Proper closure of the optic fissure requires ephrin A5-EphB2-JNK signaling

Vertebrate eye development begins when a pair of optic vesicles buds from placodal tissue on either side of the forebrain. As the optic vesicles contact the surface ectoderm, the outer part of each optic vesicle collapses inward to form an inner layer that will become the neural retina (Fuhrmann, 2010; Lamb et al., 2007; Lamba et al., 2008). This elaborate morphogenetic event, which occurs in the mouse on embryonic day (E) 10.5, results in a residual groove on the ventral surface of the optic cup and optic stalk called the optic fissure (or choroid fissure). Beginning at its proximal end, this fissure along the ventral surface of the developing eye closes by E12.5.

Studies of several null mouse mutants suggest that aberrant dorso-ventral (DV) patterning (Lupo et al., 2005; Polleux et al., 2007), cell proliferation and apoptosis (Valenciano et al., 2009) can prevent proper optic fissure closure and cause ocular coloboma. Specifically, midline-derived Shh induces Pax2, Vax1 and Vax2, which work together during the early specification of ventral eye structures to create the optic fissure (Chiang et al., 1996; Kim and Lemke, 2006; Macdonald et al., 1995; Zhang and Yang, 2001). BMP signaling is also important in optic cup patterning (Furuta and Hogan, 1998; Morcillo et al., 2006; Murali et al., 2005; Sakuta et al., 2001), because a Bmp7 null mutation was shown to impair optic fissure formation by changing cell proliferation and apoptosis (Morcillo et al., 2006). In addition, mutations in genes encoding FGF receptors 1 and 2, Frizzled 5 (Fzd5) and the retinoic acid receptor (RAR) are all associated with incomplete development or closure of the optic fissure (Cai et al., 2013; Kastner et al., 1994; Liu and Nathans, 2008; Lupo et al., 2011). Thus, the signaling pathways downstream of these receptors seem to interact with one another for ventral specification of the eye and might converge on the regulation of common transcription factors such as Pax2, Vax1 and Vax2 (Barbieri et al., 2002; Mui et al., 2005; Torres et al., 1996). For example, conditional mutations of Fgfr1 and Fgfr2 disrupt optic fissure formation in a way similar to that observed in Bmp7-null embryos (Morcillo et al., 2006) and this occurs via downregulation of Pax2 (Cai et al., 2013).

To create the neural retina, the optic cup undergoes a dynamic morphological shift from an essentially flat to a spherical shape: its edges bend downward toward the midline, touch and finally fuse (Lamb et al., 2007). Because unbalanced cell proliferation is a major force causing tissue shape change, genes with mitotic functions are expected to be essential for proper optic fissure closure. Indeed, the process of neurulation and neural tube closure, which are remarkably similar to optic fissure formation and closure, are disrupted by premature neuronal differentiation and excessive cell proliferation (Copp et al., 2003). Phosphatase and actin regulator 4 (Phactr4) is crucial for closure of the neural tube and optic fissure via its role in the regulation of cell proliferation. Phactr4-null embryos show abnormal cell cycle progression – increased cell proliferation associated with inactive PP1, hyperphosphorylation of Rb and upregulation of the targets of E2F. As a result Phactr4-null embryos display both exencephaly and retinal coloboma phenotypes (Kim et al., 2007a).

Because neurulation and other similar processes depend on a precise balance between cell proliferation and apoptosis, they can be disrupted by misregulation of apoptosis just as easily as by misregulation of cell proliferation. The neural tube defects of Apaf1-null mutant embryos are associated with both increased cell proliferation and reduced cell death (Cecconi et al., 1998; Yoshida et al., 1998). In addition, the link between apoptosis and optic fissure development has been previously described (Laemle et al., 1999; Ozeki et al., 2000; Trousse et al., 2001), although the precise role it plays in eye development should be further investigated. More recently, Bcl6 and Bcl6 interacting corepressor (Bcor) were found to act together in repressing p53-dependent apoptosis in the developing optic cup, such that disruption of their function prevents optic fissure closure (Lee et al., 2013). In addition, the JNK family of kinases, which are well known for their roles in the regulation of cellular differentiation and apoptosis, have also been associated with optic fissure closure (Weston et al., 2003). Collectively, these studies suggest that precise temporal and spatial regulation of apoptosis is required to maintain an adequate population of neural progenitors for proper closure of both the neural tube and the optic fissure.

