Eph, a Protein Family Coming of Age: More Confusion, Insight, or Complexity?

Hallmarks of a new signaling concept: The "interaction mode" determines the biological outcome

It seems likely that the characteristic Eph-ephrin signaling mechanism, which relies exclusively on cell surface–bound interaction partners to generate and relay signals, provides the key to the apparent paradox by which a given Eph-ephrin pair mediates either adhesion or repulsion (Fig. 2). By default, activation of Eph tyrosine kinases activates pathways that modulate cytoskeletal plasticity, contraction of the cytoskeleton, loss of focal adhesions, cell rounding, and cell-cell repulsion or detachment (38, 44). However, Ephs and ephrins initially form heterotetrameric complexes (10), which then assemble into large signaling clusters (45) that involve several distinct Eph-ephrin interaction sites (46, 47) and tether the interacting cells. A distinctive feature of Eph-ephrin signaling—the phenomenon of bidirectional signaling in the Eph-bearing cell (forward signaling) and in the ephrin-bearing cell (reverse signaling)—is an extensively reviewed property that is essential for the understanding of the biological functions of Ephs and ephrins (41, 42, 48). Reverse signaling, initially suggested by the presence of highly conserved cytoplasmic tyrosine residues in ephrin B (49), is initiated in ephrin clusters through their phosphorylation by associated Src kinases (50) to provide docking sites for adaptor molecules, in particular growth factor receptor–bound protein 4 (Grb4), and for the initiation of signaling pathways that modulate the actin cytoskeleton (51). Very little is known about ephrin A reverse signaling (41), likely as a result of the difficulty in dissecting pathways of GPI-anchored proteins. Genetic studies in C. elegans, however, clearly suggest roles for these ephrins as signal transducers (52, 53). It is likely that Eph-ephrin–mediated cell positioning can be viewed conclusively only by considering forward and reverse signaling as being equally important components, and there is mounting evidence for essential roles for reverse signaling in nerve guidance, synaptogenesis, and vascular patterning.

For cell-cell repulsion to proceed after Eph-ephrin interactions, the resulting multivalent molecular tethers between opposing cells must be broken (Fig. 2): a key event that not only provides a switch between cell-cell repulsion and adhesion (54–57), but also determines the fate of the signaling cluster and consequently (necessarily) the type of resulting signaling cascade. It is now evident from several studies that, whereas clustering is clearly essential for phosphotyrosine-mediated Eph and ephrin signaling, it also triggers tyrosine-independent functions (58–60), in particular, cell adhesion and migration (61–63). Considerable experimental evidence confirms that the composition and dynamic regulation of Eph-ephrin signaling cluster assembly and disassembly and signal relay (Fig. 3) determine the nature and strength of the responses [reviewed in (39, 64)].

First, Eph function is regulated by phosphorylation of the activation loop tyrosine and two juxtamembrane tyrosines, which together modulate the conformation, accessibility, and activity of the kinase domain but also provide Src homology 2 domain (SH2 domain)–docking sites for downstream molecules (65–67). Clearly, the ability to activate downstream pathways necessarily depends on Eph tyrosine kinase signaling capacity, and modulating the ratio of kinase-active to kinase-inactive receptors will switch responses from repulsion to adhesion (22). Protein tyrosine phosphatases (PTPs) will play important roles in modulating Eph function (68) (Fig. 2), although evidence for Eph-specific PTPs is currently limited. One potential regulator of Eph kinase activity is low molecular weight (LMW)–PTP, which is believed to modulate EphB2-induced cell adhesion and capillary assembly (69), to mediate the dephosphorylation of EphA2 (70, 71), and to participate in signaling downstream of EphA2 (72–74). In addition, PTP receptor type O (Ptpro) dephosphorylates EphA4 and EphB2 in neuronal cells, thereby controlling the sensitivity of neuronal axons to ephrins (75). Other PTPs also negatively regulate reverse signaling of type B ephrins (76).

Second, the abundance of Eph and ephrin in gradients directly influences the signaling outcome, and the underlying principles involved have been extensively explored in vitro (63, 69, 77–80) and in vivo [reviewed in (31, 32)]. There is also good evidence that several Eph and ephrin family members are found on the same cells during development, which raises important questions about the coexistence and regulation of forward and reverse signaling in the same cell (81, 82). Somewhat unexpectedly, the alternative scenario, in which different Eph family members that can potentially bind to a common ephrin coexist on the same cell, remains to be addressed. In light of the extensive cross-reactivity between different Eph-ephrin family members (11, 83, 84), one would expect that Ephs that can bind to common ephrins would assemble in the same signaling cluster, whereby differences in individual binding affinities would have a major effect on the composition, size, and stability of the signaling complex and its resulting signal.

