Canonical and non-canonical VEGF pathways: New developments in biology and signal transduction

Abstract

The last five years have witnessed a significant expansion in our understanding of VEGF signaling. In particular, the process of canonical activation of VEGFR tyrosine kinases by homodimeric VEGF molecules have now been broadened by the realization that heterodimeric ligands and receptors are also active participants in the signaling process. While heterodimer receptors were described two decades ago, their impact, along with the effect of additional cell surface partners and novel autocrine VEGF signaling pathways, are only now starting to be clarified. Furthermore, ligand-independent signaling (non-canonical) has been identified which occurs through galectin and gremlin binding, and upon rise of intracellular levels of reactive oxygen species. Activation of the VEGF receptors in the absence of ligand holds immediate implications for therapeutic approaches that exclusively target VEGF. The present review provides a concise summary of the recent developments in both canonical and non-canonical VEGF signaling and places these findings in perspective to their potential clinical and biological ramifications.

Introduction

Since the remarkable finding that genetic deletion of one VEGF allele impairs endothelial differentiation and vascular morphogenesis ^1,2, a large number of laboratories have focused on uncovering the signaling mechanisms activated downstream receptor binding. Much has been clarified and excellent reviews are available describing the most relevant second messengers and their relative contributions to endothelial cell migration, survival and differentiation ^3,4,5. These effects result from VEGF ligand interacting with cell surface VEGF receptor tyrosine kinases (RTKs) on the endothelium and considered to be “canonical VEGF signaling”. This mode of receptor activation deserves this classification not only because it was described first, but also because it appears to be the prevalent form by which VEGF induces proliferation, migration and vascular morphogenesis.

Here, our objective is to highlight recent developments that uncover additional modulators of the VEGF-VEGFR signaling axis. Specifically, we will discuss the biological relevance of receptor/ligand heterodimers, signaling compartmentalization, contribution of cell surface proteins to downstream cellular functions, and autocrine signaling. Furthermore, we will discuss the data on ligand-independent receptor activation or “non-canonical VEGF signaling”. This more recent mode of receptor activation appears to occur either at the cell surface, through specific galectin or gremlin, or in an intracellular compartment via Src-mediated activation. Interestingly, Src activation was recently described for two of the three VEGF RTKs and it appears to have distinct downstream consequences to ligand-mediated trans-phosphorylation in all cases ^6,7.

Although the contributions of alternative and non-canonical VEGF pathways have been technically difficult to ascertain, the biological effects of these pathways are on par with those of canonical VEGF and are critical to our understanding of VEGF effects in vivo. In addition to their impactful biological information, the findings can help explain differences in the clinical outcomes of therapies that specifically target either ligands or receptors.

Refinement of canonical signaling

Recent publications have provided additional complexity to the process of VEGF-VEGFR2 canonical signaling without altering its basic tenets. During the last five years, several studies have reinforced the contribution of receptor and ligand heterodimers and highlighted the effect of additional cell surface partners to the signaling process. These findings uncovered that small changes in the constellation of molecular players can provide a powerful impact to the signaling outcomes.

Unconventional heterodimers contribute to VEGF signaling

Canonical VEGF signaling is generally thought to be an interaction between homodimeric VEGF ligands and homodimeric VEGF RTKs (Figure 1, center). However, it has long been hypothesized that VEGF receptors could heterodimerize under physiological conditions, and in fact computational models predicted that VEGFR1/2 heterodimers make up 10–50% of active signaling complexes in the endothelium ⁸. In further support, heterodimerization has been well demonstrated for other RTKs. For example, in vascular smooth muscle cells, epidermal growth factor receptor (EGFR) becomes activated upon platelet-derived growth factor (PDGF) stimulation through basal EGFR heterodimerization with PDGFβR ⁹. Recently, experiments exploring heterodimerization of VEGF signaling components have uncovered interactions between heterodimeric ligands and heterodimeric receptors.

Figure 1. Heterodimeric VEGF ligand and receptors alter canonical signaling outputs.

Typically, VEGF signaling is thought to take place between two homodimers of VEGF which causes the dimerization of two homodimers of VEGFR2 (center), leading to ERK activation and downstream cellular effects. Artificial heterodimers of PlGF and VEGF-E have been used to study signaling outputs of VEGFR1/VEGFR2 heterodimers as PlGF binds only VEGFR1, and VEGF-E binds only VEGFR2 (left). VEGFR1/VEGFR2 signaling is distinct from VEGFR2 homodimerization as it induces tube formation and cellular migration, but not other cellular effects. Heterodimers of VEGFR2/VEGFR3 can be induced by either VEGF-A or VEGF-C ligand, contribute to sprouting, and are found endogenously at tip cell filipodia (right).