The Eph-ephrin signaling system plays a major role in the wiring of many parts of the brain, shaping the topographic projections of retinal ganglion cell (RGC) axons to their targets in the lateral geniculate nucleus and superior colliculus (McLaughlin and O'Leary, 2005). Unlike other receptor tyrosine kinases, Eph receptors and their ephrin ligands are not known to stimulate cell proliferation (Pasquale, 2010). Eph receptors and ephrins are, however, expressed abundantly in the early embryonic brain where they help regulate the apoptotic cell death of neuroepithelial cells (Depaepe et al., 2005; Kim et al., 2013; Park et al., 2013). Although the role of Eph receptors and ephrins in the early development of the optic vesicle and optic cup is not well-studied, there is some indication that Eph and ephrin are expressed during the mid-to-late development of the neural retina. Not only are ephrin B2 and EphB2 asymmetrically expressed along the DV axis at E16.5, but the Vax2 null mutation leads to ectopic expression of ephrin B2 in the ventral retina instead of EphB2 (Mui et al., 2005). In addition, the finding that EphA4 might be important for the development of astroglia and their precursors, combined with the observation of its expression in the developing retina and optic nerve at E12.5 (Petros et al., 2006), suggests that Ephs and ephrins may play more diverse roles in early retinal development than previously appreciated.

In this study, we find that the interaction of ephrin A5 with EphB2 leads to a sustained activation of JNK, which controls the balance of apoptosis and cell proliferation necessary for proper closure of the optic fissure during early eye morphogenesis.

Ephrin A5 is expressed in a high nasal to low temporal gradient during the later stages of the development of the neural retina (McLaughlin and O'Leary, 2005), but little is known about its expression during early eye morphogenesis. We thus used ephrin A5 (Efna5) BAC transgenic lines that contain sufficient cis-acting elements to drive expression of GFP in the pattern of endogenous Efna5 expression (Yoo et al., 2011). In horizontal sections along the naso-temporal axis, we observed that Efna5 expression begins in the optic vesicle at E9.5. Expression remains low until the optic vesicle invaginates to form the optic cup at E10.5 (Fig. S1A,B). Then, Efna5 is upregulated in the optic cup at E11.5 in a high ventronasal to low dorsotemporal gradient (Fig. 1A,B, Fig. S1C,C′).

Morphologically, a subset of ephrin A5⁺ retinal cells scattered in the proximoventral optic cup at E11.5 appeared to be dying (Fig. 1B, Fig. S1C,D). We thus used terminal nucleotidyl transferase-mediated dUTP nick end labeling (TUNEL) staining to detect apoptotic cells in this region (Fig. 1C, Fig. S1E-G). Although a small number of apoptotic cells in the proximoventral optic cup were TUNEL⁺, there was almost no apoptotic cell death on E12.5 after the closure of the optic fissure (Fig. 1C; data not shown). Furthermore, the TUNEL⁺, GFP⁺ cells in the proximoventral optic cup were also positive for cleaved caspase 3 (Fig. S1H-K), suggesting that they are undergoing a caspase-mediated apoptotic signaling cascade.

We next observed that Efna5-null mutant embryos had 20% fewer TUNEL⁺ apoptotic cells in the proximoventral optic cup on E11.5 than their wild-type littermate controls (Fig. 1D,F). Furthermore, by labeling E11.5 embryos with a 30 min pulse of 5-ethynyl-2′-deoxyuridine (EdU), we observed a ∼25% increase in the proliferative cell population in the proximoventral optic cup in Efna5-null mutant embryos (Fig. 1E,F). In addition, 9 of the 12 Efna5-null mutant embryos we examined at E14.5 had a deep indentation in the neural retina along the site of optic fissure closure (Fig. 1G), probably reflecting improper closure of the optic cup caused by an abnormal ventral curvature. Although the optic disc and optic nerve were not grossly disrupted in Efna5-null mutant embryos (data not shown), our results still suggest that ephrin A5 plays a role in optic cup morphogenesis at E11.5 by regulating both cell proliferation and apoptosis.

