Regulated Cleavage of a Contact-Mediated Axon Repellent

See allHide authors and affiliations

Science  25 Aug 2000:
Vol. 289, Issue 5483, pp. 1360-1365
DOI: 10.1126/science.289.5483.1360


Contact-mediated axon repulsion by ephrins raises an unresolved question: these cell surface ligands form a high-affinity multivalent complex with their receptors present on axons, yet rather than being bound, axons can be rapidly repelled. We show here that ephrin-A2 forms a stable complex with the metalloprotease Kuzbanian, involving interactions outside the cleavage region and the protease domain. Eph receptor binding triggered ephrin-A2 cleavage in a localized reaction specific to the cognate ligand. A cleavage-inhibiting mutation in ephrin-A2 delayed axon withdrawal. These studies reveal mechanisms for protease recognition and control of cell surface proteins, and, for ephrin-A2, they may provide a means for efficient axon detachment and termination of signaling.

Repulsion by direct cell contact is one of the basic mechanisms of axon guidance (1) and allows patterning of neural connections in a spatially precise manner (2, 3). However, this mechanism raises inherent questions. The binding of an axonal receptor to its cell surface ligand would be expected to favor adhesion, so how is this reconciled with repulsion? Also, how is contact-mediated repellent signaling terminated? These questions are further emphasized by the known properties of the ephrins, which are well-characterized cell surface axon repellents: The ephrins and their receptors are expressed at high density; the receptors do not appear to be down-regulated upon ligand binding; and the receptor-ligand interaction is multivalent, has a low off rate, and has a high affinity (2–5). Despite this, axon detachment and termination of signaling presumably have to be efficient because axons explore their embryonic environment in a dynamic fashion, involving both advances and withdrawals (6–8). One potential mechanism to reconcile contact-mediated signaling with repulsion could be the cleavage of ligand from the cell surface. However, because soluble truncated forms of ephrins cannot activate their receptors (9), unregulated cleavage could create problems of its own by destroying active ligand.

Many cell surface proteins undergo ectodomain shedding by proteolytic cleavage. Examples are the liberation of signaling molecules that are active in soluble forms, such as transforming growth factor–α, tumor necrosis factor–α (TNF-α), kit ligand, and Delta; adhesion molecule shedding; and shedding of the amyloid precusor protein (APP) implicated in Alzheimer's disease (10–12). ADAMs (a disintegrin and a metalloprotease) are proteases responsible for many of these shedding processes (13–15). ADAMs themselves are important for development, and Kuzbanian/ADAM10 (Kuz) was identified in a Drosophilagenetic screen as being required for normal axon extension (16). Despite rapid progress in identifying roles for the ADAMs, their regulation and ligand specificity are not well understood. Several treatments, such as protein kinase C (PKC) activators or calcium ionophores, are known to activate generalized shedding, though it is less clear whether the known pathways can target individual substrates without activating more general proteolysis. Also, there is very little sequence specificity at the substrate cleavage sites, and it is unclear how ADAM proteases specifically recognize their correct targets. In addition to a protease domain, ADAMs have disintegrin, cysteine-rich, and cytoplasmic domains, suggesting that these other domains might function in substrate binding (13, 14).

To investigate whether ephrin-A2, a membrane-bound protein with a glycosyl-phosphatidylinositol (GPI) lipid anchor, is cleaved from the membrane, we tested the effect of a soluble EphA3 receptor ectodomain fused to an immunoglobulin Fc tag (EphA3-Fc). When Neuro2a neuroblastoma cells expressing transfected ephrin-A2 were treated with clustered EphA3-Fc, ephrin-A2 disappeared from the cell membrane fraction, and a smaller form appeared in the supernatant (Fig. 1A). Unclustered EphA3-Fc had no effect. This result indicated ephrin-A2 is cleaved from the cell membrane in a mechanism regulated by binding to clustered receptor.

