Abstract
Eph/ephrin signaling is crucial for organizing retinotopic maps in vertebrates. Unlike other EphAs, which are expressed in the embryonic ventral retina, EphA4 is found in the retinal ganglion cell (RGC) layer at perinatal stages, and its role in mammalian visual system development remains unclear. Using classic in vitro stripe assays, we demonstrate that, while RGC axons are repelled by ephrinB2, they grow on ephrinB1 stripes through EphA4-mediated adhesion. In vivo, retinal axons from EphA4-deficient mice from either sex show impaired arborization in the medial, but not lateral, regions of the superior colliculus that express ephrinB1. Gain-of-function experiments further reveal that ephrinB1-mediated adhesion depends on EphA4 tyrosine kinase activity but it is independent of its sterile alpha motif. Together, our findings suggest that EphA4/ephrinB1 forward signaling likely facilitates adhesion between retinal axon terminals and cells in the medial colliculus, contributing to the establishment of proper connectivity within the visual system.
Significance Statement
The significance of our findings goes beyond unveiling the mechanisms underlying topographic visual map formation. By discovering a dual role for EphA4 in mediating both repulsion and adhesion, which challenges the conventional understanding of EphA4/ephrin interactions, this work opens up new avenues for exploring EphA4 broader implications in other cellular contexts, including cell differentiation, migration, and synaptic plasticity.
Introduction
EphA receptors are a family of receptor tyrosine kinases essential not only for guiding cell positioning and tissue organization during embryonic development but also for regulating processes such as neuronal connectivity, cell migration, and angiogenesis. During the development of the visual system, Eph/ephrin signaling plays a key role in establishing retinotopic maps in vertebrates. It is widely accepted that EphA/ephrinA signaling mediates repulsion and is required for topographic rostrocaudal mapping in the visual targets while the lateromedial axis has been proposed to be mapped, at least in part, via EphB and ephrinB signaling (Hindges et al., 2002; Triplett and Feldheim, 2012; Suetterlin and Drescher, 2014). EphA5/6 are expressed in a nasotemporal gradient in the mouse retina while EphBs are gradually expressed along the dorsoventral retinal axis (Hindges et al., 2002; Feldheim et al., 2004; Reber et al., 2004; Carreres et al., 2011). Ventral retinal axons project into medial areas of the superior colliculus (SC) that express ephrinB1 in the superficial layer at birth, and, based on the spatiotemporal patterns of EphBs/ephrinBs in the retina and SC, respectively, it was proposed that Eph/ephrinB signaling should also mediate an attractive or permissive response in RGCs (Hindges et al., 2002). In particular, ephrinB1 was proposed to act as a bifunctional guidance molecule to control the position-dependent bidirectional extension of interstitial branches of RGC axons originating from similar dorsoventral retinal sites (McLaughlin et al., 2003). Computational models have proposed a mechanism by which ephrins should exert concentration-dependent attractive and repulsive forces on axons by modulating cell–substrate adhesion and forcing axons to terminate at positions within the optimal ephrinB gradient (McLaughlin et al., 2003; Flanagan, 2006).
Both Ephs and ephrins are cell surface molecules that can signal in the forward and reverse directions (Böhme et al., 1996; Stein et al., 1998). Interactions between A and B classes of Eph and ephrins mainly occur within each subgroup (North et al., 2009), but engagement between EphA and ephrinB molecules also takes place in certain contexts (Kullander et al., 2001, 2003; Himanen et al., 2004). EphA4 exemplifies the flexibility of this family of signaling molecules. Like other EphA receptors, EphA4 was initially characterized as a binding partner of ephrinA ligands (Cheng and Flanagan, 1994; Drescher et al., 1995; Dottori et al., 1998; Walkenhorst et al., 2000; Swartz et al., 2001), but additional studies demonstrated that EphA4 actually interacts with both A and B ligands (Gale et al., 1996; Kullander et al., 2003). In fact, EphA4 can bind class A ephrins and ephrinB2 and ephrinB3 due to the conformational plasticity of the receptor in its ligand binding side (Bowden et al., 2009). In vitro, EphA4 strongly binds to ephrinB2 and ephrinB3 and has been suggested to exhibit minimal or no binding to ephrinB1 (Gale et al., 1996; Guo and Lesk, 2014).
EphA4-mediated signaling is required for many cellular processes including rhombomere-boundary formation, neuronal differentiation and migration, axon navigation, and synaptic plasticity (Conover et al., 2000; Helmbacher et al., 2000; Walkenhorst et al., 2000; Swartz et al., 2001; Dufour et al., 2003; Cooke et al., 2005; Goldshmit et al., 2006; Escalante et al., 2013; Paixão et al., 2013). Although repulsion is the most common and well-understood outcome of EphA4–ephrin interactions, indirect evidence suggested that EphA4-dependent adhesion may occur in different contexts such as hindbrain segmentation in zebrafish, cross talk between neurons and astrocytes in the hippocampus, or blood platelet aggregation (Prevost et al., 2002; Cooke et al., 2005; Murai and Pasquale, 2011). However, EphA4-triggered adhesion has not been so far demonstrated. Also, as it has been assumed that EphA4 plays a function similar to other EphA receptors (Walkenhorst et al., 2000; Reber et al., 2007) but its precise role in visual system development has not been thoroughly investigated.