In early experiments, artificial systems were used in an attempt to clarify the possible effects of heterodimeric receptors on canonical VEGF signaling. Experiments on immortalized cells expressing high levels of both VEGFR1 and VEGFR2 showed that addition of VEGF increases VEGFR1/2 association by immunoprecipitation and produces distinct signaling outputs from either receptor alone ¹⁰. VEGFR3 is co-expressed with VEGFR2 in normal lymphatic endothelium, and dimerization in response to VEGF-C was observed in both primary cells and in 293 cells over-expressing both receptors ¹¹. Although the downstream cascade is not well understood, a unique pattern of phosphorylation was observed on kinase-dead VEGFR3 when in the presence of VEGFR2, strongly suggesting direct phosphorylation of VEGFR3 by VEGFR2 within the heterodimer ¹¹.

Recently, strides have been made to detect VEGF receptor heterodimers in increasingly dynamic settings. VEGFR heterodimers have been successfully detected by immunoprecipitation and in situ proximity ligation assays, both of which rely on close physical binding of two disparate proteins ¹². Using these techniques, it was found that endothelial cells frequently co-express VEGFR2 and VEGFR3 and that heterodimers are common in developing endothelial cells actively engaged in angiogenesis, particularly at tip cell filipodia. Although VEGFA does not bind or activate VEGFR3, blockade of VEGFR3 decreases VEGFA-mediated sprouting, suggesting that VEGFR3 contributes to VEGFA response via VEGFR2/3 heterodimers ¹² (Figure 1, Right). Phosphorylation of VEGFR2 upon exposure of the VEGFR1-specific ligand Placental Growth Factor (PlGF) uncovered the presence of VEGFR1/2 heterodimers in vivo, which act to enhance angiogenic response ¹³.

In addition to receptors, VEGF ligands heterodimerize in certain conditions. Due to their close protein homology, PlGF and VEGFA were predicted to form heterodimers when co-expressed in the same cell ¹⁴, and in fact have been observed in the conditioned media of several tumor cell lines ¹⁵. While some VEGFA/PlGF heterodimers induce mitosis and chemotaxis ¹⁵, in vivo data shows VEGFA heterodimerization with the specific PlGF isoform “PlGF1” effectively antagonizes VEGFA signaling and angiogenesis ¹⁶. VEGFA/PlGF heterodimers were exploited in a tumor model where over-expression of a dysfunctional PlGF mutant acted to sequester active VEGFA in heterodimers, and therefore suppressed tumor angiogenesis ¹⁷. These results demonstrate that while some VEGF heterodimers have pro-angiogenic signaling capacity, other heterodimers act to inhibit the angiogenic signaling output.

The ability of PlGF and VEGF to heterodimerize has been used as a tool to explore the physiological function of endogenous VEGFR1/VEGFR2 heterodimers ¹⁸. In these experiments, a synthetic ligand specific to VEGFR1/2 heterodimers was created by co-expressing VEGFR2-specific ligand VEGF-E (a viral VEGF mimetic protein) and the VEGFR1-specific ligand PlGF. Application of this ligand to endothelial cells induced several angiogenic responses such as VEGFR2 phosphorylation, migration and tube formation. However VEGFR1/2 activation did not induce proliferation, ERK signaling and other VEGFR2 functions, suggesting the heterodimer holds a unique signaling function ¹⁸ (Figure 1, Left).

Direct and indirect detection of VEGF receptor and ligand heterodimers indicates that heterodimerization is a true physiological phenomenon. The experiments described here carefully target heterodimeric complexes for activation or blockade, and taken together suggest that distinct signaling effects are easily overlooked when only homodimers are expected. Incorporation of VEGF ligand and receptor heterodimers into the canonical model may help explain otherwise unpredicted signaling and developmental effects.

New developments in VEGFR1 signaling

The effects of VEGFR1 have been difficult to pin down in endothelial culture models because VEGFR1 has a ten-fold lower kinase activity than VEGFR2 ¹⁹ and it induces little detectable downstream signaling ²⁰. VEGFR1 does, however, have a clear biological and signaling impact, particularly during development, as indicated by homologous recombination ²¹ and in pathological conditions such as diabetes. Diabetes is marked by progressive nephropathy caused by a disruption in osmotic pressure which damages the specialized endothelium of the kidney glomeruli. Early progression of this disease is associated with abnormal VEGF-mediated angiogenesis ²². Peptide inhibition of VEGFR1 in a mouse model increased symptoms of nephropathy, accompanied by glomerular cell death ²³. This blockade of VEGFR1 resulted in a suppression of diabetes-induced phospho-PI3K and phospho-Akt, and up-regulation of FoxO3a. The signaling cascade resulting from VEGFR1 blockade depressed phospho-eNOS, producing an increase of oxidative stress within the kidney. These results show that normal VEGFR1 signaling provides a protective effect in the kidney, and in fact signals to stimulate NO production within endothelial cells ²³.

Although these experiments demonstrate clear effects driven by VEGFR1, it is unclear whether the primary signaling event required the kinase activity of VEGFR1. Over a decade ago, it was elegantly demonstrated that deletion of the intracellular kinase domain of VEGFR1 is fully compatible with normal angiogenesis and embryonic development ²⁴, unlike the inactivation of the full receptor ²¹. Combined these and several other findings are consistent with a model in which, VEGFR1 act as a “decoy” receptor, blocking VEGF access to VEGFR2 rather than producing an independent signaling cascade on its own. Recent experiments have helped develop our understanding of the cellular ramifications of VEGFR1 regulation on VEGFR2 signaling in conditions of stress.