Figure 1

Clustered EphA3-Fc activates cleavage of ephrin-A2. (A) Neuro-2a cells expressing mouse ephrin-A2 (33) were incubated with no addition or with clustered control Fc tag (Fc), clustering antibody (Ab), unclustered EphA-Fc, or antibody-clustered EphA3-Fc. Cell lysate (Cell) and culture supernatant (Sup.) were collected after 2 hours and analyzed by immunoblotting with antibody to ephrin-A2 (38). Ephrin-A2 cleavage was activated by clustered EphA3-Fc. (B) Time course of ephrin-A2 release after the addition of clustered EphA3-Fc, 20 nM. Cleavage was first detectable by 10 min and was largely complete by 45 min. (C) Molecular weight (as kD) of ephrin-A2 released by clustered EphA3-Fc stimulation. The product was smaller than that released by PI-PLC (17). (D) Clustered EphA3-Fc–activated cleavage of ephrin-A2 was observed in primary cultured mouse E18 posterior midbrain neurons.

In time course experiments, no cleavage was seen until ∼10 min after EphA3-Fc addition, and cleavage was largely complete within 45 min (Fig. 1B). The molecular size of the released product is smaller than the soluble form released by phosphatidylinositol-specific phospholipase C (PI-PLC) by ∼1.9 kD (Fig. 1C), indicating that ephrin release was not due to phospholipase C or D activity and was due to cleavage within the polypeptide (17). Receptor-activated release of ephrin-A2 also occurred in NG 108 neuroblastoma cells, HEK 293T kidney epithelial cells, and NIH-3T3 embryonic fibroblasts (18), as well as in primary cultured neurons from embryonic day 18 (E18) mouse midbrain (Fig. 1D), a cellular context where ephrin-A2 and -A5 are known to function in axon guidance (2, 3, 19, 20). No cleavage activation was seen when ephrin-A2 was clustered by antibody (antibody to myc tag and secondary clustering antibody) (18), suggesting that clustering is not sufficient to activate cleavage and that EphA3-Fc may induce a specific regulatory conformational change.

The ADAM metalloprotease Kuz shows an axon-guidance phenotype inDrosophila (16), has been implicated in ectodomain shedding (13, 14), and is expressed in mammalian neural cell lines (21). To assess its expression pattern in mammalian development, we performed in situ hybridization, showing that Kuz RNA was widely expressed in the nervous system of E18 mouse embryos, with high expression in the posterior midbrain, which diminished toward the anterior midbrain (Fig. 2B). This pattern is reminiscent of the graded midbrain expression of ephrin-A2 and -A5 (Fig. 2B) (19,20). The cleavage of ephrin-A2 induced by EphA3-Fc was blocked by the metalloprotease inhibitor o-phenanthroline (Fig. 2A). A dominant negative version of Kuz (Kuz-DN) lacking the protease domain (12, 22) inhibited ephrin-A2 ectodomain shedding (Fig. 2C). Conversely, full-length Kuz (Kuz-FL) activated shedding (Fig. 2D). These results indicate that ephrin-A2 cleavage can be mediated by an ADAM protease and suggest that Kuz is a good candidate for a role in this process during development.

