VEGFs belong to the VEGF/PDGF (platelet-derived growth factor) group of the cystine-knot superfamily of hormones and extracellular signaling molecules 12, which are all characterized by the presence of eight conserved cysteine residues forming the typical cystine-knot structure (named after cystine, a dimer of two cysteines linked by a disulfide bond). The VEGF/PDGF group is evolutionarily related to other groups within the cystine-knot superfamily, notably the glycoprotein hormone and mucin-like protein families and, more distantly, the transforming growth factor-β (TGF-β) family. The absence of any of these proteins in unicellular eukaryotes such as yeast suggests that the cystine-knot structure evolved to perform hormonal and extracellular-signaling functions in multicellular organisms with tissue-level organization.

The known members of the human VEGF family are shown in Table 1. VEGFs have been found in all vertebrate species so far examined and are highly conserved between species. VEGF-A has been found in teleost fish (the zebrafish Danio rerio and the pufferfish Fugu rubripes), frogs (Xenopus laevis), birds (Gallus gallus), and mammals (Table 1). The sequence and genomic organization of the vertebrate VEGF-A genes is highly conserved between teleost fish and mammals, even though separation of these two groups from their common ancestor occurred around 450 million years ago: pufferfish VEGF-A shows 68% and 69.7% amino-acid identity with human and mouse VEGF-A, respectively 13.

VEGF-like proteins emerged relatively early in the evolution of multicellular animal life, as indicated by their presence in several invertebrate species. Invertebrate VEGF/VEGFR systems have been identified in fly (Drosophila melanogaster), nematode (Caenorhabditis elegans) and, most recently, jellyfish (Podocoryne carnea). Drosophila has three PDGF/VEGF-like factors (PVFs), which act through a single receptor, PVR 141516. In C. elegans, four VEGFRs, VERs (vascular endothelial growth factor receptor related) 1, 2, 3 and 4, have been identified 17. Definitive identification of a VER ligand is awaited, although a putative homolog of Drosophila PVF1 was revealed by a survey of the C. elegans genome 17. A single VEGF/VEGFR system has been found in P. cornea 18, with the VEGF being a possible homolog of Drosophila PVF1. In all cases, the invertebrate ligands appear to be more closely related to the VEGFs than to the PDGFs.

Alignment of the VEGF/PDGF homology domains (VHD) of VEGFs, PDGFs and PVFs, encompassing the residues making up the cystine-knot structure, reveals a high degree of regional conservation (Figure 1a). The eight cysteine residues of the cystine-knot structure are highly conserved, except in Drosophila PVF2, which lacks cysteine 2, and human PDGF-C and PDGF-D, which both lack cysteine 4. Phylogenetic analysis of these sequences reveals that the VEGF/PDGF family tree is essentially composed of two branches evolved from a putative common ancestor, a VEGF branch comprising VEGFs A-D, PLGF, Orf virus encoded VEGF-Es and Drosophila PVFs 1-3, and a PDGF branch, comprising PDGFs A-D (Figure 1b). Within the human VEGF family, VEGF-A is most closely related to PLGF (53% amino-acid identity within the VHD 19). The Orf virus-encoded VEGF-Es segregate into two groups, with VEGF -E_(D1701) and VEGF-E_(NZ2) most closely related to VEGF-A and PLGF, and VEGF-E_(NZ7) more similar to VEGF-C and VEGF-D. The Drosophila PVFs are more closely related to the VEGFs than the PDGFs, albeit distantly, with PVF1 most closely related to VEGF-C and VEGF-D (Figure 1b).

The gene structures and encoded functional domains of human and Drosophila VEGFs are shown in Figure 2. The human VEGF genes are characterized by a highly conserved seven exon structure, with the exception of VEGF-A, which has eight exons. Alternative splicing of the human VEGF-A gene gives rise to at least six different transcripts (Table 2), encoding isoforms of the following lengths (in amino acids, excluding the signal peptide): 121 (120 in mouse), 145, 165 (164 in mouse), 183, 189 and 206 20. All transcripts contain exons 1-5 and 8, with diversity generated through the alternative splicing of exons 6 and 7. A hydrophobic signal sequence essential for secretion of VEGF-A is encoded within exon 1 and a small region of exon 2, and the VHD is encoded by exons 3 and 4. Human VEGF-A₁₂₁ and VEGF-A₁₆₅ and their equivalents in other species are the two major isoforms in mammals; VEGF-A₁₂₁ lacks exons 6 and 7, and VEGF-A₁₆₅ lacks exon 6 (Table 2). Exon 6 encodes a heparin-binding domain, while exons 7 and 8 encode a NRP1/heparin-binding domain; with the exception of VEGF-A₁₂₁, all isoforms are thought to bind the polysaccharide heparin. VEGF-A₁₆₅ binds to NRP1 and NRP2, whereas VEGF-A₁₄₅ binds only to NRP2 1011. Recently, another splice variant of human VEGF-A was identified, VEGF-A_165b, which lacks exon 6 and contains an alternative exon 8 encoding a novel carboxy-terminal sequence, thereby raising the possibility of the existence of a family of sister isoforms containing this novel carboxyl terminus 21.