Both prolonged VEGF exposure and prolonged cellular stress result in an increased ratio of VEGFR1:VEGFR2 in the endothelium. Endothelial cells exposed to high levels of VEGF, for example in a squamous-cell carcinoma setting, express high levels of VEGFR1, while normal endothelium expresses high levels of VEGFR2 ²⁵. Prolonged VEGF exposure induces Akt/ERK survival pathways which inhibit degradation of VEGFR1. Concurrently, VEGF signaling through the JNK/c-Jun pathway, leads to the endocytosis and degradation of VEGFR2, keeping the signaling pathway in check. VEGFR1 is required for the VEGF-induced survival advantage, most likely mediated by an increase in Bcl2 expression ²⁵.

Under serum starvation conditions, normal endothelium first elevates VEGFR2 levels, an event that is followed by its down-regulation 24hrs thereafter ²⁶. Meanwhile soluble VEGFR1 (sVEGFR1) decreases during the early phase, and then increases to above normal levels after the 24hr period. Although full-length VEGFR1 levels are not altered during this time, increased sVEGFR1 sequesters VEGF in the ECM, preventing it from accessing VEGFR2. Accordingly, the serum-starvation response increases responsiveness to VEGF and pro-survival cues at early stages; but at late stages the effect leads to a reduction in VEGF responsiveness and an increase in apoptosis ²⁶. Although the direct signaling output of VEGFR1 is unclear, it appears to be highly-valued by the endothelium as a tool to control VEGFR2 function.

Alteration of VEGF signaling outputs by cell surface proteins

Integrins are a family of ECM-binding receptors, that upon ligand engagement, induce angiogenic signaling and survival pathways within the endothelium ²⁷. Thus, it should not come as a surprise that activation of integrin receptors might be tightly-associated with VEGFR2 responses to VEGF. Addition of VEGF induces physical association of VEGFR2 with integrin subunit β3, and when integrin signaling is blocked, VEGFR2 cannot be fully phosphorylated ²⁸. Cross-talk between these two receptors classes has been demonstrated in several biological platforms, where activation of either receptor stimulates binding and activation of the other ²⁷.

More recent experiments demonstrate that VEGF binding to ECM alters VEGFR2/integrin cross-talk ²⁹. VEGF is spliced into at least nine different isoforms, which vary in their ability to bind to the ECM or diffuse freely in a soluble form ³⁰. VEGF isoforms elicit unique vascular phenotypes, but only recently has this effect been characterized at the signaling level. The “bound” or “soluble” availability of VEGF robustly alters the kinetics of VEGFR2 signaling by manipulating its relationship to β1 integrin ²⁹. Bound-VEGF increases the association of VEGFR2 with β1 integrin, which alters cell surface organization of VEGFR2 clusters and prolongs receptor activation. This provides a distinct signaling compartmentalization than the one offered by soluble ligand. Differences in signaling clusters at the cell surface translate to an extension of the downstream kinetics of the p38/MAPK pathway ²⁹. Together these results indicate that the ECM context of the endothelium affects not only direct activation of integrins, but also modulates interactions between integrins and VEGF receptors which has downstream signaling consequences.

Progressively complicated endothelial receptor clusters are being uncovered which may fine-tune angiogenic responses in different tissue beds. For example, CD63 is a transmembrane tetraspanin expressed by endothelial cells that, when silenced, results in abrogated angiogenic response to VEGF and other growth factors ³¹. CD63 binds both VEGFR2 and β1 integrin, and ablation of CD63 was found to disrupt the VEGFR2-β1 integrin complex formation and downstream signaling in response to VEGF ³¹. Another novel VEGFR2 complex important for conveying VEGF signaling requires coordination by syndecan-1 (Scd1). Scd1 organizes a complex of VEGFR2, VE-Cadherin, and αv β3 integrin, without which the endothelium cannot respond to VEGF or VE-Cadherin engagement ³². Further experiments coupling cell-surface complexes found in vitro with in vivo functional experiments may help unravel signaling disparities among different biological settings.

Interactions between receptors, ligands, ECM and intracellular signaling machinery are further muddied by the fact that these complex interactions occur in a three-dimensional environment. In a cancer setting, for instance, VEGF receptors are often expressed both on neo-angiogenic endothelium as well as on the tumors themselves, and so have an opportunity to interact with ligand and each other in opposing cell types (referred to as trans interactions).

Upon VEGF stimulation, VEGFR2 and its co-receptor Nrp1 were found to form complexes in trans at the cell-cell interface between co-cultured cells expressing either single receptor ³³. These complexes produce distinct signaling cascades in endothelial cell, in part due to improper internalization of VEGFR2. In mouse tumor models, trans expression of Nrp1 suppressed angiogenesis and tumor growth by arresting VEGFR2 internalization and therefore downstream signaling ³³. These findings further expand the circumstances that must be taken into account when studying angiogenic signaling pathways. Realistically, a two-dimensional monoculture can only reveal so much about the biology at work in a human patient.