Figure 2

Interactions of ephrin-A2 and metalloprotease. (A) OPT, a zinc chelator that inhibits metalloprotease action, blocked the cleavage activation of ephrin-A2 by clustered EphA3-Fc. (B) Whole-mount RNA in situ hybridization of E18 mouse brain; the cortex has been removed to uncover the midbrain. Probes were Kuz or ephrin-A2 antisense RNA or were Kuz sense control. SC, midbrain superior colliculus; IC, midbrain inferior colliculus; Cb, cerebellum; and Th, thalamus. Kuz expression was widespread in the nervous system and was particularly prominent in the posterior midbrain, including the posterior SC, where ephrin-A2 is most prominent, and the IC, where ephrin-A5 (3) is most prominent. (C and D) The effect on receptor-activated ephrin-A2 cleavage of Kuz-DN, lacking the protease domain, or Kuz-FL. NIH 3T3 cells were transfected with the plasmids indicated; DNA amounts are shown in μg. Cleavage of ephrin-A2 was assayed 48 hours after transfection (38). Kuz-DN inhibited the receptor-induced ectodomain shedding of ephrin-A2, whereas Kuz-FL augmented it. stim., stimulation. (E) Ephrin-A2 and Kuz-FL form a stable complex. Stimulation with clustered EphA3-Fc was for 15 min. The complex of ephrin-A2 and Kuz was not dependent on the addition of clustered EphA3-Fc. GPI-anchored native placental AP (PLAP) was used as a negative control. IP, immunoprecipitation. (F) The complex of ephrin-A2 and Kuz involves sequences outside the juxtamembrane and protease domains, respectively. Ephrin-A2 formed a complex with Kuz-DN, lacking the protease domain. AP-ephrin-A2JM, containing the entire ephrin-A2 juxtamembrane domain, formed no detectable complex. (G) A conserved motif found in proteins whose cleavage can be mediated by Kuz (37). A single motif in all vertebrate ephrins and the other proteins shown was identified with the MEME program (24). The stretch of 10 amino acids showing the strongest conservation is demarcated by spaces. The 15–amino acid oligomer peptide sequence in ephrin-A2 used to make a synthetic peptide is underlined. (H) The ephrin-A2 cleavage assay was performed with the indicated concentrations of synthetic peptide. M, 15–amino acid oligomer peptide consisting of a motif derived from ephrin-A2, underlined in (G); C, a scrambled peptide with the same amino acid composition, TEFFPPGKKLQFFSL. The ephrin-A2 15–amino acid oligomer activated ephrin-A2 cleavage in the absence of EphA3-Fc and caused, with a peak of activity at ∼10 μM, a synergistic activation of cleavage in the presence of EphA3-Fc.

Although it has been proposed that ADAM proteases might form a stable complex with the substrate to be cleaved, such a complex has not been reported (13, 14). Co-immunoprecipitation from transfected cells revealed that a complex was formed between ephrin-A2 and Kuz (Fig. 2E). This complex was seen under conditions where lipid rafts would be disrupted (23), and placental alkaline phosphatase (AP), which, like ephrin-A2, is a GPI-anchored protein, showed no coprecipitation with Kuz (Fig. 2E). The complex of ephrin-A2 and Kuz was seen in the absence of EphA3-Fc. The addition of clustered EphA3-Fc did not cause any short-term modulation of the complex (Fig. 2E), although after longer periods of time, less complex was detected as ephrin-A2 was cleaved (18). The association was not dependent on the protease domain of Kuz, because Kuz-DN, which lacks the pro-domain and protease domain, still associated with ephrin-A2 (Fig. 2F). Furthermore, although Kuz-DN associated efficiently with ephrin-A2, it formed no detectable association with AP–ephrin-A2JM, a protein containing the juxtamembrane region of ephrin-A2 but excluding the receptor-binding core region (Figs. 2F and 3A), indicating that complex formation involves sequences outside the juxtamembrane cleavage region. The complex of ephrin-A2 with Kuz could involve direct or indirect binding.

