A dual shaping mechanism for postsynaptic ephrin-B3 as a receptor that sculpts dendrites and synapses

Abstract

As the neural network becomes wired, postsynaptic signaling molecules are thought to control the growth of dendrites and synapses. However, how these molecules are coordinated to sculpt postsynaptic structures is less well understood. We find that ephrin-B3, a transmembrane ligand for Eph receptors, functions postsynaptically as a receptor to transduce reverse signals into developing dendrites of mouse hippocampal neurons. Both tyrosine phosphorylation–dependent GRB4 SH2/SH3 adaptor-mediated signals and PSD-95–discs large–zona occludens-1 (PDZ) domain–dependent signals are required for inhibition of dendrite branching, whereas only PDZ interactions are necessary for spine formation and excitatory synaptic function. PICK1 and syntenin, two PDZ domain proteins, participate with ephrin-B3 in these postsynaptic activities. PICK1 has a specific role in spine and synapse formation, and syntenin promotes both dendrite pruning and synapse formation to build postsynaptic structures that are essential for neural circuits. The study thus dissects ephrin-B reverse signaling into three distinct intracellular pathways and protein–protein interactions that mediate the maturation of postsynaptic neurons.

Main

During brain development, neurons wire together by axon–dendrite contact, and the initial connections of nerve terminals with dendritic protrusions are thought to initiate the formation of excitatory synapses. Although a number of cell surface receptors, cell adhesion molecules and intracellular cytoskeletal regulators have been implicated in the formation of neural circuits^1,2, how these molecules signal and coordinate their activities in vivo remains poorly understood.

Recent studies have shown that Eph receptor tyrosine kinases and their membrane-anchored ephrin ligands have critical roles in neuronal connectivity and excitatory synapse formation upon axon-dendrite contact^1,3,4. The Eph family is the largest group of receptor tyrosine kinases and is divided into two subfamilies, termed EphA and EphB. Their ephrin ligands are also attached to the plasma membrane either through a glycosylphosphatidylinositol linkage (ephrin-A) or by a hydrophobic transmembrane domain (ephrin-B) that is followed by a short, highly conserved cytoplasmic tail that becomes tyrosine phosphorylated and can associate with both SH2 and PDZ domain–containing proteins⁵. Key features of EphB and ephrin-B interactions are their ability to transduce signals bidirectionally into both the EphB-expressing cells (forward signaling) and ephrin-B-expressing cells (reverse signaling) on cell–cell contact^6,7, and their involvement in a wide array of developmental and adult processes⁸.

EphB receptors are known to be expressed in dendrites and transduce postsynaptic forward signals that promote the assembly, maturation and plasticity of synapses^1,3,4. Ephrin-Bs are expressed in both axons and dendrites, and can function as the ligand to induce postsynaptic EphB forward signaling^9,10,11,12 and as a receptor to transduce reverse signals that are important for axon guidance and pruning^6,13,14, presynaptic development¹⁵ and plasticity of mossy fiber–CA3 and retinal–tectal synapses^16,17. Although much less understood, ephrin-B proteins that are expressed in the postsynaptic compartment have been implicated in synapse formation in cultured neurons^18,19,20 and in the synaptic plasticity of the Schaffer/CA3–CA1 circuit in the fully developed adult brain^{21,22,23,24,25}. However, the postsynaptic roles of ephrin-Bs as revealed using in vitro cultured neuron-based assays remain ambiguous, and the involvement of reverse signals in vivo in developing dendrites and spines are largely unknown, even though ephrin-B3 (henceforth referred to as eB3) is expressed to high levels in the dendritic field of certain neurons (see below). We thus hypothesized that postsynaptic eB3 may be able to transduce reverse signals into dendrites that are necessary for the early development of synaptic structures. Here we report our in vivo analysis of gene-targeted mutant mice, which shows that eB3 transduces tyrosine phosphorylation/ SH2-dependent and PDZ-dependent reverse signals to bring about morphogenesis of the postsynaptic compartment.

Results

eB3 reverse signaling shapes dendrites and spines

We used the Thy1-GFP-M transgenic mouse²⁶ to visualize a subset of hippocampal CA1 pyramidal neurons during the critical postnatal period, postnatal day 9 (P9) to P20, and observed that the development of dendrites in wild-type mice proceeds through initial elongation and pruning of primary branches that extend from the cell body, subsequent branching of secondary dendritic arbors and finally the formation of spines and synapses (Supplementary Fig. 1). The Allen Developing Mouse Brain Atlas expression database shows the ephrin-B3 (Efnb3) gene, encoding one of the transmembrane ligands for Eph receptors, is specifically and highly expressed in CA1 pyramidal neurons (Supplementary Fig. 2). We further analyzed the expression of eB3 protein in CA1 area throughout postnatal development (P7–P42) using X-gal staining of brain sections from mice carrying the Efnb3^lacZ reporter knock-in mutation (Efnb3^lacZ/+)²⁷. This mutation encodes a membrane localized C-terminal truncated ephrin-B3–β-galactosidase (eB3–β-gal) fusion protein that is expressed in the correct temporal, spatial and subcellular pattern as the endogenous wild-type protein, and provides for a high signal-to-noise ratio reporter. Our analysis revealed that while eB3–β-gal was barely detected at P7, it strongly and specifically labeled the dendritic field of CA1 pyramidal cells afterwards (Fig. 1a). This eB3 expression is colocalized with PSD-95, a postsynaptic marker in cultured hippocampal neurons (Supplementary Fig. 3). The data are consistent with the Allen Brain Atlas database (Supplementary Fig. 2) and previous studies^21,22,24,25 that indicated a postsynaptic role in the mature adult brain for ephrin-B2 (eB2) and eB3 proteins in mediating synaptic plasticity of the Schaffer/CA3–CA1 circuit. Although roles for ephrin-B1 (eB1)-mediated reverse signaling have also been implicated in the dendrites of cultured hippocampal neurons²⁰, expression of the ephrin-B1 (Efnb1) gene is undetectable in the developing and adult hippocampus as shown in the Allen Brain Atlas database (Supplementary Fig. 2).

Figure 1: eB3 is required for dendrite pruning and spine formation in hippocampal CA1 neurons.

