PDGF- B mRNAs.7 The coordinate upregulation of multiple angiogenic factors is likely to account for the development of neovascularization. Activation upstream through HIF-1α may have benefits by turning on multiple factors and perhaps doing so in a cell type-specific manner. Does HIF-1 have specific effects within endothelial cells? Arterial endothelial cells were cultured at 1% or 20% oxygen and their ability to invade through an experimental basement membrane was measured. Hypoxia increased the invasiveness of the cells.8 Infection of the cells with AdCA5 also increased the invasiveness of the cells under non-hypoxic conditions. When cells were plated on Matrigel, the formation of tube-like networks was stimulated by hypoxia or infection of the cells with AdCA5. To determine the molecular basis for these observations,

Downloaded from iovs.arvojournals.org on 09/30/2021 29

microarray analysis was performed. Arterial endothelial cells were exposed to 20% or 1% oxygen for 24 hours and the pattern of gene expression was assessed in three independent cell isolates. A threshold of a statistically-significant 1.5-fold increase or decrease in gene expression was used. The experiment was repeated under non-hypoxic conditions using cells infected with AdCA5 or AdLacZ. 245 genes were up-regulated and 325 genes were down- regulated in response to both hypoxia and AdCA5.8 Many of the genes encoded signal transduction molecules, growth factors, collagens, and receptors, but the largest group in which changes occurred were genes encoding transcription factors. Expression of the mRNAs encoding the erythropoietin receptor, the chemokine receptor CXCR4, VEGF-A, and VEGF-C were upregulated within 8 hours of the onset of hypoxia suggesting that they are direct target genes. In fact, it is known from other experiments that CXCR4 and VEGF-A are direct target genes. PTGIS, which encodes an enzyme required for production of prostaglandins, shows bimodal activation, suggesting activation by HIF-1 early followed by the recruitment of additional transcriptional activators at later timepoints. Thus, it appears that hypoxia induces HIF-1, which binds to its primary target genes and induces their expression, some of which encode other transcription factors that contribute to the regulation of gene expression later in the hypoxic response. HIF-1 plays a critical role in tumor development. Over-expression of HIF-1α in colon carcinoma cells results in increased tumor growth in a xenograft model compared to cells transfected with empty vector. Magnetic resonance imaging showed an increase in vascular volume and vascular permeability in HIF-1α-overexpressing tumors.9 Both basal and hypoxia- induced secretion of VEGF were reduced in TMK-1 human gastric cancer cells stably transfected with an expression vector encoding a dominant-negative form of HIF-1α.10 When these cells were injected orthotopically into the gastric wall, there was a reduction in the size of tumors derived from cells expressing dominant-negative HIF-1α and the vessels were very small with almost no lumen and marked reduction in pericytes coverage. This suggests that reduction of HIF-1α in tumor cells prevents endothelial cells from recruiting pericytes. The mechanism for this is unknown. Multiple anti-cancer drugs are known to have anti-angiogenic activity that is due in part to their inhibition of HIF-1 (reviewed in Ref. 13).

Questions

1. If one were targeting HIF for anti-angiogenesis, would it be necessary to target both HIF-1 and HIF-2? The role of HIF-2 (which is a dimer of HIF-1 and HIF-2 , which is structurally related to HIF-1α but is the product of a distinct gene) is not as well-established. In some cells HIF-2 is present in the cytoplasm and is not active, suggesting that nuclear translocation may occur in response to an unidentified signal. There are many tumors in which both HIF-1 and HIF-2 are over-expressed, but there are also tumors in which just HIF-1 or HIF-2 is high. Some inhibitors have specificity for HIF-1α. It is not clear that growth factors induce HIF-2α.

2. The list of anti-tumor agents that inhibit HIF-1 is extensive. What is the mechanism for inhibition by something like Iressa? Activation of signal transduction pathways downstream of receptor tyrosine kinases stimulates HIF-1α production, so blockade of these signaling pathways reduces HIF-1. A consequence of HIF-1 inhibition by these agents is that the tumors become hypoxic, resulting in

Downloaded from iovs.arvojournals.org on 09/30/2021 30

the induction of HIF-1α through inhibition of the PHD-VHL system. This suggests that these signal transduction inhibitors would not be as efficacious when used alone as they would be if used with a HIF-1-specific inhibitor should such a small molecule inhibitor become available.

3. What are the downstream effects of HIF-1? Selective knockout of HIF-1α in endothelial cells has a negative effect on the vascularization of tumors. The effect appears to involve loss of autocrine VEGF signaling through VEGF receptors within endothelial cells. The anti-angiogenic agent angiostatin also inhibits the expression of HIF-1 target genes in endothelial cells.

4. There are hamartomatous tumors such as in tuberous sclerosis and Peutz-Jager syndrome that occur because of mutations in tumor suppressor genes that feed into mTor. Is anything known about HIF-1 in these tumors? Dysregulation of HIF-1 and VEGF is a unifying feature of hamartomatous syndromes (see Brugarolas 2003, 2004).

5. There are multiple oxygen binding heme proteins that are quite good at detecting oxygen levels and changing conformation. Is HIF-1 the only story for oxygen sensing in cells? No, it is unlikely that it’s the only story, but so far this is the mechanism for sensing oxygen that is best understood.

References

1. Hirota K, Semenza GL. Regulation of hypoxia-inducible factor 1 by prolyl and asparaginyl hydroxylases. Biochem Biophys Res Commun. 2005;Sep 8; [Epub ahead of print]. 2. Fukuda R, Hirota K, Fan F, Jung YD, Ellis LM, Semenza GL. Insulin-like growth factor 1 induces hypoxia-inducible factor 1-mediated vascular endothelial growth factor expression, which is dependent on MAP kinase and phosphatidylinositol 3-kinase signaling in colon cancer cells. J Biol Chem. 2002;277:38205-38211. 3. Laughner E, Taghavi P, Chiles K, Mahon PC, Semenza GL. HER2 (neu) signaling increases the rate of hypoxia-inducible factor 1α (HIF-1 α) synthesis: novel mechanism for HIF-1- mediated vascular endothelial growth factor expression. Mol Cell Biol. 2001;21:3995-4004. 4. Poulaki V, Joussen AM, Mitsiades N, Mitsiades CS, Iliaki EF, Adamis AP. Insulin-like growth factor-I plays a pathogenetic role in diabetic retinopathy. Am J Pathol. 2004;165:457- 469. 5. Ozaki H, Yu AY, Della N, et al. Hypoxia inducible factor-1α is increased in ischemic retina: temporal and spatial correlation with VEGF expression. Invest Ophthalmol Vis Sci. 1999;40:182- 189. 6. Iyer NV, Kotch LE, Agani F, Leung SW, Laughner E, Wenger RH, Gassmann M, Gearhart JD, Lawler AM, Yu AY, Semenza GL. Cellular and developmental control of O2 homeostasis by hypoxia-inducible factor 1 α. Genes Dev. 1998;12:149-162. 7. Kelly BD, Hackett SF, Hirota K, et al. Cell type-specific regulation of angiogenic growth factor gene expression and induction of angiogenesis in nonischemic tissue by a constitutively active form of hypoxia-inducible factor 1. Circ Res. 2003;93:1074-1081. 8. Manalo DJ, Rowan A, Lavoie T, et al. Transcriptional regulation of vascular endothelial cell responses to hypoxia by HIF-1. Blood. 2005;105:659-669.

Downloaded from iovs.arvojournals.org on 09/30/2021 31

9. Ravi R, Mookerjee B, Bhujwalla ZM, et al. Regulation of tumor angiogenesis by p53-induced degradation of hypoxia-inducible factor 1alpha. Genes Dev. 2000;14:34-44. 10. Stoeltzing O, McCarty MF, Wey JS, et al. Role of hypoxia-inducible factor 1alpha in gastric cancer cell growth, angiogenesis, and vessel maturation. J Natl Cancer Inst. 2004;96:946-956. 11. Brugarolas JB, Vazquez F, Reddy A, Sellers WR, Kaelin WG Jr. TSC2 regulates VEGF through mTOR-dependent and -independent pathways. Cancer Cell. 2003;4:147-158. 12. Brugarolas J, Kaelin WG Jr. Dysregulation of HIF and VEGF is a unifying feature of the familial hamartoma syndromes. Cancer Cell. 2004;6:7-10. 13. Semenza GL. Targeting HIF-1 for cancer therapy. Nat Rev Cancer. 2003;3:721-32.

Downloaded from iovs.arvojournals.org on 09/30/2021 32

Id1 Regulation of Angiogenesis

Supplement 13 to Campochiaro PA, First ARVO/Pfizer Institute Working Group. Ocular versus Extraocular Neovascularization: Mirror Images or Vague Resemblances. Invest Ophthalmol Vis Sci. 2006;47:462-474.

Rhoda M. Alani

The Sidney Kimmel Comprehensive Cancer Center, Johns Hopkins University School of Medicine, Baltimore, Maryland

Inhibitor of differentiation (Id) proteins are primarily involved in regulating cellular differentiation. They are dominant -negative helix-loop-helix (HLH) transcription factors. Normally, basic helix-loop-helix (bHLH) transcription factors bind DNA as dimers primarily to E-boxes and often regulate genes important for differentiation. Id proteins lack the basic domain and therefore they can’t bind DNA. Thus, Id can be taken to mean inhibitor of DNA-binding or inhibitor of differentiation. Over the past decade, Id proteins have been determined to regulate many processes involved in tumorigenesis including tumor angiogenesis, tissue invasion by regulating MMPs and CXCR4, cellular proliferation primarily by regulating the Rb pathway and p16 as well as ETS transcription factors, and dedifferentiation or anaplasia by regulating the tissue- specific HLH transcription factors (reviewed in Ref. 1). Id1, -2 and -3 are expressed ubiquitously, while Id4 is expressed in neurons. In adults, most expression is in proliferating stem cells. Id1 K0s have no obvious phenotype. Id2 K0s lack NK cells, Peyers patches, and Langerhans cells, and have defective spermatogenesis. Id3 K0s have altered humoral immunity. Id genes made headlines in October 1999 when work from Robert Benezra’s lab demonstrated that mice that lacked both Id1 and 1d3 had loss of viability with defects in neurogenesis and angiogenesis in the brain. In xenograft tumor models, heterozygous double knockouts were unable to support growth of any of three types of tumor cells due to defective angiogenesis.2 This manuscript suggested that inhibiting Id gene function would be an effective means of targeting a large variety of tumors therapeutically. How do Id genes affect the angiogenic process? In order to answer this question it is necessary to look at the downstream effectors of this transcriptional regulatory protein. What are downstream effectors of Id genes? Using PCR-based subtractive hybridization, genes were identified that are selectively expressed or downregulated in Id1 knockout compared to Id1 wild type mouse embryo fibroblasts. Genes repressed by Id1 are upregulated in null cells and include thrombospondin 1 (Tsp1), β3 , , and other adhesion molecules. The upregulation of Tsp1 was confirmed by quantitative RT-PCR. In cells transfected with a Tsp1 promoter/luciferase reporter construct, addition of Id1 results in reduction in luciferase activity. An attempt was made to map the Id1 responsive areas in the promoter, but even deleting down to 160 bp, there was still some repression, suggesting that it is an indirect effect. Conditioned medium (CM) from cell cultures derived from Id1 KOs caused less endothelial cell migration than CM from wild type cell cultures and this was reversed by antibodies to Tsp1. Matrigel implant assays were done in which VEGF- or FGF2-loaded matrigel was implanted in Id1 KOs or wild type mice. The KOs had less angiogenesis and there was Tsp1 staining around vessels.

