Ephrin-A5 potentiates netrin-1 axon guidance by enhancing Neogenin availability

Axonal growth cones are guided by molecular cues in the extracellular environment. The mechanisms of combinatorial integration of guidance signals at the growth cone cell membrane are still being unravelled. Limb-innervating axons of vertebrate spinal lateral motor column (LMC) neurons are attracted to netrin-1 via its receptor, Neogenin, and are repelled from ephrin-A5 through its receptor EphA4. The presence of both cues elicits synergistic guidance of LMC axons, but the mechanism of this effect remains unknown. Using fluorescence immunohistochemistry, we show that ephrin-A5 increases LMC growth cone Neogenin protein levels and netrin-1 binding. This effect is enhanced by overexpressing EphA4 and is inhibited by blocking ephrin-A5-EphA4 binding. These effects have a functional consequence on LMC growth cone responses since bath addition of ephrin-A5 increases the responsiveness of LMC axons to netrin-1. Surprisingly, the overexpression of EphA4 lacking its cytoplasmic tail, also enhances Neogenin levels at the growth cone and potentiates LMC axon preference for growth on netrin-1. Since netrins and ephrins participate in a wide variety of biological processes, the enhancement of netrin-1 signalling by ephrins may have broad implications.

ephrin-B2 via a molecular complex containing Unc5c and EphB2. Its binding of netrin-1 and ephrin-B2 results in the increased activation of Src family of kinase effectors of netrin-1 and ephrin signalling, beyond that evoked by the presence of netrin-1 or ephrin-B2 alone. Lateral LMC axons also respond to netrin-1 and ephrin-A5 in a synergistic fashion, but the cellular and molecular mechanism of this effect remains elusive 18 .
Netrin-1 signalling specifies a wide variety of axon guidance decisions, often by acting in concert with other guidance cues 19,20 . For example, at the developing spinal cord midline, acting through their Robo receptors, Slit proteins silence netrin-1 attraction in commissural and motor axons 21,22 . In thalamocortical (TC) growth cones, Slit1 signalling via Robo1 and FLRT3 raises the levels of DCC allowing netrin-1 attraction, such that in the absence of Slit1, TC axons are unresponsive towards netrin-1 23,24 . These and above studies suggest that netrin-1 signalling through its attractive receptors depends on the action of other axon guidance signals, prompting us to examine the mechanism of netrin-1 and ephrin-A integration by LMC axons. Our results demonstrate that ephrin-A5 induces an increase in Neogenin abundance in LMC growth cones through its receptor EphA4, sensitizing lateral LMC axons to netrin-1. This effect occurs in the absence of the intracellular signalling tail of EphA4, demonstrating that lateral LMC axon repulsion from ephrin-A5 and sensitization to netrin-1 occur through molecularly distinct pathways.

Ephrin-A5 increases Neogenin and EphA4 protein levels in LMC growth cones. Chicken motor
axon attraction to netrin-1 is mediated through its attractive receptor Neogenin 18 . To investigate the possibility that ephrin-A5 sensitises LMC axons to netrin-1 by increasing Neogenin abundance in growth cones, we measured the relative Neogenin protein levels in LMC growth cones by immunofluorescence (IF) using an anti-Neogenin affinity-purified polyclonal antiserum directed against the extracellular domain of Neogenin (anti-Neogenin polyclonal antibody) 18 . HH st. 24-25 LMC explants were incubated over-night and treated for 15′ with various concentrations of ephrin-A5 ( Fig. 2a-e). The average area of the growth cones selected for analysis did not differ between treatments ( Supplementary Fig. 1). However, compared to Fc control treatment, application of pre-clustered ephrin-A5 at 50 and 100 ng/mL, resulted in, respectively, 1.7 ± 0.3-fold and 2 ± 0.3-fold increases in the levels of Neogenin IF ( Fig. 2a; p = 0.03, and p = 0.01), while higher ephrin-A5 concentration did not cause a significant rise in Neogenin IF ( Fig. 2a; 250 ng/ml: p = 0.08, 500 ng/mL: p = 0.93). Similar Neogenin IF increase was also detected using a polyclonal anti-Neogenin antiserum directed against the cytoplasmic tail and extracellular Neogenin: ephrin-A5 100 ng/mL exposure resulted in a 1.7 ± 0.2-fold increase in intracellular Neogenin IF (Fig. 3a- To characterize the dynamics of the ephrin-A5-induced rise in Neogenin IF, explants were subject to 100 ng/ mL ephrin-A5 for 10′, 20′ and 40′. The Neogenin IF increase occurred within 10′ of exposure to ephrin-A5 and remained as such after 40′ (Fig. 2f). Furthermore, even a 5′ exposure to ephrin-A5 at 500 ng/mL resulted in a 1.7 ± 0.2-fold increase in Neogenin IF ( Fig. 2g; p = 0.02).