Here, we first demonstrate that EphA4 shows a nonredundant spatiotemporal expression pattern with other EphA receptors in the developing retina of mice. Instead of being expressed in the embryonic ventral retina as EphA5 and EphA6 (Feldheim et al., 2000), EphA4 is transiently detected along the RGCs layer at perinatal stages. In vitro, RGC axons are repelled by ephrinB2, but they grow on ephrinB1 substrates. Remarkable repulsion to ephrinB2 and affinity to ephrinB1 are both lost in the absence of EphA4. Functional experiments in vivo confirmed that ephrinB1 triggers EphA4-mediated cell adhesion. RGC axons coming from the ventral retina lacking EphA4 do not properly arborize in medial areas of the SC. Conversely, artificial activation of EphA4 in RGCs induces aberrant cell adhesion. Together these results strongly suggest that expression of EphA4 in RGCs is essential to mediate an adhesive response of retinal axons to cells in medial areas of the superior colliculus during the establishment of the visual circuit.
Materials and Methods
Mouse lines
The EphA4 mutant mice (EphA4KO) used in this study were previously described (Kullander et al., 2001). Electroporation experiments were performed on embryos from a mixed genetic background. Two inbred mice lines (C57BL/6J and DBA/2J) were bred to obtain a first generation of B6DBAF1/J hybrid females that were backcrossed with C57BL/6 males. Then, the progeny of this cross was electroporated. A similar strategy was followed to obtain EphA4KO electroporated mice.
DNA plasmids
EphA4wt; EphA4ΔC, EphA4KD; EphA4ΔSAM (Kullander et al., 2001; Egea et al., 2005); ephrinB2, ephrinB1, and ephrinB1DC were cloned in a pCAG plasmid. pCAG-GFP and pCAG-EphA6 plasmids were used previously (Carreres et al., 2011).
In utero and postnatal electroporation
In utero electroporation was performed as in Garcia-Frigola et al. (2007, 2008) and postnatal electroporation as in Benjumeda et al. (2013).
In situ hybridization
In situ hybridization was performed according to the reported methods (Schaeren-Wiemers and Gerfin-Moser, 1993). A riboprobe to detect mouse EphA4 mRNA was synthesized using the following primers: 5′-(GGGCCACTGAGCAAGAAA)-3′ and 5′-(CCTGGACCAAAGCAATG)-3′ from E14.5 mouse embryos cDNA. To detect EphrinB1 and EphrinB2, we used a specific antisense riboprobe (a gift from Oscar Marín, King's College London, London, UK, and S. Williams, University of North Carolina, USA, respectively).
Images were captured with Leica DM2500 equipped with a Leica DFC7000T camera and Leica Application Suite version 4.10.0 software.
Histology, imaging, and axon arbor quantification
Brain whole-mount immunodetection was performed at P9. After fixation with 4% PFA, brains were washed in PBS. After postfixation of perfused postnatal mice, retinas and brains were sectioned into 80 μm slides with a Leica Vibratome before being processed for immunofluorescence. The following primary antibodies were used: Neuronal Class III β-tubulin (Covance, #MMS-435P), ChAT (Advanced Targeting Systems, #AB-N3AP), Calbindin D-28K (Swant, #CB38A), Brn3a (Millipore, MAB1585), and anti-GFP antibody (Aves Lab, GFP-1020). Then the antibody solution was discarded, and the samples were washed in 2% PBS–Triton X-100 and incubated with the relevant secondary antibody for 2 h at room temperature. After washing again, samples were flat mounted onto a microscope glass slide with Mowiol.
For in situ hybridization, tissue was cryoprotected and sectioned into 25 µm slides. Retinal sections and whole-mount preparations were imaged using a Leica microscope setup with a DM2500 Leica camera (DFC7000T) or an Olympus FV1200 confocal microscope/FV10-ASW. Brain whole mounts were acquired using a Leica MZ16F stereoscope. All raw image processing and quantification were performed with ImageJ software.
Single-cell analysis
Shekhar et al. (2022)’s original bam1 files were obtained from the SRA archive (GEO ID: GSE185671). 10x Genomics Cell Ranger (v6.1.1, Zheng et al., 2017) was used to recover the original FASTQ files and output the count matrices. Raw count matrices were ambient-RNA corrected through CellBender (v0.3.0, Fleming et al., 2023). Scanpy (v1.8.2, Wolf et al., 2018) was used for all analyses of single-cell data following recommended usage guidelines (Heumos et al., 2023). Cell barcodes corresponding to retinal ganglion cells (RGCs) according to the classification of Shekhar et al. (metadata file at https://github.com/shekharlab/mouseRGCdev) were kept for further analysis. Only cells derived from E13, E14, E16, and P0 mice were analyzed. Normalization was performed through the shifted-log method and batch-effects removal and integration through scvi-tools (v0.16.1, Gayoso et al., 2022). A cell was defined as positive for gene expression when its normalized/logged level of expression was positive.
Stripe assays
Stripes assays were prepared as previously described (Seiradake et al., 2014). Subsequently, 6 mg/ml of FC recombinant protein, ephrinB1, and 12.5 mg/ml of FC recombinant protein, ephrinB2 (R&D Systems), were mixed with Alexa Fluor 594–conjugated anti-hFC antibody (Invitrogen) in HBSS for 30 min. Proteins were injected into matrices (90 mm width) placed on crystal coverslip pretreated with poli-ʟ-lysine, resulting in red-fluorescent stripes. After 30 min incubation at 37°C, coverslips were washed with HBSS, and matrices were carefully removed. The coverslips were further coated with 6 or 12.5 mg/ml of FC protein mixed with anti-hFC (no fluorescent dye) for 30 min at 37°C and then washed three times with HBSS and coated with 20 mg/ml of laminin for 2 h at 37°C. Retinal explants from E15.5 wild-type or EphA4 mutant mice were cultured for 48 h on the stripes, fixed with 4% PFA in PBS for 20 min at room temperature, and washed and incubated with the mouse monoclonal anti-β-III tubulin antibody (BioLegend). Alexa Fluor 488 anti-mouse secondary antibody was used to visualize the β-III tubulin signal. The total number of β-III tubulin pixels on red or black stripes was quantified using ImageJ and an automatic stripe assay analysis (Fleitas et al., 2021), calculated as repulsion index and represented as fold change in relation to Fc/Fc control stripes (Schneider et al., 2012).