To identify the receptor for ephrin A5 in the proximoventral optic cup at E11.5, we visualized the binding of EphA8 and ephrin A5 Fc fusion proteins. As expected, EphA8-Fc binds primarily to the nasal retina because of its high expression of ephrin A5 (Fig. 2A,B). Ephrin A5-Fc, however, binds primarily to the proximal optic cup (Fig. 2C, Fig. S2A,B), where we consistently observed apoptotic cell death. EphA mRNAs were barely detectable via RNA in situ hybridization along the proximoventral optic cup at E11.5 (data not shown). In addition, we also observed ephrin B1-Fc binding to the proximoventral optic cup (Fig. 2D), suggesting that EphB genes are potential receptors for ephrin A5. This would provide in vivo physiological relevance to the published finding that ephrin A5 is capable of binding to and activating the EphB2 receptor (Himanen et al., 2004). Consistent with this hypothesis, we observed binding of ephrin A5-Fc to both EphA8 and EphB2 in HEK293 cells (Fig. 2E-G), whereas ephrin B1-Fc is only capable of binding to EphB2 (Fig. 2H-J). In addition to EphB2, ephrin A5-Fc also binds EphB1 but not EphB3 (Fig. 2K-R). Together, these results suggest either EphB1 or EphB2 as the physiologically relevant receptor for ephrin A5 in the proximoventral optic cup at E11.5.

To further identify the authentic EphB receptor for ephrin A5, we stained the proximoventral optic cup with EphB-specific antibodies. We observed staining of EphB2, but not EphB1 or EphB3 in the proximoventral optic cup at E11.5 (Fig. 3A-C). We confirmed the specificity of the EphB2-specific antibody by confirming the previously reported presence of EphB2 in the developing diencephalon (Fig. 3D) (Magdaleno et al., 2006). For further confirmation of the EphB2 expression pattern in the optic cup, we modified the EphB2 BAC RP23-237B18 with a GFP expression cassette into a site immediately upstream of the Ephb2 start codon and used the recombinant BAC to generate Ephb2-Gfp transgenic mice (Fig. 3E). As expected, we observed intense GFP expression in the ventral diencephalon (Fig. 3F,G). In addition, we also observed intense staining in the proximoventral optic cup on horizontal sections of transgenic embryos at E11.5 (Fig. 3H-J, Fig. S2C). Importantly, EphB2 expression temporally and spatially overlapped with the apoptotic cells and ephrin-A5-positive cells observed in Fig. 1 (Fig. 3J,K). Together, these results suggest EphB2 as a physiologically relevant receptor for ephrin-A5 in the proximoventral retina at E11.5.

To investigate the role EphB2 plays in the proximoventral optic cup at E11.5, we modified the EphB2 BAC to express either a wild-type Ephb2 cDNA or a mutant Ephb2 lacking its cytoplasmic region (Ephb2-dc) with the intention that Ephb2-dc will act as a dominant negative. We also placed lacZ downstream of an IRES promoter to verify in vivo EphB2 and EphB2-dc protein expression (Fig. 4A). We were able to confirm that EphB2-dC was effective in blocking the tyrosine-phosphorylation of wild-type EphB2 in HEK293 cells, supporting its dominant-negative function against endogenous EphB2 (Fig. S3A,B). We injected the wild-type EphB2 or EphB2-dC recombinant BAC into fertilized mouse embryos and euthanized pregnant females at E11.5 to analyze transgenic embryos. Consistent with previous reports of EphB2 expression in developing limbs (Compagni et al., 2003), we observed lacZ expression in the limbs of these transgenic embryos (Fig. S3C,E). We also confirmed expression of wild-type Ephb2 or Ephb2-dC in transgenic embryo limbs using RT-PCR (Fig. S3D,F). Importantly, we also found that lacZ expression was restricted to the proximoventral optic cup of transgenic embryos expressing Ephb2 or Ephb2-dC (Fig. 4B). We were able to confirm that EphB2-dC acts as a dominant negative against endogenous EphB2 by observing that Ephb2-dC mice display axon guidance defects in the anterior commissure during postnatal brain development (Henkemeyer et al., 1996) (Fig. S3G). Using TUNEL staining, more apoptotic cells (∼20%) were observed in the proximoventral optic cup of Ephb2 transgenic embryos than that of wild-type embryos, whereas less apoptotic cells (∼15%) were found in the proximoventral optic cup of Ephb2-dC transgenic embryos than in wild-type embryos (Fig. 4C,E). Next, we performed an EdU pulse-labeling experiment at E11.5. Ephb2 transgenic embryos had ∼40% fewer actively proliferating cells in the proximoventral optic cup than wild-type embryos, whereas Ephb2-dC transgenic embryos showed a ∼45% increase in the number of proliferating cells (Fig. 4D,E). In addition, although whole retinas from Ephb2 transgenic embryos at E14.5 are significantly smaller than those of wild-type littermate control embryos (Fig. S4A-C), none showed optic fissure closure defects. By contrast, whole retinas from Ephb2-dC transgenic embryos at E14.5 were both smaller than the wild type and showed optic fissure closure defects, typically appearing as a deep indentation along the ventral pole of the neural retina (Fig. S4D-F). These results suggest that the reduction in apoptosis observed among the Ephb2-dC transgenic embryos has a significant impact on optic fissure closure.