Third, it is apparent that regulated disruption of the molecular Eph-ephrin tether between cells fulfills a gatekeeper function in the progression to either cell-cell repulsion or adhesion (Fig. 2). Considering how critical disruption of the Eph-ephrin tether is to the ensuing signaling pathways, a feedback mechanism for this process that is tightly linked to other parameters that control Eph-ephrin signaling would seem essential. Two mechanisms have been identified that achieve controlled termination of Eph-ephrin–mediated cell-cell contacts (Fig. 2). In fibroblast monolayers of opposing cells that contain either EphB2 or EphB4 and either ephrin B1 or ephrin B2, Rac-mediated ruffling of the opposing cell membranes seems to trigger "transendocytosis," whereby entire Eph-ephrin complexes, including adjacent plasma membrane components, are internalized into one of the opposing cells (55, 56). Apparently, the direction of this transendocytosis relies on the intracellular domains of the involved Eph and ephrin proteins, whereby truncation of the cytoplasmic tails of either EphB or ephrin B leads to preferential endocytosis of the Eph-ephrin complex into the ephrin- or Eph-containing cell, respectively (55, 56). Both blocking the phosphorylation of ephrin B1 (56) and exposing cells to the Rho-dependent kinase inhibitor Y-27632 [which blocks actin fiber assembly, ephrin A5–induced cell rounding, and membrane blebbing (44)] do not affect transendocytosis and cell retraction (55). The mechanism of transendocytosis is interesting, because it implies that the intact Eph-ephrin signaling cluster, which likely includes associated signaling components, would be transferred from one cell into another. Currently, very little is known about the pathways and molecules that regulate Eph internalization, but a report suggests the involvement of clathrin-mediated endocytosis (85), which is also involved in the internalization of other RTKs. It seems plausible that, similar to other RTKs, critical Eph signaling steps, in particular those leading to changes in cell morphology (86), will persist in various endocytic compartments [reviewed in (87)]. Loss of Eph or ephrin during transendocytosis would obviously preclude such persistent signaling, which suggests that morphological changes in these cells may be triggered during the initial cell-cell contact and therefore executed independently of the Eph-ephrin signaling cluster.

As for EphA-ephrin A–mediated cell-cell contacts, ephrin-shedding by the transmembrane metalloprotease ADAM10 (a Disintegrin and Metalloprotease 10), also known as Kuzbanian (54, 57), releases the molecular tether between the opposing cells (Fig. 2). In general, regulated RTK ligand cleavage by ADAM proteases fulfills essential functions during normal and pathological tissue and organ development (88), which is supported by the similarities between mice deficient in ADAM10, ADAM17, Notch, Eph, erbB1, or epidermal growth factor (EGF) (89, 90). Not surprisingly, the cleavage of ephrins is tightly regulated, whereby only the intact Eph-ephrin complex provides the critical high-affinity binding site for ADAM10, which then positions its protease domain into a conformation that allows efficient cleavage of only Eph-bound ephrin (57). Similar to other RTK ligands (89, 90), ADAM10-mediated shedding of ephrins is inhibited by tyrosine kinase inhibitors (54) and relies on Eph kinase activation, a feedback control that ensures disruption of the Eph-ephrin tether only under conditions of ongoing, repulsive Eph-ephrin signaling. To date, ADAM-mediated shedding of type B ephrins has not been demonstrated; however, the finding that rhomboid-like protein 2, a rhomboid serine protease, efficiently cleaves ephrin B3 raises the possibility that ephrin shedding may also play a part in the endocytosis of EphB–ephrin B complexes (91).

Ephs communicate with other signaling systems

Considering the interest that Eph-ephrin biology has attracted, and the large number of molecules that are known to partake in downstream signaling cascades, the understanding of the pathways that execute the various responses attributed to Eph-ephrin signaling is surprisingly limited. To some extent, this may reflect the difficulty of dissecting pathways that rely on kinase activation and the generation of SH2 domain–docking sites, as well as on the assembly of multimeric receptor clusters (even in the absence of kinase activity). Important roles for signaling components that execute Eph- and ephrin-triggered changes in cell morphology, motility, adhesion, and repulsion—including Src and Abl kinases, phosphotyrosine-binding adaptors, PDZ domain–containing proteins, the 85-kD subunit of phosphoinositide 3-kinase (PI3K), and modulators of Ras and Rho family small guanosine triphosphatases (GTPases)—have been extensively reviewed (38–41, 92, 93). Not surprisingly, there is crosstalk between Ephs and signaling mechanisms that control cell adhesion and cytoskeletal plasticity, such as integrin and PI3K pathways and Ras-ERK (extracellular signal–regulated kinase) signaling.