Intensity fluctuation analysis in the retina and superior colliculus
A curved rectangular ROI was drawn in Fiji (Schindelin et al., 2012) across the region of interest in each image. For the superior colliculus samples, a 40-µm-thick ROI was drawn, with the start of the ROI (initial ROI, iROI) located 150 µm in a straight line from the most rostral point of the colliculus and 20 µm from the surface. For the retina samples, 20-µm-thick ROI was plotted above (amacrine layer) and below (ganglion cell layer) the inner plexiform layer. For each section, a custom ImageJ macroscript was used to obtain the average gray value intensity of the GFP channel along the x-axis of the image (fluorescence intensity, FI). Data were background-corrected using a rolling average and normalized to 100 (normalized intensity, N. Int). Local maxima and minima were detected and averaged to quantify the level of fluctuation of the GFP signal across the region of interest [intensity fluctuation (IF) %]. FI from three sections from at least three different animals were averaged.
Quantitative and statistical analysis
The data are presented as a box and whisker plot summarizing a set of data (the minimum, the first quartile, median, third quartile, and maximum) and how the data are distributed. In the box plot, the box indicates from the first quartile to the third quartile. Statistical analysis of the quantitative results was conducted using RStudio. These data have been compared in accordance with Student's t test or ANOVA (Tukey post hoc, multiple comparisons of means). Minimal statistical significance was fixed at p < 0.05 for the ANOVA, and the results of the Student's t test were represented as follows: *p < 0.05, **p < 0.01, and ***p < 0.001.
Results
EphA4 is transiently expressed in the developing retinal ganglion cell layer
It has been reported that EphA4 is expressed in the ganglion cell layer of the mouse retina (Reber et al., 2004), astrocyte progenitor cells at the optic disc (Petros et al., 2006; Agrawal et al., 2014), and Müller cells in the postnatal retina (Roesch et al., 2008). In situ hybridization at embryonic and postnatal stages combined with immunostaining for the RGC marker Brn3a (akin Pou4f1) confirmed these previous findings and also revealed that EphA4 expression in the RGC layer initiates at E15.5 and peaks at perinatal stages. This receptor continues to be expressed until P10 and is downregulated between P14 and adulthood (Fig. 1A–K). Thus, unlike EphA5/6, expressed in the early embryonic retina in a high-ventral to low-dorsal gradient (Feldheim et al., 2000), EphA4 does not exhibit a gradual expression pattern in the RGC layer; instead, it is uniformly expressed across the entire retina, and its peak of expression is delayed until late embryonic stages. We also validated the late temporal expression of EphA4 by analyzing previously reported single-cell RNA-seq data from developing mouse retinas (Shekhar et al., 2022). Our computational reanalysis of the previously published retinal datasets revealed that EphA4 is expressed in a limited number of cells coexpressing other RGC markers at embryonic days E13/14, while the proportion of double-labeled cells increases at birth (Fig. 1L,M). These findings indicate that EphA4 and EphA5/6 likely fulfill different roles in the mouse retina, highlighting the necessity for further investigations to elucidate the specific functions of EphA4.
Spatiotemporal expression of EphA4 mRNA in the developing mouse retina. In situ hybridization for EphA4 mRNA in coronal sections from wild-type mice thought development and postnatal stages. A, At E14.5 EphA4 transcripts are detected in the optic disc. B–F, From E15.5 to postnatal stages, EphA4 mRNA is expressed in the RGCs layer (black arrows) and continues to be expressed until postnatal stages disappear in the adult retina. E, F, From P14 in advance, EphA4 mRNA is also expressed in the inner plexiform layer (IPL). G–K, Fluorescence in situ hybridization for EphA4 mRNA combined with immunofluorescence for Pou4f1 (Brn3a) in the RGCs layer. L, 2D visualization of single-cell RGC clusters colored by developmental stages (from E13 to P0) or according to the expression of different RGC markers. M, Plot represents the percentage of EphA4+ cells expressing different specific RGC markers from E13 to P0. Scale bar, 200 μm (A–C), 100 μm (D–F), 1 mm (G), and 50 μm (H, I, J, K). See Extended Data Figure 1-1 for more details.
Figure 1-1
EphA4KO mice do not show evident morphological retinal defects. (A-B) Tuj1 immunostaining in retinal section shows that intraretinal RGC axons are not affected in EphA4KO mice. (C-D) CHAT immunostaining in retinal sections labels amacrine cells and (E-F) calbindin labeling amacrine and horizontal cells show no gross difference in the IPL layering, where RGC dendrites arborize. (G-H) Immunohistochemistry for transcription factor Pou4f1 (Brn3a) in retinal sections labels RGCs cells and show no major differences in the RGCs layer. ONL (Outer Nuclear Layer); OPL (Outer Plexiform Layer); INL (Inner Nuclear Layer); IPL (Inner Plexiform Layer); GCL (Ganglion Cell Layer). Scale bar 100 μm. Download Figure 1-1, TIF file.