Ephb2-null mutant mice exhibit several phenotypes including abnormal axon guidance (Henkemeyer et al., 1996), hyperactivity and persistent circling behavior (Henderson et al., 2001). Therefore, we next asked whether Ephb2-null embryos (CD-1 genetic background) displayed any abnormalities in cell proliferation and apoptosis in the proximoventral optic cup. As expected, Ephb2-null mutant embryos showed ∼35% fewer apoptotic cells and 50% more proliferative cells in the proximoventral optic cup at E11.5 than their wild-type littermates (Fig. 5A-C). In addition, whole retina from Ephb2-null mutant embryos at E14.5 showed optic fissure closure defects ranging from a deep indentation to incomplete closure to severe closure defects (Fig. 5D,E). Together, these results support our hypothesis that EphB2 is a biologically relevant receptor for ephrin A5 in the proximoventral optic cup and that EphB2 signaling influences apoptosis and cell proliferation to regulate optic fissure closure.

A previous study reported JNK activation downstream of EphB signaling (Becker et al., 2000). We confirmed that EphB2 but not EphB2-dC induced strong phosphorylation of JNK in multiple cell types (Fig. S5A-F). JNK phosphorylation in HEK293 cells was unaffected, however, by EphA2, EphA4, EphA8 and EphB1 (Fig. S5G-J). Like EphB2, EphB3 also induced JNK phosphorylation (Fig. S5K,L), but, because EphB3 does not bind ephrin A5 (Fig. 2R), EphB3-mediated JNK activation is unrelated to ephrin A5. To determine whether ephrin A5 stimulates EphB2-mediated JNK activation, we generated HEK293 or P19 cell lines stably expressing EphB2. Phosphorylated JNK was clearly detectable after these EphB2-expressing cells were treated with ephrin A5 for 2 h (Fig. 6A). Western analysis confirmed that 2 or 6 h of ephrin A5 treatment significantly elevated the levels of phosphorylated JNK, phosphorylated c-Jun and cleaved caspase 3 (Fig. 6B,C). This ephrin A5-induced JNK activation does not occur, however, in cells stably expressing either EphB2-dC or EphB3 (Fig. S6A,B). Using the TUNEL assay and cleaved caspase 3 staining, we also observed that 2 h of ephrin A5 treatment induced apoptosis in EphB2-expressing cells to the same extent as anisomycin, a JNK activator. Ephrin A5 and anisomycin-induced apoptosis were both blocked, however, by the JNK inhibitor SP600125 (Fig. 6D,E, Fig. S6C,D). By staining for Ki67, we also observed a strong reduction in proliferation in cells expressing EphB2 when they were treated for 2 h with ephrin A5. This anti-proliferative effect of ephrin A5 was also blocked by treatment with the JNK inhibitor (Fig. 6F,G, Fig. S6C,E). This pro-apoptotic and anti-proliferative activation of JNK by prolonged treatment with ephrin A5 did not occur in cells stably expressing Ephb2-dC or Ephb3 (data not shown). Together, these results strongly suggest that prolonged engagement of EphB2 with ephrin A5 induces a sustained activation of JNK that promotes apoptosis and reduces cell proliferation.