Initially, the notion that Eph activation seems to inhibit mitogen-activated protein kinase (MAPK) signaling was interpreted to mean that Eph signaling is largely independent of this "classical" mitogenic RTK signaling pathway (42). However, the finding that cyclic adenosine monophosphate response element–binding protein (CREB)–mediated activation of ephrin B2 transcription [downstream of MAP/ERK kinase (MEK) and ERK activation] leads to increased EphB activity, which in turn elevates N-methyl-d-aspartate (NMDA) receptor phosphorylation and activity, thus causing epileptic seizures, has brought crosstalk between Eph and MAPK signaling back into the limelight (94). The Ras-MAPK pathway is a central component of many RTK signaling mechanisms, usually activated when RTK autophosphorylation through the recruitment of Grb2 and SOS1 leads to Ras- and Raf-mediated phosphorylation of MEK1 and MEK2 and the activation of ERK1 and ERK2 (95). The integration of Eph and MAPK signaling pathways seems to be highly conserved, because this crosstalk is critical for the regulation of oocyte maturation in C. elegans (96, 97). The activity of the single C. elegans Eph receptor VAB-1 inhibits MAPK signaling and, in parallel with the C. elegans NMDA receptor, inhibits oocyte maturation—a block that is relieved upon the recruitment of VAB-1 to the actin homolog, major sperm protein (96). Likewise, mammalian Ephs, in particular EphB2 and EphA2, inhibit MAPK signaling, which was initially described in neuronal and epithelial cells (98, 99). In the case of EphB2, recruitment of the GTPase activating protein (GAP) p120-Ras (p120-RasGAP) to the activated receptor (100, 101), which leads to reduced GTP-bound Ras and subsequent inhibition of Ras-MAPK signaling, is necessary for ephrin-induced neurite retraction (98). EphA2-mediated inhibition of the MAPK pathway in endothelial and epithelial cell lines even attenuates MAPK activation by growth factor receptor signaling (99), which agrees with the inefficient mitogenic signaling previously observed in cells that contain activated Eph proteins (27). Although subsequent studies revealed that ablation of both p120-RasGAP–binding sites and the introduction of an exogenous Grb2-docking site were needed to convert ephrin-B1–activated EphB2 from a MAPK-signaling repressor into an activator of ERK in neuronal cells (102), transient expression of wild-type EphB2 alone causes MAPK activation in human embryonic kidney 293T cells (67). Adding to the debate, other authors have suggested that a Shc-mediated interaction between EphA2 and Grb2 leads to ERK activation and cell detachment in breast and prostate cancer cell lines (103), whereas ephrin A1–dependent stimulation of endogenous EphA2 in wild-type mouse embryo fibroblasts (MEFs), but not in p120-RasGAP^−/− MEFs, robustly inhibits ERK phosphorylation and activation (102). However, in P19 mouse embryonic carcinoma cells, recruitment of Grb2 and Shc to activated EphB1, which leads to phosphorylation of Src and Shc and to ERK activation, seems necessary for the attachment and directed migration of these cells (104). Taken together, these findings confirm a definite involvement of Ras-MAPK signaling downstream of several Ephs in the execution of a range of disparate Eph functions, in particular the inhibition of MAPK activity. There is now also very good evidence that Ephs act downstream of MAPK signaling. In primary cortical neurons, ephrin B2–activated EphB (through associated Src) promotes NMDA receptor phosphorylation to potentiate Ca²⁺ influx and activate signaling pathways involved in the specification and maturation of synaptic connections (105). It is now known that one of the genes whose transcription is activated after MAPK activation and phosphorylation of CREB is ephrin B2 itself, which suggests that, in this case, positive feedback leads to epileptic seizures (96).

So how is it possible for Ras-MAPK signaling to execute such a diverse array of Eph receptor functions as well as to mediate similarly complex signaling outcomes for many other RTKs, such as EGFR and nerve growth factor receptor (NGFR)? The first clues to answer this question have come from the application of systems biology strategies (106) to unravel ERK responses in PC12 neuronal cells to inputs by EGFR and NGFR that promote proliferation and differentiation, respectively (107). These studies suggest that, depending on the type of signal relayed from the activated RTK, transient Ras-MAPK activation (by EGFR) triggers a negative feedback mechanism that results in proliferation, whereas sustained activation results in a positive feedback loop that leads to cell differentiation (107). It is tempting to speculate that a similar concept might explain the seemingly conflicting range of biological responses that emanate from activated Eph receptors, because it is highly plausible that their well-established ability to trigger opposing signaling pathways (such as MAPK activation or inhibition) reflects the capacity for dynamic regulation of the size and composition of their signaling clusters.

Another example of crosstalk that may well affect Eph function in normal and oncogenic development (clearly in the intestinal epithelium) involves the communication between the Eph and Wnt signaling pathways (34, 108). Wnt signaling controls a complex program of precursor cell proliferation and renewal, Paneth cell differentiation and compartmentalization, and the ordered migration of epithelial cells along the colonic crypts, a stage that is controlled by the graded abundance of both EphB and ephrin (109). The abundance of EphB and ephrin B is tightly controlled through the β-catenin:TCF (T cell factor) transcription factor complex and, not surprisingly, EphB2^−/−EphB3^−/− mice exhibit pronounced intermingling of differentiated and precursor cell populations. It seems very likely that this role of Ephs and their regulation by Wnt signaling have major implications for tumor progression, which suggests the potential role of EphB receptors as tumor suppressors.

Retinotopic patterning: More than axon repulsion

According to the chemoaffinity hypothesis (110), the assembly of topographic neural maps relies on the action of axon guidance molecules, which are present in spatially restricted, complementary gradients on projecting axons and their targets and serve to position axonal termination zones that faithfully reflect the origin of individual axons. Ephs and ephrins appear to be the only protein families to date that fulfill these criteria, and their presence and function are absolutely essential for accurate topographic map assembly, a role that has been conserved from C. elegans (6, 7) to vertebrates (64).

The original thinking was that Eph- and ephrin-mediated guidance operated in a straightforward manner. The pattern of the low (anterior) to high (posterior) abundance of the retinal ganglion guidance ligands ephrin A5 (15) and ephrin A2 (14, 111) in the chick brain tectum or the mouse brain superior colliculus matched the low-to-high, nasal-to-temporal (NT) graded abundance of EphA3 (112) in the chick retina (111). In vitro studies that showed that these ephrins strongly repel axons from the temporal retina (in which EphA3 is highly abundant), but not from the nasal retina [in which EphA3 is low in abundance (15)], and that targeted deletion of ephrin A2 and ephrin A5 results in severe anterior-posterior (AP) mapping defects (21, 113) ended the 20-year-long search for the elusive affinity labels that had been predicted to guide retinotectal projection (110). In an apparent variation of this concept, functional analysis of EphB2- and EphB3-deficient mice revealed that ventral retinal axons, which have a high abundance of EphB2, EphB3, and EphB4, project to targets in the medial colliculus that contain a high abundance of ephrin B1, whereas dorsal axons that contain a low abundance of EphB project to the lateral part of the collicular gradient, in which the abundance of ephrin B1 is low (114). This suggested chemoattraction rather than repulsion as the underlying force that guides the projection of dorso-ventral (DV) graded retinal axons into lateral-medial (LM) positions in the tectum or superior colliculus (115, 116).