Downloaded from iovs.arvojournals.org on 09/30/2021 33

These data suggest that Tsp1 is a downstream effector of Id1; the effects of Id1 on Tsp1 are indirect and Tsp1 is a major determinant of Id effects on angiogenesis.3 Id made headlines again in November 2001, when work from Shahin Rafii’s lab showed that tumor growth and angiogenesis could be re-established in Id KOs by bone marrow transplantation.4 How does the bone marrow do this? There is evidence to suggest that these cells are endothelial cells. This is the pathway felt to be involved in endothelial cell development from hemangioblasts. However, is the tumor xenograft model representative of human cancer? Autochthonous tumor systems in which a tumor develops in a mouse is more relevant. One such model system is two step tumor induction of skin cancer in mice using DMBA/TPA treatment. In this setting, mice develop papillomas which ultimately develop into invasive squamous cell carcinomas. Id1 KO mice treated with DMBA and TPA developed more papillomas than wild type mice and there was no difference in angiogenesis. This data was counter to the previous results using tumor xenografts which made us question whether Id1 KO mice also possessed a defect in tumor surveillance, Since γδ T cells play an important role in skin tumor surveillance and receptor KO mice have increased skin cancer susceptibility,5 γδ T cells were assessed in the skin of Id1 KO mice and found to be markedly decreased. Is this because they are not being made in the thymus or are they just not getting out to where they should be? The embryonic thymus was normal; there was just a difference in γδ T cells in skin. Is there a defect in migration or premature death of γδ T cells? Thymocytes from wild type or Id1 KO mice were injected into Rag KO mice and while wild type thymocytes could produce γδ T cells in skin, those from Id1 KOs could not, suggesting either a migration defect or cell survival defect in Id1 KO thymocytes. The Id1 KO thymocytes had reduced migration to SDF-1 and this was determined to be due to reduced expression of CXCR4 in Id1 KO thymocytes.6 While these studies have identified a new Id-mediated pathway involving CXCR4, this work further points out an important difference between xenograft and autothenous tumor models. Since tumor xenografts are unable to grow in Id1 KO mice due to defects in tumor- associated angiogenesis while autochthonous tumors show no angiogenic defects in these same mice, we suggest that the sudden appearance of a large load of cells in xenograft models results in a large release of angiogenic factors stimulating angiogenesis that is Id-dependent and bone- marrow derived. In contrast, the gradual growth of tumors in autochthonous tumor models is accompanied by slow induction of angiogenesis as tumors go from in situ to invasive disease. Angiogenic factors rise slowly and oxygen and glucose are slowly depleted and results in angiogenesis that is Id-independent and derived from neighboring vessels.7 A recent study investigated the source of endothelial cells in tumors that developed in patients who had had bone marrow transplants. Tumor biopsies from six patients that had received transplant cells derived from a member of the opposite sex were examined with q-fish and immunohistochemistry to determine the origin of endothelial cells. It was found that 4-12% of endothelial cells in tumors were derived from the bone marrow. Tumor types were unusual in that there were no epithelial tumors, but regardless, it suggests that the bone marrow can contribute, but only a very small component of the endothelial cells are derived from the bone marrow.8 We suggest that bone marrow angiogenesis is essentially Id-dependent angiogenesis. There is acute, severe hypoxia, which results in recruitment of cells from the bone marrow, and then gradually there is remodeling and the bone marrow components are decreased. This suggests that it might be possible to use bone marrow stem cells as therapeutic agents in hypoxic disease.

Downloaded from iovs.arvojournals.org on 09/30/2021 34

Question

1. What is the mechanism for Id-dependent bone marrow associated angiogenesis? CXCR4 may be involved and Id regulation of CXCR4 may be important for bone marrow cells, the hemangioblasts and precursor cells, to get out of the bone marrow. Studies are currently underway to test this hypothesis.

References

1. Sikder HA, Devlin MK, Dunlap S, Ryu B, Alani RM. Id proteins in cell growth and tumorigenesis. Cancer Cell. 2003;3:525-530. 2. Lyden D, Young AZ, Zagzag D, et al. Id1 and Id3 are required for neurogenesis, angiogenesis and vascularization of tumour xenografts. Nature. 1999;401:670-677. 3. Volpert OV, Pili R, Sikder HA, et al. Id1 regulates angiogenesis through transcriptional repression of thrombospondin-1. Cancer Cell. 2002;2:473-483. 4. Lyden D, Hattori K, Dias S, et al. Impaired recruitment of bone-marrow-derived endothelial and hematopoietic precursor cells blocks tumor angiogenesis and growth. Nat Med. 2001;7:1194-1201. 5. Girardi M, Oppenheim DE, Steele CR, et al. Regulation of cutaneous malignancy by gammadelta T cells. Science. 2001;294:605-609. 6. Sikder HA, Huso DL, Zhang H, et al. Disruption of Id1 reveals major differences in angiogenesis between transplanted and autochthonous tumors. Cancer Cell. 2003;4:291-299. 7. Alani RM, Silverthorn CF, Orosz K. Tumor angiogenesis in mice and men. Cancer Biol Ther. 2004;3:498-500. 8. Peters BA, Diaz LA, Polyak K, et al. Contribution of bone marrow-derived endothelial cells to human tumor vasculature. Nat Med. 2005;11:261-262.

Downloaded from iovs.arvojournals.org on 09/30/2021 35

Inflammation and NF-κB

Supplement 14 to Campochiaro PA, First ARVO/Pfizer Institute Working Group. Ocular versus Extraocular Neovascularization: Mirror Images or Vague Resemblances. Invest Ophthalmol Vis Sci. 2006;47:462-474.

Rainer de Martin

Department of Vascular Biology & Thrombosis Research, Medical University of Vienna, Austria

Endothelial cells can be activated by chemical, biological and physical stimuli such as hypoxia, VEGF, IL-1, TNF, LPS, advanced glycation endproducts, oxidized LDL, and many others. Depending upon the stimulus, they will respond with altered gene expression and produce a number of gene products, e.g.,interleukins and adhesion molecules. During the inflammatory response, these genes mediate the chemoattraction of leukocytes and their adhesion to the endothelium. NFκB is a transcription factor that plays a central role in the regulation of expression of these genes. The NFκB family consists of a group of both activating and inhibiting factors. The activators contain a Rel homology domain for DNA binding and transactivation, and a homo- and heterodimerization domain. The p65 or RelA subunit is the most prominent activator. It can form heterodimers with p50, which like p52 is made from larger precursors and proteolytically cleaved. The inhibitory c-terminal fragment is called IκBγ. The most prominent NFκB heterodimer is p65/p50, whereas the most prominent inhibitor is IκBα. It contains ankyrin repeats, which are protein-protein interaction domains. How is NFκB regulated? Under unstimulated conditions, NFκB resides in the cytoplasm bound to IκBα. When cells are stimulated, phosphorylation of IκBα occurs on two serine residues, which is a signal for ubiquitination and degradation of IκBα by the proteosome. The p65/p50 subunit is then free, a nuclear localization sequence that was masked is now exposed, and NFκB can translocate to the nucleus resulting in gene expression. This phosphorylation of IκBα takes place within the signalosome, a high-molecular weight complex that contains two IκBα kinases, IKK1 (IKKα) and IKK2 (IKKβ). They are bound to a noncatalytic subunit, IKKγ (NEMO), which couples the signalosome to upstream activating molecules, such as the receptors for IL1 or TNF, via MAP3 type kinases, e.g., MEKK1 and TAK1. Many other stimuli can activate NFκB e.g., viruses, γ-irradiation, or UV light; however, all signals appear to converge on the signalosome. One exception is lymphotoxin, which activates NF-kB through an alternative pathway utilizing NIK and IKK1 and leads to phosphorylation and to the proteolytic cleavage of the precursor p100 to yield p52. p52, in association with RelB is important for dendritic cell activation and maturation. In the context of inflammatory signals, it is predominantly IKK2 that transmits the signals. NFκB is a stress-response factor that is tightly regulated to ensure that its activation is transient. This is accomplished through a regulatory loop by which NFκB induces the expression of its own inhibitor, IκBα. IκBα limits nuclear translocation of NFκB, as it can also enter the nucleus and remove NFκB from its binding site(s). There is a sort of oscillation of NFκB from cytoplasm to nucleus and back; the dampening of this oscillation depends on the combination of

Downloaded from iovs.arvojournals.org on 09/30/2021 36

the IκB species that are present.1 Also, there is a constitutive nuclear to cytoplasmic shuttling of NFκB, IκB, NIK, and IKKα, suggesting additional ways of activation of NF-kB in the nucleus.2 Adenoviral vectors used to transfect IκBα into endothelial cells results in reduced expression of inflammatory markers such as VCAM, IL1, IL6, and IL8. Over-expression of a dominant-negative form of IKK2 has the same effect and also inhibits tube formation in endothelial cells. TNF results in rapid and transient degradation of IκBα, but VEGF does not. Endothelial cell stimulation with VEGF was compared to stimulation with IL1 using microarrays. Expression of Down syndrome critical region (DSCR1) was increased by both VEGF and IL1. DSCR1 is important in NFAT signaling. It is an inhibitor of calcineurin, which is a phosphatase that dephosphorylates NFAT leading to its nuclear localization. Therefore, DSCR1 may be part of another negative feedback mechanism that is operative in endothelial cells, in response to VEGF, leading to shut down NFAT activity (also, DSCR1 expression itself is NFAT-dependent). So this resembles the situation with NFκB/IκBα where a transcription factor regulates the expression of its own inhibitor. There are genes that are specifically regulated by inflammatory, not by angiogenic, stimuli, and vice versa, but also genes that are regulated by both stimuli, e.g., MAF2C is a transcription factor upregulated by VEGF, but not IL1. TSLP (tumor stromal lymphopoietin) is upregulated by IL1, but not VEGF. There is 40% overlap between genes upregulated by VEGF and IL1 and there is 30% overlap between genes upregulated by VEGF, IL1, and EGF. Can regulatory information be delineated from microarray data? We have used a bioinformatics approach to answer this question. The starting point is a set of co-clustered genes, which means that their kinetics of expression is similar. We then postulate that there are common regulatory elements in the promoters; however, caution is needed, because expression of genes is not only regulated transcriptionally, but also on the level of RNA stability; but if you have a sufficient number of genes, you may find transcription factor binding sites that are statistically over-represented. We then extract the promoter regions from all the modulated genes from databases and search for known transcription factor binding sites, and for protein modules, which are combinations of transcription factor binding sites (because usually it is not one single transcription factor that induces a gene, but rather a combination of transcription factors), and also for sequence patterns which can encompass known transcription factor binding sites or novel ones. About 50% of genes upregulated by IL1 contain NFκB binding sites within 1 kb of the transcription start site.3 Signaling pathways are not linear, but often branched and there is cross-talk between individual pathways. One question that we have addressed recently is the link between NFκB and other signaling pathways. IKK2 has served as a starting point for these studies. IKK2 goes through a number of phosphorylations that regulates and ultimately also shuts down its activity. IKKs contain a kinase and a leucine zipper domain, which mediate dimerization and a HLH domain. The association of the HLH domain with the kinase domain has regulatory function, so the HLH domain was used as bait in a yeast 2-hybrid screen. One gene that we found was α-catulin. α- Catulin has its name from its homology to vinculin and α-catenin. The interaction was confirmed by co-immunoprecipitation. Both IKK1and IKK2 bind to α-catulin, and the interaction is stronger with IKK1. When α-catulin is transfected into endothelial cells, it localizes to lamellipodia, the leading edges of migrating cells. α-catulin also interacts with Lbc, a Rho-guanine nucleotide exchange factor (Rho-GEF). This is important because cross-talk between the NFκB pathway and the Rho pathway has been described previously, but the

Downloaded from iovs.arvojournals.org on 09/30/2021 37

molecular basis was not understood. α-Catulin was transfected into cells and the ability of Lbc to induce NFκB activity was assayed in luciferase assays using an NFκB-dependent reporter construct. Transfection with wild type α-catulin enhanced Lbc-stimulated NFκB activity, whereas a dominant-negative construct reduced NFκB activity. When cells were stimulated with lysophosphatidic acid (LPA), an inducer of Rho, and compared to TNF stimulation, it is seen that α-catulin enhances NFκB activity in response to both stimuli. If α-catulin is indeed a link between the NFκB and Rho pathways, then small increases of α-catulin levels should facilitate the cross-talk, but large increases should disrupt the connection between the pathways. Indeed, transfection experiments using this setup supported this scenario. In summary, NFκB is of central importance in endothelial cell biology, not only for inflammation, but also for several other functions including aspects of angiogenesis, which may include protection from apoptosis and cytoskeletal reorganization during cell migration.

References

1. Hoffmann A, Levchenko A, Scott ML, Baltimore D. The IkappaB-NF-kappaB signaling module: temporal control and selective gene activation. Science. 2002;298:1241-1245. 2. Birbach A, Gold P, Binder BR, Hofer E, de Martin R, Schmid JA. Signaling molecules of the NF-kappa B pathway shuttle constitutively between cytoplasm and nucleus. J Biol Chem. 2002;277:10842-10851. 3. Mayer H, Bilban M, Kurtev V, et al. Deciphering regulatory patterns of inflammatory gene expression from interleukin-1-stimulated human endothelial cells. Arterioscler Thromb Vasc Biol. 2004;24:1192-1198.