EphA4 is essential for proper axonal arborization in the medial superior colliculus
To assess the role of EphA4 in RGCs, we examined the retina and the visual projections of EphA4KO mice along the retinofugal pathway. We first analyzed cell stratification in the absence of EphA4 by performing immunohistochemistry in the retina of P9 wild-type and EphA4KO mice using markers for neurons located in the RGC layer. We did not observe notable defects in the number or anatomy of ganglion or amacrine cells, indicating that EphA4 is not necessary for their differentiation (Extended Data Fig. 1-1). Next, we electroporated plasmids encoding the green fluorescence protein (GFP) in the retinas of control and EphA4KO newborn mice and analyzed targeted RGCs (Fig. 2A). Retinal electroporation at birth allows targeting of individual RGCs and permits the visualization of their dendrites and axons (Benjumeda et al., 2013; Dhande et al., 2013; Marcucci et al., 2016; Murcia-Belmonte et al., 2019). Single-labeled RGCs showed a similar dendritic morphology in both EphA4KO and control mice (Fig. 2B,C), and their axons reached the correct area at the SC in both EphA4KO and control mice (Fig. 2D–G), indicating that EphA4 is not necessary for axon growth or for reaching the correct terminal zone. However, although axon terminals of dorsal RGCs from EphA4KO mice arborized properly in a similar collicular location than control axons (Fig. 2D,E,H), terminals coming from the ventral retina showed fainter arborization at the medial SC region compared with the controls (Fig. 2F,G,H). Sagittal sections from the medial SC further confirmed this phenotype (Fig. 2J,K).
EphA4 is required for the establishment of RGC arbors in the medial superior colliculus. A, Experimental approach used to analyze RGC terminals at the superior colliculus (SC). Retinas were electroporated at P0 with GFP encoding plasmids to target single RGCs. Electroporated mice were analyzed at P9, when retinal terminals reached their termination zone in the SC (R, rostral; C, caudal; L, lateral; M, medial). Retinas were examined to confirm single-cell electroporation, and the image color was inverted. B, C, Representative image of a single RGCs electroporated in the retina. D, E, Representative example of an RGC axon coming from the dorsal retina targeting the lateral SC in control mice or in EphA4KO mice. F, Representative example of a retinal axon coming from the ventral retina targeting the medial SC in control mice. G, Representative example of a retinal axon coming from the ventral retina in an EphA4KO mouse. Note that the arborization is highly faded compared with the control. H, Quantification of axonal arborizations in whole-mount SCs. The dots represent the area (μm2) occupied by RGC arborizations from EphA4KO and control mice in the lateral and medial SC. t test *p < 0.05. SC from at least four animals/genotype were analyzed. I, Drawing of the sectioning plane used to analyze RGC axon arborizations in the SC. J, K, Representative sagittal sections across the SC confirm that retinal axons from both EphA4KO and control mice reach the target area (arrows). Unlike in control mice (arrowheads), axon terminals do not properly arborize in medial collicular regions of EphA4KO mice (empty arrowheads). Sections from at least four animals/genotype were analyzed. Scale bar, 50 μm (B, C), 500 μm (D–G) and 100 μm (J, K). See Extended Data Figure 2-1 for more details.
Figure 2-1
Spacio-temporal expression pattern of mRNA ephrinB1 and ephrinB2 in the retina and in the SC. (A-D) In situ hybridization in retinal sections shows the expression of ephrinB1 transcripts through development. (E-F) mRNA ephrinB2 expression in a coronal section of the retina (E16.5) and superior colliculus at P0. (G-J’’) mRNA ephrinB1 expression in a coronal section of the superior colliculus at E16.5 (H- H’), E18.5 (I- I’) and P0 (J- J’) showing that ephrinB1 mRNA is expressed in a high medial-low lateral gradient throughout the superior colliculus. Scale bar 200 μm (A-C, E, H’- J’) and 500 μm (D, F, H-J). Download Figure 2-1, TIF file.
These findings suggested that EphA4 does not participate in RGC axon growth, guidance, or targeting at the SC, but it is essential to establish proper RGC arborizations at medial collicular areas.
EphrinB1 but not ephrinB2 induces EphA4-dependent adhesion
Because it was proposed that ephrinB1 expression in the SC acts as a ligand to mediate axon retinocollicular mapping of RGCs (Hindges et al., 2002), we wondered whether this ligand may activate the signaling mediated by EphA4 in RGC axons. We first confirmed that ephrinB1 and ephrinB2 are not expressed in the RGC layer of wild-type mice from E14 to P0 and observed that ephrinB1, but not ephrinB2, shows a high-medial to low-lateral gradient in the superficial layer of the superior colliculus at perinatal stages (Extended Data Fig. 2-1).
To investigate whether ephrinB1 mediates RGC axonal responses, we performed in vitro assays by exposing retinal explants from the dorsal or ventral retina to ephrinB1 or control (Fc) stripes. As ephrinB2 has been reported to mediate EphA4 repulsive signaling (Brors et al., 2003; Cooke et al., 2005; G. Zimmer et al., 2011), we also included ephrinB2 stripes as an extra control of the assay. As expected, RGC axons from the dorsal retinal explants were more repelled by ephrinB2 than by control stripes and this repulsion was significantly stronger in axons from the ventral retina (Fig. 3A,B,I). In contrast, RGC axons from either dorsal or ventral retina showed no repulsion toward ephrinB1, and instead, they preferentially grew on ephrinB1 stripes (Fig. 3C,D,J). Strikingly, both the repulsion for ephrinB2 and the preference for ephrinB1 were abolished in EphA4 null mice explants (Fig. 3E–H,I,J). These findings revealed that, while ephrinB2 acts as a repellent for EphA4 RGC axons, ephrinB1 triggers EphA4-mediated RGC axon adhesion.
EphA4/ephrinB2 signaling mediates RGC axonal repulsion but EphA4/ephrinB1 promotes axon adhesion. A–H, Representative dorsal and ventral retinal explants from wild-type and EphA4KO E15 embryos growing on alternate stripes containing Fc (black) and ephrinB2 or B1 (red). The explants were stained with anti-β-III tubulin (Tuj1) to visualize axons (gray). Images were digitalized, and Tuj1+ pixels (green) on red stripes were quantified and normalized to the Fc/Fc control (n = 4 independent experiments). I, J, Each dot in the boxplots represents the percentage of pixels (axons) on red stripes (n = 11–46 explants/condition from 4 independent experiments). Statistical differences (one-way ANOVA) are indicated after Tukey's (post hoc) corrected multiple comparison: *p < 0.05; **p < 0.01; ***p < 0.001. Scale bar, 200 μm.