(a) Developmental expression of eB3 in the CA1 pyramidal cell layer visualized by X-gal staining of coronal sections for the eB3–β-gal fusion protein (blue) in Efnb3^lacZ mice from postnatal week 1 (PW1) to PW6. Structure of the hippocampus is visualized by eosin counterstaining (red). Arrowheads indicate the localization of eB3 protein in the CA1 dendritic field. There is also strong expression of eB3 in granule cells of the dentate gyrus (DG) where the eB3–β-gal fusion localizes to mossy fiber axons and dendrites. Scale bars represent 300 μm (top) and 150 μm (bottom). (b) Efnb3^−/− and Efnb3^−/−; Efnb2^lacZ/6FΔV mutants at P12 showed excessive dendrites and reduced spine density in CA1 pyramidal neurons. Top left, dashed white circle indicates the crossed primary dendrites. Green, Thy1-GFP-M fluorescence; blue, NeuroTrace to stain CA1 pyramidal cell layer; white arrowheads, individual spines. Scale bars represent 20 μm (top) and 10 μm (bottom). (c–f) Quantification of primary dendrites (c), spine density (d), percentage distribution of spine length (e) and spine head diameter (f) in wild-type, Efnb3^−/− and Efnb3^−/−; Efnb2^lacZ/6FΔV neurons at P12 and P20 (n = 12 in each group). Error bars represent mean ± s.e.m. *P < 0.05, **P < 0.01, ***P < 0.001.

To determine if eB3 has a role in the initial development of dendrites and synapses, we examined the early morphological changes of dendritic branches and the density of spines along the dendrites in various Efnb mutants crossed with the Thy1-GFP-M transgenic mouse line to label a subset of CA1 neurons (Fig. 1b). The neurons in Efnb3^−/− mice showed a significant increase in the number of primary dendrites (P = 0.00006) and total dendritic branches (P = 0.00013) compared to the Efnb3^+/+ wild-type littermates at P12 (Fig. 1c,d and Supplementary Fig. 1). This result was confirmed using Golgi stain (Supplementary Fig. 4), and together these two methods indicate that eB3 negatively regulates dendrite growth and arborization of the developing CA1 pyramidal neurons. We carefully examined the postnatal time course of dendrite arborization in the CA1 pyramidal neurons of wild-type and Efnb3^−/− null mutants (Supplementary Fig. 1). The data showed that there are approximately 50% more primary dendrites extending from wild-type neurons at early postnatal days (that is, P9) than what is observed at later stages (P12 to P20), indicating that exuberant processes normally become eliminated or pruned between P9 and P12. By contrast, the number of primary dendrites in Efnb3^−/− mice did not become reduced at later postnatal stages. The Efnb3^−/− mice also showed more total dendritic branches at P12 compared to the wild-type controls; however, both mutant and wild-type groups eventually reached a stable level after P16 when no difference was noted between the two groups. As wild-type and Efnb3^−/− mutant mice both showed similar numbers of primary dendrites at P9, we conclude that the initial outgrowth of dendrites is not affected by the lack of eB3 expression. However, the failure to eliminate exuberant primary dendrites observed in the Efnb3^−/− mutant as the animal ages indicates a role for eB3 in postnatal pruning of dendritic branches.

We next asked whether the cytoplasmic domain of eB3 is required for dendritic pruning. As described previously²⁷, the Efnb3^lacZ mutation expresses a truncated eB3–β-gal fusion protein that retains the extracellular and transmembrane domains to provide ligand-like activities to initiate forward signaling but lacks the cytoplasmic segment and thus cannot transduce reverse signals. Efnb3^lacZ/lacZ mutants showed approximately 50% more primary dendrites and total dendritic branches that were similar to the Efnb3^−/− mice (Supplementary Fig. 5), suggesting that the role of eB3 in dendrite elimination is due to its ability to act like a receptor and transduce reverse signals. We next generated compound mutants using Efnb2^lacZ (C-terminal truncated²⁸) and Efnb2^6FΔV (point mutant eliminating both SH2 and PDZ binding²⁹) alleles that interfere with reverse signaling of the related eB2 protein, which is also expressed in CA1 neurons (Supplementary Fig. 2). The Efnb3^−/−; Efnb2^lacZ/6FΔV compound mutants exhibited an enhanced phenotype with even more primary dendrites, particularly at P20 (Fig. 1b,c).

The Thy1-GFP-M transgene also allowed for analysis of dendritic spines on the CA1 neurons in the various mutants (Fig. 1b). By contrast to the increase in primary dendrites and dendritic branches, the density of spines was greatly reduced in mutants that were analyzed at P12 compared with that of wild-type mice (Fig. 1d). In wild-type neurons, most spines are mushroom shaped (with a thin neck and large head), whereas spines in the mutant have protrusions with smaller head diameters and different lengths, either very short (<0.5 μm) or very long (>2 μm) (Fig. 1e,f), suggesting that the change in density of spines is attributable to an alteration of spine maturation. The strongest reduction in spine density was noted in the Efnb3^−/−; Efnb2^lacZ/6FΔV compound mutants, in which significant differences with wild type remained even at P20 (P = 0.022). The data are consistent with a role for ephrin-B reverse signaling in neuron maturation.

As EphB receptors are highly expressed in CA3 pyrimadal neurons¹⁴, it raises the question of whether these molecules act as ligands to stimulate ephrin-B reverse signaling for CA1 maturation. We therefore analyzed dendritic morphogenesis in protein-null mutants of EphB1, which is specifically expressed in CA3 neurons, and compound knockouts of both EphB1 and EphB2 to determine whether these molecules are biological partners for eB3 (Supplementary Fig. 6). We found increased CA1 branches and less matured spines in EphB1^−/− single mutants and an even more severe phenotype in EphB1^−/−; EphB2 ^−/− compound mutants, suggesting that EphB molecules are important for dendrite elimination and spine maturation. Notably, the deficit was recovered in EphB1^−/−; EphB2^{lacZ/ lacZ} mice to the level of EphB1^−/− single mutants, indicating that the intracellular segment of EphB2 is not required. These data confirm the results with eB3 mutants, and indicate that the role for EphB receptors here is to act as CA3-expressed ligands to stimulate ephrin-B reverse signals into dendrites of CA1 neurons.

eB3 is required and sufficient for synaptic function

We next examined the role of eB3 in synaptic function by performing whole-cell voltage clamp recordings in brain slices collected at P12 and P20 to assess spontaneous miniature excitatory postsynaptic currents (mEPSCs) occurring in CA1 pyramidal neurons in acute wild-type and Efnb3 mutants. Spontaneous mEPSCs are a measurement of the random release of presynaptic vesicles at different synaptic sites and are used to assess numbers of functional synapses. Although no change in amplitude was observed, both Efnb3^−/− (Fig. 2a,b) and Efnb3^lacZ/lacZ mutants (data not shown) exhibited a greatly reduced mEPSC frequency compared with the wild type (17.5% and 34.1%, respectively). Combined with the morphological data that indicated reduced numbers of dendritic spines in the mutants (Fig. 1d), the electrophysiology confirms a reduction in the number of functional Schaffer/CA3–CA1 synapses in the absence of postsynaptic eB3 reverse signaling.