Downloaded from iovs.arvojournals.org on 09/30/2021 38

Mechanisms in Vessel Pathfinding

Supplement 15 to Campochiaro PA, First ARVO/Pfizer Institute Working Group. Ocular versus Extraocular Neovascularization: Mirror Images or Vague Resemblances. Invest Ophthalmol Vis Sci. 2006;47:462-474.

Peter Carmeliet

University of Leuven, Leuven, Belgium

Pathfinding occurs through attraction and repulsion, breaking long distances down into smaller, more manageable distances. There are only a few classes of molecules that can do this. In zebrafish, the intersomitic vessels (ISV) initially sprout from the dorsal aorta. It occurs in absence of flow and hypoxia; it is genetically pre-programmed. The vessels take a route to a checkpoint or guidepost, and then they are directed to another guidepost until it reaches its destination where it helps to form the DLAV. At each of these guideposts there are attractive and repulsive molecules that tell the vessels where to go – not to stray into the somites, but rather to take the correct ventral to dorsal navigational route. There are four classes of molecules that provide directional cues: (1) netrin binding to deleted in colorectal carcinoma (DCC) or Unc5, (2) semaphorins binding to neuropilins and plexins, (3) ephrins binding to Ephs, and (4) slits binding to roundabout (Robo). In vessels, the primary Robo expressed is Robo4. How can so few signals do so much? They do so by using other tricks. An example is the role of netrin-DCC and slit-Robo signaling in the crossing of commissural axons at the spinal cord midline in mammals. Netrins at the midline attract axons by activating the DCC receptor, while Rig1/Robo3 silences Robo, which would otherwise repel axons from entering the midline. So by timed expression of another binding partner (Robo3), the repulsive signal of slits is silenced. A similar phenomenon modulates the attractant activity of netrins- when Unc5 is expressed, the combination of netrin, DCC, and Unc5 mediate repulsion. What happens when netrin and Unc5 are knocked down? In Unc5b knockout mice there are more vessel branches suggesting that Unc5b acts as a repellant. In zebrafish, netrin is expressed at the horizontal myoseptum (this is the first guidepost) and also at the floorplate in the neural tube. If netrin1a is knocked down, the initial trajectory is normal, but at the horizontal myoseptum it becomes abnormal so that the vessels go off in many different directions. This is true when you knockdown the Unc5b receptor; the vessels no longer know where to go when they reach the horizontal myoseptum. So in wild type fish, netrin is expressed at two guideposts and directs the vessels on their path, and in the absence of netrin the vessels go in many directions. Another ligand and receptor pair in the netrin family is repulsive guidance molecule A (rgmA) and neogenin (another type of receptor of the DCC family). In situ hybridization for neogenin shows expression in the brain, but also in the lateral plate mesoderm where the angioblasts develop. In splanchnic mesoderm of the chick there is co- localization of neogenin with VEGFR2 (in angioblasts). RgmA is expressed in the somites at the time that these angioblasts start to migrate. Angioblasts are born in the lateral plate mesoderm, express neogenin 1, are attracted to the midline by netrin

Downloaded from iovs.arvojournals.org on 09/30/2021 39

1B, and are kept from entering the somites by rgmA expressed in the somites. That ultimately leads to the formation of the dorsal aorta and the posterior cardinal vein. In neogenin knockdown zebrafish, the angioblasts stall, fail to migrate to the midline, and do not stay within the boundaries of the somites. The same is true for knockdown of the ligand, rgmA. Another ligand/receptor pair is slits/Robos. In the retina, netrin attracts the axons of the retinal ganglion cells to the optic nerve and then they must stay within the nerve. The slits and neuropilins provide a repulsive corridor that prevents the axons from going outside the optic nerve. The axons are attracted to the midline of the optic chiasm by netrins. At that point the repulsive signal of the slits is silenced by the expression of Robo3 silencing Robo1. Once at the midline, the Robos silence DCC so the axons can be expelled across the midline and not stalled. Once the axons have crossed the midline, Robos repel it and prevent it from recrossing. Does this also occur in vessels? It is not yet resolved, but it appears that slit2 and Robo1 can attract endothelial cells in tumors. Robo4 is an endothelial cell-specific receptor and when it binds slit 2, it repels endothelial cells, but a Robo4-trap inhibits angiogenesis. Slit 1A is expressed in the hypercord, which is located centrally and involved in attracting angioblasts toward the midline. It is also expressed in somites and in the floorplate. Knockdown of slit 1A results in malformation of the dorsal aorta and the intersomitic vessels branch at the correct locations, but then don’t follow the correct path. Like the neogenin knockdowns some angioblasts stall in the lateral plate. So it appears that slit 1A expressed by the hypercord is involved in attracting Robo4-positive angioblasts toward the midline and when that is interrupted, it would explain why the dorsal aorta is underdeveloped. Also, it is involved in preventing migration into the somites, because in its absence vessels migrate aberrantly into somites. All of this is phenocopied by Robo4 knockdowns. Robo3 is not expressed in vessels in humans and mice. In Robo3 knockdowns, the intersomitic vessels form properly but then begin branching abnormally. So, the ISVs produce slit 2 that can bind to Robo3. Robo3 is expressed in the neural tube and in the somites. So it appears that guidance receptors may be expressed on other cells besides endothelial cells. It is possible that guidance receptors on somites or neurons can be activated by a guidance ligand produced by the migrating endothelial cells and then feedback to the navigating vessel. In summary, the hyperchord produces slit 1A and netrin 1B, and they bind to their respective receptors, Robo4 and neogenin; they are important for angioblast migration from the lateral plate mesoderm to the midline. RgmA, which binds to neogenin is important to keep the angioblasts on their path and prevent migration into somites. In terms of ISV guidance up to the dorsal roof to form the dorsal aorta, rgmA helps the neogenin-positive angioblasts to migrate to the midline and at the midline, netrin 1B and slit 1A attract the cells and then there are guideposts, where netrin 1B helps to continue guidance and slit 1A is expressed in the somites and keep the Robo4-positive ISVs in between the somite boundaries.

Questions

1. There are a variety of human diseases in which there are vascular malformations. Have any of these mechanisms been implicated in those? Not known.

2. Is it know which signals induce these receptors on endothelial cells? No

Downloaded from iovs.arvojournals.org on 09/30/2021 40

3. How do you separate out direct from indirect effects? You have to look at the timing and pattern of expression and do tissue-specific knockdowns.

4. What is the current thinking about the role of PlGF in neovascular diseases? PlGF KO mice develop normally, so PlGF is not needed for physiological development. But when the mice are challenged with tumors or ischemia, they don’t develop as much NV, so PlGF plays a role in pathologic NV. There have been some contradictory results both in vitro and in vivo. The explanation for the different results in vitro is that endothelial cells produce a very large amount of PlGF in culture and their responsiveness to VEGFR1 becomes saturated, because endothelial cells from PlGF knockouts don’t lose the responsiveness. When overexpressed in tumors, PlGF may form heterodimers with VEGF and thereby lower the levels of VEGF. But Peter Carmeliet has done experiments that are not entirely consistent with that. PlGF KOs have been mated with mice that get spontaneous tumors, such as RipTag mice or K14-HPVB16 mice, and the initial results suggest that the growth of these tumors is reduced. He has generated antibodies against murine and human PlGF and when injected into humans or in the eye, there is significant inhibition of angiogenesis. One area where the anti-angiogenic activity of anti-PlGFs differs from some others, is that there have not been any major toxicities. With several VEGF inhibitors there is pruning of normal vessels in the trachea and some other organs. That is not seen with anti-PlGF treatment, nor is there any hypertension. If you measure PAI1 levels, which may be a measure of thrombotic risk, there is an increase with anti-VEGF treatments, but not anti-PlGF. Also, anti-PlGF treatments don’t interfere with pregnancy, like anti-VEGF treatments.

Downloaded from iovs.arvojournals.org on 09/30/2021 41

Tumor Angiogenesis

Supplement 16 to Campochiaro PA, First ARVO/Pfizer Institute Working Group. Ocular versus Extraocular Neovascularization: Mirror Images or Vague Resemblances. Invest Ophthalmol Vis Sci. 2006;47:462-474.

Donald McDonald

University of California, San Francisco, CA

The formation of normal blood vessels is very complex. The appropriate factors are needed in the appropriate amounts and at the right time. That is one reason why blood vessels formed during pathological neovascularization are abnormal. These abnormalities provide targets for specific treatments. It is important to define the abnormalities of tumor blood vessels and determine the effects of angiogenesis inhibitors that exploit these abnormalities. The elements of tumor blood vessels are endothelial cells, pericytes, and extracellular matrix in the form of basement membrane that envelopes endothelial cells and pericytes. Endothelial cells in tumors have abnormal gene expression. Abnormalities in intercellular junctions lead to vessel leakiness. The vessels are also sprouting and active in ways that normal endothelial cells are not. Importantly, a dependency on growth factors provides the rationale for treatment with growth factor antagonists. Pericytes are abnormal in tumor vessels; they have an unusual association with endothelial cells. The vascular basement membrane is also abnormal in tumor vessels, having multiple layers and exposed cryptic sites. Normal blood vessels in all organs have a hierarchical organization, with arterioles branching down to capillaries, which in turn are connected to venules that become progressively larger and eventually connect to veins. This hierarchy is missing in tumor vessels. Blood vessels of different sizes are interconnected in a seemingly random fashion; none has characteristics of arterioles, capillaries, or venules. The endothelium of tumor vessels is so thin in some regions that erythrocytes are visible through the wall. Filopodia are abundant on the external surface of endothelial cells. Filopodia are normally most common at the tips of growing vessels, but in tumors they occur not only at the tip but also scattered over the vessel surface in a disorganized pattern. Tumor blood vessels have an abnormal pattern of protein expression. Injection of antibody to α5β1 shows the expression of this integrin on tumor vessels, but not normal vessels. Expression of the integrin on the luminal surface as well as abluminal surface of tumor vessels indicates loss of endothelial cell polarity. Pericytes on normal vessels form a cable overlying endothelial cells and are very intimately associated, particularly at junctions. This is not the case in tumor vessels, where pericytes have a loose association. What about the effect of angiogenesis inhibitors? The effect on tumor vessels of a variety of VEGF antagonists was examined to determine the effects of this class of agent on the normal microvasculature of adult mice. The inhibitors examined included: (1) adenoviral vectors that express soluble VEGFR1 or VEGFR2 (Calvin Kuo, Stanford), (2) VEGF Trap (Regeneron), (3) DC101 antibody directed against VEGFR2 (ImClone), (4) a tyrosine kinase inhibitor that blocks VEGFR1, VEGFR2, and VEGFR3 and at a 10-fold higher concentration blocks PDGF receptors (Pfizer). Not only are many tumor vessels dependent on VEGF for survival, but the normal microvasculature also has some distinct properties with respect to VEGF