Since JNK was previously implicated in the closure of the optic fissure (Weston et al., 2003), we asked whether JNK is specifically activated in the proximoventral optic cup at E11.5. As predicted, JNK phosphorylation was clearly detectable in the proximoventral optic cup of wild-type embryos. This JNK activation, however, was significantly reduced (∼2.5-fold) in both Efna5- and Ephb2-null mutant embryos (Fig. 7A-C). We next looked for dysregulation of apoptosis and cell proliferation in the proximoventral optic cups of E11.5 JNK-deficient embryos. We compared JNK1^+/+; JNK2^+/− (JNK1 and JNK2 are also known as mitogen-activated protein kinases Mapk8 and Mapk9, respectively) embryos with JNK1^−/−; JNK2^+/− embryos because the latter show the retinal coloboma phenotype (Weston et al., 2003). JNK1^−/−; JNK2^+/− embryos show significantly fewer apoptotic cells (∼50%) and more proliferative cells (∼35%) in the proximoventral optic cup than control littermates (Fig. 7D-F). We also observed that JNK1^+/−; JNK2^−/− embryos showed reduced apoptosis and enhanced cell proliferation in the proximoventral optic cup (data not shown). Unlike the ventral patterning defects observed in JNK-deficient mice (Weston et al., 2003), we found that both Vax2 and Pax2 expression was not significantly altered in either the Efna5- or Ephb2-null mutants (Fig. S7A-C).

Finally, we asked whether inhibition of apoptosis in the proximoventral optic cup could recapitulate the phenotype induced by disruption of ephrin A5-EphB2-JNK signaling. To do so, we inserted a human XIAP expression cassette into the EphB2 BAC and used it to generate Ephb2-XIAP transgenic embryos. We confirmed via RT-PCR analysis using cDNA from the limbs of the Ephb2-XIAP transgenics that the human XIAP mRNA is ectopically transcribed from the EphB2 BAC (Fig. 7G). As expected, apoptotic cell death was strongly inhibited in the proximoventral optic cup (Fig. 7H,I), but cell proliferation was unaltered (data not shown). Strikingly, ectopic expression of XIAP in the proximoventral optic cup induced a severe coloboma phenotype (100% penetrance), implicating apoptosis in the proximoventral optic cup as a crucial morphogenetic process for inducing proper closure of the optic fissure (Fig. 7J). Together, our results strongly support a role for the ephrin A5-EphB2-JNK signaling axis in the proximoventral optic cup in facilitating optic fissure closure via induction of apoptotic cell death.

We present three novel findings in this report. First, EphB2 is a biologically relevant receptor for ephrin A5 in the proximoventral optic cup while the optic fissure is closing. Second, prolonged engagement of EphB2 with ephrin A5 in the proximoventral optic cup induces a sustained activation of JNK that inhibits cell proliferation and enhances apoptosis. Third, this ephrin A5-EphB2-JNK signaling axis plays a crucial role in maintaining the appropriate number of cells for proper closure of optic fissure.

Most Eph-ephrin interactions tend to be restricted to a single subclass – EphA receptors are activated by ephrin A ligands, EphB receptors by ephrin B ligands (Pasquale, 2005). We show, however, that early eye morphogenesis and optic fissure closure require the cross-subclass interaction of EphB2 with ephrin A5 working upstream of JNK signaling. One previous report showed that EphB2 binds ephrin A5 with high affinity and that EphB2 activation by ephrin A5 is sufficient to induce autophosphorylation and neurite retraction in neuroblastoma cells (Himanen et al., 2004). However, no physiological significance has been assigned to the cross-subclass interaction of EphB2 with ephrin A5, nor is there any evidence that this interaction occurs in vivo.

Our present study provides strong evidence that ephrin A5 is a physiologically relevant ligand for the EphB2 receptor in the proximoventral optic cup at E11.5. Although ephrin A5 expression was stronger in the ventronasal region of the cup, ephrin A5 expression did partially overlap with EphB2 in the proximoventral optic cup at E11.5 (Fig. 8A). We show that ephrin A5 binds specifically to EphB2 on the surface of cultured cells and that this interaction is sufficient to modulate cell proliferation and apoptosis via JNK activation. In addition, both ephrin A5 and Ephb2-null mutant embryos show reduced apoptosis and enhanced cell proliferation in the proximoventral optic cup, whereas Ephb2 gain-of-function mutants induce opposite results.