Expanding biological and signaling effects of autocrine VEGF

Canonical VEGF signaling occurs in a paracrine manner, where cell-surface endothelial VEGF receptors are activated by VEGF ligand originating from a secondary cell type. However, recent explorations have expanded our understanding of VEGF pathways to include autocrine signaling in the canonical pathway, meaning activation of receptor by ligand produced in the same cell type. We also discuss the evidence for intracrine signaling, which is cell-autonomous autocrine signaling that occurs entirely within the cell in a unique, non-plasma membrane compartment.

Autocrine VEGF signaling in the endothelium

Endothelial cells constitutively express VEGF receptors, and a subset of endothelial cells also co-express VEGF in homeostatic conditions. Aortic endothelial cells in particular exhibit the highest levels of VEGF expression when compared to other endothelial cell types and also showed phospho-VEGFR2 ³⁴. Besides the aorta, arterial endothelia of the adult has been shown to express VEGF in a salt-and-pepper pattern in larger vessels where it is induced by shear stress ³⁵. Genetic experiments ultimately provide the best evidence that autocrine VEGF signaling occurs within the endothelial compartment itself. Genetic excision of VEGF from the endothelium uncovers a broad need for autocrine VEGF in the vasculature. Endothelial VEGF-ablation results in degeneration of the endothelia in multiple vascular beds, resulting in a sudden death phenotype that is lethal in 55% of the mutant mice by 6 months of age ³⁶.

The autocrine VEGF pathway in the endothelium is poorly understood, likely because it occurs through alternative pathways or at a low background level. However, some information about autocrine regulation has emerged. For example, heterotypic cell-cell interactions with pericytes have been shown to increase VEGF-mediated survival pathways in the endothelium ³⁷. In this case, pericytes secrete the ECM protein vitronectin which is a ligand for endothelial αv integrin, and this induces VEGF up-regulation and downstream survival factor Bcl-w expression (Figure 2). Another molecular player in this pathway is the transmembrane protein Crim1 which enhances autocrine VEGF signaling through VEGFR2 in an intracellular compartment ³⁸. In this case, autocrine VEGF potentiated proliferation and survival of endothelial cells that was associated with the expression of anti-apoptosis factor Bcl-2 ³⁸ (Figure 2).

Figure 2. Autocrine VEGF signaling in the endothelium.

Pericytes secrete vitronectin which is a ligand for integrin αV. Integrin signaling increases NFkB activity, which activates expression of VEGF. Intracellular VEGF signaling via VEGFR2 in an endomembrane compartment is enhanced by the presence of the transmembrane protein Crim1. Autocrine VEGF signaling causes increased expression of pro-survival factors Bcl-w and Bcl-2 and decreased expression of pro-apoptotic BAX, promoting cell survival.

Autocrine VEGF signaling in cancer

It is well documented that tumor cells evade apoptotic signals by co-opting proliferation and survival machinery, the VEGF pathway is no exception. Many non-endothelial tumors express VEGF as well as VEGF receptors, and autocrine signaling has been identified in, to name a few, breast cancer ³⁹, colorectal cancer ⁴⁰, epidermal tumors ⁴¹ and precursor lesions to esophageal cancer ⁴².

In some instances autocrine tumor signaling takes place via the major endothelial VEGF receptor (VEGFR2) ^43,41, however the downstream signaling cascades may be distinct from classic angiogenic signaling. Glioblastoma multiform, a malignant brain tumor, expresses high levels of VEGF and VEGFR2 in gliomoa stem-like cells. In this CD133+ population, constitutively-active phospho-VEGFR2 is found primarily in the tumor cytoplasm ⁴⁴, which is atypical of stimulated endothelium. On the other hand, the autocrine VEGF signaling cascade observed in pre-cancerous Barrett’s esophagus cells is somewhat analogous to that observed in VEGF-stimulated endothelial cells ⁴². In Barrett’s cells, autocrine VEGF signals through VEGFR2, inducing proliferation via PLC-gamma/protein kinase C/ERK signaling ⁴².

Increasingly it has been found that the VEGF “co-receptor” Neuropilin-1 (Nrp1) is highly correlated with tumor malignancy, and is in fact a major player in autocrine VEGF pathways in cancer ⁴⁵. Even in glioblastomia stem cells, where proliferative signaling takes place through VEGFR2, the presence of Nrp1 modifies VEGFR2 recycling and boosts stores of intracellular VEGFR2 to abnormally high levels ⁴⁴. Autocrine VEGF signaling through Nrp1 is essential for maintenance of cancer stem cells in squamous tumors of the skin, but furthermore, deletion of Nrp1 from normal epidermis prevents tumor initiation entirely ⁴⁶. RhoA ⁴⁷ and Rac-1 ⁴⁸ have been reported as effectors of Nrp1. In renal cell carcinoma, Nrp1 fell upstream of Ras activation and phosphorylation of ERK1/2 and Akt, even in the absence of other VEGF receptors ⁴⁹. In human melanoma cells, Nrp1 also induces Akt activation, which is responsible for an invasive migratory cellular phenotype ⁵⁰.