Figure 3

Receptor activation of ephrin-A2 cleavage is not accompanied by general cleavage of membrane proteins and is blocked by protein tyrosine kinase inhibitors. (A) Schematic diagram of ephrin-A2 constructs (33). Core, receptor-binding core domain conserved throughout the ephrin family; JM, juxtamembrane domain of ephrin-A2 with arrowhead denoting the cleavage site; AP, alkaline phosphatase; and GFP, green fluorescent protein. The horizontal line denotes the GPI membrane anchor, and the dark boxes indicate myc epitope tags. (B) The indicated plasmids were transfected into NIH-3T3 cells and incubated for 2 hours with clustered control-Fc (open bar), clustered EphA3-Fc (shaded bar), 1 μM PMA (hatched bar), or PI-PLC (solid bar). Culture supernatants were collected, and AP activity was quantitated. Clustered EphA3-Fc treatment activated cleavage of ephrin-A2 but not coexpressed AP-ephrin-A2JM. (C) Neuro-2a cells expressing mouse ephrin-A2 were surface labeled with sulfo-NHS-biotin (0.2 mg/ml for 30 min at 4°C in PBS) and stimulated with either clustered control Fc or clustered EphA3-Fc. Culture supernatants were blotted with HRP-streptavidin (upper panel) or anti-ephrin-A2 (lower panel). Clustered EphA3-Fc induced ephrin-A2 release (lower panel), and a band at the size of ephrin-A2 is visible in the sulfo-NHS-labeled fraction (asterisk in upper panel), but generalized ectodomain shedding was not seen. Molecular weights are shown as kD. (D toF) Localized shedding of GFP–ephrin-A2 fluorescence. Fluorescence micrographs are on the left, and phase-contrast images are on the right (39). The paired images were taken within 30 s, although they are not precisely superimposable because of axon motility. At the site of axonal contact (black arrowhead), there was a localized dispersal of fluorescence from the target cell, seen in (D) at approximately the time of growth cone collapse. As the axon withdrew 15 min later (E), the dispersed fluorescence was no longer easily seen. However, in a brighter version of the same image (F), faint staining could be seen over the axon (white arrows); this was not seen in axons that had not contacted a fluorescent cell. (G) PMA (1 μM), an activator of PKC, activates cleavage of ephrin-A2. However, H-7 (20 μM), a powerful inhibitor of PKC, does not suppress the receptor-induced cleavage of ephrin-A2 caused by clustered EphA3-Fc. (H) Genistein, a protein tyrosine kinase inhibitor, suppressed the receptor-induced activation of ephrin-A2 cleavage caused by clustered EphA3-Fc.

A motif search with the MEME program (24) was applied to see if proteins that are cleaved by Kuz might have common sequences. A conserved motif was found in all of the proteins tested, including all eight vertebrate ephrins, Delta, TNF-α, and APP (Fig. 2G). This conserved motif is located roughly in the middle of the receptor-binding core domain of the ephrins. A 15–amino acid oligomer synthetic peptide containing the sequence in ephrin-A2 activated ephrin-A2 shedding up to five times that of a scrambled-sequence peptide or no peptide. Some activation was seen in the absence of EphA3-Fc, and there was a further synergistic activation when the peptide was added together with EphA3-Fc (Fig. 2H). The effect of this peptide might involve interaction with ephrin-A2, metalloprotease, or a separate protein that could mediate their interaction. Because this motif is in the middle of the receptor-binding domain, it could be involved in the mechanism of receptor-induced cleavage activation.

Several agents, such as PKC activators, calcium ionophores, or tyrosine phosphatase inhibitors, induce a generalized ectodomain shedding (10, 11). Also, several leukocyte surface markers are down-regulated after leukocyte activation or antibody cross-linking, although the effect of cognate biological ligands and the specificity of this process is unclear (25–27). To assess whether the cleavage of ephrin-A2 induced by EphA3-Fc receptor is specific to the cognate ligand, we first tested for cross-activation of cleavage of the AP–ephrin-A2JM fusion protein (Fig. 3A). Although native placental AP was not shed by stimulation with the PKC activator phorbol myristate acetate (PMA), the presence of the ephrin-A2 juxtamembrane region in AP–ephrin-A2JM allowed cleavage in response to PMA (Fig. 3B). However, when ephrin-A2 cleavage was activated by EphA3-Fc, cleavage of coexpressed AP–ephrin-A2JM was not cross-activated (Fig. 3B). Likewise, EphA3-Fc activation of ephrin-A2 cleavage failed to activate cleavage of coexpressed ephrin-B1, which does not bind EphA3, even though ephrin-B1 cleavage was efficiently activated by PMA (18). To further assess specificity, we labeled cell surface proteins of Neuro2a cells with sulfo-NHS-biotin (NHS, N-hydroxysuccinimide) before treatment with EphA3-Fc. Cleaved ephrin-A2 appeared in the supernatant, but no other biotin-labeled cleavage products were evident, indicating that there was not a generalized activation of ectodomain shedding (Fig. 3C).