Figure 2: eB3 is required and sufficient for excitatory synaptic function in hippocampal CA1 neurons.

(a,b) Efnb3^−/− mutant showed reduced mEPSC frequency in CA1 pyramidal neurons at P12 and P20. Scale bar represents 20 pA (vertical) × 1 s (horizontal). Error bars, mean ± s.e.m. **P < 0.01, ***P < 0.001. (c) Generation of conditional loxP-flanked knock-in Efnb3^neo mutant. Exons (coding and non-coding segments shown by dark gray and light gray boxes) and introns (lines) are flanked by a diphtheria toxin (DT-A) and thymidine kinase (TK) expression cassette for negative selection. EcoRI (E) and NsiI (N) indicate restriction sites. The primary targeted allele, Efnb3^neo homozygous was crossed with CAGG-Cre^ERT2M driver and a tdTomato reporter to identify cells exposed to active Cre. eB3 expression in the initial Efnb3^neo/neo mutant was restored on tamoxifen administration, which was used to induce Cre-mediated excision of the loxP-flanked PGK-neo cassette. (d) Schedule of tamoxifen treatment and hippocampus dissection during postnatal development. (e) Cre-mediated recombination in CA1 area was detected in P12 hippocampal sections by tdTomato fluorescence (arrowheads) in cre⁺; tdTomato⁺; Efnb3^neo/neo mice but not in cre⁻; tdTomato⁺; Efnb3^neo/neo mice following tamoxifen administration at P5. Scale bars represent 100 μm (left) and 20 μm (right). (f) mEPSC recordings were performed in tdTomato-positive (indicated by glass electrodes) and tdTomato-negative neurons in hippocampal CA1 area of cre⁺; tdTomato⁺; Efnb3^neo/neo mice after tamoxifen treatment. Scale bars represent 10 μm (left) and 20 pA (vertical) × 1 s (horizontal) (right).

To further investigate the sufficiency of eB3 in synaptic development, we designed an experiment to rescue expression of eB3 in a genetically null mouse. To accomplish this, a conditional knock-in mouse line was generated, termed Efnb3^neo, by inserting a loxP-flanked PGK-neo cassette into the intron between the forth and the fifth exons of Efnb3 (Fig. 2c). We previously used a similar strategy in our original generation of Efnb3 mutations²⁷ and found that such an insertion results in a protein-null allele, as judged by a hopping locomotion in adult homozygotes. Efnb3^neo/neo mice containing an ubiquitous CAGG-Cre^ERT2M transgene³⁰ and a loxP-flanked Stop tdTomato reporter³¹ (cre⁺; tdTomato⁺; Efnb3^neo/neo) were treated with tamoxifen at P5 (Fig. 2d,e) or at P8 (Supplementary Fig. 7) to activate Cre recombinase and delete the PGK-neo cassette, which allowed eB3 expression, for our purpose, in a subset of CA1 pyramidal neurons. At P12, mEPSCs were recorded from tdTomato-positive, which indicates eB3 rescued neurons or negative CA1 neurons in acute brain slices (Fig. 2f). The frequency of mEPSCs in tdTomato-positive neurons (0.52 ± 0.08 Hz) that were treated with tamoxifen at P5 was significantly higher than that of the tdTomato-negative neurons (0.11 ± 0.03 Hz , P = 0.00097), with no difference observed in amplitude between two groups (Fig. 2f). Thus, eB3 expression initiated at P5 is sufficient to rescue the Schaffer/CA3–CA1 synaptic defects associated with the mutants.

SH2 and PDZ binding mediate eB3 reverse signals

Reverse signaling that is mediated by ephrin-B proteins may involve at least two distinct pathways that are governed by the ability of its intracellular domain to form protein–protein interactions with both SH2 and PDZ domain-containing intracellular proteins. Following interactions with their cognate EphB receptors (and EphA4) and formation of circular tetramers and higher-order clusters, the ephrin-B cytoplasmic tail becomes tyrosine phosphorylated and recruits the SH2/SH3 adaptor protein GRB4 (ref. 32) to bridge the ephrin-B molecule with a number of intracellular proteins including the Rac guanine nucleotide exchange factor DOCK180 and downstream effector PAK1 (ref. 14). The ephrin-B proteins also bind numerous PDZ domain proteins through their conserved C-terminal tail⁵. We generated germline point mutations in the Efnb3 gene that eliminate either tyrosine phosphorylation, termed Efnb3^3F and Efnb3^5F (ref. 14), and/or the PDZ interaction motif by deleting the extreme C-terminal valine residue that is essential for PDZ binding, termed Efnb3^ΔV and Efnb3^3FΔV, respectively (Fig. 3a and Supplementary Fig. 8a,b). Using these mutations, we aimed to precisely determine which signals eB3 needs to transduce into the CA1 neurons for dendritic morphogenesis and synaptic transmission.

Figure 3: eB3 tyrosine phosphorylation and PDZ binding are differentially required for dendrite morphogenesis and synaptic function.

(a) Point mutations in Efnb3 that change tyrosines (Y) to phenylalanines (F) to eliminate tyrosine phosphorylation and SH2 binding (3F and 5F), the C-terminal valine (V) required for PDZ binding (ΔV) or both SH2 and PDZ binding (3FΔV). (b) Thy1-GFP-M fluorescence (green) was used to visualize the morphology of CA1 neurons at P12. White arrowheads indicate individual spines. Scale bars represent 20 μm (left) and 10 μm (right). (c–f) Quantification of primary dendrites (c), spine density (d), and percentage distribution of spine length (e) and spine head diameter (f) in wild-type, Efnb3^3F/3F, Efnb3^5F/5F, Efnb3^ΔV/ΔV and Efnb3^3FΔV/3FΔV mutants at P12 (n = 25 for each group). Error bars, mean ± s.e.m. *P < 0.05, **P < 0.01, ***P < 0.001.