To assess the specificity of these effects, we also quantified the consequence of ephrin-A5 treatment of LMC growth cones on IF levels of the ephrin-A5 receptor EphA4, the netrin-1 repulsive receptor Unc5c, as well as the surface glycoprotein BEN 10,18,28 . Ephrin-A5 addition significantly increased EphA4 IF, possibly be due to an increase in EphA4 trafficking to the growth cone plasma membrane resulting from ephrin-A5-EphA4 interactions 29 . It is also possible that the oligomerization of EphA4 when bound to clusters of ephrin-A5 could contribute to increasing EphA4 IF 30 (1.9 ± 0.2-fold, p = 0.0001) without affecting Unc5c or BEN IF levels (Fig. 2h). Ephrin-A5 treatment did not significantly alter the area of growth cones included in the IF analysis ( Supplementary Fig. 1a-d). Together, these data suggest that the increased LMC growth cone sensitivity to netrin-1 following ephrin-A5 treatment could be due to an increase in Neogenin abundance.

Ephrin-A5 increases neogenin protein abundance on the surface of growth cones. Ephrin-A5
induced LMC axon sensitization to netrin-1 may result from increased abundance of Neogenin on the surface of growth cones. To test this possibility, we applied the anti-Neogenin polyclonal antiserum with either ephrin-A5 at 50 ng/mL or MN media as control, to live LMC growth cones for 20′, followed by standard fixation. Application of an antibody against the intracellular protein β III tubulin, did not result in any labelling, suggesting that this treatment results in specific detection of cell surface epitopes ( Supplementary Fig. 2a-h). A 20′ treatment with ephrin-A5 resulted in a 1.6 ± 0.2-fold increase in surface Neogenin IF (Fig. 4a-c; p = 0.04), arguing that ephrin-A5 application results in increased of Neogenin protein levels on the surface of LMC growth cones.
Ephrin-A5 enhances netrin-1 binding in growth cones. We next sought to determine whether increased LMC growth cone surface Neogenin IF levels might result in increased netrin-1 binding to LMC growth cones. LMC explants were incubated with netrin-1 and ephrin-A5 or netrin-1 alone as control. After fixation, a monoclonal antibody against netrin-1 was used to detect relative netrin-1 binding (Fig. 4d-g). Compared to a netrin-1 treatment, the addition of ephrin-A5 resulted in increased levels of netrin-1 IF in LMC growth cones ( Fig. 4d-g, 1.5 ± 0.1-fold increase, p = 0.0003), without a change in growth cone size ( Supplementary Fig. 1). These result suggest that the sensitization of LMCl axons to netrin-1 by ephrin-A5 (Fig. 1a,b) may be a consequence of enhanced netrin-1 binding in LMC growth cones.

Ephrin-A5 induces an increase in the co-localization of Neogenin and EphA4 IF in LMC growth
cones. Following ephrin-A5 treatments, we noticed the formation of dense EphA4 IF puncta in LMC growth cones (Fig. 2n), likely resulting from ligand-induced EphA4 clustering 30 . To investigate the possibility that the ephrin-A5 increase in EphA4 and Neogenin IF may be spatially correlated, we analysed the co-localization of Neogenin and EphA4 IF in LMC growth cones. LMC neuron explants were treated with either MN media, ephrin-A5 or netrin-1 (at 100 ng/mL) followed by immunostaining for EphA4 and either Neogenin, BEN or Unc5C in LMC growth cones. Treatment with ephrin-A5 resulted in an increase in the co-localisation of Neogenin with EphA4 IF, but not with Unc5c or BEN IF (Fig. 6a  www.nature.com/scientificreports www.nature.com/scientificreports/ p = 0.08; 1.03 ± 0.1-fold, p = 0.45 respectively), suggesting that the increase in Neogenin protein levels may be occurring in close proximity to ephrin-A5-EphA4 clusters ( Fig. 6b-i).