To evaluate a potential differential response of retinal axons to ephrinB1 and ephrinB2 in vivo, we monocularly electroporated ephrinB1 and ephrinB2 encoding plasmids in wild-type mouse embryos and analyzed the phenotypes at P9 (Fig. 4A). The analysis of the electroporated retinas showed an aberrant RGC distribution after electroporation of ephrinB1. Targeted cells were not evenly located and grouped forming patches in the RGC layer (which expresses EphA4 endogenously), but not in the inner nuclear layer (Fig. 4C–C’’; Table 1). This phenotype was not observed in the retinas ectopically expressing ephrinB2 or control retinas electroporated with GFP alone (Fig. 4B–B’’,D–D’’; Table 1). Moreover, overexpression of ephrinB1 in the retina of mice lacking EphA4 did not produce the same phenotype as in control retinas (Extended Data Fig. 4-1), and the electroporation of a mutated form of ephrinB1 that specifically lacks the intracellular domain (from 305 to 1180aa; ephrinB1-ΔC), showed a phenotype similar to the retinas electroporated with control GFP plasmids alone (Extended Data Fig. 4-1). Together, these results strongly suggested that ephrinB1, but not ephrinB2, activates EphA4 signaling in RGCs to induce cell–cell adhesion and that this adhesive phenotype is mediated through forward signaling rather than reverse signaling.
EphA4/ephrinB1 signaling promotes cell adhesion in vivo. A, Experimental approach carried out to analyze the retinas (whole mounted and sections) of P9 mice electroporated at E13.5 with GFP, ephrinB1, or ephrinB2 encoding plasmids. B–D, Flat-mounted views of P9 retinas electroporated with GFP, ephrinB1, or ephrinB2 plasmids. B’–D’, High magnification of electroporated retinal regions. Notably, electroporation of ephrinB1 but not ephinB2 or GFP alone produced patches of targeted cells in control retinas. B’’–D’’, Sections of electroporated retinas with ephrinB1 revealed that targeted cells form patches in the RGC but not in the inner nuclear layer, while electroporated retinas with GFP or ephrinB2 plasmid did not show patchy phenotype neither in the RGCs nor the inner nuclear layer. Below each retinal section, two-dimensional graphs show fluorescence intensities (FI) and normalized intensity plots (N. Int.) obtained from the areas delineated with a dashed rectangle across the inner nuclear (top) and RGCs layer (bottom). At least three different animals were analyzed for each condition showing a consistent phenotype. Scale bar, 1 mm (B–D), 100 μm (B’–D’), and 200 μm (B’’–D’’). See Extended Data Figure 4-1 for more details.
Figure 4-1
Ectopic expression of ephrinB1 in EphA4KO mice and ephrinB1 lacking the intracellular domain in a control mice. (A) Representative section of a P9 postnatal retina electroporated with ephrinB1 in a EphA4KO mice did not produce any patchy phenotype neither the RGCs nor the inner nuclear layer. Two-dimensional plot shows fluorescence intensities (FI) and normalized intensity plots (N. Int) obtained from the areas delineated with a dashed rectangle show a similar phenotype to GFP expression in the RGC layer and in the inner nuclear layer. (B) Representative section of a retina electroporated with the mutant form of ephrinB1 lacking the intracellular domain produces a phenotype similar to that observed for GFP expression in RGCs layer. Two-dimensional plot shows fluorescence intensities (FI) and normalized intensity plots (N. Int.) obtained from the areas delineated with a dashed rectangle represent the fluorescence distribution in the RGC layer and in the inner nuclear layer. Three independent experiments consistently showed the same result. Scale bar 100 μm. Download Figure 4-1, TIF file.
Intensity fluctuation (IF) of ephrinB-electroporated cells in the retina
In vivo activation of EphA4 in RGCs leads to cell adhesion
To further investigate EphA4 function in vivo, we performed gain-of-function experiments by electroporating EphA4-encoding plasmids in wild-type embryos and analyzed the retinas and the axons of targeted RGCs at P9. EphA6 encoding plasmids were also electroporated as controls to explore potential differences in axon targeting and arborization between EphA4 and other EphA receptors. As previously reported, electroporation of EphA6 sends RGC axons to rostral areas of the SC (Carreres et al., 2011; Extended Data Fig. 5-1). In contrast, RGC axons ectopically expressing EphA4 projected in the correct topographic area, but, once there, they formed well-defined patches or columns contrasting with the RGCs electroporated with control GFP plasmids that showed a homogeneous distribution of the terminals (Fig. 5A–C). A patchy distribution of GFP cells was also observed in the corresponding flat-mounted retinas and in retinal sections of mice electroporated with EphA4-encoding plasmids in both the RGC and the inner nuclear layers (Fig. 5E–E’’; Table 2) compared with control GFP-expressing RGC cells (Fig. 5D–D’’; Table 2). These results support the idea that EphA4 mediates cell adhesion also in vivo.