It seems, however, that these two apparently contradictory findings reflect aspects of an underlying common axon guidance mechanism. Generally, all retinal axons in rodents or chicken embryos, in contrast to those in fish and frogs, initially cross the superior colliculus or tectum in a nontopographic manner, substantially overshooting their correct AP and LM termination zones. Accurate positioning of the axon arbors is then established by promoting and inhibiting (back-)branching toward the correct and aberrant termination zones, respectively [reviewed in (64)]. For both axes of the projection map, this axon branching is guided by ephrins that act as "ligand-density sensors" (78, 117), which direct branch distribution and directional bias toward the correct position by eliciting repulsive or attractive responses according to the abundance of the Eph receptors that they encounter [reviewed in (30, 64, 116)].

Important insights into the ligand-density sensor concept have been gained from functional genetics approaches, which compared perturbations in the NT gradient of retinal EphA5 and EphA6, through the ectopic expression of EphA3, on a random proportion of retinal ganglion cells of either wild-type or EphA4-deficient mice (20, 32). Analysis of their retinotopic patterning defects demonstrated that the relative rather than the absolute abundance of Eph and ephrin, together with signaling activity, determines the projection position of retinal axons (31). Retinal axons find their correct positions by competing with all other axons for the available collicular space. Thus, axons with the highest or lowest abundance of EphA always project to either the anterior-most or posterior-most positions, respectively; ectopic overexpression of EphA3 does not lead to arborizations outside of the map. These studies also demonstrate that the slope of the gradient, which reflects the combined activities of Ephs at each position in the gradient, determines the fidelity, or the "discrimination limit," at which two neighboring retinal axons project to distinct collicular positions. By considering that the profile of axon guidance molecules in the retina and superior colliculus is far more complex than initially anticipated, and that most Ephs (apart from EphA1, EphA2, EphB5, and EphB6) and ephrins (apart from ephrin A1, ephrin 3, and ephrin 4) are present in overlapping gradients on retinal ganglion cells and on collicular or tectal targets [reviewed in (64)], the finding that the relative abundance of Eph or ephrin determines axon projections has considerable implications.

First, current studies and derived models have focused on selected Eph-ephrin interactions that are considered relevant because of high-affinity binding data derived in vitro. It is now obvious that the characteristic property of ephrins to promiscuously activate a range Ephs within, or outside of, a given subclass (84) and to act as signal transducers in their own right (41, 81, 118, 119) has to be taken into account to understand Eph- and ephrin-controlled positioning. As an example, retinal axons in ephrin A2^−/−ephrin A5^−/− mice show unexplained DV mapping defects (21) that are typical for EphB2^−/−EphB3^−/− mice, which suggests a potentially relevant interaction between ephrin A5 and EphB2 during map formation.

Second, the presence of overlapping countergradients, and thus the coexpression on the same axon of ephrins and Ephs that would normally interact in trans, will likely affect the overall response of such an axon in the target zone. This is a topic of current research and dispute. The notion that retinal and tectal countergradients of ephrins and Ephs may help to establish and maintain the graded abundance of Eph and ephrin in these sites currently lacks experimental support. The phenotypes of ephrin A2^−/−ephrin A5^−/− mice revealed apparently undisturbed retinal gradients of EphA5 and EphA6, and likewise the ephrin B2 gradient seemed unaffected in EphB2^−/−EphB3^−/− mice (21, 114). However, analysis of ephrin A2^−/−ephrin A3^−/−ephrin A5^−/− mice revealed that these ephrins, in addition to their role in axon positioning (120), are essential for the establishment and internal organization of the thalamocortical projection map (121).

Although the presence of different guidance signals and their integrated regulation on migrating cells seems ideally suited as the basis for high-fidelity vertebrate axon positioning, its faithful analysis in vitro with axon repulsion as readout (15, 122) provides a considerable experimental challenge. Thus, the apparently increased sensitivity of ephrin A2^−/−ephrin A5^−/− nasal axons to repulsive signals in "stripe assays" (21, 122) led to a model whereby retinal ephrin A2 and ephrin A5 act to inhibit repulsive signaling by EphA4 (which is present uniformly across the retina) by interacting in an antagonistic manner on the same axon membrane in cis (82, 123). Overall, this would result in a NT gradient of nonfunctional EphA4 receptors on retinal axons, complementing and thus sharpening the repulsive gradient of EphA5 and EphA6 (82). According to this model, targeted deletion of EphA4 should not affect the projections of nasal axons, because it is nonfunctional in the retinal axons of wild-type mice. However, this is not the case, and pronounced abnormal projections of these axons are observed in EphA4^−/− mice (32), which suggests a very significant contribution of EphA4 forward signaling to the gradient. Notably, potential axon guidance effects of ephrin reverse signaling (41, 81, 118, 119), and of interactions between EphA4 and ephrin B2 on the dorsal half of nasal axons (83) or between ephrin A5 and EphB2 on the lateral half of anterior targets (11), were not considered in the model. Likewise, structural and kinetic observations, suggesting that the binding domains of Ephs and ephrins preferentially interact in trans (11, 46, 47, 124), had not been contemplated in the context of this cis-interaction model.