Ephrin-A5 -EphA4 binding is required for the growth cone elevation of Neogenin by ephrin-A5.
The ephrin-A5 induced increase in EphA4 IF and Neogenin-EphA4 IF co-localization in LMC growth cones suggests that the increase in Neogenin IF may be dependent on ephrin-A5-EphA4 interactions. To test this idea, ephrin-A5-EphA4 interactions were blocked using a 12-amino-acid peptide (KYL) that selectively binds EphA4 at its ephrin binding domain and inhibits ephrin-A5/EphA4 interactions without affecting ephrin-A5 binding to other Eph receptors 33 . LMC explants were treated with either MN media, KYL peptide at 50 µM or KYL peptide at 12.5 µM for 20′ prior to a 20′ exposure to either MN media or ephrin-A5 at 100 ng/mL ( Fig. 6j-m). The pre-incubation of LMC explants with either KYL at 12.5 or 50 µM abolished the ephrin-A5 induced increase in Neogenin IF, EphA4 IF and in the co-incidence of Neogenin and EphA4 IF ( Fig. 6j-m). These results argue that the LMC growth cone increase in Neogenin IF caused by ephrin-A5, requires its binding to EphA4.

EphA4 potentiates the ephrin-A5-led increase in neogenin abundance. Since blocking EphA4
function resulted in the attenuation of ephrin-A5-induced Neogenin IF increase, we reasoned that increasing EphA4 expression levels in LMC growth cones would increase the ephrin-A5-mediated induction of Neogenin. To test this idea, GFP expression plasmids alone or together with mouse EphA4 expression plasmids were www.nature.com/scientificreports www.nature.com/scientificreports/  www.nature.com/scientificreports www.nature.com/scientificreports/ introduced into chick neural tubes at HH st. 18-19 by in ovo electroporation 10,34 . GFP-expressing LMC neurons were explanted at HH st. 24-25, treated with MN media, netrin-1 or ephrin-A5 (either at 100 mg/mL) and Neogenin IF levels were quantified in GFP+ growth cones (Fig. 6n-t). Ephrin-A5 treatment of EphA4 and GFP overexpressing LMC neurons led to a 3 ± 0.3-fold increase in Neogenin IF compared to LMC neurons overexpressing GFP only (Fig. 6n,q,t, p = 0.007), arguing that EphA4 is sufficient to potentiate the ephrin-A5-induced increase in Neogenin IF.
To gain further insights in the mechanism behind the ephrin-A5-dependent increase in Neogenin IF, we used pharmacological inhibitors of specific cellular processes, subjecting LMC explants overexpressing EphA4 to these for 20′ prior to ephrin-A5 and netrin-1 treatment, followed by growth cone Neogenin IF quantification ( Supplementary Fig. 2i). To assess the possible role of PKA, we used the PKA inhibitor KT5720 (5 µM) 35 , to block protein synthesis we used anisomycin (80 µM) 36 . Proteasomal and lysosomal Neogenin degradation was blocked by MG132 (20 µM) 37 and chloroquine (20 µM) 38 , respectively. Since both EphA4 and Neogenin can be cleaved by γ-secretase 39,40 , LMC explants were treated with the γ-secretase inhibitor DAPT 41 . Finally, since Src family kinases are activated downstream of both EphA4 and Netrin-1 receptors 42-46 , LMC explants were treated with the Src family inhibitor SU6656 47 . None of these treatments resulted in a significant increase in Neogenin IF when treated with MN media or attenuated the increase in Neogenin levels following ephrin-A5 treatment ( Supplementary Fig. 2i).