Ectopic expression of EphA4 leads to RGC aggregation. A, Experimental approach carried out to analyze the superior colliculus (SC) of P9 mice electroporated at E13.5 with GFP- or EphA4-encoding plasmids. The asterisk marks the most rostral point of the SC in the section. The arrowheads indicate the initial point of the ROI (iROI). B, C, Representative sagittal sections of the SC from P9 mice electroporated at E13 with GFP or EphA4/GFP encoding plasmids. While control RGC axon arborizations are evenly distributed through the SC, terminals from EphA4-electroporated mice form patchy arborizations. Images at the right show two-dimensional plots representing fluorescence intensity (FI) signal and normalized intensity plots (N. Int.) obtained from the areas delineated by the dashed rectangle. D, E, Whole-mount retinas from P9 mice electroporated an E13.5 with GFP or EphA4/GFP encoding plasmids. D’, E’, High magnification of the retinas in D and E. D’’, F’’, Representative sections of electroporated retinas revealed that EphA4 but not control targeted cells form patches in both the RGC and inner nuclear layer. Below each retinal section, two-dimensional graphs show fluorescence intensities (FI) and normalized intensity plots (N. Int.) obtained from the areas delineated with a dashed rectangle drawn in the amacrine layer (top) and the RGCs layer (bottom). At least three different animals were analyzed for each condition showing a consistent phenotype. Scale bar, 100 μm (B, C, D’, E’), 1 mm (D, E), and 200 μm (D’’, E’’). See Extended Data Figure 5-1 for more details.
Figure 5-1
Ectopic expression of EphA6 leads to misarborization of central RGCs into rostral collicular areas. (A) EphA6 electroporated RGCs from the central retina project into the rostral SC. Two-dimensional plot shows fluorescence intensities (FI) and normalized intensity plot (N. Int) confirms the EphA6 axonal projection in the rostral SC. (B, B’, B’’) Wholemount retinas and a section of electroporated retinas with EphA6 did not produce any patchy phenotype in the RGCs layer. Two-dimensional plot shows fluorescence intensities (FI) and normalized intensity plots (N. Int.) obtained from the areas delineated with a dashed rectangle confirm the absence of retinal patches in the EphA6 electroporated RGCs. Scale bar 100 μm (A, B’), 1 mm (B) and 200 μm (B’’). Download Figure 5-1, TIF file.
Intensity fluctuation (IF) of EphA4-electroporated cells in retina and SC
The tyrosine kinase but not the SAM domain is required for EphA4-mediated cell adhesion
To investigate the EphA4 domain responsible for mediating cell adhesion, we next electroporated several mutant forms of EphA4 into the retinas of E13.5 embryos and analyzed both the retinas and SCs of P9 mice. First, we electroporated EphA4ΔC, a truncated form of the receptor that retains its ability to bind the ligand outside the cell but lacks the cytosolic domain and therefore does not convey information through forward signaling. As expected, the phenotype observed after electroporation of this mutated form of EphA4 was similar to electroporation of control GFP encoding plasmids, showing no disturbances in the organization of retinal neurons when observed in whole mount of retinal or SC sections (Fig. 6, compare A–A’’, B,B’’; Table 3). The sterile alpha motif (SAM) motif is known to mediate interactions among EphA receptors (Kullander et al., 2001), and electroporation of a mutated form of EphA4 that lacks the SAM domain (EphA4ΔSAM) showed a similar phenotype to the electroporation of full-length EphA4 plasmids (Fig. 6, compare A–A’’, C–C’’; Table 3), indicating that this motif is not required to mediate cell adhesion. Then, we electroporated a mutated form of EphA4 with a point mutation that changes the lysine residue K653 of the kinase domain to a methionine (EphA4KD; Kullander et al., 2001). The electroporation of EphA4KD resulted in an intermediate phenotype more closely resembling that observed with control plasmids, rather than the phenotype associated with the EphA4 full-length plasmid (Fig. 6, compare A–A’’, D–D’’; Table 3).
EphA4 kinase activity of K653R is required for the adhesive response. A–D’’, Schematic of EphA4 full-length and mutated versions of EphA4 and representative sections of retinas electroporated with the different EphA4 mutated forms as well as SC section belonging to the same samples confirmed the phenotypes observed in whole-mount preparations. Two-dimensional graphs below each retinal section or SC section show fluorescence intensities (FI) and normalized intensity plots (N. Int.) obtained from the areas delineated with a dashed rectangle drawn in the RGC layer or in the SC. A–A’’, Flat-mounted retina, retinal section electroporated with a full-length form of EphA4 and their corresponding SC section showing the patches phenotype and below the plots representing fluorescence intensity signal (FI) and normalized intensity plots (N. Int.) obtained from the areas delineated with a dashed rectangle. B–B’’, Flat-mounted retina, retinal section, and SC showed that, contrasting to electroporation of full-length EphA4, an EphA4 mutated version lacking the cytoplasmic domain leads a phenotype similar to control GFP plasmids. C–C’’, Electroporation of a mutated form of EphA4 lacking the sterile alpha motif (SAM) leads to a phenotype similar to electroporation of the EphA4 full-length receptor. D–D’’, Electroporation of an EphA4 form containing a point mutation in the Tyr653 residue results in an intermediate phenotype between the control and EphA4 full length. At least three different animals were analyzed for each condition showing a consistent phenotype. Scale bar, 100 μm (A’–D’) and 200 μm (A’’–D’’).
Intensity fluctuation (IF) of electroporated cells in the retina and SC
These findings suggest that the SAM domain of EphA4 is not required for mediating the adhesive response, as is the case for EphA4 signaling in the corticospinal tract, where the SAM domain is also dispensable (Kullander et al., 2001). On the contrary, the kinase activity of the K653 residue in the tyrosine kinase domain is crucial for mediating the adhesive axonal response observed in the retina and the SC. It is known that the phosphorylation of different tyrosine residues within the EphA4 juxtamembrane region controls the kinase activity of this receptor (Egea et al., 2005). The intermediate phenotype observed after electroporation of the EphA4KD plasmid suggests that phosphorylation of other residues, likely in the juxtamembrane domain, may be also important to activate cell adhesion.