As expected, the new Efnb3^ΔV and Efnb3^3FΔV mutations were found to be viable as homozygotes at Mendelian ratios, and the adults seemed healthy, fertile and long lived as previously shown for all the other Efnb3 mutants. Quantitative real-time PCR analysis revealed that all the mutants have similar Efnb3 mRNA level compared to that of wild-type mice (Supplementary Fig. 8c). Notably, like the previously described Efnb3^lacZ/lacZ, Efnb3^3F/3F and Efnb3^5F/5F mutants, the Efnb3^ΔV/ΔV and Efnb3^3FΔV/3FΔV adults also exhibit normal locomotion (as opposed to the Efnb3^−/− mice, which show a hopping gait), indicating that all these mutations express proteins that traffic to the plasma membrane and can still function as ligands to stimulate Eph forward signaling that is important for locomotor circuitry²⁷. To confirm the plasma membrane localization of the new eB3 mutant proteins, primary hippocampal neurons from wild-type, Efnb3^−/−, Efnb3^3F/3F, Efnb3^5F/5F, Efnb3^ΔV/ΔV and Efnb3^3FΔV/3FΔV mice were cultured for 12 d and cell-surface proteins were labeled with biotin⁶. Biotinylated proteins were then pulled down from lysates with streptavidin and immunoblotted with antibodies to eB3, revealing that all mutants expressed eB3 protein on the cell surface (Supplementary Fig. 8d). We then exposed neuron cultures to soluble preclustered EphB2-Fc extracellular domains and observed obvious tyrosine phosphorylation of eB3 in neurons from wild-type and Efnb3^ΔV/ΔV cultures (Supplementary Fig. 9a) but not Efnb3^5F/5F or Efnb3^{3F ΔV/3FΔV} neurons. Furthermore, although cultures from wild-type or Efnb3^5F/5F mice showed clear colocalization of the PDZ domain-containing protein PICK1 with eB3, this was not observed in Efnb3^ΔV/ΔV or Efnb3^{3F ΔV/3FΔV} neurons (Supplementary Fig. 9a). We further tested for a physical interaction between eB3 and PICK1 using co-immunoprecipitation from brain protein lysates. The data showed that eB3 was co-immunoprecipitated with PICK1 from wild-type or Efnb3^5F/5F but not that from Efnb3^ΔV/ΔV or Efnb3^3FΔV/3FΔV mutants (Supplementary Fig. 9b). Therefore, while the Efnb3^ΔV and Efnb3^{3F ΔV} mutations do express their respective eB3-ΔV and eB3-3FΔV proteins on the surface of hippocampal neurons and can be clustered by EphB-Fc treatment, they are not able to interact with PDZ domain-containing proteins.

Dendritic morphogenesis of the various Efnb3 point mutants was then visualized using the Thyl-GFP-M transgene at P12 to determine the roles of distinct signaling motifs of eB3 in arborization and spine formation. All the Efnb3 point mutants showed similar excessive primary dendritic branches comparable to that of the Efnb3^−/− or EfnbB3^lacZ/lacZ homozygotes, although only the Efnb3^ΔV/ΔV and Efnb3^3FΔV/3FΔV mutants showed reduced spine density along the dendrites (Fig. 3b–d). In the mutants with valine deleted, the percentage of matured spines (0.5–1 μm) was decreased, whereas that of long dendritic protrusions (>2.0 μm) was increased (Fig. 3e,f). Recordings of Schaffer/CA3–CA1 synapses in P12 brain slices revealed that the frequency and amplitude of mEPSCs in Efnb3^3F/3F and Efnb3^5F/5F mutants were not different from that of wild-type littermates. However, both Efnb3^ΔV/ΔV and Efnb3^3FΔV/3FΔV mutants showed a reduced mEPSC frequency with no change in amplitude (Fig. 4). These data indicate that both tyrosine phosphorylation and PDZ interactions of the eB3 cytoplasmic tail are required for the normal pruning of developing dendrites, and that only PDZ binding activities are essential for the development of spines and normal synaptic activity.

Figure 4: PDZ binding of eB3 are required for synaptic function.

(a) mEPSCs were recorded in CA1 pyramidal neurons from wild types, Efnb3^3F/3F, Efnb3^5F/5F, Efnb3^ΔV/ΔV and Efnb3^3FΔV/3FΔV mutants at P12. Scale bars represent 20 pA (vertical) × 1 s (horizontal). (b,c) Quantification of mEPSC frequency (b) and amplitude (c) for CA1 pyramidal neurons from the different Efnb3 mutants (n = 15). Error bars represent mean ± s.e.m. *P < 0.05, ***P < 0.001.

We next studied dendritic branching and synaptogenesis in cultures of primary hippocampal neurons from wild-type, Efnb3^−/− and Efnb3^ΔV/ΔV mutants collected at birth (P0). Following 8 d of growth in vitro, the cultures were transfected with a membrane-anchored farnesylated enhanced green fluorescent protein (f-EGFP) marker and then analyzed 4 d later with antibodies to GFP and synapsin, a presynaptic marker, to visualize dendrites and synapses. The data showed that neurons from Efnb3^−/− hippocampi had more primary dendritic processes (8.3 ± 1.3) and less spines and synapsin-positive synapses on spines compared to the wild-type neurons, which exhibited fewer primary dendrites (4.8 ± 0.8) and greater density of spines and synapses (Supplementary Fig. 10a,b and Fig. 5a,b). Similarly, neurons from Efnb3^ΔV/ΔV cultures also exhibited more primary dendrites (9.4 ± 0.7), and fewer spines and synapses, on both spines and shafts (Supplementary Fig. 10a,b and Fig. 5a,b). We further transfected eB2, eB3 or eB3ΔV into cultured Efnb3 ^−/− hippocampal neurons to see if the role of eB3 is rescued. The data showed that both eB2 and eB3 expression were able to compensate Efnb3 deficit, whereas eB3-ΔV mutant failed to rescue the deficit (Supplementary Fig. 10b). Electrophysiological recordings further revealed a reduced frequency of mEPSCs in Efnb3 ^−/− and Efnb3^ΔV/ΔV neuron cultures compared to wild-type littermates (Fig. 5c), which is consistent with the results obtained from acute hippocampal slices and indicates that fewer functional synapses form in the absence of the eB3 PDZ binding motif.

Figure 5: PICK1 and syntenin mediate eB3 reverse signaling through PDZ binding to control spine and synapse formation and synaptic function.