Downloaded from iovs.arvojournals.org on 09/30/2021 42

dependency in adult mice. Each microvascular bed in the body has a different degree of VEGF- dependency, ranging from none (independent) to fairly dependent. An example of the latter is the thyroid gland in which about half the capillaries in adult mice are VEGF-dependent. This is a dose- and time-dependent phenomenon. After treatment with a VEGF antagonist for about 3 weeks, about half the capillaries in the thyroid have regressed. It appears that capillaries that are fenestrated are particularly sensitive to VEGF inhibitors in the adult. About 20% of the capillaries in the trachea of adult mice are VEGF-dependent. In the immediate postnatal period, 100% of tracheal capillaries are VEGF-dependent, but the proportion gradually drops to 20% by 16 weeks of age. The tracheal microcirculation of mice lends itself to detailed analysis of the effects of VEGF antagonists because the vasculature is highly organized, distributed as a monolayer, and readily viewed in whole mount preparations. Injection of a fluorescent lectin into living mice shows which vessels are patent, because blood flow is required to carry the lectin to the vessels it labels. Under baseline conditions, all tracheal blood vessels are painted with the lectin as shown by the tight correlation between lectin staining and postmortem staining with CD31, which stains all endothelial cells. Two days after the onset of treatment with a VEGF antagonist, there is mismatch between lectin and CD31 staining, indicating that some vessels are not perfused. By 10 days, some capillaries are neither perfused with the lectin nor stained with CD31. Apoptotic endothelial cells, identified by activated caspase 3 immunoreactivity are present in regressing capillaries. The sequence seems to be, first, loss of vascular lumen, followed by apoptosis of endothelial cells, and then disappearance of endothelial cells, without loss of pericytes or the vascular basement membrane. In pancreatic islet tumors in RIP-Tag2 transgenic mice, all vessels have both lectin and CD31 staining under baseline conditions, although there is some variability because of heterogeneous blood flow. Twenty-four hours after onset of treatment with a VEGF antagonist, many tumor vessels are not perfused, but CD31 staining is little changed. By 2 days after initiation of treatment, there is a significant reduction in both lectin staining and CD31 staining, with greater reduction in perfusion. At 7 days, there is a match again, with about 70% reduction in both, indicating that all the remaining vessels are perfused. These findings show that the pruning occurs very quickly, first by vessel closure and then by apoptosis and removal of endothelial cells. Once flow is lost, the stimulus for vessel maintenance is lost and the vessels regress. In the trachea, at baseline there is complete colocalization of endothelial cells and basement membrane, but after 10 days of treatment, about 20% of basement membrane sleeves have no endothelial cells. The same thing happens in tumors, but a much larger proportion of basement membrane sleeves are left behind when vessels regress. Once treatment is stopped, explosive vascular regrowth occurs. The regrowth appears to take advantage of existing basement membrane sleeves, which act like railroad tracks. As long as basement membrane sleeves persist and the VEGF drive for regrowth remains, the entire microcirculation of tumors can be reconstructed within a few days. PDGF is an essential growth and survival factor for pericytes. In tumor vessels, many endothelial cells are sensitive to VEGF inhibition, but most pericytes are not. In RIP-Tag2 tumors treated with a VEGF antagonist for 7 days, the microcirculation is pruned. Staining with α-smooth muscle actin (αSMA), which stains pericytes, shows that pericytes are not just located on the surviving tumor vessels but are also associated with strands of basement membrane that lack endothelial cells. Basement membrane and pericytes appear to be key elements for vascular regrowth.

Downloaded from iovs.arvojournals.org on 09/30/2021 43

The effects of several PDGF antagonists have been examined: (1) an anti-PDGF aptamer (Archemix), (2) adenoviral vectors expressing soluble PDGF receptors (Calvin Kuo, Stanford), (3) Glivec, a kinase inhibitor that blocks PDGF receptors, abl and c-kit. Systemic delivery of the anti-PDGF aptamer to mice with implanted Lewis lung carcinomas resulted in rapid decrease in number of pericytes with the tumors. A 50% reduction in pericytes on tumor vessels after 24 hours progressed to about 80% reduction after 4 weeks. About 4 days after loss of pericytes begins, endothelial cells start to die. By 4 weeks, there was equal reduction of pericytes and endothelial cells. It appears that many endothelial cells are dependent upon pericytes in this particular tumor. One reason is that the pericytes are a source of VEGF. As the tumor cells themselves do not make a lot of VEGF, loss of pericytes eliminates a major source of VEGF. Three other markers for pericytes, NG2, desmin, and PDGFRβ, have been examined in addition to αSMA. In Lewis lung carcinomas treated with anti-PDGF-B aptamer, all of the markers show about the same 80% reduction over 4 weeks of treatment. Disappearance of pericytes, not just a change in phenotype, was confirmed by electron microscopy. Because pericyte phenotype can change after treatment with VEGF antagonists, loss of a single marker, such as αSMA, may reflect downregulation of that molecule rather than loss of pericytes. In addition, pericytes that remain may undergo normalization as reflected by closer association with endothelial cells. A similar phenomenon is seen after treatment with VEGF antagonists, where many tumor vessels are eliminated and the remaining ones appear more normal, both by overall geometry and by close association with endothelial cells. In summary, endothelial cells, pericytes, and vascular basement membrane are all abnormal in tumor vessels. Treatment with a VEGF inhibitor results in rapid pruning of susceptible blood vessels, leaving a population of VEGF-resistant blood vessels behind. Blood vessel pruning also leaves behind the pericytes and basement membrane of the pruned vessels. When the VEGF antagonist is stopped, tumor vessels rapidly regrow along the remaining sleeves of basement membrane and pericytes. A second round of VEGF inhibition can take out the regrown tumor vessels, indicating that the new vessels are VEGF-dependent. PDGF antagonists can reduce the population of pericytes and secondarily the population of endothelial cells in susceptible tumors. Still unanswered is the question of whether the entire vasculature of some tumors can be eliminated by a combination of VEGF and PDGF antagonists.

Questions

1. Is Lewis Lung carcinoma a special target for pericytes? It is unclear whether Lewis lung carcinoma is a special case, but it clearly differs from RIP-Tag2 tumors. Like beta cells in normal pancreatic islets, tumor cells in RIP-Tag2 tumors are VEGF factories. PDGF inhibitors by themselves have very little effect on the microvasculature of RIP-Tag2 tumors. Some pericytes are removed, but most endothelial cells survive. Apparently tumor cell production of VEGF in these tumors is sufficient to support the survival of the endothelial cells and hence the microvasculature. An important, unanswered question is, “Do tumor vessels become more vulnerable to VEGF inhibitors after their pericytes are lost?”

2. Does loss of pericytes eventually result in vessel occlusion? That is not yet known.

Downloaded from iovs.arvojournals.org on 09/30/2021 44

3. Have you looked at early time points to determine why VEGF antagonists result in occlusion of tumor vessels? Vascular occlusion is not simply the consequence of endothelial cell swelling. Loss of vessel patency may result from loss of a signal that keeps the lumen open. It is important to identify the molecular basis of what keeps the lumen open and what changes to result in closure, but it is not known at this time.

4. Given the importance of integrin binding in regulating growth factor signaling, in terms of specificity and extent of the signal and the upregulation of α5β1, do you see any functional role for α5β1 in VEGF antagonist-induced vascular regression? It is a reasonable possibility, but we don’t have a function blocking antibody for mouse α5β1 to be able to answer that question.

5. Do you see upregulation of other integrins on the cell surface? There is upregulation of αvβ3 on endothelial cells of many tumors, but it is much more spotty than α5β1 in RIP-Tag2 tumors.

6. The same sequence of events occurs in regression of hyaloid vessels during development. The vessels narrow, become occluded, and then there is synchronous apoptosis and loss of endothelial cells. Then macrophages come in and remove the basement membrane. Do you know the status of macrophages in the tumors you have studied? Macrophages are present in the tumors we have studied, but their association with blood vessels has not been systematically examined. This needs to be done. We have followed tumors treated with VEGF antagonists for as long as 4 weeks. Basement membrane strands are still present at that point, so indicative of slow clearance of the basement membrane sleeves. I suspect that the process of vessel regression in tumors is similar to that occurring elsewhere. It will be important to identify the mechanism of vessel closure after inhibition of VEGF signaling. Fibrin accumulates within the lumen of some regressing vessels before closure, but it is unknown whether this is a primary or secondary event. One possibility is that phospholipids (phosphatidylserine) in the luminal plasma membrane of regressing endothelial cells flip in polarity, making the luminal surface thrombogenic.

7. Do pericytes of normal vessels depend on PDGF for survival? This does not seem to be the case under baseline conditions. Subtle changes may occur, but few pericytes are lost in the normal organs we have examined.

Downloaded from iovs.arvojournals.org on 09/30/2021 45

Wound Repair and Angiogenesis in Skin

Supplement 17 to Campochiaro PA, First ARVO/Pfizer Institute Working Group. Ocular versus Extraocular Neovascularization: Mirror Images or Vague Resemblances. Invest Ophthalmol Vis Sci. 2006;47:462-474.

Michael Detmar

Swiss Federal Institute of Technology, Zurich, Switzerland

In skin there is avascular tissue (epidermis) overlying richly vascularized tissue (dermis) that also contains many lymphatics. Two major topics that will be addressed are (1) the role blood vessels as important targets for anti-inflammatory therapy and (2) recent aspects of tumor lymphangiogenesis and in particular a new concept of lymph node lymphangiogenesis and its contribution to tumor spread. Psoriasis is an inflammatory skin disease with unknown etiology characterized by epidermal hyperplasia consisting of finger-like projections, chronic inflammation, and pronounced angiogenesis. VEGF is highly upregulated in the epidermis and VEGF receptors are upregulated in the dermis. There are two independent mechanisms of angiogenesis in psoriasis. (1) Epidermal keratinocytes hyper-proliferate in psoriasis (but also in wound healing and squamous cell carcinoma) and they release growth factors of the EGF family such as TGFα or amphiregulin, which in an autocrine way stimulate the cells to release VEGF and PlGF. The VEGF and PlGF then stimulate cutanous vessels. This mechanism makes sense, because whenever you have proliferation of an avascular tissue, you need a way to ensure enhanced vascular support. (2) Direct hypoxia-induced release of VEGF and PlGF. So VEGF and PlGF are upregulated in psoriasis, but can the phenotypic characteristics of psoriasis be induced by persistent activation of the dermal vessels? Transgenic mice that over- express PlGF or VEGF under control of the K14 promoter were generated. These mice did not develop the chronic inflammation of psoriasis. The major phenotype was enhanced acute inflammation. In a delayed type hypersensitivity model, mice are made allergic and then challenged leading to inflammation and edema of the ear. The swelling of the ear is measured as a parameter of the amount of inflammation. In a normal mouse, there is initial inflammation that peaks at 48 hours and within a week it resolves. In PlGF overexpressors, there was increased inflammation and edema, but just as in normal mice, it did not persist. In contrast, in VEGF overexpressors, there is persistent inflammation; they are completely unable to downregulate inflammation once it has been induced. The epidermis becomes thicker and has fingerlike projections something like that seen in psoriasis. Homozygous VEGF transgenic mice develop spontaneous inflammatory lesions at 6 months of age that look very much like psoriasis with inflammation, induration, scaly skin. Histologically they are indistinguishable from psoriasis with hyperplastic epidermal keratinocytes with fingerlike projections, inflammation with CD4 cells in the dermis and CD8 cells in the epidermis, neutrophils, and prominent angiogenesis. This suggests that chronic overexpression of VEGF in the skin is sufficient to generate a psoriasis phenotype. It is also necessary to maintain the disease, because treatment with VEGF- trap results in normality within 2 weeks.

Downloaded from iovs.arvojournals.org on 09/30/2021 46

There is also genetic data implicating VEGF in psoriasis. The Vegf gene is located near the major psoriasis susceptibility locus. There have been several SNPs described in the VEGF gene including a G to C polymorphism at 634 in the 5’ UTR, which has been reported to be associated with enhanced VEGF levels in normal individuals and the frequency of this genotype is significantly increased in psoriasis patients that have severe disease. There are certain SNPs or haplotypes that can predict whether a patient will respond to retinoid therapy. The hypothesis is that there is an angiogenic constitution that might determine susceptibility to chronic inflammation and also response to retinoids. Does that apply just to the skin disease or to other chronic inflammatory diseases as well? VEGF is not the only agent that contributes. The VEGF transgenics were crossed with PlGF KO mice. Mice that are VEGF transgenic and wild type for PlGF have persistent inflammation in the skin, while VEGF transgenics that lack PlGF have inflammation that is a bit prolonged compared to wild type mice, but it goes down to background levels within 2-3 weeks. There are 2 SNPs in the PlGF gene that are associated with severe psoriasis and there is a possibility of genetic interaction between the VEGF and PlGF genes in patients with psoriasis. There are certain combinations of SNPs that are significantly increased in early onset or more severe psoriasis. So both VEGF and PlGF may be good targets for treatment in psoriasis and other chronic inflammatory diseases. Inflammation was induced in the skin of wild type mice and then they were treated systemically with an antibody directed against VEGF receptor 1 or 2 (Imclone), or both together. Either antibody alone did not have a significant effect, but both together significantly inhibited the inflammatory response. Blocking lymphangiogenesis in this model is also beneficial.