Discussion
We report here that the tyrosine kinase receptor EphA4 plays an essential role in shaping visual axon arborizations at the medial areas of the superior colliculus. In contrast to previous studies in vitro suggesting that EphA4/ephrinB1 complexes can not form, our results reveal that EphA4/ephrinB1 binding may occur in vivo to trigger an adhesive cell–cell response. Indeed, they support the notion that the adhesive signaling mediated by EphA4 is likely accomplished through interactions with ephrinB1 and involves the tyrosine kinase but not the SAM domain. We propose a model in which retinal EphA4 signaling is activated by collicular ephrinB1, facilitating the anchoring of interstitial branches to cells within collicular regions near the midline (Fig. 7).
Working model. The diagram illustrates the retinotopic map within the superior colliculus (SC) in control and EphA4 knock-out mice. Axons expressing EphA4 exhibit proper arborization and can reach both the lateral and medial regions of the SC. In mice lacking EphA4, axons originating in the dorsal retina also project correctly to lateral collicular areas but RGCs from the ventral retina reach the medial SC but once there they fail to form appropriate arbors. Our findings propose a crucial role for EphA4 expressed in the retinal ganglion cell terminals. Axons coming from the ventral retina bind to ephrinB1, expressed by SC cells promoting an adhesive response that anchors visual axons from the ventral retina to lateral SC cells.
EphA4 mediates adhesion rather than repulsion in retinal ganglion cells
Although it has been long assumed that EphA4 plays a redundant function with other EphA receptors for the establishment of the retinotopic map (Walkenhorst et al., 2000), its expression pattern already suggested a role for EphA4 in RGC axons distinct from other EphAs. Previous studies have reported that during early embryonic development, EphA4 transcripts are detected in the optic nerve but not in the retina (Petros et al., 2006; Agrawal et al., 2014), and by birth, EphA4 is expressed in RGCs without showing any notable gradient (Feldheim et al., 1998; Reber et al., 2004). Here, we confirmed these findings and observed that EphA4 expression in RGCs begins later than other Eph receptors. The expression of EphA4 in RGCs is downregulated at postnatal stages, once the retinotopic map is completely established. EphA3 is responsible for the guidance of the RGCs along the rostrocaudal axis of the tectum in chick (Cheng et al., 1995; Feldheim et al., 1998, 2004; Brown et al., 2000), and EphA5/A6 play similar functions in mammals (Carreres et al., 2011). Contrary to the results obtained when EphA6 or EphA5 are ectopically introduced in RGCs, electroporation of EphA4 did not change the targeting of visual axons to rostral positions. Instead, interstitial branches were attached to each other and projected in a fasciculated manner forming columns. Adhesion was also induced in the cell body of retinal cells ectopically expressing EphA4. Together with the genetic removal of EphA4 that led to impaired terminal arborizations of RGC axons in the medial region of SC that express ephrinB1, these observations strongly suggest that EphA4 receptors induce cell adhesion between axons from the ventral retina and cells from the medial SC. The results also suggest that, contrary to previous proposals, EphA4 is not interchangeable with other EphA receptors during visual system development.
The reanalysis of previously published RNA sequencing data (Shekhar et al., 2022) also confirmed this late EphA4 retinal expression. In situ hybridization data in retinal sections suggest that a high percentage of RGCs simultaneously express EphA4. Although the percentage of RGCs expressing both EphA4 and specific RGC markers peaks at birth, it does not reach 100% according to the reanalysis of single-cell RNA-seq data. Several nonexcluding explanations could account for this apparent discrepancy: (1) EphA4 may be transiently expressed in RGCs, resulting in a progressive increase in the percentage of double-positive cells as development advances; (2) the identification of RGCs in the scRNA-seq dataset using various markers may overestimate the actual number of RGCs; or (3) EphA4 expression may not be present in all RGCs. The in situ data show abundant EphA4 mRNA expression throughout the entire RGC layer, supporting the first and/or second hypothesis. However, it remains to be determined whether EphA4 is indeed expressed by all RGCs at some stage of their maturation or is restricted to only a subset of RGCs.
EphrinB1 is detected in a high-medial to low-lateral gradient in the mouse SC at perinatal stages (Hindges et al., 2002; Schmitt et al., 2006; Extended Data Fig. 2-1). Ventral axons express high levels of EphB receptors and project into medial collicular areas. Thus, it was proposed that EphB/ephrinB signaling should be mediating an attractive or permissive response (Hindges et al., 2002). However, EphB/ephrinB signaling triggers repulsion in most of the cell types they act, and, so far EphB/ephrinB signaling has not been reported to mediate attraction or adhesion. In addition, a potential implication of EphBs in the arborization of retinal axons along the mediolateral axis does not match with their temporal expression pattern because they are expressed earlier. While EphBs reach their highest expression levels around E15 in the ventral retina, EphA4 is expressed predominantly at perinatal stages, aligning with the phases of arborization and pruning. These expression patterns together with the stripe assays and the functional experiments favor the idea that EphA4/ephrinB1 signaling acts later than signalings mediated by other EphA receptors, once RGC axons have reached their termination zones.
Coinciding with ephrinB1, Wnt3 is also expressed in a mediolateral decreasing gradient in the superior colliculus. It has been proposed that Wnt3 mediates repulsion through the Ryk receptor, expressed in a ventrodorsal decreasing gradient in the retina, but also mediates attraction of dorsal axons at lower Wnt3 concentrations through frizzled receptors (Schmitt et al., 2006). On the other hand, other adhesion molecules have been implicated in mediolateral mapping. In particular, activated leukocyte cell adhesion molecule (ALCAM) appears to play a role in mediolateral mapping as a substrate for incoming RGC axons within the SC due to its colocalization with ephrinB1 in the neuropil during perinatal stages (Buhusi et al., 2009). The arborization defects observed in medial collicular areas of ALCAM knock-out mice (Demyanenko and Maness, 2003) closely resemble those seen in EphA4KO mutants, suggesting a potential cooperation between ALCAM and ephrinB1 to activate EphA4 signaling in conjunction with Wnt3/Fz signaling and counterbalancing Wnt3/Ryk-mediated repulsion.