In many tumor models, autocrine VEGF signaling relies on abnormal interactions between unrelated receptors that work synergistically to enhance survival and proliferation. Ablation of VEGF or VEGFR1 in epidermal tumor cells slows tumor proliferation, but this ablation in an EGFR-deficient background inhibits tumor formation entirely, suggesting a contribution of both EGF and VEGF pathways ⁴¹.

The autocrine VEGF signaling pathway was dissected in great detail in triple-negative breast cancers. Here it was found that VEGF/Nrp2 interaction stimulates α6 β1 integrin, which in turn activates focal adhesion kinase (FAK) ⁵¹. FAK induces expression of the Hedgehog effector Gli1, although canonical Hedgehog signaling activation was not observed ⁵¹. Instead, FAK activation of the Ras/MEK pathway is required for induction of Gli1 expression, which in turn activates the stem-cell factor BMI-1. A feedback loop is then enacted, where Gli1 enhances Nrp2 and VEGF expression, thus amplifying the signaling pathway and inducing an aggressive cancer stem cell phenotype ⁵¹.

The case for intracrine VEGF signaling

In theory, autocrine signaling may occur through secretion of VEGF which then interacts with cell surface receptors on the same cell type, if not the same exact cell. But several key experiments indicate that autocrine signaling can actually be intracrine in nature- that is, it occurs within an intracellular compartment and it does not require release of the ligand from the cell.

Studies on VEGFR2 have uncovered unusual RTK trafficking patterns that may be compatible with compartmentalized intracellular signaling. Early reports showed a large portion of total VEGFR2 resides within the cell in endosomal storage compartments which translocate to the cell surface upon VEGF stimulation ⁵². Later studies find the majority of endothelial VEGFR2 resides in the Golgi compartment at any one time ⁵³, suggesting endosomes and/or Golgi as potential sites of intracrine ligand/receptor interaction. Although we observe that a small number of endosome-like compartments contain VEGF, the majority of intracellular VEGF (visualized with a YFP-VEGF construct) co-localizes to a perinuclear compartment with VEGFR2 (Figure 3A, E), which is confirmed to be the Golgi compartment (Figure 3C). VEGF and VEGFR2 do indeed appear to interact in this compartment, as intracellular phosphorylation of VEGFR2 can be observed within the Golgi when YFP-VEGF is expressed (Figure 3B-C). Super-resolution microscopy shows VEGF localization to the lumen of the Golgi (Figure 3D), and is clumped in tight association with VEGFR2 (Figure 3E).

Figure 3. Autocrine VEGF and VEGFR2 interactions in the Golgi.

Confocal image shows YFP-VEGF co-localization with VEGFR2 (A) and phospho-Tyr1175 VEGFR2 (B) in a perinuclear region of the cell. (C) Giantin staining confirms that YPF-VEGF and phosphorylated VEGFR co-localize in the Golgi compartment. (D) Super-resolution images show YFP-VEGF localized to pockets within the lumen of the Golgi compartment (E) and in close proximity to intracellular VEGFR2 (arrowheads). See supplemental for detailed experimental methods and video showing 3D view of panel D and E.

Much of the compelling functional evidence for intracrine signaling relies on the comparison of extracellular antibody blockade of VEGF to intracellular VEGFR-inactivation with small molecule inhibitors. The autocrine signaling effects of VEGF in the endothelium can be inhibited by SU5416 (VEGFR2 inhibitor), but not anti-VEGF antibody ³⁶, which is also the case in the context of endothelial Crim1 deficiency ³⁸. Tumor stem cells ⁴⁴ and hematopoietic stem cells ⁵⁴ were shown to require VEGF intracellularly using a similar technique. Consistent with these experiments, meta-analysis of patients treated with RTK inhibitors found slightly more severe vascular side-effects than those treated with extracellular VEGF-traps ⁵⁵.

One caveat of the “external vs internal” pharmaceutical approach is the promiscuity of small-molecule RTK inhibitors. Although small molecule “VEGFR inhibitors” strongly suppress VEGFR phosphorylation in response to ligand, direct action of inhibitors on other non-VEGF receptors has been reported. For instance, the VEGFR2 inhibitor SU5416 (which has gone to Phase III clinical trials), is also a potent agonist of the aryl hydrocarbon receptor which regulates immune function ⁵⁶. Similarly, SU4312, another VEGFR2 inhibitor, directly inhibits neuronal nitrogen-oxygen synthase in cerebellar granule neurons ⁵⁷. Because several of the VEGFR inhibitors used in basic research have been approved for human trials, these molecules are now being closely scrutinized for other functions. Potentially, as study of these inhibitors becomes increasingly detailed, blockade of “intracrine” VEGF signaling by kinase inhibitors may be partially attributable to an off-target effect, particularly in animal models.

Another set of experiments that lend support to intracrine VEGF signaling involves the supplementation of VEGF-deficient cell lines with exogenous VEGF. Presumably, addition of extracellular VEGF mimics autocrine signaling that occurs at the plasma membrane, but in several cases this supplement fails to rescue a VEGF-deficient phenotype. This effect is observed in colorectal cancer ⁴⁰, as well as in the VEGF-deficient endothelial compartment, which does not respond to extracellular VEGF ³⁶.