(a) Spine and synapse formation were indicated with a transfected f-EGFP reporter and presynaptic marker synapsin in 12-d cultured hippocampal neurons from wild-type, Efnb3^−/− and Efnb3^ΔV/ΔV mice. Arrowhead, synapse on spine; arrow, synapse on shaft. Scale bar represents 10 μm. (b) The density of spines and synapses on spines (arrowhead in a) and on shafts (arrow in a) were quantified. n = 10. (c) mEPSCs were recorded in 12–14-d cultured neurons from Efnb3 mutants and wild-type (left), and in Efnb3^−/− hippocampal neurons that were infected with lentivirus packaged eB3-PICK1 or eB3-syntenin (Syn) expression vectors. Quantification of mEPSC frequency (top) and amplitude (bottom) is shown. n = 15–20. Error bars, mean ± s.e.m. (d) Expression of Flag-tagged wild-type eB3-PICK1 or Syn chimeric fusion proteins (left panels) in transfected Efnb3^−/− neurons rescues spine and synapse formation, as visualized with f-EGFP and synapsin in chimeric protein-expressing neurons labeled with anti-Flag antibodies (upper panels). Expression of the eB3-PICK1-ΔPDZ or eB3-Syn-ΔPDZ1+2 fusion proteins deleted for the respective PDZ domains (right panels) had little if any effect on spine and synapse formation in Efnb3^−/− neurons. Arrowhead, synapse on spine; arrow, synapse on shaft. Scale bar represents 10 μm. (e) Examples of mEPSC recorded in Efnb3^−/− neurons expressing eB3-PICK1 or Syn fusion proteins and their PDZ-deleted mutant forms. Scale bars represent 40 pA (vertical) × 2 s (horizontal). *P < 0.05, **P < 0.01, *** P < 0.001.

PICK1 and syntenin transduce eB3 signals in dendrites

Among the downstream PDZ domain-containing molecules that may participate in ephrin-B reverse signaling, PICK1 and syntenin are expressed in the developing hippocampus (Supplementary Fig. 11), are localized to dendritic spines and synapses^33,34,35 and associate with ephrin-B proteins^36,37,38. PICK1 is a homo-oligomerizing bin–amphiphysin–rvs (BAR) domain protein that contains a single N-terminal PDZ domain that binds many target proteins³⁹. Syntenin has two PDZ domains with different binding affinity to ephrin-B and other membrane associated molecules such as syndecan, glutamate receptors and neurofascin⁴⁰, and has been implicated to participate with eB1 and eB2, but not eB3, in presynaptic development¹⁵. To determine whether PICK1 or syntenin is involved in postsynaptic development of dendrites and synapses, we overexpressed in primary hippocampal neurons the full-length versions of PICK1 or syntenin and various mutants, including the isolated PICK1 PDZ domain (PICK1-PDZ), and PDZ deletion mutants PICK1-ΔPDZ, syntenin-ΔPDZ1 (Syn-ΔPDZ1) and syntenin-ΔPDZ2 (Syn-ΔPDZ2), along with f-EGFP to label the transfected cells. Dendritic branches, spine density and synapse formation were visualized with antibodies to GFP and synapsin. The data showed that overexpressing wild-type syntenin in Efnb3 ^−/− neurons reduced the number of primary dendrites and increased the density of spines and synapses, whereas overexpressing PICK1 did not (Supplementary Figs. 12 and 13), suggesting that eB3 is essential for PICK1-mediated, but not syntenin-mediated, function in neurons. In wild-type neurons, overexpressing wild-type syntenin, but not its mutants, reduced the number of primary dendrites compared to untransfected control neurons, whereas no change in the number of primary dendrites was observed for any of the PICK1-expressing neurons (Supplementary Fig. 12). Moreover, expression of either PICK1-PDZ or syntenin mutants Syn-ΔPDZ1 or Syn-ΔPDZ2, which have one PDZ domain left, reduced the density of spines and synapses (Supplementary Fig. 13). Consistent with this observation, electrophysiological recordings in neuron cultures determined that expression of PICK1-PDZ or syntenin mutants reduced the frequency of mEPSCs in wild-type neurons, although wild-type PICK1 and syntenin led to an increase (Supplementary Fig. 14). These data suggest roles for PICK1 and syntenin and their PDZ domains in regulating the number of functional synapses.

To further determine the sufficiency of PICK1 and syntenin in ephrin-B signaling, we expressed in Efnb3^−/− primary hippocampal neurons Flag-tagged eB3-PICK1 and eB3-syntenin chimeric fusion proteins, in which the eB3 C-terminal valine residue was deleted and replaced with either full length PICK1 or syntenin (Flag-eB3-PICK1 or Flag-eB3-Syn) or their ΔPDZ mutants (Flag-eB3-PICK1-ΔPDZ or Flag-eB3-Syn-ΔPDZ1+2) (Fig. 5d,e). The idea behind these constructs is to bypass the requirement for PDZ mediated protein–protein associations of PICK1 or syntenin with eB3 and to exclude other possible PDZ associations by covalently linking PICK1 or syntenin directly to the eB3 C terminus but still keeping their ability to interact with either downstream PICK1- or syntenin-binding proteins. For the control fusion proteins Flag-eB3-PICK1-ΔPDZ or Flag-eB3-Syn-ΔPDZ1+2, lack of PDZ domains will disrupt their binding to other proteins. The expression vectors were co-transfected with f-EGFP into 8-d cultured primary hippocampal neurons from Efnb3^−/− pups collected at P0 and scored 4 d later using immunofluorescence with antibodies to GFP, synapsin and Flag. We found that Efnb3^−/− neurons expressing either eB3-PICK1 or eB3-Syn, which is indicated by antibody to Flag immunofluorescence (Fig. 5d), exhibited higher density of spines and spine synapse than the neurons expressing their PDZ-deleted forms (Fig. 5b,d), which were similar to the control untransfected Efnb3^−/− neurons. Notably, eB3-Syn-expressing neurons also showed reduced primary dendritic branches, which were similar to that of wild-type neurons, whereas eB3-PICK1-expressing neurons remained unchanged (Fig. 6). This is in contrast with the effects of gain-of-function eB3-GRB4-SH3 chimeric fusion proteins for axon pruning in transfected neurons¹⁴, which caused a reduction in primary dendritic branches but did not affect the amount of spines and synapses compared with loss-of-function eB3-GRB4-SH3Mut fusion proteins with inactivated SH3 domains (Fig. 6a,c). Electrophysiological recordings further confirmed that both eB3-PICK1 and eB3-Syn–expressing neurons exhibited remarkably higher mEPSC frequency than those expressing the PDZ domain deleted mutants although no change in amplitude was observed (Fig. 5c,e). These results indicate that both eB3-PICK1 and eB3-Syn chimeric fusion proteins provide for a gain-of-function of eB3 to induce spine maturation and synapse formation in the absence of other PDZ protein binding to eB3. Consistent with this, their PDZ domain deleted forms act as loss-of-function.

Figure 6: GRB4, PICK1 and syntenin mediated distinct reverse signaling to prune primary dendrites in cultured hippocampal neurons.