The ephrin-A5-induced upregulation of neogenin does not depend on protein synthesis or proteolysis. The above lack of inhibitor effects could be because these were used in the context of LMC growth cones overexpressing EphA4, whose elevated signalling might be impervious to a partial pharmacological blockade, prompting us to re-examine protein and protease inhibition in the context of wild type LMC explants. Thus, if either process is involved in the augmentation of Neogenin protein levels by ephrin-A5, their blockade may also affect Neogenin levels. LMC explants were pre-treated for 20′ with either DMSO or the protein synthesis inhibitor cycloheximide (CHX) 50 µM, followed by a 5′ treatment with either Fc or ephrin-A5 500 ng/mL (Fig. 7a). The ephrin-A5-induced increase in Neogenin IF was evident even in the presence of CHX and did not significantly differ from DMSO pre-treatment (Fig. 7a, p = 0.418). To investigate the short-term effects of blocking proteolytic cleavage on Neogenin and EphA4 levels, LMC explants were subjected to a broad-spectrum commercial protease inhibitor cocktail 48 and after a 20′ treatment, F-actin, EphA4 and Neogenin IF were not significantly altered relative to DMSO treatment (Fig. 7b, 1.19 ± 0.1-fold, p = 0.08; 1.23 ± 0.1-fold, p = 0.06; 1.28 ± 0.1-fold, p = 0.08 respectively). The ectodomain cleavage of Dcc, a member of the immunoglobulin domain family closely related to Neogenin 32 , can be blocked by the broad-spectrum metalloprotease inhibitor GM6001 49 . However, relative to DMSO, a 20′ treatment with GM6001 failed to increase EphA4 and Neogenin IF ( Supplementary Fig. 2m,  1.3 ± 0.1-fold p = 0.1; 1.5 ± 0.2-fold, p = 0.07 respectively). Furthermore, since potential proprotein convertases cleavage sites are present in both EphA4 and Neogenin 50 , we treated LMC explants with ephrin-A5 and the broad spectrum proprotein convertase inhibitor RVKR 51 . This resulted in 1.3 ± 0.1-fold increase in Neogenin IF without affecting EphA4 IF ( Supplementary Fig. 2l, p = 0.03 and p = 0.2 respectively). Altogether, these results suggest that a 20′ inhibition of protein synthesis or proteolytic cleavage does not substantially increase Neogenin abundance in LMC growth cones.
The cytoplasmic tail of EphA4 is dispensable for potentiating the ephrin-A5-induced increase in neogenin abundance. Next, we asked if the intracellular domain of EphA4, required for the relay of signals in ephrin:Eph forward signalling 52 , is also required for the EphA4-induced potentiation of ephrin-A5-mediated Neogenin IF increase. To do so, we explanted LMC neurons from chick spinal cords electroporated with either expression plasmids encoding an EphA4 and GFP fusion protein (EphA4-GFP) or a truncated EphA4 missing the intracellular domain and GFP fusion protein (EphA4ΔICD-GFP) 10 . EphA4 immunofluorescence and GFP fluorescence was used to confirm fusion protein expression levels (Fig. 8b,c). Both sets of explants were treated with either MN media or ephrin-A5, and Neogenin IF levels were quantified in growth cones (Fig. 8a). In line with ephrin-A5-led induction of EphA4 receptor clustering, in EphA4-GFP-expressing growth cones, ephrin-A5 exposure led to the formation of patches of GFP signal, that also co-localized with Neogenin ( Fig. 8e-g). In contrast, in growth cones expressing EphA4ΔICD-GFP and treated with ephrin-A5, GFP signal was more diffuse, in line with the requirement for the intracellular domain of EphA4 for ephrin-A5 induced EphA4 clustering (Fig. 8e,k) 30 . Surprisingly, EphA4ΔICD-GFP-expressing growth cones exposed to ephrin-A5 showed a robust increase in Neogenin IF when compared to controls (Fig. 8a,j,m, 6.6 ± 0.6-fold induction, p = 0.0006), which was indistinguishable from that observed in EphA4-GFP expressing growth cones (Fig. 8a,d,g, 5.1 ± 0.5-fold induction; p = 0.3). These results suggest that the ephrin-A5 increase in Neogenin IF does not require canonical EphA signalling through the EphA4 cytoplasmic tail.