It should be noted that several assumptions are made in VEGF-supplementation experiments. A major assumption is that all VEGF is equal. If a tissue of interest secretes a specific VEGF isoform, but is supplemented with another (supplementation is almost universally done with recombinant VEGF164), then a one-to-one comparison of “rescue” cannot be made. Similarly, post-translational modifications to VEGF, such as glycosylation patterns, may be specific to each cell type and may interfere with signaling in ways we do not yet understand.

An experiment that partially addresses the issues raised by VEGF-supplementation, is to use VEGF derived directly from the compartment of interest. To determine if autocrine signaling is intracrine, cells from VEGF-deficient and WT populations can be co-cultured so that the WT cells can supplement extracellular autocrine VEGF to neighboring mutant cells. If the VEGF-KO phenotype persists in these conditions (these cells proliferate less or die more than their WT counterparts), then it can be concluded that this cell type undergoes intracrine signaling.

These types of experiments have been conducted in several model systems. Fluorescent-labeled WT and VEGF-deficient tumor cells were mixed in equal number and injected into a mouse xenograft model ⁴⁹. In this context, it was found that VEGF-deficient cells had a growth deficiency despite close contact with WT cells ⁴⁹. Similarly, co-culture experiments were performed which allowed labeled WT and VEGF-KO endothelial cells to intermingle. In this case the endothelial VEGF produced by neighboring cells was unable to rescue VEGF-KO cell death phenotype ³⁶. These results have been questioned however, namely because autocrine VEGFR2 phosphorylation is lower in sparse HUVECs and higher in confluent monolayers, which is interpreted to indicate that crowded cells have more access to VEGF secreted from neighboring cells ³⁵.

As the biological ramifications of autocrine VEGF signaling become increasingly understood, the underlying intracellular mechanism require additional investigation. Most studies assume that autocrine VEGF signaling occurs through VEGFR2 activation, and some direct evidence of VEGFR2 phosphorylation has been observed in the presence of phosphatase inhibitors ^36,38. However, the downstream signaling cascades triggered by this phosphorylation event have not been investigated in-depth, nor the involvement of other VEGF receptors.

Biologically, in the endothelium, autocrine VEGF signaling causes a strikingly different cellular outcome (homeostasis) than the addition of high-levels of VEGF (endothelial activation: proliferation, migration and angiogenesis). It is well known that receptors at the plasma membrane encounter a distinct lipid composition and pH environment from those in intracellular compartments ⁵⁸. These compartments therefore provide distinct chemical environments that may themselves alter signaling kinetics, but also allow interaction with a unique set of spatially-restricted partners. The RTK EGFR, for example, has been shown to localize to the mitochondria upon ligand stimulation ⁵⁹. Here EGFR interacts with a unique set of mitochondrial proteins to regulate bioenergetics and cell death, distinct from the players involved in canonical EGFR signal transduction ⁵⁹. Because discoveries in spatially isolated signaling are ongoing due to technical and paradigm advances, major differences in autocrine VEGF signaling from paracrine VEGF signaling are anticipated.

Non-canonical signaling: VEGF-independent activation of VEGF receptors

One of the major recent advances in our understanding of VEGF signaling has been the realization that VEGF RTKs can be activated in a VEGF-independent manner, which we refer to as “non-canonical” VEGF signaling. This can occur through alternative “ligands” outside of the VEGF-family or intracellularly by Src kinases.

Alternative “ligands”: Galectin and gremlin binding to VEGF receptors

VEGF receptors, and nearly all cell surface proteins, must be glycosylated for proper function on endothelial cells ⁶⁰. Once dismissed as a simple chaperone for protein-folding, glycosylation is an often underestimated post-translational modification that in fact regulates a wide variety of biological functions once proteins are secreted. Glycosylation alters the biological function of proteins in three major ways: 1) Stabilization of protein folds and ECM interactions, 2) Direct modulation of protein function and 3) Provision of binding sites for glycan-binding proteins ⁶¹. It is this third function that is most relevant to VEGF signaling, where glycan-binding proteins, called galectins, can act as alternative “ligands” to glycosylated receptor tyrosine kinases.

Galectin-1 (Gal1) binds the VEGF co-receptor Nrp1 via its carbohydrate-binding domain in tumor-associated endothelial cells ^62,63. In the absence of VEGF, Gal1-Nrp1 interactions directly contribute to endothelial adhesion, and through the co-activation of VEGFR2, Gal1-Nrp1 induces JNK-mediated cellular migration ⁶² (Figure 4A). Vascular permeability, a cellular effect highly regulated by VEGF, can be induced by Gal1 signaling through VEGF receptors ⁶³. In this case, phosphorylation of VEGFR1 triggers a PI3K/Akt/RhoA signaling cascade that reduced the amount of VE-cadherin at cell-cell junctions (Figure 4B). Both Nrp1 and VEGFR1 are required for this permeability effect, but VEGFR2 was found to be unnecessary ⁶³.