Third, the coexistence of various Ephs and ephrins on a single axon raises intriguing questions about how signaling on such a cell surface is regulated to achieve specific outcomes. In this context, it is important to consider that opposing axons with different Eph and ephrin family members on each of their surfaces will be in sufficient contact to allow the assembly of signaling complexes between neighboring axons in trans to occur, both in vivo (that is, in axon bundles entering the superior colliculus) and in vitro [such as in stripe assays in the monitoring of outgrowth from retinal ganglion cell explants (15, 122)]. Do these interactions actually take place, and if so, do they occur throughout the axon, or are there mechanisms in place that bias signaling activity and the resulting responses toward the axon growth and termination zones? The first answers to these questions were provided in a study that convincingly demonstrated that EphA receptors and GPI-anchored ephrin A ligands, coexpressed on the membrane of the same motor axon, separate into distinct domains. This lateral segregation of repulsive EphA and attractive ephrin A signaling clusters allows distinct signaling outcomes to occur in separate areas of the axon membrane (81). Importantly, the same study confirmed that overexpression of ephrin A in cells that contain both ephrin A and EphA modulates reverse ephrin A signaling and function in controlling growth cone spreading, but has no effect on EphA forward signaling–mediated growth cone repulsion and collapse. Together, these findings offer an insight into how the effective uncoupling of Ephs and ephrins into distinct functional cell-membrane domains (i) provides a mechanism that controls concurrent, yet distinct, cell-morphological responses on the same cell surface, and (ii) (by mutual exclusion) effectively prevents Eph-ephrin interactions in cis.

Functions of Ephs during vascular patterning

The first evidence to associate Eph signaling with angiogenesis, which was discovered before any other function of Ephs had been established, came from experiments that showed that recombinant ephrin A1 acts as a trigger of rat corneal angiogenesis in vivo and that demonstrated the chemotaxis of EphA2-containing bovine adrenal capillary endothelial cells (125) and capillary-like assembly (126) in vitro. Although the roles of EphA2 and ephrin A1 in adult angiogenic remodeling and some of the underlying signaling pathways have been confirmed (127, 128), it is now clear that type B Eph receptors and ephrins are the essential regulators of the assembly, maturation, and maintenance of blood vessels and the patterning of the vascular system in vertebrates [reviewed in (37, 40, 41)]. Similar to the concepts first described for Eph-guided axonal mapping, but less dependent on countergradients, initial models were based on the preferential abundance in functionally distinct regions (129, 130), in which ephrin B2 and EphB4 were thought to demarcate arterial and venous endothelium, respectively. Given the very similar defects that result from targeted deletion of either gene and lead to embryonic lethality (129, 131)—which include the failure of angiogenic remodeling in the yolk sac and head, disrupted cardinal veins, and insufficient or missing myocardial trabeculation of the heart (formation of protrusions on the luminal wall of the early heart tube that maintain blood flow before the development of cardiac valves)—it was tempting to speculate that bidirectional signaling between ephrin B2 (reverse signals) on the arterial endothelium that contacts the neighboring EphB4 on venous endothelial cells (forward signals) is essential for appropriate remodeling of the embryonic vasculature (132).

Indeed, transgenic mice that express a mutant form of ephrin B2 (ephrin B2-ΔC), in which its entire cytoplasmic domain is replaced by the hemaggultinin (HA)–tag sequence, show a vascular phenotype very similar to that of the classical ephrin B2^−/− mouse (133). This suggests that reverse signaling of ephrin B2 is essential for angiogenic remodeling (133). However, comparison of this transgenic mouse with a mouse that expresses ephrin B2 in which the cytoplasmic domain was replaced with the β-galactosidase (β-Gal) protein domain revealed lethal vascular defects only in the ephrin B2-ΔC mouse. Expression of the ephrin B2–β-Gal fusion protein allowed for normal angiogenesis and normal embryonic development to live-born pups, which suggests that ephrin B2 signaling is not essential for vascular development and angiogenic remodeling (118). It transpires that deletion of the cytoplasmic domain of ephrin B2 effectively inhibits its trafficking to the plasma membrane. This results in ephrin B2 being undetectable at the cell surface and acting as a null mutation that replicates the phenotype of ephrin B2^−/− mice (118). Thus, it seems likely that, during embryonic vascular remodeling, ephrin B2 functions only as a ligand to trigger EphB4 forward signaling.

It is now evident that ephrin B2 is more widely abundant than was initially appreciated. Its coexistence with EphB2 on mesenchymal cells, including pericytes and vascular smooth muscle cells (130, 134–135), and the inappropriate vascularization of somites in ephrin B2^−/− and in EphB2^−/−EphB3^−/− mice suggested a role for ephrin B2 in endothelial and mesenchymal cell-cell communication to regulate blood vessel growth into the intersomitic space (130, 131, 134). Indeed, the inappropriate ectopic expression of ephrin B2 in Xenopus (137) or mouse embryos (135) results in the abnormal migration of intersomitic veins into the adjacent somitic tissue and atypical recruitment of vascular smooth muscle cells to the ascending aorta (135). This is in line with EphB–ephrin-B2 signals guiding endothelial and mesenchymal cell-cell interactions and blood vessel maturation during embryogenesis. It is tempting to extrapolate this finding to predict a mechanism whereby ephrin B2 expression in mesenchymal elements is crucially important in vascular patterning. However, endothelial-specific deletion of ephrin B2 in a targeting approach that leaves mesenchymal ephrin B2 intact, phenocopies the angiogenic remodeling defects in the yolk sac, head, trunk, and heart of conventional ephrin B2^−/− mice with 100% penetrance (138). This finding clearly suggests an additional and distinct role for mesenchymal ephrin B2, which does not compensate for the loss of endothelial remodeling by ephrin B2 (138).