(a) Expression of eB3-GRB4-SH3 or eB3-syntenin (eB3-Syn) chimeric fusion proteins in transfected Efnb3^−/− hippocampal neurons reduces primary dendrites as visualized with f-EGFP. Expression of eB3-PICK1 or its mutant eB3-PICK1-ΔPDZ or mutants of eB3-GRB4 or eB3-Syn fusion proteins had little, if any, effect on the number of primary dendrites in Efnb3^−/− neurons. Scale bar represents 10 μm. (b) Cultured hippocampal neurons at P12 from Efnb3^−/− mice show more primary dendritic branches than wild-type littermates, as visualized with f-EGFP. Scale bar represents 10 μm. (c) Analysis of the number of primary dendrites in neurons expressing different chimeric fusion proteins. Error bars, mean ± s.e.m. n = 10–12. *P < 0.05, **P < 0.01.

To further clarify how the various wild-type and PDZ deleted eB3-PICK1 and eB3-Syn fusion proteins result in gain-of-function and loss-of-function effects, we tested whether these chimeric fusion proteins were able to form protein–protein complexes with syndecan-2, a syntenin-associated protein expressed in the hippocampus^36,41, or additional PICK1, which is thought to homo-oligomerize³⁹, using immunofluorescence co-localization and co-immunoprecipitation in transfected cells (Fig. 7 and Supplementary Fig. 15). The data showed that the wild-type form eB3-PICK1 was able to recruit more PICK1 than the eB3-PICK1-ΔPDZ form did, but no obvious interaction of syndecan-2 was found with either eB3-PICK1 fusion proteins. By contrast, the wild-type form eB3-Syn was able to interact with syndecan-2 but not PICK1, whereas Syn-ΔPDZ1+2 did not bind to either. These data suggest that although both PICK1 and syntenin are capable of mediating eB3 reverse signaling through a PDZ-dependent mechanism, these two proteins transduce ephrin-B signals through two distinct pathways involving different protein–protein interactions that either enable both dendrite pruning and spine maturation (syntenin) or only spine maturation (PICK1).

Figure 7: Association of eB3-PICK1 or eB3-Syn with downstream signaling molecules.

(a) Following EphB2-Fc treatment for 16 h in cultured hippocampal neurons to cluster the eB3 proteins to spots on the plasma membrane, wild-type Flag-eB3-PICK1 co-localized with PICK1, whereas wild-type Flag-eB3-Syn co-localized with syndecan-2 (SDC2) (arrowheads). The PDZ domain-deleted proteins showed highly diminished or no ability to form protein–protein interactions with PICK1 or SDC2. Scale bar represents 5 μm. (b) In transfected Cos-1 cells, HA-PICK1 or HA-SDC2 was co-immunoprecipitated with wild-type Flag-eB3-PICK1 or Flag-eB3-Syn, respectively, but little if any was precipitated with the eB3-PICK1-ΔPDZ or eB3-Syn-ΔPDZ1+2 PDZ deleted counterparts. IB, immunoblot; IP, immunoprecipitation. The blots presented were cropped, and the full-length blots are presented in Supplementary Figure 15.

Discussion

Ephrin-B mediates reverse signaling at postsynaptic structures

The current study is in contrast with previous extensive studies of postsynaptic Eph receptors on dendritic morphogenesis^1,3,4 that found ephrin-Bs to be involved presynaptically in axons and nerve terminals^{13,14,15,16,17}, and indicates that eB3 also functions as a postsynaptic receptor to transduce reverse signals that are required for both long-scale dendrite pruning and short-scale spine maturation. Conversely, as interacting molecules, EphB1 and EphB2 act as presynaptic ligands to stimulate eB3 reverse signals. Such a postsynaptic role for eB3 in the CA3–CA1 circuit is consistent with its intense expression in CA1 neurons initiated after the first postnatal week and coinciding with robust expression of EphB1 and EphB2 molecules in CA3 neurons¹⁴.

The present study provides evidence that upon axon–dendrite contact followed by interaction with presynaptic EphB receptors, postsynaptic eB3 is able to recruit GRB4 and PICK1/syntenin through SH2 and PDZ binding motifs, respectively. These downstream molecules bridge eB3 with different cytoskeletal regulators, receptors and ion channels by protein–protein interaction and altering either their subcellular targeting and/or surface expression^5,39,40. This provides a mechanism by which both dendrite pruning molecules and functional synaptic structure–related molecules are recruited to the postsynaptic sites on axon-dendrite contact at the earliest developmental stages. The dual function of postsynaptic eB3 is attributable to its ability to interact with three different downstream proteins, the SH2/SH3 adaptor GRB4 and two PDZ domain-containing proteins, syntenin and PICK1, which we propose are able to recruit additional proteins that help prune exuberant dendritic processes, sculpt dendritic spines and form functional synapses.

GRB4, PICK1 and syntenin transduce distinct signals

GRB4 has been shown to couple eB3 to DOCK180 and PAK to bring about guanine nucleotide exchange and signaling downstream of RAC and CDC42 (ref. 14), which may be involved in the regulation of dendrite growth and remodeling⁴². PICK1 contains a PDZ domain in the N terminus and a BAR domain in the C terminus³⁹. The PDZ domain of PICK1 binds to a large number of membrane proteins including ephrin-Bs, whereas the BAR domain binds lipid molecules and may oligomerize with other BAR domains³⁹. Our results show that the PDZ domain also contributes to the oligomerization and dimerization of PICK1, which is also supported by evidence that the PDZ domain enhances the BAR domain's lipid binding⁴³. Therefore, both BAR and PDZ domains are crucial for the formation of PICK1 homodimer on the cell membrane to bridge eB3 with various postsynaptic associated molecules.

Unlike GRB4 or PICK1, which function solely for dendritic pruning or spine maturation, respectively, syntenin affects both pruning and spine maturation, and both functions require its PDZ domains. Mechanistically, syntenin might interfere with actin dynamics by interacting with focal-adhesion kinase (FAK)⁴⁴ and by facilitating receptor subcellular trafficking, such as syndecan, which is regulated by aquaporin PIP2 and ARF family GTPase⁴⁵, and has been implicated in the regulation of dendritic filopodia and spine formation via the neurofibromin-PKA-Ena/VASP pathway⁴⁶. Furthermore, syntenin interacts with adhesion molecules such as synCAM, neurofascin and syndecan, all of which contribute to synaptic structure⁴⁰. Thus, the dual function of syntenin in dendrite morphogenesis that relays its surprising variety and diversity of interacting partners⁴⁰ contributes to the eB3 reverse signals that mediate synaptic maturation and formation of neural circuits.