How tumors metastasize to lymph nodes The traditional model of tumor progression is that tumors induce angiogenesis to promote their growth and this increases metastasis, but how do tumor cells get into lymph nodes? The dogma has been that lymphatic vessels play a passive role and tumor cells happen upon these pre- existent lymphatics and then are washed to the lymph nodes. It was almost impossible to image the lymphatic vessels because markers weren’t available and lymphangiogenic factors were unknown. Now it is known that at a certain stage lymphatics develop from pre-existent embryonic veins. The transcription factor PROX1 is switched on by unknown stimuli, and this results in cells budding off from the veins and migrating and during this process the cells lose expression of blood vessel-specific genes and begin expressing lymphatic-specific markers: live1, VEGF receptor 3, and PROX1. Lymphatic endothelial cells can be cultured and they maintain their lineage-specific differentiation over many passages. Transcriptional profiling of these cells compared to vascular endothelial cells can help to reveal other lymphatic-specific genes. Both the Affimetrix platform and Applied Biosystems Gene Array platform gave comparable results. Some previously known genes such as VEGF receptor 3 and PROX1 were upregulated, but many novel genes were also upregulated. Vascular endothelial cells can be reprogrammed by infection with Kaposi sarcoma virus carrying lymphatic-specific genes. VEGF-C and -D are relatively specific lymphangiogenic growth factors that work through VEGF receptor 3 on lymphatic endothelial cells. In an orthotopic breast cancer model, lymphatic vessels can be seen to proliferate and spread into tumors. The degree of lymphangiogensis correlates with the amount of metastasis, suggesting it might be a prognostic feature and a therapeutic target. It has been suggested that

Downloaded from iovs.arvojournals.org on 09/30/2021 47

this could be an artifact of xenografts and that human tumors do not have lymphangiogenesis. Malignant melanomas metastasize very early to lymph nodes. Blood vessels occur homogenously throughout tumors, but lymphangiogenesis occurs in hot spots. Lymphangiogenesis seems to be a sensitive prognostic feature for future metastasis of melanomas. Primary tumors were graded for amount of lymphangiogenesis and three groups were identified: (1) no lymphangiogenesis, (2) abundant lymphangiogenesis, and (3) moderate lymphangiogenesis. The results were striking. None of the group 1 patients developed lymph node metastasis over the next 10 years and were still alive, compared to metastasis to lymph nodes in all of the patients in group 2 and only 10-15% survived at 10 years. So lymphangiogenesis does occur in humans and has prognostic significance. VEGF-C and -D can induce lymphangiogenesis mostly surrounding tumors but sometimes inside of tumors. The big question mark is the role of VEGF-A in this process; it has been considered just an angiogenic factor, but lymphatic vessels express VEGF receptor 2. Transgenic mice that express GFP under control of the K14 promoter have green skin. When skin cells become carcinogenic in a skin carcinogenesis model, they remain fluorescent, and they are still fluorescent after metastasis. It is very easy to detect metastasis and can also homogenize lymph nodes and quantify metastasis. The GFP mice were crossed with VEGF transgenics. In wild type mice, angiogenesis occurs very early in benign tumors and it was enhanced when VEGF was overexpressed, but what was surprising was that there were many more lymphatics in the tumors and they were proliferating as shown by BRDU labeling. So VEGF-A is a lymphangiogenic agent. All of the VEGF overexpressors developed metastasis to regional lymph nodes. These VEGF-producing tumor cells continue to promote lymphangiogenesis in the lymph nodes, which results in more metastasis to distant sites. The tumors induce lymphangiogenesis in the draining lymph nodes even before they metastasize, which is a new twist to the seed and soil hypothesis. The seed can actively modify the soil to enhance the chances of metastasis. So there is lymphangiogenesis in the primary tumors promoting metastasis to the sentinel lymph nodes where the tumor cells can cause more lymphangiogenesis promoting metastasis to distant lymph nodes and organs.

Questions

1. Can you exclude two confounding variables in the VEGF-A skin carcinogenesis model? You showed previously that VEGF-A leads to chronic inflammation and in the carcinogenesis model you paint with TPA, which is a very strong irritant and causes the release of many inflammatory agents. It would be expected that the VEGF-A mice would have much more inflammation and inflammatory cells are sources of lymphangiogenic stimulators (VEGF C and other agents). Secondly, in the metastasis when you have VEGF-A producing tumor cells, which grow faster, how can you separate the role of lymphangiogenesis from the advantage that these cells have in metastasis. For the first question, TPA does induce some inflammation. We looked at inflammatory cells in tumors from VEGF-A transgenics and wild type mice and we didn’t see major differences in CD11b-positive macrophages, CD4-positive lymphocytes, and others. We also performed in situ hybridization for VEGF-C and –D and the major source of VEGF-C was tumor cells or epidermal keratinocytes, but do not see any difference in stromal cells or inflammatory cells VEGF-C in VEGF-A overexpressing mice compared to wild type. So with the tools that

Downloaded from iovs.arvojournals.org on 09/30/2021 48

we have, we cannot see that there is enhanced inflammation in the VEGF-A overexpressors in this model. For the second question, when the cells are cultured, the VEGF-A expressing tumor cells do not proliferate faster than the wild type tumor cells. When you look at these lymph nodes containing metastases and you see these induced lymphatic channels inside of lymph nodes, it is suggestive that it may facilitate distant spread.

2. You showed that lymphatic density predicts melanoma progression. Is that independent of vessel density? Yes. We did multivariant analysis.

3. If you treat melanomas with VEGF inhibitors, you would predict that lymphatics would also respond- what have you seen in that setting? In fact, melanoma may not be very VEGF-dependent. It may be more driven by FGF2 or other factors. We don’t think the lymphangiogenesis in melanoma is VEGF-A driven. We think there is another lymphangiogenic factor overexpressed in melanoma.

Downloaded from iovs.arvojournals.org on 09/30/2021 49

Role of VEGF in Maintenance of the Pulmonary Microcirculation and the Etiology of Emphysema

Supplement 18 to Campochiaro PA, First ARVO/Pfizer Institute Working Group. Ocular versus Extraocular Neovascularization: Mirror Images or Vague Resemblances. Invest Ophthalmol Vis Sci. 2006;47:462-474.

Rubin Tuder

Johns Hopkins University, Baltimore, MD

Alveoli are uniquely designed to maximize exposure of alveolar capillaries to oxygen. Destruction and injury to alveolar septa may result in emphysema. There is a high prevalence of emphysema, with 3 million affected patients in the world and 120,000 in the United States alone. Smoking is the most greatest factor for emphysema. The fundamental problem in emphysema is destruction of alveolar structures resulting in reduction of surface area for exchange of gases. When the pulmonary arteries are outlined using a polymer cast, one realizes that the destroyed alveolar units follow the destruction of pulmonary vessels (also known as the vascular hypothesis of emphysema). The realization that endothelial cells play a central role in the organization and function of alveoli led us to the hypothesis that collapse of the circulation may be the basis for the loss of lung tissue and that smoking leads to the death of blood vessels by downregulation of VEGF and/or VEGF receptors. The lung has high levels of VEGF that can be upregulated by hypoxia. Using low melt agarose injected under constant physiological pressure, a gel forms that preserves the alveolar structure, providing good specimens for histology. The alveoli have three types of cells organized around the alveolar septae and myofibroblasts. Myofibroblasts form little bridges with pericytes or endothelial cells, or they attach to epithelial cells and communicate signals so if something happens to endothelial cells it can affect the other cells. After 3 weeks of VEGF receptor inhibition, there is loss of alveolar tissue and this is associated with apoptosis of cells in the alveolar septum. Treatment with a broad spectrum apoptosis inhibitor prevents the loss of tissue. The paradigm of apoptotic loss of alveolar cells, including endothelial cells, has been recently extended to the rodent cigarette smoke model of emphysema. This observation caused a paradigm shift in the thinking regarding the pathogenesis of emphysema. The traditional hypothesis is that smoking causes oxidative damage that results in upregulation of proteases that dissolve the alveolar tissue. In adult mice, blockade of both VEGFR1 and R2 results in a marked reduction in alveolar tissue. The mechanism of cell death involves a lipid mediator, ceramide. It is the balance between the proapoptotic ceramide and the pro-survival metabolite of ceramide, sphingosine-1- phosphate that determines whether the cells live or die. Ceramide is hydrophobic lipid that can be synthesized in two ways. One is by condensation of serine and palmityl CoA in mitochondria and microsomes. There is a sequential action of two enzymes, serine palmityl transferase and ceramide synthase. The second is by the action of sphingomyelinase (enzyme that is absent in Niemann-Pick disease). Sphingomyelinase takes an abundant lipid in the cell membrane, sphingomyelin, and generates ceramide.

Downloaded from iovs.arvojournals.org on 09/30/2021 50

The sequence of events appears to be that VEGFR blockade for even a few days results in activation of ceramide synthase, which produces ceramide. This early increase in ceramide stimulates sphingomyelinase to produce more ceramide. If ceramide synthase is blocked and ceramide is reduced, and supplemental sphingosine-1-kinase is given, the lung tissue is preserved. If ceramide is instilled directly into lung, alveolar destruction and emphysema results. There is a second wave of ceramide synthesis from sphingomyelinase. Application of a soluble form of ceramide to cultures of fibroblasts from sphingomyelinase null cells results in lower production of ceramide and less apoptosis than seen in wild type cells. Using phage display, a peptide, lung homing peptide (LHP), that selectively binds to lung vascular endothelial cells, was identified. The peptide binds to the cell surface and is internalized. The peptide was linked to a pro-apoptotic molecule and when this is injected systemically in a mouse, it results in disruption of the lung endothelial cells and reduced regenerative capacity. In summary, the following pathogenic scheme is proposed for emphysema. Smoking induces oxidative stress and inflammation, but also blocks VEGF signaling which promotes endothelial cell death through a ceramide-dependent pathway, which results destruction of lung tissue.

Questions

1. Is there any evidence of lung damage in any of the patients who have been on systemic anti- VEGF medications in cancer trials? Some of these patients have been treated for more than a year. Are pulmonary function tests done? It is not known and pulmonary function tests are not done.

2. There are recent studies in the eye implicating ceramide in cell death. In arrestin KO mice and some models of retinitis pigmentosa, increased levels of ceramide have been associated with photoreceptor cell death. In RPE cells, ceramide causes cell death by causing oxidative damage to mitochondria and there is a secondary amplification through down-regulation of catalase. Ceramide has also been implicated in cell death in Alzheimer’s disease.

Downloaded from iovs.arvojournals.org on 09/30/2021 51

Angiogenesis in Rheumatoid Arthritis

Supplement 19 to Campochiaro PA, First ARVO/Pfizer Institute Working Group. Ocular versus Extraocular Neovascularization: Mirror Images or Vague Resemblances. Invest Ophthalmol Vis Sci. 2006;47:462-474.

Alisa Koch

University of Michigan, Ann Arbor, MI

Rheumatoid arthritis (RA) is large social and economic burden in the United States with an estimated annual cost of 9 billion dollars. The average annual cost per patient is $19,000, including the indirect costs of loss of productivity. So it is a major health problem. The histology of normal synovium shows that it is very loose tissue. Rheumatoid synovium is much denser, because it is packed with blood vessels often surrounded by leukocytes. Important stimulators of angiogenesis in RA include vascular endothelial growth factor (VEGF) and several proangiogenic cytokines such as interleukin-8, tumor necrosis factor- α, and members of the CXC chemokine family. VEGF-A, -C, and –D have been identified in joints from RA patients and high serum levels of VEGF or FGF-2 are correlated with poor prognosis. Induction of VEGF may be by transforming growth factor-β (TGF-β), IL-1 and hypoxia in RA joints. In a rodent model of RA, preventive administration of fumagillin (TNP- 470) was found to reduce serum VEGF and vascularity in the synovium. In mouse collagen- induced arthritis, soluble VEGFR1 or anti-VEGFR1 reduce inflammation and new vessels in joints. Cultured human synovial fibroblasts have been used to study how VEGF expression might be regulated in joints. IL-1, TNF-α, and TGF-β each increase VEGF expression by synovial fibroblasts. Three independent studies have shown that antagonists of αvβ3 integrin attenuate joint inflammation and neovascularization in various models. A proapoptotic αvβ3 antagonist composed of an RGD peptide linked to a heptapeptide dimer selectively homes to mouse joints with inflammatory arthritis and not normal joints and provides therapeutic benefit. It had been hypothesized that soluble adhesion molecules such as sE-selectin might bind leukocytes and reduce inflammation in joints; however it was found that soluble E-selectin and soluble vascular cell adhesion molecule-1 each stimulate angiogenesis in joints; sE-selectin acts through sialyl Lewisx (Lex) on endothelial cells and sVCAM-1 does so through VLA-4. Cells were isolated from the joints of RA patients and only a subset of macrophages was found to promote angiogenesis. Monoclonal antibodies (mAbs) to the cells were generated and many were found that only recognized macrophages, but one recognized only endothelium and epithelium. Further characterization showed that it reacted with select endothelium in synovium, skin, and lymph nodes, but did not detect myeloid cells. The antigen, 4A11, is up-regulated on cultured endothelium by cytokines, such as TNFα, within minutes, but by 2 hours it is back to baseline. In vivo, there is some basal expression on endothelium in skin, and expression is increased 6-24 hours after administration of uroshiol, poison ivy extract. Compared to normal synovium, 4A11 is more highly expressed in RA synovium. It was found that mAb 4A11 recognizes the related glycoconjugates Lewisy (Ley) and H-2.