As the defects of the optic fissure closure were more severe in Ephb2-null mutants than Efna5 mutants, we cannot rule out the possibility that other ephrin ligands also interact with EphB2 in the proximoventral optic cup. Unfortunately, apart from ephrin A5 and EphB2, we were unable to detect expression of any other ephrin or Eph genes in the E11.5 proximoventral optic cup by in situ RNA hybridization (H.N., unpublished observation). In addition, neither Ephb1- nor Ephb3-null mutant embryos showed defects in optic fissure closure. Nevertheless, according to an RT-PCR analysis, ephrin A2 was weakly expressed in the optic cup at E11.5. Consistent with this result, the optic fissure closure defect was more severe in Efna2^+/−; Efna5^−/− embryos, reaching the level of Ephb2-null mutant embryos. Still, Efna2^−/−; Efna5^+/− embryos showed a deep indentation along the ventral pole of the neural retina similar to ephrinA5-knockout embryos. In addition, Efna2-null mutant embryos, like wild-type embryos, had no defects in optic fissure closure. These results strongly suggest that although ephrin A2 can partially supplement ephrin A5, it is ephrin A5 that acts as the primary ligand responsible for EphB2-mediated optic fissure closure. In addition, since Vax2 and Pax2 expression are not significantly altered in the proximoventral optic cup of ephrin A5-knockout mice, EphB2-dC transgenic mice or EphB2-knockout mice, it is likely that these ventral patterning transcription factors act upstream of ephrin A5 and EphB2 (Fig. S7A and B).

Using cell surface binding assays with Fc-fusion proteins, we showed that ephrin A5 also binds EphB1 but not EphB3. This is rather unexpected because ephrin A5 is reported not to bind EphB1 (Himanen et al., 2004). Although this issue needs to be resolved in the future, our findings suggest that ephrin A5 binds at least two different EphB receptors, EphB1 and EphB2, to regulate totally different signaling pathways – only EphB2, not EphB1, activates JNK-mediated pro-apoptotic signaling.

A previous study reports that EphB1 or EphB2 binds to the Nck adaptor to stimulate the Nck-interacting kinase (NIK; also known as MAP3K14), which, in turn, activates JNK (Becker et al., 2000), but it was not determined whether ephrin B1-stimulated JNK activation in P19 cells is mediated by EphB1 or EphB2. Our intensive analyses with various Eph receptors have revealed that JNK is not activated by EphB1, EphA2, EphA4 or EphA8. Instead, we found that JNK is activated by EphB2 in at least three different cell types (i.e. HEK293, P19 and Neuro2a cells). EphB2 was also reported to induce downregulation of the Ras-MAPK pathway, which is essential for ephrin-induced neurite retraction (Elowe et al., 2001). This inhibition of MAPK activity was induced by short-term ephrin treatment (i.e. 15-30 min). When EphB2-expressing cells are treated with ephrin A5 for prolonged periods of between 2 and 6 h, MAPK activity is not significantly altered (H.N., unpublished observation). By contrast, JNK is significantly activated in EphB2-expressing cells exposed to 1-2 h of ephrin A5 treatment, but short-term ephrin A5 treatment (i.e. 30 min or less) does not activate JNK. We show that JNK is highly activated in the E11.5 proximoventral optic cups of wild-type embryos but not in Efna5- or Ephb2-null mutant embryos. We also show that although EphB3 is capable of activating JNK in cultured cells, it is neither expressed in the proximoventral optic cup at E11.5 nor does it bind ephrin A5. On the basis of these findings, we propose that a prolonged interaction of EphB2 with ephrin A5 in the E11.5 proximoventral optic cup induces a sustained activation of JNK. This JNK activation regulates the processes of cell proliferation and apoptosis in a way that is crucial for proper optic cup morphogenesis (Fig. 8B).

Although prolonged JNK activation is known to play a role in TNF- and UV-induced apoptosis (Davis, 2000; Dhanasekaran and Reddy, 2008; Liu and Lin, 2005), it has not been studied in Eph-expressing cells. Our findings suggest caution in studying the mechanisms by which JNK acts downstream of Eph receptor because it seems to be constrained to a small subset of Ephs and ephrins. Our results also suggest that the reduced cell proliferation observed downstream of ephrin A5-EphB2-JNK signaling is coupled to enhanced apoptosis because these two processes are simultaneously impaired by JNK inhibition in cultured cells. We also observed the coupled dysregulation of cell proliferation and apoptosis consistently in the proximoventral optic cup of ephrin A5-EphB2-JNK signaling mutants (Fig. 8B). It is possible that sustained JNK activation causes a general pro-apoptotic shift that has different outcomes in different cells (e.g. in progenitor cells versus terminally differentiated cells) depending on the cell's threshold for entering apoptotic cell death. A cell that receives this pro-apoptotic ‘push’ while well below an apoptotic threshold might respond by halting proliferation, exiting the cell cycle and further differentiating without initiating the cell death program. Further studies should address the mechanism by which EphB2-mediated JNK activation inhibits cell proliferation.