To investigate whether ephrin-A2 shedding is limited to sites of cell-cell contact, we made a fusion protein that was composed of green fluorescent protein and ephrin-A2 (GFP–ephrin-A2) (Fig. 3A). GFP–ephrin-A2 retained the ability to bind EphA3 receptor (18) and was expressed over the entire surface of transfected cells (Fig. 3D). These cells were challenged with axons of medial hippocampal neurons, which are repelled by ephrin-A2 (28,29) and can be grown easily as a dispersed culture and observed in real time. Where cells expressing GFP–ephrin-A2 were contacted by an axon, fluorescence dispersed locally from the target cell surface and was seen over the axon and surrounding area (Fig. 3, D to F). However, there was no indication of a generalized dispersal of fluorescence from the rest of the cell.

Although PMA can activate ephrin-A2 cleavage, a strong PKC inhibitor, H-7, did not inhibit cleavage activated by EphA3-Fc (Fig. 3G), indicating that the receptor-regulated mechanism does not involve PKC activation. In contrast, protein tyrosine kinase inhibitors, such as genistein and herbimycin-A, did inhibit EphA3-Fc activation of cleavage (Fig. 3H). PP2, an inhibitor specific for Src-family kinases (30), had no obvious effect, however. These results appear consistent with the ability of vanadate, a tyrosine phosphatase inhibitor, to cause a general activation in cleavage of surface proteins (31), including ephrin-A2 (18), indicating that tyrosine kinases are either involved in or can modulate the receptor-induced pathway of ephrin cleavage.

To determine whether ephrin-A2 cleavage affects axon behavior, we created a proteolysis-blocking mutation in ephrin-A2. As with other proteins that undergo ectodomain shedding, point mutations, insertions, or deletions in the juxtamembrane region of ephrin-A2 typically had little or no effect on cleavage. However, two constructs, ephrin-A2IS2 and ephrin-A2IS3, which have a short insertion in the juxtamembrane region (Fig. 4A), were not cleaved by EphA3-Fc stimulation or PKC activation, although their binding affinity for the EphA3 receptor was unaffected (Fig. 4B) (18). Stable transformant cells expressing wild-type or uncleavable ephrin-A2IS3 were isolated. These lines showed similar ephrin-A2 protein expression, and doxycycline removal allowed expression levels to be induced and controlled (32, 33). Axon responses to these target cells were assessed by using medial hippocampal neurons. “Growth cone collapse” was defined as full collapse with no obvious growth cone structure remaining, and “withdrawal” was defined as no discernible connection between axon and target cell. Cells that did not express ephrin-A ligands caused no growth cone collapse (n = 10 axon trials). Upon ephrin expression, both the ephrin-A2 and ephrin-A2IS3 cell lines caused growth cone collapse within 10 min of first contact (n = 21 axons in each group). With wild-type ephrin-A2, axon withdrawal followed growth cone collapse with an average time of 26.6 ± 14.8 min (mean ± SD) (Fig. 4, C, E, and F). With ephrin-A2IS3 cells, axon withdrawal was greatly delayed, with an average time of 72.5 ± 36.0 min (P < 0.0001, unpaired t test) (Fig. 4, D to F). In about half of these cases, the axon remained collapsed and stuck to the target cell for over 80 min (Fig. 4, D and E), whereas all of the control axons had withdrawn by this time (Fig. 4, C and E). It is unlikely that the differences in time course can be explained by the expression level of ephrin-A2, because doxycycline modulation of the expression levels over a several-fold concentration range had no discernible effect on axon withdrawal (18). A delay was seen even when the initial contact leading to collapse involved only one to three filopodia (wild-type ephrin-A2, n = 10, 30.3 ± 15.1 min; ephrin-A2IS3, n = 9, 73.9 ± 29.9 min; P < 0.005) (Fig. 4F) and similarly when it was more extensive (wild-type ephrin-A2, n = 8, 26.1 ± 16.4 min; ephrin-A2IS3, n = 11, 64.8 ± 36.0 min; P < 0.01). The kinetics of growth cone collapse and withdrawal when wild-type ephrin-A2 was used in this assay appear consistent with the rate of ephrin-A2 cleavage induced by clustered EphA3-Fc, especially considering that the rate of receptor-ligand association has to be factored into the EphA3-Fc experiments. The kinetics also appear comparable to real-time studies of axon behavior in vivo [for example, (6, 8)] as well as to ephrin-mediated withdrawal induced by primary cultured neurons or glia in vitro (34). Our results show that a mutation that blocks ephrin cleavage does not prevent signaling leading to growth cone collapse but does inhibit axon withdrawal. Because growth cones never recovered from collapse while they remained bound to a target cell (Fig. 4D), the cleavage-inhibiting mutation evidently interferes with their ability to terminate ephrin signaling. In addition, because axons can withdraw rapidly in response to soluble clustered ephrin-Fc protein (35) or membrane suspensions bearing ephrins, including ephrin-A2IS3 (18, 36), it is likely that at least part of the delay in withdrawal seen here with intact target cells is due to an impairment of axon detachment.