Discussion
Our experiments provide evidence that ephrin-A5:EphA4 interaction sensitizes spinal motor axons growth cones to netrin-1 by increasing the abundance of Neogenin, thus enhancing their netrin-1 binding and guidance responses. Further evidence suggests that these effects occur outside of canonical ephrinA:EphA signalling. Below, we discuss the potential molecular mechanisms underlying these findings, and, their in vivo relevance to motor axon guidance, and other functions of Neogenin and netrin-1 signalling.

Mechanisms of ephrin-A5 induced neogenin upregulation. While our experiments strongly support
the idea that ephrin-A5 exposure of LMC growth cones results in increased surface levels of Neogenin, we have yet to develop a complete mechanistic understanding of this phenomenon. The time-frame of the effect is on the order of minutes, and compatible with the idea of an axon guidance cue causing a rapid increase in growth cone protein translation 53,54 . However, the ephrin-A5-evoked increase in growth cone Neogenin occurs in the presence of protein synthesis inhibitors, arguing against the involvement of protein synthesis. Suppression of Neogenin degradation also does not appear to be involved since a variety of protease, proteasome and lysosomal degradation inhibitors failed to increase Neogenin levels. One remaining possibility is that cell surface Neogenin levels are controlled through its trafficking. Forskolin, an activator of adenylate cyclase, causes an increase in cAMP, in turn enhancing PKA activity 55 . In the growth cones of cultured mouse cortical neurons, Forskolin induces an increase in DCC IF 56 . The Forskolin induction in DCC IF can be blocked by tetanus toxin (TeNT), a potent inhibitor of a subset of VAMP SNARE proteins that mediate vesicle fusion during exocytosis 56,57 . In chick dorsal root ganglion neurons in vitro, Inhibition of VAMP2 with the use of TeNT abolishes growth cone attraction 58 . In LMC growth cones overexpressing EphA4 and treated with ephrin-A5, inhibition of PKA through treatment with KT5720 did not alter Neogenin IF. Furthermore, Forskolin treatment failed to increase Neogenin IF in LMC growth cones ( Supplementary Fig. 2i,k). Pretreatment of explants with TeNT also failed to block the ephrin-A5 induction in Neogenin IF (Supplementary Fig. 2j). Taken together our data suggests that the ephrin-A5 induction in Neogenin IF occurs independently from cAMP, PKA and tetanus toxin sensitive exocytosis. Interestingly, botulinum toxin, by cleaving the SNARE protein Syntaxin 1, abolishes netrin-1 axon guidance and neural migration in vitro 59,60 . The possible blockade of the ephrin-A5 increase in Neogenin IF through botulinum toxin should be investigated in the future. In Cos-7 cells overexpressing a fluorescently-tagged EphA2, ephrin-A1 induces a rapid translocation of EphA2 from Rab11+ recycling endosome to the plasma membrane, followed by a reduction in EphA2 cell membrane levels and its appearance in Rab5+ late endosomes 29 . Considering that this experiment was performed using ephrin-A1 at 2 µg/mL whereas our experiments were mostly done at an ephrin-A concentration 20 times lower, it is possible that treatment with lower concentrations of ephrin-As would result in a prolonged increase in plasma membrane levels of EphA4 and less robust endocytosis 61 . In this context, we can hypothesize that in LMC growth cones, ephrin-A5 could induce the trafficking of EphA4-carrying vesicles to the growth cone cell membrane, and if such vesicles also contain Neogenin, this might result in increased cell surface Neogenin levels.