Downloaded from iovs.arvojournals.org on 09/30/2021 52

Lewis glycoconjugates, Lex, Ley, and H-2 constitute the chemical basis for several blood group systems, but they also have been implicated in other processes. They are expressed by some carcinomas and have prognostic significance. They are expressed in gastric mucin and by Helicobacter pylori, constituting molecular mimicry that can lead to autoimmune gastritis if antibodies generated against H. pylori also recognize gastric mucin. Soluble forms of the antigens, H-2-glucose (H-2g) and Ley- glucose (Leyg), which are critical components of soluble 4A11, were tested for angiogenic activity. They stimulated chemotaxis of cultured endothelial cells and stimulated angiogenesis in corneal pocket assays, which was blocked by mAb 4A11. Synovial fluid from patients with RA were found to have substantially higher levels of soluble 4A11 than synovial fluid from patients with osteoarthritis. So, soluble 4A11 reacts with an unknown receptor on endothelial cells to stimulate angiogenesis and is found in high levels in the joints of patients with RA. Soluble 4A11 is also pro- inflammatory; it is highly potent in recruitment of leukocytes after intraperitoneal injection. It also mediates adhesion by upregulating ICAM-1 on endothelium. It signals in endothelium via a JAK-STAT pathway. It seems to be involved in joint-specific homing. The action of the pro-angiogenic agents in joints is balanced by inhibitors. The inflamed synovium is adjacent to avascular cartilage. Marsha Moses has shown that a cartilage-derived factor Troponin-I has antiangiogenic activity. In collaboration with Olga Volpert and Noel Bouck, it was noted that IL-4 is strongly angiostatic in the cornea and this effect could be blocked by antibody to IL-4. Adenoviral vector (Ad)-mediated gene transfer of IL-4 to the ankle joint of rats was followed by M. batrycium injection, which causes arthritis. Compared to null vector-injected joints, those injected with AdIL-4 showed less swelling and inflammation. Gene transfer to joints of thrombospondin, angiostatin, endostatin, and troponin-I each cause significant benefit in arthritis models.

Downloaded from iovs.arvojournals.org on 09/30/2021 53

Developmental versus Pathologic Retinal Neovascularization

Supplement 20 to Campochiaro PA, First ARVO/Pfizer Institute Working Group. Ocular versus Extraocular Neovascularization: Mirror Images or Vague Resemblances. Invest Ophthalmol Vis Sci. 2006;47:462-474.

Marcus Fruttiger

Institute of Ophthalmology, University College London, UK

Developmental retinal angiogenesis occurs quite late in rodents, starting at postnatal day (P) 0 when vessels emerge from the optic nerve. In situ hybridization using probes for vascular endothelial growth factor receptor (VEGFR) 1 or 2, which label endothelial cells, platelet- derived growth factor receptor (PDGFR) β, which labels pericytes, and PDGFRα, which labels astrocytes, shows astrocytes leading the way while endothelial cells and pericytes follow together. Astrocytes are crucial for retinal vascularization, because they express VEGF and provide a template for vascular development. In situ hybridization for VEGF shows that it is most strongly expressed by astrocytes in the not-yet-vascularized part of the retina. This part of the retina is hypoxic as revealed by EF5 staining. (EF5, injected in live animals, is reduced by an oxygen-inhibitable reductase resulting in permanent protein adducts that, after tissue fixation, can be stained with an antibody to visualize hypoxic tissue.) The central, vascularized part of the retina displays higher oxygen levels and VEGF mRNA is downregulated in this area. There is a sharp transition from high to low VEGF mRNA at the leading edge of the growing vascular network, which is likely to result in a VEGF protein gradient at this location. Endothelial cells at the leading edge, so-called tip cells, display a number of features not found on other endothelial cells (stalk cells). They are distinguished by high expression of PDGF-B compared to other endothelial cells (this stimulates pericytes to migrate and cover the growing vessels). Another marker for tip cells is the notch ligand delta like 4 (Dll4). Tip cells also differ in morphology from other endothelial cells. They have long filopodia, similar to growth cones on axons, which might be important in sensing VEGF gradients. The filopodia are eliminated by soluble VEGFR1 suggesting that they are sensitive to VEGF levels. When the eye is injected with VEGF164, flooding all VEGF gradients, filopodia are also disturbed and the expansion of the vascular network is halted. In summary, during development, negative feedback between VEGF expression and vessel mediated tissue oxygenation leads to a VEGF gradient. Tip cells at the distal end of growing vessels read the gradient and guide vessel growth. How does this scenario compare to pathologies such as diabetic retinopathy where hypoxia and VEGF expression lead to retinal neovascularization? Oxygen-induced ischemic retinopathy (OIR) in mice can serve as a suitable model to study this. In this model one-week- old mice are exposed to hyperoxia which causes regression of newly developed, VEGF- dependent retinal vessels. This primarily occurs around arteries because oxygen can diffuse directly through the vessel walls into surrounding tissue and downregulate VEGF in these regions. In the center of the retina around the central hyaloid artery this effect is most

Downloaded from iovs.arvojournals.org on 09/30/2021 54

pronounced and leads to a large capillary-depleted area. When mice are placed back into normoxia, this capillary-free zone becomes hypoxic, visualized by intense staining with EF5, and increases VEGF production. Two days after return to room air the localized high expression of VEGF in the avascular areas results in initial sprouting from remaining vessels but fails to induce successful revascularization of the capillary-free zones. Instead, vessels start to form neovascular tufts towards the vitreous. Why is the hypoxic, VEGF expressing region not vascularized normally as it occurs during development in the peripheral retina? Could it be that pericytes surrounding new vessels express high levels of VEGF promoting uncontrolled growth of sprouts? This is not what is seen. Instead the greatest VEGF expression is by astrocytes. Astrocytes survive the hyperoxia and hypoxia, although GFAP gets downregulated. The receptors for VEGF are expressed in the endothelial cells in the tufts. Are tip cell markers expressed in the tufts? In situ hybridizations for PDGFB or Dll4 show that they are expressed in tufts, which indicates that tip cells are present, but their location is abnormal in that they are not located adjacent to the avascular areas where the revascularization is supposed to occur. Is there a VEGF gradient? It is very difficult to show at the protein level, but one hypothesis to explain the improper orientation of tip cells in tufts is that the VEGF gradient is disturbed. The following four observations support this view: 1) When mice are placed into10% oxygen at P4, they become globally hypoxic and there is increased expression of VEGF in the retina. These increased levels of VEGF, which alter the normal gradients, result in vascular growth where it is not usually seen, such as early sprouting downward to form deep capillaries.

2) Mice that are genetically engineered to express only VEGF121 and not the other VEGF isoforms, have reduced survival, but those that survive show neovascular tufts in the retina. These mice have VEGF that cannot localize properly to form normal gradients, and the result is abnormal sprouting to form neovascular tufts. 3) If mice are put in hyperoxia at P7 and kept there instead of being returned to room air at P12, the large avascular areas gradually revascularize without neovascular tufts. While gradients may be flattened in this situation, they are not completely eliminated. 4) Op/op mice have a mutation in colony stimulating factor 1 (CSF-1) resulting in CSF-1 deficiency. This leads to a lack of macrophages. In the OIR model, these mice revascularize the avascular areas without development of neovascular tufts. A hypothesis that could reconcile these observations is that VEGF gradients are needed for normal vascular development. If the gradients are flattened, vascular development is delayed, but still occurs without abnormal sprouting. But if gradients are severely disturbed or eliminated, then abnormal sprouting occurs. Macrophages contribute to disturbances of gradients in hypoxic retina. In their absence, the increased expression of VEGF by astrocytes and other retinal cells may not disturb the gradient sufficiently to promote abnormal sprouting.

Questions 1. Op/op mice get revascularization without sprouting. Is it possible that in the absence of macrophages there is no degradation of the basal lamina, which allows more rapid revascularization? The collagen tubes are gone within 48 hours in all mice in the hyperoxia model.

Downloaded from iovs.arvojournals.org on 09/30/2021 55

2. Has anyone looked at whether the tip cells are associated with a specific MMP, like MMP9? Holger Gerhardt has been investigating this and finds that it is not the tip cells that express MMPs, but macrophages. This helps to liberate VEGF from the ECM and may be a mechanism by which macrophages contribute to disturbance of the VEGF gradient. 3. Are expression of PDGF-B and Dll4 in tip cells dependent upon each other? It is not definitely established, but there is some evidence suggesting that PDGF-B is downstream of Dll4. 4. How do tip cells find each other? They don’t find each other; the stalk cells proliferate and the branching occurs behind the tip cells.

Downloaded from iovs.arvojournals.org on 09/30/2021 56

Ocular Neovascularization

Supplement 21 to Campochiaro PA, First ARVO/Pfizer Institute Working Group. Ocular versus Extraocular Neovascularization: Mirror Images or Vague Resemblances. Invest Ophthalmol Vis Sci. 2006;47:462-474.

Peter Campochiaro

Johns Hopkins University, Baltimore, MD

There are many indications in the literature suggesting that some investigators view angiogenesis as a generic process that is the same in all tissues, pathologies, and stages of development. What is the evidence for an alternative view? One piece of evidence stems from transgenic mice in which Rhodopsin promoter drives expression of VEGF in photoreceptors.1,2 The Rhodopsin promoter begins expression at postnatal day (P) 7 and by P10 endothelial cells begin to migrate from the deep capillary bed into the outer nuclear layer, and by P14 there are vessels extending from the deep capillary bed to the subretinal space. At later time points there is an extensive network of new vessels throughout the subretinal space. Why is it that only the deep capillary bed participates in the sprouting? Is it that the VEGF does not get to the other capillary beds in sufficient amounts? Is it that there is a developmentally regulated gene product that is expressed only in the region of the deep capillary bed that sensitizes it to effects of VEGF? To address these questions, the tetracycline-inducible promoter system was used.3 In this system two sets of transgenic mice are generated, the driver line and the effector line. For driver mice, a promoter is coupled to the Reverse Tetracycline Transactivator. We generated two types of driver mice, one using the Rhodopsin promoter and another using the Interphotoreceptor retinoid binding protein (IRBP) promoter. The IRBP promoter begins expression much earlier than the Rhodopsin promoter, making them useful for assessing effects of expression between P0 and P7. For the effector line of mice, the Tetracycline Response Element was coupled to Vegf165. Driver and effector mice are crossed to yield double transgenic mice that have doxycycline-inducible expression of VEGF in photoreceptors. The two types of double transgenic mice are referred to as Tet/opsin/VEGF and Tet/IRBP/VEGF. For either of these two types of double transgenic mice, the retinas are completely normal in absence of doxycycline. However, when adult Tet/opsin/VEGF and Tet/IRBP/VEGF are treated with doxycycline, neovascularization begins to sprout from the deep capillary bed of the retina within 4 days. The phenotype is the same as the mice with constitutive expression of VEGF in the retina; neovascularization originates only from the deep capillaries of the retina. The level of expression of VEGF is higher, resulting in severe proliferative retinopathy and retinal detachment, but all of the sprouting and proliferation originates from the deep capillary bed. Thus, expression of VEGF in photoreceptors of adult mice is similar to expression of VEGF at P7. What happens if you express VEGF between birth and P7? This can be done by giving doxycycline to mothers because it is secreted in breast milk, and this is supplemented by subcutaneous injections of doxycycline in neonates. This results in increased expression of VEGF in the retina of neonatal double transgenic mice. When doxycycline is started at birth and the mice are examined at P10, there is a neonatal phenotype consisting of sprouting from the