Figure 4. Non-canonical ligand independent signaling by Galectins.

(A) In the absence of VEGF ligand galectin-1 binds Nrp1, inducing endothelial adhesion and signaling through VEGFR2 which activates a migration cascade. (B) Gal1 binding to VEGFR1 produces a signaling cascade that degrades VE-Cadherin increasing vascular permeability. (C) Upon canonical VEGF ligand binding, a short burst of angiogenic signaling occurs and VEGFR2 is quickly degraded. (D) Gal3 multimerizes to form a cell surface lattice of VEGFR2 which resists degradation, allowing prolonged angiogenic signaling. (E) Gal3 also induces VEGFR2 signaling in the complete absence of ligand.

VEGFR2 is susceptible to binding by Galectin-3 (Gal3) ⁶⁴, a galectin capable of multimerizing and forming cell-surface lattices with glycoprotein binding targets ⁶⁵. Again, in the absence of VEGF, Gal3 binding induces a VEGF-like signaling at the cell surface, but also amplifies the response of VEGFR2 to canonical VEGF ligand ⁶⁴ (Figure 4C–D). Gal3-mediated alteration to VEGFR2 signaling is likely due to the incorporation of these proteins in plasma membrane lattice structures, which increases VEGFR2 retention at the cell surface ⁶⁴ (Figure 4C–D). Previously, it was demonstrated that VEGF-induced internalization of VEGFR2 plays an important role in potentiation of ERK signaling by spatially separating the receptor from de-phosphorylation by the phosphatase Ptp1b at the plasma membrane ⁶⁶. While internalization seems to be relevant for potentiation of canonical signaling, alternative ligands, such as galectins, do not seem to require this internalization process.

Combined treatment of Gal1 and Gal3 enhances angiogenic tube formation at the cellular level, which is not observed upon addition of either single galectin ⁶⁷. While the addition of single (or combined) galectins induced equivalent phospho-VEGFR2 levels, only combined Gal1 and Gal3 treatment uniquely induces VEGFR1 phosphorylation ⁶⁷. Similar to the effects of Gal3 on VEGFR2 trafficking, combined treatment decreased endocytosis of VEGF1, trapping it at the plasma membrane. The downstream effects of this were the induction of Hsp27, and amplification of ERK phosphorylation ⁶⁷.

Because the enzymes responsible for protein glycosylation vary from cell to cell, and are highly reactive to cellular stimuli, the ability of Gal1 to bind to VEGF receptors is context-dependent. It was found that endothelial cells cultured in a tumor-like environment (mimicked in vitro with hypoxia and immunosuppresive cytokines) produced a high level of Gal1 glycan-epitopes ⁶⁸. In this setting, VEGFR2 is highly-decorated with Gal1 epitopes and direct VEGFR2-Gal1 binding is enhanced, even in the absence of VEGF ligand. Gal1 induces VEGFR2 autophosphorylation and triggers a Akt-Erk1/2 signaling cascade that closely resembles canonical VEGF signaling, although with altered VEGFR2 internalization kinetics ⁶⁸. This interaction is not exclusive to the tumor environment. Gal1-VEGFR2 signaling occurs in other specialized contexts, such as vessel growth in the developing placenta and maternal spiral artery remodeling ⁶⁹. Further investigation into the cellular glycosylation signatures of specific organs and pathological conditions may unveil differences in the angiogenic response due to galectin/VEGF receptor interactions.

Another alternative ligand that has been shown to induce VEGFR2 signaling in the absence of canonical VEGF is the protein gremlin, which belongs to the same cystein-knot superfamily as VEGF. Addition of purified gremlin, which is a BMP antagonist, was unexpectedly found to induce sprouting, migration and invasion of endothelial cells in angiogenic assays ⁷⁰. Gremlin directly binds VEGFR2, induces VEGFR2 auto-phosphorylation upon stimulation ⁷¹, and further induces a complex of αv β3 integrin with VEGFR2 ⁷². These functions are highly-analogous to the effects of canonical VEGF ligand, suggesting a mode of action as a VEGF-mimic.

Ligand-independent VEGF receptor signaling

VEGF receptors can participate in non-canonical ligand-independent signaling cascades in specific circumstances. For example, shear stress has been shown to promote VEGF-independent, Src-mediated VEGFR2 activation leading to regulation of Akt and eNOS function in vitro ⁷³. However, demonstration of ligand-independent VEGFR2 phosphorylation in vivo was lagging until recently.

Diabetes is characterized by endothelial dysfunction that is in part due to hyperglycemia-induced reactive oxygen species (ROS) ⁷⁴. Low levels of ROS are a necessary contributor in normal signaling cascades ⁷⁵. Excessive ROS, on the other hand, induces aberrant phosphorylation of VEGFR2, even in the absence of autocrine or paracrine VEGF ligand ⁷. Upon ROS-induced phosphorylation, VEGFR2 does not undergo auto-phosphorylation observed in canonical signaling. Rather, similar to shear stress conditions, VEGFR2 is phosphorylated by the Src family of kinases, which induces downstream activation of PLC-γ, but not p38 ⁷. ROS-induced signaling is intracrine and occurs in the Golgi compartment, after which VEGFR2 is degraded and subsequent response to VEGF addition is lessened due to lack of receptor ⁷. For this reason, diabetic animals exhibit decreased VEGFR2 phosphorylation in response to VEGF, an effect that is strikingly rescued by blockade of ROS production by NAC treatment ⁷.