So what is the role of ephrin B2 on pericytes and smooth muscle cells? A complementary approach for the targeted deletion of ephrin B2 in mural cells that express platelet-derived growth factor receptor β on their surface has provided some important answers (139). Loss of ephrin B2 expression by this approach is lethal shortly after birth as a result of hemorrhaging of several organs, including the skin, intestine, kidney, and lung, as well as the vascular plexus of the central nervous system, which is suggestive of disrupted microvessel architecture. Detailed analysis revealed compromised or missing endothelial and smooth muscle cell (pericyte) interactions resulting from defects in the adhesion, spreading, and directional migration of ephrin B2^−/− smooth muscle cells. Overall, it appears that, during angiogenic remodeling, ephrin B2 acts in an adhesive manner to facilitate adequate and effective cell-cell contacts between pericytes and smooth muscle cells as well as with the underlying endothelial cell layer: a model that is also consistent with the expression pattern of the relevant EphB receptors on these cells (139).

Finally, emerging evidence suggests that an essential role in vasculogenesis and angiogenesis is not limited to ephrin B2 and EphB4. Several additional Ephs and ephrins are found in less-restricted arterial and venous expression patterns, including the presence of ephrin B1, EphB3, and EphB4 on venous endothelial cells and of EphB3 on some arteries (130). Also, ephrin B1, EphB2, and EphB3 are found in addition to ephrin B2 and EphB4 on the primitive vasculature: EphB2 is found on embryonic mesenchyme, whereas EphB3 and EphB4 are found on all major veins and on aortic arches, and ephrin B1 mRNA is detected in all major blood vessel primordia. Apparently, these additional Eph signals fulfill partially overlapping functions in vascular development, and targeted deletion of EphB2 and EphB3 results in a 30% penetrance of vascular defects that are similar but not identical to those of ephrin B2^−/− mice. In particular, defective endothelial cell guidance leads to absent or abnormally shaped primordial vessels, poorly developed head vasculature, less extensively folded traberculae in the heart, and premature death at embryonic day 11.5 as a result of these vascular remodeling defects (130).

In this context, it is noteworthy that the gene expression profile and cell behavior in the endothelial lining of the endocardium are reminiscent of those of the rest of the vasculature. In the heart, cell contact–dependent and tissue-dependent morphological changes result in the typical two-chambered morphology, the development of the atrioventricular, aortic, and pulmonary valves, and endothelial cell–lined myocardial trabeculations. Consistent with their preferential expression in the endocardium, targeted deletion of either EphB2, EphB3, EphB4, or ephrin B2 results in arrested development and little or no myocardial trabeculation (129–131), whereas other typical heart structures seem unaffected. Intriguingly, characterization of the perinatal-lethal phenotype of mice that lack EphA3, an Eph family member not previously implicated in cardiovascular development, reveals its critical role in the development of the atrioventricular valves of the heart (140). During cardiogenesis, EphA3 is present in the mesenchymal cells of the endocardial cushions, whereas in a complementary fashion ephrin A1 is present on the surrounding endothelial lining. Of all the EphA3-binding ephrins, ephrin A1 has the lowest affinity for EphA3 (12, 13, 141), which could indicate that such a low-affinity interaction is required for specific cell-cell adhesion to establish the contact between these two cell layers, similar to the interaction between mesenchymal ephrin B2 and endothelial EphB4 discussed earlier. In this context, an earlier study revealing distinct localization in neonatal rat cardiomyocytes of EphA3, whose abundance is reduced after exposure to the proinflammatory cytokine interleukin-1 (IL-1) (142), is of note. This study, which analyzed the modulation in expression of cytokine-responsive genes upon injury (heart failure), revealed the IL-1–mediated increase in the abundance of ADAM10 concurrent with a decrease in the abundance of EphA3. In light of ADAM10’s role in cell-cell contact–dependent cleavage of EphA3-bound ephrin (57), it is tempting to speculate that an increase in the ratio between ephrin A1 and EphA3 concurrent with an increase in ADAM10 abundance leads to repulsive signaling and increased mobility between mesenchymal and endothelial cells, which is required during regenerative tissue remodeling.

Weak EphA4 interactions mediate axon guidance in the spinal cord

Based on structural features (143, 144), EphA4 is a typical class A receptor. In particular, the H-I loop of the N-terminal β-sandwich jellyroll structure, which is critically involved in ligand binding, consists of only 7 amino acid residues in EphA receptors, as compared with 17 residues in EphB receptors (143). This region appears to mediate the low-affinity ephrin interaction, which is required for heterotetramer formation (47, 124). As for other Ephs, EphA4 binds to both ephrin A and ephrin B ligands, although with much lower affinity to the latter (83, 145).

In contrast to the countergradient mode of action discussed above, in several situations, EphA4 and its interacting ligands appear to function in discrete zones that exclude one cell population from another or act as repulsive barriers to pioneering axons. As another striking feature, EphA4 interacts in these developmental processes involving discrete zones or repulsive barriers with ephrin B ligands for which it has only very modest affinities.

The exclusion of cell populations is perhaps best exemplified by the critical role of EphA4 in hindbrain development. Indeed, EphA4 (also known as segmental Eph-like kinase-1) was initially isolated in a search for genes involved in hindbrain segmentation (146). Apart from its expression in alternating hindbrain rhombomeres, a dynamic pattern of expression was observed in the forebrain, spinal cord, and presomitic mesoderm. In the hindbrain, the presence of EphA4 in the r3 and r5 rhombomeres is contrasted with that of ephrin B in the alternating rhombomeres. It was demonstrated that EphA4 interacts preferentially with ephrin B2 to establish rhombomere boundaries through a contact repulsion mechanism (147).