The process of retinal histogenesis occurs through a distinct proximal-to-distal gradient during retinal development (Martinez-Morales and Wittbrodt, 2009). Consistent with this, we observed fewer mitotic cells in the proximoventral region of the optic cup at E11.5 than in the peripheral region. In mouse embryos with mutations in the ephrin A5-EphB2-JNK signaling axis, however, apoptosis was inhibited and the number of mitotic cells in the proximoventral optic cup was consistently elevated (Fig. 8B). As a result, these embryos displayed morphological abnormalities related to the closure of the optic fissure. For example, Efna5-null mutant embryos and embryos expressing the dominant negative EphB2-dC develop a deep indentation along the site of optic fissure closure. Some Ephb2-null mutant embryos have even more severe defects, including persistently open optic fissures. We did not, however, observe any optic fissure closure defects in embryos overexpressing EphB2. Their retinas were smaller owing to enhanced apoptosis in the proximoventral optic cup, but there were no closure defects. We did observe extreme closure defects coupled to small retinal size induced by overexpression of XIAP. XIAP expression strongly inhibits apoptosis but does not affect cell proliferation in the proximoventral optic cup. Thus, the inhibition of apoptotic cell death in the proximoventral optic cup is more detrimental to optic fissure closure than changes in cell proliferation. Despite the fact that both lead to an overabundance of cells in the proximoventral optic cup, we have shown that a reduction in apoptosis is not equivalent to an increase in cell proliferation. An appropriate balance is crucial. Excessive cells might affect the ventral curvature of the optic cup and the precise apposition of the approaching optic fissure edges. It is important to note that the optic fissure of Ephb2-null mutants is not efficiently closed by E14.5, but it is closed by E18.5. Nevertheless, all Ephb2-null mutants observed at E18.5 have a deep indentation along the ventral pole (H.N., unpublished observation). By contrast, the optic fissures of XIAP transgenic embryos never close, even in the latest developmental stages. It may be the fact that Ephb2-null mutant embryos show both enhanced cell proliferation and reduced apoptosis in the proximoventral optic cup that explains their ability to eventually close the optic fissure. Their additional proliferating cells somehow allow the optic fissure to close in later stages, even if the resulting retina is marred by a deep indentation along its ventral pole. The exact role these extra proliferating cells play in supplementing the process of optic fissure closure will require further study. It also seems that the way optic fissure closure affects the development of the neural retina depends on genetic background. The retinas of EphB2-dC embryos in the C57BL6 background are much smaller than those of wild-type embryos, whereas the retinas of Ephb2-null mutant embryos in CD-1 background are only slightly reduced when compared with wild-type controls. This suggests that the development of the neural retina in the C57BL6 background is more susceptible to changes in apoptosis. Because loss of BMP7 has been reported to reduce apoptosis (Morcillo et al., 2006), it is possible that the ephrin A5-EphB2 and BMP7 signaling pathways converge on apoptotic regulation to direct proper closure of the optic fissure. It will thus be interesting to investigate whether BMP7 signaling stimulates JNK activity during optic fissure closure.

In summary, our results raise the interesting possibility that a cross-subclass interaction of EphB2 with ephrin A5 provides a signal that allows retinal progenitors in the proximoventral optic cup at E11.5 to exit the cell cycle. We have shown that JNK activation downstream of this ephrin A5-EphB2 interaction plays a central role in biochemically linking the inhibition of cell proliferation with an increase in apoptotic cell death. This ephrin A5-EphB2-JNK signaling axis in the proximoventral optic cup is thus essential for the selective reduction of retinal progenitors so that an appropriate pool of cells is maintained for precise closure of the optic fissure (Fig. 8).