In addition to VEGFR2, VEGFR1 undergoes ligand-independent signaling in macrophages in the context of atherosclerosis ⁷⁶. VEGFR1 can be expressed at high levels in macrophages, and addition of low-density lipoprotein (LDL) induced VEGFR1 endocytosis in complex with LDL receptor (LDLR). VEGFR1 autophosphorylation was observed in response to LDL treatment, and triggered a similar downstream pathway to that found in VEGF-stimulated endothelial cells ⁷⁶. This signaling pathway regulated macrophage migration in response to LDL, and may take part in macrophage recruitment and activation in atherosclerotic plaques ⁷⁶.

The role of Nrp1 in VEGF signaling has been difficult to understand, but a point of concordance is that Nrp1 acts as a co-receptor that binds and presents VEGF ligand to VEGFR2 to enhance VEGFR2 signaling effects ⁷⁷. However, Nrp1 has been reported to have some capacity as an independent signaling entity. Nrp1’s three C-terminal (cytoplasmic) amino acids are required for ligand-dependent migration and survival, independent of the presence of VEGFR2 ^47,78. New work shows Nrp1 also contributes to angiogenesis in an alternative, ligand-independent fashion. Stimulation of endothelial cells with the integrin ligand fibronectin induces Nrp1-dependent phosphorylation of paxillin ⁷⁹. It was found that the presence of Nrp1 mediates formation of a complex between integrins and Abl1 responsible for cytoskeletal remodeling and angiogenesis on fibronectin ⁷⁹, entirely in the absence of a specific Nrp1 ligand.

VEGFR3, the canonical receptor for VEGFC, also undergoes ligand-independent signaling in non-pathological conditions. VEGFR3 associates with β1 integrin, and becomes activated upon addition of β1 ECM ligands, even in the absence of VEGFC ⁸⁰. Somewhat similar to ROS-mediated VEGFR2 signaling, ligand-independent phosphorylation of VEGFR3 does not occur by auto-phosphorylation, but instead is mediated by c-Src in complex with VEGFR3 and β1 integrin ⁸¹. c-Src phosphorylation allows recruitment of the adaptor proteins CRKI/II and SHC, and induces downstream JNK phosphorylation ⁸¹.

Treatment with VEGFC is pro-angiogenic, and antibody blockade of VEGFR3-VEGFC interactions decrease angiogenesis, suggesting pro-angiogenic canonical signaling by VEGFR3 receptors. It was therefore unexpected to find that ablation of VEGFR3 in endothelial cells results in excessive angiogenic sprouting ⁶. This result and others suggests that VEGFR3 mediates bimodal signaling: “active” pro-angiogenic signaling in response to ligand, and “passive” anti-angiogenic signaling in the absence of ligand. In vivo, passive ligand-independent signaling is responsible for inducing Notch1 transcriptional targets, maintaining a non-angiogenic “stalk-cell” phenotype. This ligand-independent signaling pathway takes place by way of PI3K and deregulation of transcription factor FoxC2 ⁶.

Perspective

For decades studies on VEGF signaling have converged on a straightforward model: paracrine VEGF interacts with specific RTKs on the surface of endothelial cells inducing angiogenesis. Importantly, targeting canonical VEGF signaling has produced successful therapeutic strategies against diseases where VEGF-mediated vessel outgrowth is the major contributor to the pathology, such as in macular degeneration and some forms of cancer.

Hidden in the strength of the canonical model are a wealth of alternative modes of VEGF signaling. These include partner receptors and ligand-independent activation that have proven to alter, enhance and/or convey signaling to a subsets of downstream effectors not as obviously impacted by canonical signaling. The papers outlined in this review have produced compelling data to indicate that the VEGF field is moving beyond the straight forward paracrine model.

Significance.

While over forty years of vascular research have created a model of canonical paracrine VEGF signaling, recent advances in the field continue to uncover important additional nuances in VEGF biology. Highlighted in this review are alternative modes of VEGF signaling including: heterodimeric ligands and receptors, cross activation of VEGFR’s by partner receptors, and autocrine/intracrine VEGF signaling. Also explored is non-canonical signaling, which occurs in the absence of VEGF ligand altogether. These results modify the VEGF signaling paradigm and provide a broader picture of VEGF biology.

Abbreviations

VEGF

Vascular Endothelial Growth Factor

RTK

Receptor Tyrosine Kinase

VEGFR

VEGF receptor

EGFR

Epidermal Growth Factor Receptor

PDGF

Platelet-Derived Growth Factor

PlGF

Placental Growth Factor

ECM

Extracellular Matrix

Nrp1

Neuropilin-1

Gal-1

Galectin-1