Perhaps the best-characterized developmental role of EphA4 is in the formation of motor tracts in the spinal cord. The observation that EphA4^−/− mice have an abnormal gait led to the discovery of abnormal lateralization of developing corticospinal tract (CST) axons in these mice (148). Generally, once pioneering CST axons cross the midline, they do not recross it. The motor interneurons that regulate central pattern generation—the regulators of gait—are similarly patterned through the interaction of EphA4 with ephrin B3 (149). Using EphA4 mutant mice, researchers demonstrated that kinase-dependent EphA4 signaling triggered by ephrin B3, which is highly abundant in a uniform band at the midline, acts as an impenetrable barrier to prevent axons from recrossing the midline by causing growth cone collapse and repulsion (150). Axon collapse and repulsion involve the differential regulation of Rho family proteins (151) and, perhaps not surprisingly, the same CST axon guidance phenotype was recently observed in a spontaneous mouse mutant, termed miffy, in which disruption of the gene encoding the Rac-GAP α-chimerin results in defects similar to those observed in EphA4^−/− and ephrin B3^−/− mice (152). Elaboration of the underlying defect led to the identification of α-chimerin as a downstream target of ephrin-induced EphA4 signaling in motor neurons (152–154). Interestingly, knock-out of another downstream target of EphA4 signaling, the Rho guanine nucleotide exchange factor ephexin1, did not produce a neural phenotype, which suggests that there may be functional redundancy within the ephexin family (155).

A secondary phenotype in EphA4^−/− mice, which manifests as a partially penetrant abnormality in the formation of the anterior commissure (148), is not evident in mice that express a kinase-deficient mutant EphA4 (150) or in ephrin B3^−/− mice (156). Based on mRNA expression data, ephrin B2 is the ligand most likely to be found on the neurons (150). Given that the kinase function of EphA4 is not needed for this developmental process, this phenotype has been used as a paradigm for an axon guidance function that is mediated by ephrin B reverse signaling. However, a KI mouse that expresses mutant, constitutively active, EphA4 has no apparent abnormalities in any of the developmental processes elaborated in these KO studies, but instead shows defects in thalamocortical projections and subtle abnormalities in the central pattern generator mechanism (59). This phenotype has been interpreted as reflecting the need for ephrin-induced higher-order signaling clusters in the recruitment of adequate downstream signaling components, which does not occur with constitutively active receptors.

It seems somewhat counterintuitive that the most prominent examples for EphA4 function are relayed through its interactions with ephrin B ligands, despite EphA4 having much lower affinities for these ligands as compared with its affinities for ephrin A proteins (83, 145). However, there are examples for EphA4–ephrin-A–mediated regulation of developmental processes, which include neural crest migration (157) and limb bud development (158). Also, the interaction of EphA4 with ephrin A3, which has a modest affinity for EphA4, is critical for the development of correct hippocampal dendritic spine morphology (159).

Do Ephs affect proliferation?

One characteristic that is often regarded as a sine qua non for the cancerous state is uncontrolled cell proliferation. At the time of the discovery of Eph (EphA1) in cancer cells, it was therefore assumed that Eph RTKs, in common with other prominent RTK family members, would have direct effects on proliferation and be candidate oncogenes. Indeed, early reports suggested that EphA1 (Eph) behaved as a classical oncogene, which is consistent with a role in proliferation (164). Similarly, other studies suggested that EphA2 and ephrin A1 had an autocrine proliferative effect in malignant melanoma (165). However, many other studies failed to demonstrate any proliferative effects, and the overwhelming evidence suggested a prominent role for signaling of Ephs in cell positioning. This focused most of the attention on their role in regulating cell position and motility during tumor metastasis and neo-angiogenesis (39, 160, 162, 163, 166). Accordingly, the conspicuous expression of Ephs in stem cell or precursor cell populations in the adult brain, colon, and skin (23, 109, 167), and the proliferative effect of ephrin B2-Fc or EphB2-Fc on adult neuroblasts (23), were initially interpreted as secondary to a cell-positioning effect. However, recent studies suggesting that the absence of ephrin A2 reverse signaling in ephrin A2^−/− mice leads to an increased number of proliferating neural progenitor cells without affecting cell migration (25) clearly point to a direct antimitogenic role for ephrin A2 signaling. Other reports have suggested positive effects of Eph-ephrin signaling on the proliferation of neural precursors (168) and for intestinal stem cells (167). However, most studies have reported antiproliferative effects of Eph-ephrin signaling in both normal cells (168) and cancer cells (99, 170, 171). As with other aspects of Eph-ephrin biology, it appears that it may depend on the nature of the interacting proteins and the cellular context if Ephs and ephrins affect proliferation in a positive or negative manner.

Do Ephs promote or suppress tumors?

The high abundance of Ephs and ephrins in many human cancers—including various carcinomas, melanoma, sarcoma, kidney, and brain tumors—has been previously reviewed (160, 162, 163). In some cases, there is a clear link between this increased abundance and tumor progression, which is perhaps best exemplified in malignant melanoma (44, 172, 173), where the expression of at least one of three Eph genes is increased during melanoma progression. A mutational screen of melanoma, glioblastoma, and pancreatic cancer reported a potential functional mutation in EphA3 in melanoma and further mutations in glioblastoma (174). In glioblastoma, the increased abundance of EphA2, EphB2, and ephrin B3 is directly and functionally linked to cellular processes that mediate invasiveness (175–178). Reverse signaling by ephrin B1 is also critically important for tumor invasiveness of gastric scirrhous cancers, which is dependent on the activation of Rac1, as was observed for ephrin B3 signaling in glioblastoma (179).