Downloaded from iovs.arvojournals.org on 09/30/2021 57

superficial capillary bed and dilation of the superficial vessels.4 If doxycycline is started a little later, both the neonatal and adult phenotype occur; the neonatal phenotype occurs in the peripheral retina, both occur in the midperipheral retina, and the adult phenotype occurs in the posterior retina. Because development occurs from posterior to anterior, the posterior retina is at a later developmental stage than the anterior retina. This supports the presence of a developmentally regulated permissive factor. Several lines of evidence suggest that angiopoietin 2 (Ang2), the endogenous antagonist of Tie2, is this permissive factor. One line of evidence relates to the expression profile of Ang2 during development in the retina.5 Ang2 increases during the first week after birth, peaks around P8 during development of the deep capillary bed, and then decreases, but persists in adult mice. Ang2/LacZ knockin mice show the pattern of expression of Ang2.6 Between P0 and P7 there is Ang2 expression along the surface of the retina, and at P8 there is intense expression in the region of the deep capillary bed. In adult mice, there is persistent expression of Ang2 in horizontal cells adjacent to the deep capillaries in the retina. In mice with oxygen-induced ischemic retinopathy, the retina becomes hypoxic when the mice are removed from oxygen at P12 and within 6 hours, ectopic expression of Ang2 occurs along the surface of the retina. Within days, sprouts of neovascularization occur along the retinal surface and there is intense expression of Ang2 within the sprouts. Homozygous Ang2 knockouts show very poor development of the superficial capillary bed and almost no development of the deep capillary bed, as well as persistence of the hyaloid vessels. Additional insights regarding the effects of Ang2 in the retina are provided by double transgenic mice with inducible expression of Ang2 in photoreceptors, Tet/opsin/ang2 and Tet/IRBP/ang2 mice.7 Expression of Ang2 during retinal vascular development results in increased density of the deep capillary bed at P12, but it normalizes by P18. Expression of Ang2 in adult mice has no effect on normal retinal vessels. In ischemic retinopathy, induction of Ang2 expression during the hypoxic period, P12 to P17, there is a marked increase in the amount of retinal neovascularization. In this model, VEGF levels are high until around P19 and then begin to decrease. If expression of Ang2 is begun at P20, there is rapid regression of the neovascularization. In mice in which Bruch’s membrane is ruptured by laser, choroidal neovascularization develops at the rupture sites.8 If the choroidal neovascularization is allowed to grow for 7 days and then expression of Ang 2 is induced, the neovascularization regresses, whereas without induction of Ang2, the neovascularization continues to grow. ELISAs for VEGF show an initial spike in VEGF levels within a day of rupture of Bruch’s membrane and then a rapid decrease. In triple transgenic mice that express VEGF in photoreceptors and also have inducible expression of Ang2, induction of Ang2 expression at P21 results in regression of the neovascularization. This was somewhat unexpected, because the regression occurs when there is enough expression of VEGF to stimulate neovascularization from the deep capillary bed when there is no added expression of Ang2. It was predicted that the added expression of Ang2 would enhance the neovascularization, but the opposite occurred. This suggests that it is not just the presence of VEGF that seems to modulate the effect of Ang2, but it is the ratio of Ang2 to VEGF. If the level of Ang2 is far above that for VEGF, then regression occurs. In Tet/opsin/ang2 mice, in the absence of doxycycline when there is only constitutive expression of Ang2, injection of an adenoviral vector that expresses VEGF results in NV of the cornea and iris, but no NV of the retina except for a few sprouts in the area of needle penetration. This is due to a combination of VEGF and the injury. When the mice are treated with

Downloaded from iovs.arvojournals.org on 09/30/2021 58

doxycycline and given an intravitreous injection of the adenoviral vector expressing VEGF, the co-expression of VEGF and Ang2 results in florid retinal neovascularization that originates from both the superficial and deep capillaries.9 These experiments show that VEGF by itself can only stimulate neovascularization from the deep capillary bed and this is because there is constitutive expression of Ang2 in that area. With coexpression of VEGF and Ang2, sprouting occurs from all capillary beds in the retina. If a vector expressing constitutively active HIF-1 is injected, there is upregulation of both VEGF and Ang2, and neovascularization sprouts from all capillary beds in the retina. The situation regarding choroidal neovascularization is even more complex than that regarding retinal neovascularization. The VMD2 promoter is an RPE-specific promoter10 and Tet/VMD2/VEGF mice have doxycycline-inducible expression of VEGF in RPE cells. In the presence of doxycycline, they have a normal retina and choroid and no identifiable phenotype.11 Tet/VMD2/VEGF/ang2 mice express both VEGF and Ang2 in RPE cells in the presence of doxycycline, and they have no phenotype. But if expression of VEGF is followed by rupture Bruch’s membrane, extensive choroidal neovascularization occurs, and if co-expression of VEGF and Ang2 is followed by rupture of Bruch’s membrane, then even more extensive choroidal neovascularization results. It appears that an intact Bruch’s membrane provides a barrier to neovascularization that must be perturbed before the stimulatory effects of VEGF and Ang2 on choroidal vessels become manifested. Ang1 is an agonist of Tie2; how do its effects in the retina compare to those of Ang2? Unlike Ang2, Ang1 does not show context-dependent effects. In all situations that have been tested, increased expression of Ang1 suppresses neovascularization in the retina or the choroid, but it cannot cause regression of neovascularization.12,13 This suggests that Ang1 may be useful for long-term prevention of neovascularization, but not as useful for treatment of established neovascularization. The following observations support the contention that angiogenesis has important commonalities, but also has aspects that are specific to particular tissues and/or disease processes. 1) As explained in detail above, in adult animals, increased expression of VEGF will stimulate neovascularization in the cornea and iris and deep capillary bed of the retina, but not in the superficial capillaries or the choriocapillaris. 2) Increased expression of Ang1 in skin and some other tissues stimulates neovascularization,14 but in the retina it always suppresses neovascularization. 3) TIMP1 suppresses neovascularization in several tissues throughout the body, but increased expression of TIMP1 in the retina stimulates retinal neovascularization15 possibly because it inhibits the proteolytic cleavage of proteins that require cleavage to become inhibitors. So it is not just a matter of difference between vascular beds in the eye and those in other tissues, but even different vascular beds within the eye can respond differently to angiogenic stimuli, and this is because of differences in constitutive expression of other proteins that play a role. The following are implications of the tissue- and disease-specific aspects of neovascularization. 1) It should not be assumed that experiments in chick chorioallantoic membrane or the cornea or tumor models predict what will happen in retinal and choroidal neovascularization. They may or they may not. It is possible to get hints from

Downloaded from iovs.arvojournals.org on 09/30/2021 59

experiments in other vascular beds, but such hints should be definitively tested appropriate models. 2) One cannot assume that retinal vascular development is a model of retinal neovascularization. There are important differences, so the effect of a drug on retinal neovascularization cannot be predicted from its effects on retinal vascular development. A corollary of this is that there is no validity to the claim that because one VEGF antagonist inhibits retinal vascular development and another one doesn’t, the latter one is safer in adults. 3) The potential for developmental stage-specific effects should not be ignored. Expression of VEGF in RPE cells during embryonic life results in thickening of the choroids,16 which is likely to represent increased developmental growth, because early on the choroidal vessels are responsive to VEGF alone (similar to the superficial capillaries of the retina). But if VEGF is expressed in the RPE in adult animals, there is no phenotype.11 4) Clinical trials should not be performed in patients with retinal or choroidal neovascularization based upon clinical observations or preclinical results in other vascular beds. This was done for interferon α2a. Based upon the observation that interferon α2a provides benefit in patients with cutaneous hemangiomas and results in a monkey model in which interferon α2a inhibited iris neovascularization,17 it was assumed that it would inhibit choroidal neovascularization and a large trial was done. Patients with neovascular AMD treated with interferon α2a fared worse than patients treated with placebo.18

Thus, different capillary beds can respond differently to angiogenic stimuli and it is important to sort out these local differences from the commonalities.

Questions

1. Could it be that the ability of Ang1 to inhibit ocular neovascularization is due to inhibition of leukocyte involvement? The role of leukocytes in the Ang1 effects has not been investigated, but is something that would be worth investigating in the future.

2. In mice with oxygen-induced ischemic retinopathy, we also see 2 peaks of Ang2 expression, one at day 14 when VEGF levels are highest and neovascularization is initiating, and one after day 17 when VEGF levels are lower and regression occurs. This is consistent with the hypothesis that the ratio of Ang2/VEGF determines whether growth or regression of neovascularization occurs.

3. Inflammation can appear subtle when you examine slides of retinal or choroidal neovascularization, and may be relatively less than with neovascularization in some other tissues, but it is there. It may be less impressive because of the blood-retinal barrier or relative immune privilege of the eye, but may still play an important role. I agree that there is evidence that it probably plays a role in VEGF-induced neovascularization in the eye as it clearly does in VEGF-induced neovascularization in other tissues, and it should be investigated in greater detail.

Downloaded from iovs.arvojournals.org on 09/30/2021 60

4. In postmortem human specimens, retinas from diabetics with no identifiable retinopathy show more immunohistochemical staining for VEGF than retinas from age-matched controls. It has also been shown in rodents with diabetes that Vegf mRNA is upregulated very early. How can the lack of neovascularization be explained? The upregulation of VEGF that occurs in the retina early in diabetes is localized in the inner retina. I would postulate that it is not associated with an increase in expression of Ang2 and this is why it does not induce neovascularization. I suspect that the levels of VEGF in these situations in the region of the deep capillary bed is not sufficient to take advantage of the constitutive expression of Ang2 in that region.

5. Like you I’m a firm believer in the heterogeneity of the microvasculature and that there are unique aspects of the retinal vasculature, but in the spirit of this diverse group, what are similarities and/or what are the properties of the retinal vessels that make them unique. I think it is very important to identify the common features of neovascularization that occurs in different tissues and different diseases. But the only way those similarities can be identified is through experimentation. My point is that there is not a good way to know a priori what the similarities are and so it is better not to make assumptions, but rather to do comparative studies to determine the similarities and differences. One similarity is that VEGF plays a central role as a stimulator in retinal and choroidal neovascularization and most other types of neovascularization. Another similarity is that HIF-1 seems to be a master regulator of pro- angiogenic factors in most tissues, although as pointed out by Gregg Semenza, the precise constellation of factors that it upregulates can differ in different cell types. Those are two examples and there are others that would require another lecture to discuss, but your point is an important one. By defining both similarities and differences among different types of angiogenesis important insights with therapeutic implications may be uncovered.

6. Has expression of Ang2 been noted in tip cells? Not to my knowledge.

References

1. Okamoto N, Tobe T, Hackett SF, et al. Transgenic mice with increased expression of vascular endothelial growth factor in the retina: a new model of intraretinal and subretinal neovascularization. Am J Pathol. 1997;151:281-291. 2. Tobe T, Okamoto N, Vinores MA, et al. Evolution of neovascularization in mice with overexpression of vascular endothelial growth factor in photoreceptors. Invest Ophthalmol Vis Sci. 1998;39:180-188. 3. Ohno-Matsui K, Hirose A, Yamamoto S, et al. Inducible expression of vascular endothelial growth factor in photoreceptors of adult mice causes severe proliferative retinopathy and retinal detachment. Am J Pathol. 2002;160:711-719. 4. Oshima Y, Deering T, Oshima S, et al. Angiopoietin-2 enhances retinal vessel sensitivity to vascular endothelial growth factor. J Cell Physiol. 2004;199:412-417. 5. Hackett SF, Ozaki H, Strauss RW, et al. Angiopoietin 2 expression in the retina: upregulation during physiologic and pathologic neovascularization. J Cell Physiol. 2000;184:275-284.