Presynaptic EphBs act as ligands

In addition to the role of eB3 at postsynapses, the present study also provides evidence that presynaptic EphB1 and EphB2 localized in CA3 axon terminals serve as ligands to mediate trans-synaptic signaling. Although the postsynaptic requirement of EphB forward signaling for spine maturation and synapse formation in CA3 neurons has been demonstrated previously¹², our data show that the specific CA3-expressed receptor EphB1 is also required and CA3/CA1-expressed EphB2 receptor intercellular domain is not required for spine maturation of developing CA1 pyramidal neurons. These data suggest that EphB is involved in the development of dendrites and spines as presynaptic, CA3-expressed ligands that bind CA1-expressed eB3 and stimulate reverse signaling.

Our study does not rule out the possibility of a cis interaction of eB3 and EphB2 in CA1 neurons. EphB2 has been shown to interact with the NMDA receptor through its extracellular domain in adult animals, which does not require the intracellular segment^9,11,47. However, for subsequent steps in postsynaptic specialization, the trans-synaptic binding with ephrin-Bs and the kinase activity of EphB are important^9,48. This scenario is in contrast to what we show here: ephrin-B molecules are specifically expressed in the postsynaptic compartment of CA1 neurons and the intracellular segment of EphB2 is not required for spine maturation. Therefore, the deficits in CA1 dendrites that are observed in EphB mutants are unlikely to be caused by interaction with EphB in a cis manner but are due to a ligand-like role of EphB receptors in presynaptic CA3 axons and nerve terminals.

Taken together, on axon–dendrite contact, presynaptic EphB molecules bind postsynaptic eB3 which interacts with GRB4, PICK1 and syntenin to transduce distinct reverse signals into developing dendrites to control both long-scale branch pruning, and short-scale spine maturation and synapse formation (Supplementary Fig. 16). eB3 was also shown to regulate spine formation using in vitro cultured cortical neurons through the MAPK pathway¹⁹ and has been implicated in promoting shaft synapse formation through GRIP¹⁸. Therefore, further clarification of the mechanisms by which ephrin-B, GRB4, PICK1, syntenin and other signaling pathways are coordinated upon axon–dendrite contact to regulate cytoskeletal elements should lead to an increased understanding of the complex dynamics of dendrite morphogenesis, spine maturation and synaptic plasticity in the developing and adult brain.

Methods

Mice and sample preparation.

Efnb3⁻, Ephb3^3F, Ephb3^5F, Efnb3^lacZ, Efnb2^lacZ, Efnb2^6FΔV, EphB1^−/− and EphB2^−/− knockout and knock-in mutant mice and genotyping methods have been described^{6,14,27,28,29,49}. Mice were crossed with the Thy1-GFP M transgenic mouse line²⁶. Consecutive backcrosses to the CD1 strain were performed to move the mutations to CD1 background. Mice were anesthetized (ketamine, 450 mg per kg), perfused transcardially with 0.1 M PBS followed by 4% paraformaldehyde (wt/vol) in phosphate buffer. The brains were then removed, post-fixed and sectioned at 50 μM using a vibratome. All experiments involving mice were carried out in accordance with the US National Institutes of Health Guide for the Care and Use of Animals under an Institutional Animal Care and Use Committee approved protocol and at an Association for Assessment and Accreditation of Laboratory Animal Care approved Facility at the University of Texas Southwestern Medical Center.

X-gal staining, immunofluorescence and Golgi stain.

To detect the β-gal expression by X-gal stain, mouse brain sections were processed as described⁶. For immunofluorescence, vibratome sections were incubated with a blue fluorescence dye NeuroTrace 640/660 (Molecular Probes) for visualization. For Golgi stain, freshly dissected P12, P20 mouse brains were incubated in Golgi solution containing 1.25% potassium dichromate, 1.25% HgCl and 1% KCl (wt/vol) in distilled water for 12 d. After incubation, mouse brains were embedded and sectioned at 100 μM using a vibratome and mounted on 3% gelatin-coated slides. For staining, the slides were washed with 20% ammonium hydroxide (vol/vol) and fixed in Kodak fixer solution.

Generation of Efnb3 mutant mice.

The Efnb3 targeting vectors used incorporated loxP site-specific recombination sequences and were designed to generate two-step mutations as described previously¹⁴ from a single gene targeting event in mouse embryonic stem cells (Fig. 3a and Supplementary Fig. 8). The initial insertions, which are called Efnb3^neo, eB3^neoΔV or eB3^neo3FΔV, insert into the Efnb3 fourth intron (171 bp upstream of the splice acceptor for the fifth exon) a phosphoglycerate kinase (PGK) promoter–driven neomycin resistance (neo) cassette that was flanked with loxP sequences. As designed, this replaces wild-type Efnb3 exon 5 encoding amino acids 205–340 including the cytoplasmic domain with wild type, or engineered mutations ΔV or 3FΔV in which tyrosine codons at positions 311, 318, 323 (for 3FΔV) are changed to phenylalanines and/or the valine codon at position 340 is deleted (for 3FΔV or ΔV). Chimeric mice were made from two lines of each mutation and germline transmission was obtained, leading to the recovery of the initial targeted Efnb3^neo, Efnb3^neoΔV or Efnb3^neo3FΔV lines. Because this strategy was based on our initial gene targeting of Efnb3 that showed that insertion of a neo cassette into the fourth intron disrupts the gene and leads to the null phenotype of a hopping locomotion²⁷, the initial, Efnb3^neoΔV or Efnb3^neo3FΔV lines—when made homozygous—also resulted in a loss-of-function hopping phenotype as expected. These mice were then crossed to a ubiquitous CAGG-creER^T2M driver followed by tamoxifen administration (for Efnb3^neo), or a germline Cre-expressing transgenic mouse (for Efnb3^neoΔV or Efnb3^neo3FΔV), to remove the loxP-flanked neo cassette and convert the locus to restore eB3 expression conditionally or to generate the intended Efnb3 mutant alleles that express the intended eB3 protein with select point mutations in the cytoplasmic domain. Animals were genotyped by Southern blot with the probes, as indicated, and PCR with forward primer TCCCATCTTCAGGTCCCCGAG and reverse primer TGGAAATCCAGGTGTCCGGCC for wild type (380 bp), and forward primer GGTGCTTCTGCGAGTGG and reverse primer GCATACATTATACGAAGTTATATTAAGGG for mutant (530 bp) (Supplementary Fig. 8b). The ΔV and 3FΔV mutations in the Efnb3 exon 5 were further confirmed by sequencing PCR amplified products from Efnb3^neoΔV and Efnb3^neo3FΔV mice. The expression level of Efnb3 and Efnb2 mRNA in hippocampus from Efnb3 mutants were further examined by using quantitative real-time PCR with forward primer GTGCCAGACAAG AGCCATGAA and reverse primer GGTGCTAGAACCTGGATTTGG for Efnb2, and forward primer CGTAGTCCCCTTCTGCCCTCAC and reverse primer AGATGTTCGGAGGGCTCTGG for Efnb3.