Human PLGF exists in four isoforms, PLGF-1 to PLGF-4, with PLGF-1 and PLGF-2 believed to be the major isoforms. The PLGF-1 and PLGF-2 transcripts encode isoforms (excluding signal peptide) of 131 and 152 amino acid residues, respectively. PLGF-2 is able to bind heparin and NRP1 through an exon 6 encoded heparin-binding domain 22; PLGF-1 lacks exon 6 and is thus unable to bind heparin 19. PLGF-3 also lacks exon 6 but additionally contains a 216-nucleotide insertion between exons 4 and 5. PLGF-4 consists of the same sequence as PLGF-3, plus the heparin-binding domain encoded by exon 6. PLGF-3 and PLGF-4 may function similarly to the larger VEGF-A isoforms, VEGF-A₁₈₉ and VEGF-A₂₀₆. In mice, PLGF-2 is the only PLGF isoform identified so far.

Alternative splicing of the human VEGF-B gene gives rise to two transcripts, encoding isoforms (excluding signal peptide) of 167 and 186 amino acid residues, differing only in their carboxy-terminal domains 2324. VEGF-B₁₈₆ transcripts contain the entire exon 6 and encode a soluble isoform. In VEGF-B₁₆₇ transcripts, the use of an alternative splice acceptor site in exon 6 introduces a frameshift, resulting in an alternative exon 6 (referred to as exon 6b in 23), encoding an NRP1/heparin-binding domain similar to that encoded by exons 7 and 8 in VEGF-A₁₆₅.

Little is known about alternative splicing of human VEGF-C and VEGF-D, although multiple isoforms of mouse VEGF-D have been described 25. VEGF-C and VEGF-D are closely related, both structurally and functionally. Both are ligands for VEGFR2 and VEGFR3 and are initially synthesized as disulfide-linked polypeptides containing amino- and carboxy-terminal propeptide extensions not found in other VEGF proteins, flanking a central receptor-binding VHD. The unprocessed full-length forms preferentially bind VEGFR3 and have low affinity for VEGFR2, whereas the fully processed forms have increased affinity for VEGFR2 2627. VEGF-C and VEGF-D lack the NRP/heparin-binding domain found in some VEGF isoforms and appear to be unable to bind NRPs.

The VEGFs are all secreted proteins. VEGF-A₁₂₁ and VEGF-A₁₆₅ are secreted as covalently linked homodimeric proteins, whereas the larger isoforms, VEGF-A₁₈₉ and VEGF-A₂₀₆, although believed to be secreted, are not readily diffusible and may remain sequestered in the extracellular matrix (Table 2). VEGF bioavailability may be regulated by plasmin-mediated proteolysis in the carboxy-terminal domains of the larger matrix-bound VEGF isoforms, such as VEGF-A₁₈₉, to release more diffusible, biologically active species 32. Human VEGF-A₁₆₅, the most abundant and biologically active form, is glycosylated at Asn74 and is typically expressed as a 46 kDa homodimer of 23 kDa subunits. VEGF-A₁₂₁ has biological activity in endothelial cells, but has lower potency than VEGF-A₁₆₅. The amino- and carboxy-terminal propeptide domains of VEGF-C and VEGF-D are proteolytically cleaved, possibly by plasmin, releasing the VHD during or after secretion to generate a fully processed mature form, which forms noncovalent homodimers of approximately 21 kDa that bind VEGFR2 with greatly increased affinity 2627.

Most information on the localization and expression of VEGFs has been derived from studies on VEGF-A. During embryogenesis in the mouse, VEGF-A can be detected from embryonic day 7 (E7) in the extra-embryonic and embryonic endoderm, and by E8.5 it is present at high levels in the trophoblast surrounding the embryo and in the embryonic myocardium, gut endoderm, embryonic mesenchyme and amniotic ectoderm. Later in development, VEGF-A is expressed in the mesenchyme and neuroectoderm of the head 33. VEGF-A expression declines in most tissues in the weeks after birth and is relatively low in most adult organs, except in a few vascular beds, including those of the brain choroid plexus, lung alveoli, kidney glomeruli and heart. VEGF-A expression is also upregulated during specific physiological processes such as development of the endocrine corpus luteum in pregnancy, wound healing and tissue repair, and in diseases associated with neovascularization (formation of new blood vessels). VEGF-A is produced by diverse cell types, including aortic vascular smooth muscle cells, keratinocytes, macrophages and many tumor cells 34.