The adhesive response mediated by EphA4 depends on ephrinB1
Early studies reported the inability of ephrinB1 to initiate EphA4 signaling (Bowden et al., 2009), but more recent computational analyses supported a different view (Guo and Lesk, 2014). This latter study highlights the larger size of Tyrosine at position 121 in ephrinB1 compared with leucine in ephrinB2 and ephrinB3, leading to a potential steric clash with lysine 165 in EphA4. The adjacent residue (methionine in ephrinB1 and tryptophan in ephrinB2) may introduce a reduction in complementarity quality that will weaken the binding affinity between ephrinB1 and EphA4 without abolishing the formation of the ligand–receptor complex (Guo and Lesk, 2014). Our findings support this view confirming that EphA4/ephrinB1 binding occurs and elicits an adhesive response.
The dual functionality of EphA4, leading to adhesion when it binds to ephrinB1 and repulsion with ephrinB2, could be explained by the disengagement (or lack thereof) of the transinteraction between ephrins and EphA4 in the two cells. Ephrin–Eph interactions may be disrupted through endocytosis of the ephrin–Eph complex in one of the two attached cells or via proteolytic cleavage of ephrins or Ephs (M. Zimmer et al., 2003; Marston et al., 2003; Janes et al., 2005). In the absence of endocytosis or cleavage processes, cell adhesion is maintained. These two alternative EphA4-mediated responses may potentially be contingent on distinct receptor–ligand stoichiometry, as it has been previously hypothesized (Durbin et al., 1998; Mellitzer et al., 1999; Xu et al., 1999; Klein, 2004; Pasquale, 2010). However, if this would be the case, how RGC bodys and axons stick together in the gain-of-function experiments if ephrinB1 is not in the equation? The adhesive signaling triggered by EphA4 upong ephrinB1 binding needs to be further characterized. Computational modeling predicted a mechanism by which ephrins would exert concentration-dependent attractive and repulsive forces on retinal axons by modulating cell–substrate adhesion and force axons to terminate at positions within the optimal ephrinB gradient (Flanagan, 2006). This hypothesis may be possible as our results show that ephrinB2 and ephrinB1 trigger different responses in EphA4-expressing cells indicating that ligand concentration is not the only factor important to elicit a repulsive or adhesive response and the ligand itself determines the response.
EphA4/ephrinB1 signaling in other contexts
EphA4 combined with different ephrinBs had diverse effects in a variety of cells and tissues. While ephrin/Eph signaling is typically considered to induce a repulsive response, there was a body of evidence supporting an opposite role in cell–cell adhesion in certain contexts. For instance, the results observed in EphA4 loss-of-function experiments during the formation of the rhombomeres boundaries in zebrafish were not easy to explain by a simple model based on ephrin/EphA4-mediated repulsion (Cooke et al., 2005). If EphA4 would only mediate repulsion, EphA4-depleted cells should mix with ephrinB cell populations, creating a continuum between the even and odd rhombomeres. However, in the EphA4 mutants, the segmentation was still conserved. Thus, it was proposed that EphA4 should play an intrasegmental cell–cell adhesion role. Transinteractions between EphA4-expressing cells and cells strongly expressing ephrinBs would lead to repulsion, whereas interactions between EphA4 and weakly expressed A-type or B-type ligands within rombomeres 3 and 5, should promote adhesion. Unfortunately, the expression of ephrinB1 in zebrafish rombomeres has not been reported, but it would be interesting to know whether ephrinB1 is actually mediating intrasegmental EphA4-dependent adhesion in this context.
On the other hand, recent investigations have implicated EphA4 in Alzheimer's disease. EphA4 has been identified as a receptor for Aβ oligomers (AβOs) and activates c-Abl intracellular signaling, resulting in synaptic spine alterations in Alzheimer's disease (Vargas et al., 2018). Notably, the inhibition of EphA4 signaling increases Aβ levels, while EphA4 overexpression signalizing through ephrin ligands leads to a decrease in Aβ levels (Poppe et al., 2019; Tamura et al., 2020). EphrinB1 also plays a crucial role in synaptic maturation (Lim et al., 2008), and it is implicated in craniofrontonasal syndrome, which is characterized by severe hypertelorism, frontonasal dysplasia, craniosynostosis, and developmental delays (Twigg et al., 2004; Wieland et al., 2004, 2007; Arvanitis et al., 2014), although it is unclear the precise mechanism by which mutations in this ligand produces this condition. Gaining a comprehensive understanding of the biology and mechanisms of action of EphA4 and the different responses triggered by its ligands is crucial for a better understanding of these and other pathological conditions.
Footnotes
We thank Macarena Herrera and Daniel Becerra for their technical assistance and Daniel del Toro for the stripe assay quantification method. G.C. holds a Consolider-Ingenio Fellowship from the Spanish Government, and V.M-B. holds a JdC-Incorporation Postdoctoral Fellowship from the Spanish Government. This work was supported by grants from the Valencia Regional Government (PROMETEO/2020/007; CIGE/2021/148; CIGE/2022/84), the Spanish Government (PID2022-138245NB-I00), Fundación La Caixa (HR21-00824), and Fundación ICAR (CelMa-ENVEJECE). The Instituto de Neurociencias is a Severo Ochoa Excellence Center CEX2021-001165-S.
- Correspondence should be addressed to Eloísa Herrera at e.herrera{at}umh.es or Verónica Murcia-Belmonte at vmurcia{at}umh.es.