Our results show that the ephrin-A5-induced Neogenin upregulation is potentiated by EphA4 and depends on its ephrin ligand binding domain but, surprisingly, not its intracellular domain. Although our experiments do not formally rule out the possibility that endogenous EphA4 is participating in increasing Neogenin IF when explants overexpressing truncated EphA4 (EphA4ΔICD-GFP) are treated with ephrin-A5, we provide strong evidence that the cytoplasmic tail of EphA4 is dispensable. The magnitude of the increase in Neogenin IF in growth cones overexpressing EphA4ΔICD-GFP is as high as in growth cones overexpressing full-length EphA4 (EphA4-GFP) and higher than in any experiments where only endogenous EphA4 is present. If EphA4ΔICD-GFP were not sufficient to increase Neogenin IF, we would expect an attenuation of the increase in Neogenin IF compared to full-length overexpression. Furthermore, in the context of the stripe assay, ephrin-A5 has the same potentiating effect for netrin-1 growth preference in axons overexpressing EphA4ΔICD-GFP and EphA4-GFP. However, when axons overexpressing either EphA4-GFP or EphA4ΔICD-GFP are challenged by Fc or ephrin-A5 stripes, the avoidance of ephrin-A5 stripes by explants expressing EphA4ΔICD-GFP is lost, suggesting that endogenous EphA4 does not signal efficiently with truncated EphA4. Ephrin-A5 increased the co-localisation of Neogenin and EphA4 in LMC growth cones suggesting that the elevation of surface Neogenin may be local and occurring preferentially in the vicinity of ephrinA5:EphA4 interactions. The EphA4 intracellular domain contains a signalling-essential tyrosine kinase domain 52 and is essential for ephrin-A5-induced EphA4 clustering 30,62 . EphA4 lacking this domain did not form clusters when exposed to ephrin-A5, but its expression was sufficient to elevate Neogenin levels in LMC growth cones suggesting that the EphA4 intracellular domain and clustering are dispensable for Neogenin elevation. Furthermore, since the cytoplasmic tail of Eph receptors is also required for their efficient signalling through endocytosis 63 it is likely that the ephrin-A5:EphA4 signalling events leading to increased Neogenin, are initiated at the plasma membrane and are distinct from a canonical Eph signalling cascade. One possibility is that EphA4 participates in a molecular complex that includes a receptor that can transfer a signal initiated by ephrin-A5-EphA4 binding through its own cytoplasmic tail. An example of this, is the reverse Eph:ephrin signalling that involves signal transfer to the c-Ret receptor, found in a complex with ephrin-A5 and required for normal LMC axon guidance 12,13 .
Despite their relatively high levels of EphA4, dorsal spinal cord growth cones did not respond to ephrin-A5 by increasing Neogenin levels, suggesting that EphA4 is not sufficient for this effect. Ephrin-A5 also failed to increase EphA4 levels in dorsal spinal cord growth cones suggesting that the molecular machinery behind the ephrinA5:EphA4 increase in EphA4 present in LMC neurons may be absent in dorsal spinal cord explants, possibly due to the much higher basal levels of EphA4 in growth cones of dorsal explants in vitro. It is thus feasible that the increase in Neogenin seen in LMC growth cones may be molecularly linked to the increase in EphA4 levels. This idea is in line with the hypothesis of ephrin-A5 inducing the trafficking of EphA4+/Neogenin+ vesicles towards the cell membrane in LMC growth cones. Combinatorially-expressed growth cone guidance signal receptors can modulate netrin-1 axon guidance. For instance, most thalamocortical axons do not normally respond to netrin-1, however, a subset of them can be attracted towards netrin-1, if it is provided in combination with the Robo ligand Slit1 23,64 . This effect depends on PKA, and the expression of Robo1 and its co-receptor FLRT3 whose signalling causes an increase in the levels of DCC on the growth cone surface 24 . A similar requirement for a combination of receptors may apply to the signals that result in ephrin-A5-mediate induction of Neogenin levels. A biochemical fraction of embryonic brain cell membranes, termed Netrin-synergising activity (NSA), containing a protein (or proteins) of a molecular weight of 25-35 kDa, synergises with netrin-1 to induce axonal outgrowth from dorsal spinal cord explants 65 . NSA and ephrin-A5 share a predicted molecular weight of ~26 kDa, but ephrin-A5's predicted isoelectric point of 5.97, which does not align with the prediction that the NSA is a basic protein 66 .
Ephrin-A5-induced neogenin upregulation in motor axon trajectory selection. The choice of limb trajectory made in vivo by LMC neurons is also consistent with the induction of Neogenin by ephrin-A5. LMC motor neurons express Neogenin and DCC attractive netrin receptors (NetrinR attractive ) and can be divided into two populations according to their EphA4 expression. Lateral LMC neurons express high levels of EphA4, are attracted to netrin-1 in vitro, and grow towards netrin-1-expressing dorsal limb mesenchyme, away from ephrin-A5-expressing ventral limb mesenchyme. Based on our findings, it is plausible that in addition to repelling EphA4-expressing lateral LMC axons into the dorsal limb, ephrin-A5 in the ventral limb induces a high level of Neogenin expression in lateral LMC axons that stray there by mistake, making them more responsive to netrin-1, and allowing them to choose the correct limb trajectory even in mice with Neogenin or netrin1 hypomorphic mutations 18 . In this context, it is worth noting that our experiments also demonstrate a Neogenin induction by ephrin-B2, which is normally present in the dorsal limb and is consistent with lateral LMC axons targeting this limb domain.