Oxygen tension is a key physiological regulator of VEGF-A gene expression 35. The VEGF-A gene contains hypoxia-responsive enhancer elements (HREs) in its 5' and 3' UTRs 3637, the 3' enhancer being similar to sequences within the HRE of the gene encoding the hormone erythropoietin. Transcriptional regulation of the VEGF-A gene by hypoxia is mediated by binding of the transcription factor HIF-1 (hypoxia-inducible transcription factor 1) to the HRE. HIF-1 is a heterodimer composed of HIF-1α and HIF-1β subunits, both of which are members of the basic helix-loop-helix-PAS family 38. HIF-1α is normally very labile, but under hypoxic conditions, it accumulates because proteasomal degradation is inhibited: at normal oxygen tension, proline hydroxylation targets HIF-1α for proteasomal degradation, but is inhibited by hypoxia because of the requirement of the responsible prolyl hydroxylases for molecular dioxygen. The product of the Von Hippel-Lindau (VHL) tumor-suppressor gene is also required for proteasomal proteolysis: a genetic deficiency of this protein causes VHL disease, a condition characterized by retinal and cerebellar capillary hemangioblastomas (small, highly vascular tumors). In addition, VEGF-A mRNA is stabilized under conditions of low oxygen tension as a result of binding of unidentified factors to its 3' UTR. VEGF-A gene expression is also upregulated by a variety of growth factors and cytokines, including PDGF-BB, TGF-β, basic fibroblast growth factor (FGF-2), interleukin-1β and interleukin-6, some of which can act synergistically with hypoxia1.

All of the vertebrate VEGFs and their cognate receptors studied so far are able to regulate angiogenesis, and several have key biological roles in the formation of vascular structures either during development or in the adult. VEGFR function and signaling is reviewed extensively elsewhere 13940 and is not discussed in this article. The pivotal role of VEGF-A in embryonic vascular development was demonstrated by the remarkable discovery that targeted inactivation of a single VEGF-A allele in mice caused a lethal impairment of angiogenesis, resulting in death between E11 and E12 4142. The importance of larger VEGF-A isoforms, including VEGF-A₁₆₅, was confirmed by the finding that mice expressing only VEGF-A₁₂₀ - and lacking the longer heparin-binding isoforms - die within 2 weeks of birth owing to haemorrhage and ischemic cardiomyopathy (heart failure due to lack of blood supply to the heart muscle) 43. A cardiomyocyte-specific VEGF-A gene knockout generated using Cre-lox technology results in reduced body weight and thin-walled, dilated, poorly vascularized hearts 1.

Studies involving inducible VEGF-A gene inactivation or administration of soluble (s) forms of the receptor Flt-1 to inhibit VEGF-A function have established that VEGF-A continues to be critically important during post-natal growth and organ development 1. Inducible Cre-lox-mediated disruption of the VEGF-A gene in early post-natal life causes increased mortality, reduced body growth, and impaired organ development, particularly of the liver. Inhibition of VEGF-A by treatment of mice with sFlt-1 between 1 and 8 days after birth results in a more severe effect, characterized by growth arrest and lethality, but the effect of VEGF-A inhibition became less drastic if initiated at progressively later times in post-natal life. Inhibition of VEGF-A with sFlt-1 shows that VEGF-A-driven vascularization is also essential for endochondral bone formation and development of the corpus luteum during pregnancy 1.

VEGF-A-driven angiogenesis has a major role in the patho-genesis of diverse human diseases, including cancer, eye disorders and rheumatoid arthritis 44. Recognition of the importance of VEGF-A for the development of several important classes of cancer recently culminated in the approval of Avastin, a humanized monoclonal antibody to VEGF-A, for the treatment of metastatic colorectal cancer 45. There has also been great interest in using VEGF-A for the treatment of ischemic heart disease, where the aim is to promote blood-vessel formation and thereby provide a 'biological bypass' for diseased arteries. Despite abundant preclinical data suggesting that VEGF-A protein or gene therapy could be effective in treating ischemic heart disease, clinical trials have not so far yielded definitive evidence in support of this approach 1.

VEGF-A was originally identified as vascular permeability factor (VPF) as a result of its potent ability to increase vascular permeability, resulting in leakage of proteins and other molecules out of blood vessels 234. The physiological significance of the permeability-increasing effect of VEGF-A remains unclear, but it is important in mediating some pathogenic consequences of VEGF-A overexpression in disease, an example being brain edema (swelling and build-up of fluid) following cerebral ischemia 1.