Downloaded from iovs.arvojournals.org on 09/30/2021 61

6. Hackett SF, Wiegand SJ, Yancopoulos G, Campochiaro P. Angiopoietin-2 plays an important role in retinal angiogenesis. J Cell Physiol. 2002;192:182-187. 7. Oshima Y, Oshima S, Nambu H, et al. Different effects of angiopoietin 2 in different vascular beds in the eye; new vessels are most sensitive. FASEB J. 2005;19:963-965. 8. Tobe T, Ortega S, Luna JD, et al. Targeted disruption of the FGF2 gene does not prevent choroidal neovascularization in a murine model. Am J Pathol. 1998;153:1641-1646. 9. Oshima Y, Takahashi K, Oshima S, et al. Intraocular gutless adenoviral vectored VEGF stimulates anterior segment but not retinal neovascularization. J Cell Physiol. 2004;199:399-411. 10. Esumi N, Oshima Y, Li Y, Campochiaro PA, Zack DJ. Analysis of the VMD2 promoter and implication of E-Box binding factors in Its regulation. J Biol Chem. 2004;279:19064- 19073. 11. Oshima Y, Oshima S, Nambu H, et al. Increased expression of VEGF in retinal pigmented epithelial cells is not sufficient to cause choroidal neovascularization. J Cell Physiol. 2004;201:393-400. 12. Nambu H, Nambu R, Oshima Y, et al. Angiopoietin 1 inhibits ocular neovascularization and breakdown of the blood-retinal barrier. Gene Ther. 2004;11:865-873. 13. Nambu H, Umeda N, Kachi S, Oshima Y, Nambu R, Campochiaro PA. Angiopoietin 1 prevents retinal detachment in an aggressive model of proliferative retinopathy, but has no effect on established neovascularization. J Cell Physiol. 2005;204:227-235. 14. Suri C, McClain J, Thurston G, et al. Increased vascularization in mice overexpressing angiopoietin-1. Science. 1998;282:468-471. 15. Yamada E, Tobe T, Yamada H, et al. TIMP-1 promotes VEGF-induced neovascularization in the retina. Histol Histopath. 2001;16:87-97. 16. Schwesinger C, Yee C, Rohan RM, et al. Intrachoroidal neovascularization in transgenic mice overexpressing vascular endothelial growth factor in the retinal pigment epithelium. Am J Pathol. 2001;158:1161-1172. 17. Miller JW, Stinson W, Folkman J. Regression of experimental iris neovascularization with systemic alpha-interferon. Ophthalmology. 1993;100:9-14. 18. Pharmacological Therapy for Macular Degeneration Study Group. Interferon alfa-2a is ineffective for patients with choroidal neovascularization secondary to age-related macular degeneration. Results of a prospective randomized placebo-controlled clinical trial. Arch Ophthalmol. 1997;115:865-872.

Downloaded from iovs.arvojournals.org on 09/30/2021 62

Vascular Targeting Agents and Strategies

Supplement 22 to Campochiaro PA, First ARVO/Pfizer Institute Working Group. Ocular versus Extraocular Neovascularization: Mirror Images or Vague Resemblances. Invest Ophthalmol Vis Sci. 2006;47:462-474.

Philip Thorpe

University of Texas Southwestern, Dallas, TX

Vascular targeting agents are made up of two parts, a targeting moiety that is usually an antibody or growth factor that binds to specific vessel markers in diseased tissue and an effector moiety that induces the therapeutic effect, typically damage to vessels or occlusion of vessels. To limit toxicity, surface markers are needed that allow recognition of blood vessels in diseased tissues, and not vessels in normal tissues. In tumors, neoplastic cells, stromal cells, or inflammatory cells impose a hostile environment on the tumor endothelium by secreting cytokines, secreting acidic metabolites that cause low pH, consuming oxygen resulting in hypoxia, and causing oxidative stress. These stresses cause upregulation of genes that are not normally expressed in endothelium. Tumor endothelium is dividing and remodeling, so it expresses proliferation and remodeling markers. Also, the basement membrane around angiogenic vessels is different (e.g., the ED-B domain of fibronectin is deposited around tumor vessels but is absent from the basement membrane of mature vessels). The endothelium of tumor vessels is discontinuous and allows antibodies that are in the blood to gain access to basement membrane components that are normally concealed. The tumor endothelium can also acquire markers from the environment, such as neutrophil- derived markers or bind growth factors (e.g., VEGF) to form complexes that are excellent markers of angiogenic vessels. Markers that have been used successfully for vascular targeting of tumors include: (1) angiogenesis and remodeling markers, e.g., VEGF receptors, complexes of VEGF with receptors, αvβ3 integrin, fibronectin ED-B domain, endoglin, anthrax toxin receptor, MMP2 and 9, CD13, and prostate specific membrane antigen (PSMA); (2) cell adhesion molecules, e.g., VCAM1, E-selectin, VE cadherin; (3) markers of oxidative stress, e.g., the lipid, phosphatidylserine (PS). PS is normally confined to the internal surface of the plasma membrane of vascular endothelium but gets transposed to the external surface in response to oxidative stresses in tumors. Also, blood proteins that bind to the exposed PS, such as the , can provide good tumor vessel markers. Effectors that have been used successfully in rodent model systems include: (1) toxins, e.g., diphtheria toxin, , and gelonin; (2) coagulation-inducing proteins, e.g., tissue factor; (3) cytokines IL2, IL12, and TNFα; (4) apoptosis-inducing factors, such as the Raf1 gene product, and the mitochondrial membrane disrupting peptide used by Pasqualini and coworkers; (5) radioisotopes, both α and β emitters; (6) liposomal encapsulated effectors such as doxorubicin; and (7) photosensitizers such as SnChe6. Effectors are not always needed, because naked antibodies can recruit the body’s own defense systems to destroy tumor vessels. Tarvacin is a naked anti-PS antibody which is currently in Phase I clinical trials in patients with various solid tumors. Tarvacin is currently in the form of a chimeric antibody. It has a murine Fv region that binds to PS in a β2-glycoprotein

Downloaded from iovs.arvojournals.org on 09/30/2021 63

1-dependent fashion. The Fv region is linked to human IgG1 constant regions. It is a nontoxic antibody at therapeutic doses even though antibodies of this type are sometimes associated with antiphospholipid syndromes. PS is usually confined to the internal surface of endothelium and is unavailable for binding to antibodies. But, in tumors, the endothelial cells are exposed to reactive oxygen species generated in response to hypoxia, cytokines, neutrophils, , or low pH. Lipid peroxidation probably occurs, generating calcium fluxes into the endothelium that activate the transporters that export PS to the luminal surface of the vessels. PS is one of the cleanest markers for tumor vessels yet discovered. It is absent from normal vessels including those in the ovary, a site of physiological angiogenesis, and the pancreas, a site of high vascular permeability. Treatment of mice with Tarvacin markedly reduces tumor growth in multiple tumor models, especially when combined with irradiation or cancer chemotherapy. It causes vascular damage resulting in disappearance of vessels in the center of the tumor with a surviving outer rim of tumor. This pattern of vascular damage is typical of vascular targeting agents and is thought to occur because vessels in the outer rim are subjected to lower interstitial pressures. Tarvacin recruits macrophages that bind to and damage the tumor endothelium, causing occlusion. This action is primarily an Fc-dependent process, macrophages and NK cells being the predominant effector cells. In addition, Tarvacin appears to cause a blockade of PS-mediated anti-inflammatory signals by switching the cytokines made by macrophages from TGFβ, a quiescence signal, towards TNFα and IL1, which are inflammatory mediators. Normally the macrophages produce TNFα and very little TGFβ, but when they are exposed to intact or apoptotic cells with PS on their external surface, they switch to producing TGFβ and little TNFα. This mechanism may explain why macrophages engulf PS-expressing apoptotic cells but do not engage in inflammatory responses to them. The TGFβ may also stimulate VEGF production by tumor cells as part of the proangiogenic response of macrophages. So, Tarvacin homes to tumor vessels, induces host cells to attack the tumor vessels, and may then re-educate macrophages to mount cytotoxic responses rather than proangiogenic responses towards PS- expressing tumor vessels and tumor cells. A second application of vascular targeting is ligand targeted liposomal chemotherapy being developed in the Ponzoni laboratory in Genoa, Italy in collaboration with the Allen laboratory in Alberta, Canada. Liposomes are PEGylated to provide stealth and then NGR peptides are attached that recognize CD13 on activated endothelial cells and pericytes. The liposomes are loaded with doxorubicin. Upon intravenous injection, the liposomes enter the tumor vasculature, bind to CD13, get taken up, and doxorubicin is released into endothelial cells and pericytes causing regression of the tumor vasculature. In addition, because the tumor vessels are leaky, the liposomes leak into the tumor interstitium and form a depot of drug. So, NGR- liposomes localize to tumor vessels and pericytes, extravasate into stromal tissue, and are taken up by endothelial cells and pericytes causing regression of tumor vessels. A third example of vascular targeting is an application that is being explored for treatment of choroidal neovascularization (CNV). This is a collaboration between the Campochiaro laboratory at Johns Hopkins and the Rosenblum laboratory at M.D. Anderson in Houston, TX. The drug is a fusion protein generated by ligation of VEGF121 cDNA to cDNA for gelonin, a plant toxin that inhibits protein synthesis. The construct is expressed in E. coli and forms a disulfide bonded VEGF/rGel dimmer. VEGF/rGel binds to VEGFR2 receptors resulting in internalization into endocytic vesicles and transport to the Golgi-ER complex. Gelonin is released into the soluble phase of cytoplasm where it turns off protein synthesis by damaging ribosomes. Either intravenous or intravitreous injection of VEGF/rGel causes regression of

Downloaded from iovs.arvojournals.org on 09/30/2021 64

CNV, ischemia-induced retinal neovascularization, or neovascularization in rhodopsin promoter/VEGF transgenic mice. The basis of the targeting is the increased expression of VEGFR2 on endothelial cells participating in retinal or choroidal neovascularization. This results in homing of VEGF/rGel and selective damage to new vessels within the eye.

Downloaded from iovs.arvojournals.org on 09/30/2021 65

Targeting Specific Vascular Beds

Supplement 23 to Campochiaro PA, First ARVO/Pfizer Institute Working Group. Ocular versus Extraocular Neovascularization: Mirror Images or Vague Resemblances. Invest Ophthalmol Vis Sci. 2006;47:462-474.

Dario Neri

Institute of Pharmaceutical Sciences, Swiss Federal Institute of Technology, Zurich, Switzerland

Vascular targeting employs a ligand for homing, often a single chain Fv fragment, which consists of a variable head and a variable light chain hooked together by a polypeptide linker. The antibodies are generated by antibody-phage technology. Phage libraries in which antibody fragments are displayed on the surface of filamentous phage are used to pan on the antigen and fish out specific antibodies. The process requires 5 days and works best if large antibody libraries are used. A library containing 4 billion human antibody clones was established for this purpose by brute force electroporation. The EDB domain of fibronectin has been identified as an excellent target for homing to new vessels. Fibronectin is a very abundant protein that normally occurs without this extra domain B. However, EDB gets inserted by alternative splicing whenever there is tissue remodeling. It is a small domain of 91 amino acids that is identical in mouse, rat, rabbit, dog, monkey, and man. In a normal adult, the EDB domain occurs only in the endometrium in the proliferative phase and some vessels of the ovary. A high affinity antibody to EDB (L19) specifically localizes to tumor vessels. L19 antibodies have been used in clinical trials to image solid tumors. In collaboration with Luciano Zardi, more than 30 derivatives of L19 have been produced (eg., fused to TNFα, VEGFs, IL12 interferon γ, truncated tissue factor, IL2, drugs, enzymes, photosensitizers). IL2 is a drug sold by Chiron (Proleukin) that is first line treatment for renal cell carcinoma. When you inject IL2 in tumor bearing mice, the IL2 does not preferentially accumulate at the tumor site. But the targeted version, L19-IL2, rapidly accumulates at the tumor site with tumor:blood ratios of 30:1 at 24 hours post-injection. So, this is an example of taking an existing drug and making it better. Three derivatives of L19 are in the clinic or will be by the end of the year. L19 can be used to localize the photosensitizer, Sn4ClorinE6, to choroidal neovascularization. Identification of novel targets is still a high priority. This is done by terminal perfusion of tumor-bearing animals with an active ester derivative of biotin resulting in an in vivo biotinylation of normal accessible proteins and accessible target proteins that are specifically expressed in a particular tissue or disease process. The target tissue is dissected and biotinylated proteins are purified on streptavidin. Several selective markers have been identified with this technique.

Downloaded from iovs.arvojournals.org on 09/30/2021