Medial LMC neurons, have negligible EphA4 expression and are repelled from netrin-1 through its receptor Unc5c, resulting the choice of a ventral limb trajectory 18 . Expression of EphA4 in medial LMC neurons is sufficient to redirect them towards the dorsal limb and has been explained as resulting from EphA4-mediated repulsion from ventral limb ephrin-A5. This effect could be potentiated by ephrin-A5-induced upregulation of Neogenin in medial LMC neurons, resulting in their attraction to dorsal limb netrin-1. Medial LMC neurons also express Unc5c which can mediate repulsion from netrin-1 by itself or in conjunction with NetrinR attractive receptors 14 . Elevation of Neogenin levels could result in increased abundance of NetrinR attractive -only complexes and fewer Unc5c/NetrinR attractive complexes, leading to greater attraction towards dorsal limb netrin-1.

Potential developmental functions of ephrin-A5-induced neogenin upregulation. Loss of
Neogenin, the Neogenin ligands RGMa, RGMb, and netrin-1 cause neural tube closure defects and have been proposed to be due to decreased cell-cell adhesion at the dorsal aspect of this structure [67][68][69] . Similar defects are also present in EphA7 and ephrin-A5 null embryos 70 and were hypothesised to reflect the adhesive function of an EphA7 splice isoform that lacks the intracellular domain 70 . However, in light of our results, ephrin-A5-EphA7 interactions in the neural folds could be contributing to neural tube closure by increasing Neogenin cell surface levels and thus promoting Neogenin-mediated adhesive interactions.
Post-synaptic EphA4 and ephrin-A3 expressed by astrocytes are required for synaptic plasticity by modulating long term potentiation (LTP) in the hippocampus 71 . This requirement for EphA4 for LTP occurs independently from its cytoplasmic tail, since LTP deficits seen in EphA4 null mice are absent in mice expressing EphA4 lacking its cytoplasmic tail 72 . Loss of EphA4 increases the abundance of glial glutamate transporters and LTP deficits can be rescued by blocking glial glutamate transporters 71 . The mechanism underlining the requirement for EphA4 in LTP is unknown 71 . Interestingly, Dcc null mice also show LTP deficits in the hippocampus, proposed to originate from decreased levels of Dcc-dependent Src activation of NMDA receptors 73 . The overlapping functions between chicken Neogenin and mouse Dcc as well as the expression of Neogenin in the hippocampus in mice raises the possibility that ephrin:EphA4 interactions may in part promote LTP by increasing the abundance of post-synaptic Dcc/Neogenin. The Ephrin-A induced increase in post-synaptic Dcc/Neogenin could result in higher levels of Src dependent NMDA receptor phosphorylation and LTP induction 18,32,74 .

Conclusion
Together, our data reveal a novel interaction between classical axon guidance ligands and their receptors. Ephrin-A5-directed increase of Neogenin levels in motor neuron growth cones does require EphA4 but does not appear to proceed through a canonical Eph signalling mechanism. It also results in augmented axon guidance responses to netrin-1 that are consistent with the genetically-tested requirements for ephrin and netrin signalling in motor axon guidance in vivo. Given the importance of netrin and Neogenin signalling inside and outside the nervous system, this new paradigm could have implications in a wide variety of biological processes.

Materials and Methods
Explant culture. All  Explant treatment reagents. For drug treatments, half of the motor neuron media in the explant cultures was replaced with media containing the drug or dimethyl sulfoxide (DMSO; Sigma-Aldrich, cat. no. D2650) and incubated for 20′ prior to cue treatment or fixation. The following drugs were used: cycloheximide, P1860 protease inhibitor cocktail, KT5720, γ-Secretase Inhibitor IX (DAPT), Tetanus toxin, Anisomycin and SU6656 were purchased from Millipore Sigma. MG132, Chloroquine diphosphate, RVKR (Deconoyl-RVKR-CMK), Forskolin and the KYL peptide were purchased from Tocris.