In addition to its major role in angiogenesis, VEGF-A probably has functions that are independent of both endothelial cells and blood-vessel formation. A growing body of evidence indicates that VEGF-A has neurotrophic and neuro-protective activities in vitro and in vivo 4647. It has also been implicated in amyotrophic lateral sclerosis (ALS), an incurable degenerative disorder of motor neurons. Reduced VEGF-A expression resulting from deletion of the HRE from the VEGF-A promoter predisposes mice to ALS-like motor-neuron degeneration, and mice can be protected against ALS by treatment with VEGF-A 48. Furthermore, humans with particular VEGF-A promoter haplotypes have an increased risk of ALS associated with lower circulating levels of VEGF-A 49.

The VEGFR1-specific ligand, PLGF-₁, appears to be weakly angiogenic when acting alone, but VEGF-A-PLGF heterodimers can bind to VEGFR2, are mitogenic for endothelial cells, and stimulate angiogenesis in vivo 50. Though mice lacking PLGF are viable and develop normally, they have reduced angiogenesis in pathophysiological situations such as ischemia. PLGF-deficient mice also have delayed collateral artery growth following blockage of an artery, and PLGF stimulates collateral vessel growth. PLGF stimulates monocyte chemotaxis through VEGFR1, and there is increasing evidence that the biological effects of PLGF are mediated by mobilization of bone-marrow-derived haematopoietic progenitors.

A biological role for VEGF-B has not yet been clearly established. VEGF-B knockout mice are viable, healthy and fertile, but whereas Bellomo et al. 51 reported that VEGF-B-null mice have smaller hearts and recover more slowly from cardiac ischemia than wild-type littermates, Aase et al. 52 observed no effect of loss of VEGF-B on cardiac size or development and instead found a specific defect in atrial conduction in the adult. VEGF-B-deficient mice also have impaired development of pathophysiology when arthritis or hypoxic pulmonary hypertension are experimentally induced 53.

VEGF-C and its receptor, VEGFR3 (Flt-4), are strongly implicated in the formation of the lymphatic endothelium (lymphangiogenesis). Transgenic mice overexpressing VEGF-C in keratinocytes of the skin epidermis develop enlarged lymphatic vessels, while mice overexpressing VEGF-A₁₆₄ in the same location show only blood-vessel hyperplasia 54. VEGF-C also stimulates angiogenesis in the mouse cornea 55, however, and also in rabbit models of ischemia in the hindlimb. VEGF-D is mitogenic in endothelial cells and promotes angiogenesis in vitro and in several models of angiogenesis in vivo 56. VEGF-D also stimulates lymphangiogenesis in mice when overexpressed in skin keratinocytes and tumors 57, and it induces the survival and migration of lymphatic endothelial cells.

The viral VEGF-Es encoded by different strains of the parapoxvirus Orf appear to be important for viral infection and its associated pathology. Viruses of the Orf genus cause a contagious pustular dermatitis in sheep and goats, which is transmissible to humans, and produces lesions characterized by extensive neovascularization, vascular dilation, and epidermal proliferation. VEGF-E_(NZ2) induces dermal vascularization and epidermal proliferation in sheep, and disruption of the VEGF-E_(NZ2) gene resulted in a marked decrease in the vascularization of viral lesions without impairing viral replication in the early stages of infection 58.

Drosophila PVFs and their receptor, PVR, have key roles in cell migration during two developmental processes 141516. Firstly, PVR is expressed by the border cells, a cluster of somatic follicle cells that migrate towards the oocyte during oogenesis; PVF1 is produced by oocytes and acts as a guidance cue for the PVR-expressing border cells during their migration 14. Secondly, though devoid of endothelial cells or blood vessels, Drosophila does possess blood cells or hemocytes, and the PVF/PVR system is involved in the migration of these cells. PVR is expressed in the developing hemocytes during Drosophila embryogenesis, whereas PVF1, PVF2 and PVF3 are expressed along the hemocyte migratory route; inactivating mutations in either PVR or all three PVFs arrests hemocyte movement 16.

In C. elegans, which lacks a vascular system, the VEGFR-like VER proteins are localized to cells of neural origin, suggesting a role in neurogenesis 17. The recently identified VEGF and VEGFR homologs in the jellyfish P. carnea 18 are expressed in tubular structures of the gastrovascular system and in the endoderm during development at the stage when undifferentiated cells migrate and differentiate into plate cells. In this process, the differentiating plate cells interact with matrix and smooth muscle cells, a process analogous to the interaction of endothelial and vascular smooth muscle cells in angiogenesis. As nematodes and jellyfish lack both a vascular circulatory system and blood cells, the discovery of VEGF and VEGFR-like molecules in these species suggests that these proteins performed primordial functions in tubulogenesis and neurogenesis at an early evolutionary stage and only later developed more specialized roles in hematopoiesis and vascular development in more complex organisms. The role of VEGFs and VEGFRs in cell migration appears to be fundamental to their biological functions in invertebrate and vertebrate species.