Ephrin A5 BAC transgenic mice expressing GFP have been previously described (Yoo et al., 2011). To generate the EphB2-GFP BAC targeting vector, we amplified homologous arms A (1008 bp) and B (909 bp) flanking the mouse Ephb2 translation start site (ATG) using PCR with the following primers: 5′-GGCCAAGTCGGCCGAGTAGGGGCTGTCGCTA-3′ (forward primer for A arm); 5′-GAGCTCTGCTGCGCTGCCCGGAGCCT-3′ (reverse primer for A arm); 5′-ATGCATCATCTTGCCGAGGCTTTGTA-3′ (forward primer for B arm); and 5′-ATGCATAATTCGACCCATCTGGCTTC-3′ (reverse primer for B arm). We then inserted homologous arms A and B into pGEM11z, a vector containing a GFP reporter with the SV40 polyadenylation site and a FRT-Kana-FRT cassette. The resulting vector was digested with SfiI and inserted into the Ephb2 BAC genomic DNA (RP23-237B18) using a published bacterial homologous recombination method (Kim et al., 2007b). The targeting vectors for the EphB2-EphB2 (a full-length mouse Ephb2 cDNA clone) BAC, the EphB2-EphB2-dC (a mouse Ephb2 cDNA clone lacking the coding sequence for amino acids 589-986) BAC, and the EphB2-XIAP (a full-length human XIAP cDNA clone) BAC were constructed as described for EphB2-GFP, except that the GFP reporter and FRT-Kana-FRT cassette were replaced by the indicated cDNA and an IRES-lacZ-FRT-Kana-FRT cassette. The modified recombinant BACs were injected into fertilized C57BL/6 mouse eggs as previously described (Kim et al., 2007b).

Efna2^tm1Jgf Efna5^tm1Ddmo/J (005992) (Feldheim et al., 2000), Mapk8^tm1Flv/J (004319) (Dong et al., 1998) and Mapk9^tm1Flv/J (004321) (Yang et al., 1998) mouse lines were purchased from The Jackson Laboratory. The Ephb2^tm1Paw Ephb3^tm1Kln (RBRC03899) mouse line was purchased from the RIKEN BioResource Center (Henkemeyer et al., 1996). All mice were generated and maintained in accordance with institutional guidelines approved by the Sookmyung Women's University Animal Care and Use Committee.

HEK293, Neuro2a and P19 cells were cultured as previously described (Gu et al., 2005). Ephb1 (BC057301), Ephb2 (BC043088.1), Ephb3 (BC014822), Efna2 (BC037166.2), Epha4 (NM_007936), Epha8 (NM_007939) and Ephb2-dC cDNAs were subcloned into the pIRES2-EGFP vector (Invitrogen) and transfected into cells as previously described (Choi and Park, 1999). Transient transfections were performed using Metafectin (Biontax) according to the manufacturer's instructions. Western blot analysis was performed as previously described (Shin et al., 2007).

Immunohistochemistry experiments are described in the supplementary Materials and Methods. Immunofluorescence staining was performed as previously described (Shin et al., 2007). Cells were incubated with SP600125 (10 μM, Sigma) for 30 min and then stimulated with pre-clustered ephrin A5-Fc ligands, Fc control or anisomycin (30 μg/ml, Sigma) for the indicated time. Then, they were fixed (4% paraformaldehyde and 2% sucrose in PBS) for 30 min at room temperature and rinsed in PBS three times for 5 min each. A Zeiss LSM700 confocal microscope was used for imaging.

TUNEL assays were performed using the In situ Cell Death Detection Kit with TMR red (Roche Diagnostics) according to the manufacturer's instructions. For X-gal staining, developing limbs were fixed in 0.2% glutaraldehyde for 10 min, washed three times and stained as previously described (Kim et al., 2013).

Polyclonal rabbit cleaved caspase 3 (9661s; 1:200), monoclonal rabbit phosphorylated c-Jun- (2361s; 1:200) and JNK-specific (4668s; 1:200) antibodies were purchased from Cell Signaling Technology. Polyclonal rabbit anti-ki67 (ab15580; 1:200) and polyclonal chicken anti-GFP (ab13970; 1:500) antibodies were purchased from Abcam. Polyclonal goat anti-EphB2 (AF467; 1:500) and anti-EphB3 (AF432; 1:100) were purchased from BD Biosciences. Polyclonal rabbit anti-EphB1 (sc926; 1:100) and mouse anti-JNK (sc7345; 1:200) antibodies were purchased from Santa Cruz Biotechnology.

We are indebted to Yujin Kim for the ephrin A5-GFP BAC transgenic mice.