Biochemistry analysis.

Brain tissues from wild type and Efnb3 mutants were dissected and homogenized in ice-cold lysis buffer (50 mM Tris-HCl pH 8.0, 300 mM NaCl, 20 mM HEPES pH 7.2, 1% Triton X-100, 10% glycerol, 50 mM NaF, 1% bovine serum albumin (wt/vol), 1 mM PMSF and protease inhibitor mixture; Roche Molecular Biochemicals). Following immunoprecipitation with goat anti-PICK1 (Novus) overnight and incubation with protein G beads for 2 h at 4 °C, bound proteins were separated by SDS–PAGE, transferred to nitrocellulose membranes, and then immunoblotted with rabbit-anti-eB3 (Invitrogen). For biotinalytion of cell surface proteins, primary neurons dissected from P0 hippocampus were cultured for 12 d and surface proteins were isolated by using biotinylation kits (Pierce).

For immunopreciptation in Cos-1 cells, Flag-eB3-PICK1, Flag-eB3-Syn and their mutant form were transfected to assay protein–protein interaction with HA-PICK1 or HA-syndecan-2. The procedure was performed as described previously¹⁴.

DNA constructs.

To express various eB3 proteins by transfection and infection, Flag tagged wild-type and mutant Efnb3 coding sequences were generated from the full-length cDNA expression vector pEXPRmELF3 and ligated in the EcoR1 and Nhe1sites of pFUGW vector. Various syntenin and PICK1 coding regions were produced by PCR and ligated into pFUGW vector. Constructs were transfected directly into neurons or packaged into lentivirus to infect cells.

Primary neuron culture, transfection and lentiviral infection.

Hippocampal neurons from dissected P0 hippocampi were grown as described previously¹⁴. Lentiviruses were produced by transfecting HEK293 cells with pFUGW vectors and two helper plasmids (pVSVg and pCMVΔ8.9). Viruses were harvested 48 h after transfection by collecting the medium from transfected cells. Neurons were infected with 0.2–0.5 ml conditioned HEK293 cell medium for each 24 wells of high-density neurons at 5 and 8 d in vitro (DIV) culture, and the medium was exchanged to normal growth medium 1 d later, then kept until 12 DIV–14 DIV electrophysiological analyses. For immunocytochemistry, the neurons were transfected by using polyethylenimine.

Electrophysiology.

Mutant mice and their wild-type littermates aged P12 were anesthetized by isoflurane inhalation and decapitated. Brains were quickly dissected in 5% CO₂ and 95% O₂ ice-cold artificial cerebrospinal fluid (119 mM NaCl, 2.5 mM KCl, 1 mM NaH₂PO₄, 26.2 mM NaHCO₃, 11 mM glucose, 2 mM CaCl₂ and 2 mM MgCl₂) after bubbling with 95% O₂ and 5% CO₂. Brains were vibratome-sectioned in the same solution at 400 μm and transferred to a chamber with bubbling with 95% O₂ and 5% CO₂ at 37 °C for 30 min and then maintained at 22–25 °C. Neurons were targeted for whole-cell patch-clamp recording with borosilicate glass electrodes having a resistance of 5–8 MΩ. The electrode internal solution was composed of 125 mM potassium gluconate, 10 mM HEPES, 2.6 mM KCl, 0.1 mM EGTA, 1.3 mM NaCl, 0.07 mM CaCl₂, 15 mM Sucrose, 17.3 mM Tris phosphocreatine, 0.36 mM Na-GTP and 5 mM Mg-ATP. CA1 pyramidal neurons were selected from the dorsomedial hippocampus. For mEPSC recording, tetrodotoxin (1 μM) and picrotoxin (100 μM) were included in the external solution. Cells were held at −70 mV. Miniature responses were acquired with a Multiclamp 700B at 10 kHz. Prior to mEPSC detection and analysis, current traces were low-pass filtered at 5 kHz. Events having amplitude of ×2 root mean square noise were detected using Mini Analysis (Synaptosoft).

Whole-cell patch-clamp recordings on cultured neurons were made at 21–23 °C from 12–14 DIV cultured neurons, with 4–6 MΩ patch pipettes filled with an internal solution containing 120 mM CsCl, 2 mM MgCl₂, 5 mM EGTA, 10 mM HEPES, 0.3 mM Na₃-GTP, 4 mM Na₂-ATP (pH 7.4). Cultures were continuously superfused with external solution containing of 140 mM NaCl, 10 mM HEPES, 5 mM KCl, 10 mM glucose, 2 mM CaCl₂, 2 mM MgCl₂, 1 μM tetrodotoxin and 100 μM picrotoxin (pH 7.4). Miniature responses were acquired with PCLAMP 10 software (Molecular Devices) at 10 kHz.

Quantification of dendrites, spine density and synaptic puncta.

Confocal images of Thy1-GFP expressed neurons from P9–P20 mice were obtained with sequential acquisition settings at the maximal resolution of the microscope (1024 × 1024 pixels). Each image was a Z-series of 7 images, each averaged two times. The resulting z-stack was flatted into a single image using maximum projection. A total of 20 cells for wild-type, Efnb3 mutants; four animals were used per genotype. Quantification of the number of primary dendrites (defined as dendrites longer than 21 μm emanating directly from the soma) and total dendrites (defined as the amount of all dendritic branches) was done on images acquired with ×25 objective, in which a circle was drawn around the cell body and the number of dendrites crossing each circle (primary dendrites) and the total branches were manually counted. Quantitative analysis for spines was performed by using NeuronStudio⁵⁰ and ImageJ in images acquired with ×100 objective. The number of spines with length of 0.5–2.0 μm and wide head diameter (>0.2 μm) were accounted and divided by the total length of GFP-labeled dendritic branches. For quantitative analysis for cumulative plot of spine length and head width, the protrusions with length 0.2–3.0 μm and maximum width 3 μm were counted. For assessment of synapses on spine or shaft in cultured neurons, only synapsin puncta with directly adjacent GFP labeled spine buttons or shafts were scored as colocalized synapses on spine. Acquisition of the images as well as morphometric quantification was performed under 'blinded' conditions.

Statistical analysis.

The results are presented as mean ± s.e.m. Statistical differences were determined by Student's t test for two-group comparisons or ANOVA followed by Tukey test for multiple comparisons among more than two groups.