Figure 4

Uncleavable mutant of ephrin-A2 delays axon withdrawal. (A) Amino acid sequence of ephrin-A2 insertion mutants (37). Inserted sequences are in brackets. All constructs here, including the wild-type sequence, were myc epitope tagged at the NH2-terminus (33). (B) To test for cleavage, we transfected the mutant plasmids into NIH-3T3 cells and challenged them with EphA3-Fc. The ephrin-A2IS1insertion mutant, as well as several point mutants and deletions not shown here, was still cleaved. However, ephrin-A2IS2 and ephrin-A2IS3 were not detectably cleaved. WT, wild type. (C and D) Time-lapse videomicroscopy (39) of typical mouse hippocampal neurons contacting TetOff cells that were stably transfected with ephrin-A2 (C) or ephrin-A2IS3 (D). Time is given in minutes from the first contact. In (C), the neuronal growth cone collapsed by 6 min and had fully withdrawn by 26 min; in (D), the growth cone collapsed by 8 min and remained collapsed but had not withdrawn from the target cell by 120 min. Scale bars, 25 μm. (E and F) Histograms of the interval between the times of growth cone collapse and axon withdrawal. Open bars indicate ephrin-A2–expressing cells, and solid bars indicate ephrin-A2IS3–expressing cells. (E) Axons were divided into five groups with different withdrawal times. (F) Mean withdrawal times (error bars show standard errors) for all axons and for the subset of axons where the initial contact leading to collapse involved only one to three filopodia. The ephrin-A2IS3 mutation delayed axon withdrawal.

A stable complex formed by ADAM protease and ephrin in the absence of Eph receptor is likely to play a role in the specificity of protease-substrate recognition, as well as in enhancing the efficiency and localization of subsequent cleavage. The triggering of proteolytic activity by clustered receptor provides a mechanism to tightly coordinate cleavage with repulsion, because Eph receptor signaling is triggered by clustered ligand. The specificity of the receptor-activated cleavage mechanism for the cognate ligand provides a way to limit cleavage to molecules at the site of cell-cell contact, thus allowing the target cell to retain its repellent properties, sparing other cell surface proteins. The proposed mechanism can also account for the termination of repellent signaling because truncated soluble ephrins fail to activate Eph receptors (9). In addition to axon repulsion, the cleavage mechanism described here could be involved in contact-mediated axon attraction by ephrins or other molecules and could potentially provide a way to control whether cell contacts are repellent or adhesive. The regulatory mechanisms described here may also provide new ways to understand or manipulate other cleavage events mediated by ADAM proteases with roles in development, physiology, and disease.

  • * Present address: Department of Molecular Neurobiology, Institute of Medical Science, University of Tokyo, Minato-ku, Tokyo, Japan.

  • To whom correspondence should be addressed. E-mail: flanagan{at}


View Abstract

Navigate This Article