However, emerging evidence suggests that the notion of the high expression of Eph and ephrin genes conferring a more invasive or metastatic phenotype on cancer cells is probably an oversimplification, particularly in the case of epithelial tumors (carcinomas) (34, 171, 180). A prominent example is EphB4, which seems to play contradictory roles in breast and colon cancer (26). On the one hand, the increased abundance of EphB2 and EphB4 and their ligand ephrin B2 has been implicated in breast and colon cancer (181–184), whereas on the other hand, EphB4 has a tumor-suppressive role in vitro (171), in mouse models (185), and in human colorectal cancer (180). Analysis of the EphB–ephrin B2 interaction in breast cancer cells and tumor xenografts revealed that activation of ephrin B2 by cytoplasmic truncated signaling-compromised EphB receptors may play a pro-tumorigenic role by promoting tumor angiogenesis (151). By contrast, retroviral overexpression of EphB4 in tumor xenografts and activation of EphB4 signaling switch the angiogenic program from one of sprouting to one of circumferential blood vessel growth and reduced vessel permeability, obvious features of a non-neoplastic vascularization program (186).

It is highly likely not only that the expression per se will determine pro- or anti-neoplastic roles of Ephs, but also that modulation of Eph function by somatic mutation may enhance their pro-oncogenic properties. This was shown for EphA3, which, alongside K-Ras and APC, was identified in a large-scale screen of somatic mutations as one of the three top colon cancer genes (187). Importantly, the modulation of Eph gene transcription during tumor progression also critically affects oncogenic outcomes. Expression of the genes encoding EphB2, EphB3, and their ligands is transcriptionally regulated by β-catenin and TCF to direct the sorting and migration of precursors and differentiating cell types within the normal and malignant gut epithelium (109). A gradient of decreasing concentrations of EphB2 and EphB3 from the colon crypt to the villus is countered by an increasing concentration gradient of ephrin B, and cell-repulsive signaling helps to regulate the flow of mature cells from the crypt to the villus surface. Although further data are required to fully understand the complex role of Eph and ephrin functions in this epithelial cancer, one model that fits the current data argues for a tumor-promoting role of Ephs and ephrins in the genesis of premalignant colon tumors (polyps) and in early cancers. However, as tumors progress, heterogeneity of Eph and ephrin gene expression and activating or inactivating mutations result in apparently contradictory outcomes. In some cases, the increased abundance of Ephs or ephrins persists, while in other tumors a loss of the increased Eph or ephrin abundance correlates with increased invasiveness and metastatic behavior. A recent paper now presents data compatible with such a model (188). The study reveals that the presence of EphB prevents invasiveness by mediating tumor cell repulsion away from normal cells that contain ephrin B. The neoplastic cells form tumors but do not invade. Upon loss of EphB protein, the tumor cells become capable of invading surrounding tissues (Fig. 4). Whether this reflects a difference in Eph-ephrin signaling in individual tumors, which in some cases is cell-repulsive and thus is likely to block invasiveness and in other cases is cell-adhesive and enhances tumor spread into surrounding tissue, remains unclear. However, this explanation would fit with the heterogeneity of Eph signaling found in different developmental processes. How Eph gene expression is lost during tumor evolution remains unclear. One mechanism that has been demonstrated in colorectal cancer for both EphA7 (189) and EphB2 (190) is epigenetic gene silencing. An observation that may serve to provide a further mechanism is inhibition of Wnt signaling by hypoxia (191). Here it was argued that hypoxia induces the activity of hypoxia-inducible factor 1α, which competes with TCF1 for binding to β-catenin, thus effectively suppressing transcription of Wnt-target genes such as EphB2 and EphB3. Such a mechanism would also make the EphB2 locus susceptible to permanent silencing by gene methylation. Whether this could be seen as a general mechanism for loss of Eph expression is unlikely, because EphA2, EphA3, and EphB4 appear to be up-regulated by hypoxia (192, 193).

• Download high-res image
• Open in new tab
• Download Powerpoint

A biphasic model, in which high Eph gene expression and oncogenic capacity early in tumor development give way to a loss of Eph gene expression to allow tumor invasion and metastasis, implies that Eph signaling is tumor-suppressive at this stage and fits well for epithelial cancers. However, this is not the case for mesenchymal tumors. Perhaps the most studied example of a mesenchymal tumor is melanoma. Melanocytes are neural crest derivatives that migrate throughout the embryo before differentiating into typical subepithelial melanocytes. In the skin, for example, these cells provide pigment to overlying keratinocytes. Eph-ephrin signaling plays a well-characterized role in the migration of neural crest derivatives (157, 194–196) and specifically of neural crest melanoblasts (197). The melanocyte-to-melanoma transition occurs initially in the skin but, as the tumor develops, it reacquires the high expression of Eph genes (44, 172, 198) and the migratory properties of the melanoblast. In studies of murine melanoma, high Eph abundance and signaling have been directly linked to tumor invasiveness. Thus, in this situation, Eph-ephrin signaling is clearly tumor-promoting (Fig. 5). Very limited evidence is available for sarcomas, but these reports also suggest a tumor-promoting role for Ephs in these tumors (199, 200).

• Download high-res image
• Open in new tab
• Download Powerpoint

Thus, as in other situations, Eph-ephrin signaling in cancer is not simple, and much more information is required to understand the heterogeneity of Eph and ephrin functions in neoplasia. The two scenarios posed here rest on the very different biology of epithelial tumors (carcinomas) as compared with that of "mesenchymal" tumors (melanoma, sarcoma, leukemia, and lymphoma). However, it seems certain that this two-pathway model of tumor development is an oversimplification and will require refinement in the face of new evidence.