Chick in ovo electroporation. Chicken in ovo electroporations were carried out as previously described 34 at HH st. [18][19] and harvested at HH st. 24-25. Chicken embryos were electroporated with either the pN2-eGFP (Invitrogen) expression plasmid alone or in a 1:4 molar ratio combination with pCAGGS-mEphA4 75 or with pN2-chEphA4-GFP or pN1-chEphA4ΔICD-GFP (cytoplasmic domain deleted) plasmids 10 . In vitro stripe assays. In vitro stripe assay using explants of spinal motor columns were performed as described 27 . In brief, carpets of alternating stripes of Netrin-1, ephrin-A5-Fc, or Fc only as controls were prepared using silicon matrices with micro-well system (provided by Dr. Martin Bastmeyer's laboratory). E5 chick spinal motor column was dissected using sharp tungsten needles (World Precision Instruments) and collected in MN medium. The excised motor column was then trimmed into explants with the size of 1/4 width of motor column, and 20 explants were plated on laminin coated culture dishes containing different combinations of stripe carpets in motor neuron medium and incubated overnight. Following incubation, motor column explants were fixed with 1:1 mixture of 4% paraformaldehyde (Sigma) and 30% sucrose in PBS for 5 minutes followed by 4% paraformaldehyde for 5 minutes. After PBS washes, explants were incubated with selected primary antibodies diluted in blocking solution containing 20% serum in 0.3% Triton-X/PBS(Sigma) for 2 hrs at room temperature (RT). Following PBS washes, explants were incubated with secondary antibodies diluted in the same blocking solution for 2 hrs at RT.
Image acquisition and quantification. 16-bit TIFF images were acquired using LSM700 and LSM710 Zeiss confocal microscopes. Focus was adjusted for maximum fluorescence intensity and images were collected in plane mode with the pinhole set at 1 airy unit (0.8 μm section) with a 63× oil immersion lens and a 2× digital zoom. Gain, digital offset and laser intensities were kept constant for each experiment. Laser intensities were set to minimize pixel saturation. When possible, all images within an experiment were collected in the same session. Phalloidin or GFP fluorescence in growth cones was used to delimitate the regions of interests (ROIs) which were generated with the use of a semi-automated ImageJ macro. ROI area and mean IF (mean grey value) were measured in the raw TIFF images using ImageJ (version 2.0.0, NIH, https://imagej.nih.gov/ij/). Images of growth cones depicted in figures were subject to a rainbow RGB lookup table in ImageJ followed by the overlay of a difference mask in Adobe illustrator CS2 (Adobe Inc.). To assess fluorescence distribution within the growth cones (Fig. 4g), a MATLAB (MathWorks) application was designed and programmed by Dr. Dominic Fillion (IRCM). Using the ROIs generated in ImageJ, the application divides growth cones into 100 bins with bin 1 at the centre and bin 100 following the outer perimeter of the growth cone (bwdist function). The mean fluorescence value for each bin was determined and plotted, bin#100 was omitted from the analysis due to the region of interest (ROI) occasionally slightly over representing the growth cone resulting in the absence of signal in bin#100. To analyse fluorescence overlap (Fig. 6a,m), the fluorescence signal in individual channels was thresholded and an image representing signal overlap between two channels was generated and the mean pixel intensity was quantified.
To determine the change in overlap, the data was treated as such (% signal overlap with cue treatment)/(%signal overlap control treatment). Statistical analysis. Data from the experimental replicate sets were evaluated using Microsoft Excel and Aabel (Gigawiz), statistical significance was set at 0.05. For stripe assays, statistical significance was computed using Mann-Whitney U tests, for quantifications of IF, significance was computed using unpaired-sample t-tests. Standard error of the mean values, p values and the number growth cones analysed can be found in the Supplementary Excel File.

Data Availability
The MATLAB application used for the analysis of IF distribution within growth cones as well as the ImageJ macro used to create the ROIs may be provided upon request.