|Home | About | Journals | Submit | Contact Us | Français|
The mechanisms that promote excitatory synapse formation and maturation have been extensively studied. However, the molecular events that limit excitatory synapse development so that synapses form at the right time and place and in the correct numbers are less well understood. We have identified a RhoA guanine nucleotide exchange factor, Ephexin5, which negatively regulates excitatory synapse development until EphrinB binding to the EphB receptor tyrosine kinase triggers Ephexin5 phosphorylation, ubiquitination, and degradation. The degradation of Ephexin5 promotes EphB-dependent excitatory synapse development and is mediated by Ube3A, a ubiquitin ligase that is mutated in the human cognitive disorder Angelman syndrome and duplicated in some forms of Autism Spectrum Disorders (ASDs). These findings suggest that aberrant EphB/Ephexin5 signaling during the development of synapses may contribute to the abnormal cognitive function that occurs in Angelman syndrome and, possibly, ASDs.
A crucial early step in the formation of excitatory synapses is the physical interaction between the developing presynaptic specialization and the postsynaptic dendrite (Jontes et al., 2000; Ziv and Smith, 1996). This step in excitatory synapse development is thought to be mediated by cell surface membrane proteins expressed by the developing axon and dendrite and appears to be independent of the release of the excitatory neurotransmitter glutamate (reviewed in Dalva et al., 2007). Several recent studies have revealed an important role for Ephrin cell surface-associated ligands and Eph receptor tyrosine kinases in this early cell-cell contact phase that is critical for excitatory synapse formation (Dalva et al., 2000; Ethell et al., 2001; Henkemeyer et al., 2003; Kayser et al., 2006; Kayser et al., 2008; Lai and Ip, 2009; Murai et al., 2003). Ephs can be divided into two classes, EphA and EphB, based on their ability to bind the ligands EphrinA and EphrinB, respectively (reviewed in Flanagan and Vanderhaeghen, 1998). EphBs are expressed postsynaptically on the surface of developing dendrites, while their cognate ligands, the EphrinBs, are expressed on both the developing axon and dendrite (Grunwald et al., 2004; Grunwald et al., 2001; Lim et al., 2008). When an EphrinB encounters an EphB on the developing dendrite, EphB becomes autophosphorylated, thus increasing its catalytic kinase activity (reviewed in Flanagan and Vanderhaeghen, 1998). This leads to a cascade of signaling events including the activation of guanine nucleotide exchange factors (GEFs) Tiam, Kalirin, and Intersectin, culminating in actin cytoskeleton remodeling that is critical for excitatory synapse development (reviewed in Klein, 2009). Consistent with a role for EphBs in excitatory synapse development, EphB1/EphB2/EphB3 triple knockout mice have fewer mature excitatory synapses in vivo in the cortex, and hippocampus (Henkemeyer et al., 2003; Kayser et al., 2006). In addition, the disruption of EphB function postsynaptically in dissociated hippocampal neurons leads to defects in spine morphogenesis and a decrease in excitatory synapse number (Ethell et al., 2001; Kayser et al., 2006). Conversely, activation of EphBs in hippocampal neurons leads to an increase in the number of dendritic spines and functional excitatory synapses (Henkemeyer et al., 2003; Penzes et al., 2003). These findings indicate that EphBs are positive regulators of excitatory synapse development.
While there has been considerable progress in characterizing the mechanisms by which EphBs promote excitatory synapse development, it is not known if there are EphB-associated factors that restrict the timing and extent of excitatory synapse development. We hypothesized that neurons might have evolved mechanisms which act as checkpoints to restrict EphB-mediated synapse formation, and that the release from such synapse formation checkpoints might be required if synapses are to form at the correct time and place and in appropriate numbers.
We considered the possibility that likely candidates to mediate the EphB-dependent restriction of excitatory synapse formation might be regulators of RhoA, a small G protein that functions to antagonize the effects of Rac (Tashiro et al., 2000). In previous studies we identified a RhoA GEF, Ephexin1 (E1), which interacts with EphA4 (Fu et al., 2007; Sahin et al., 2005; Shamah et al., 2001). E1 is phosphorylated by EphA4 and is required for the EphrinA-dependent retraction of axonal growth cones and dendritic spines (Fu et al., 2007; Sahin et al., 2005). While E1 does not appear to interact with EphB, E1 is a member of a family of five closely related GEFs. Of these GEFs, Ephexin5 (E5) (in addition to E1) is highly expressed in the nervous system. Therefore, we hypothesized that E5 might function to restrict the EphB-dependent development of excitatory synapses by activating RhoA.
In this study we report that EphB interacts with E5, that E5 suppresses excitatory synapse development by activating RhoA, and that this suppression is relieved by EphrinB activation of EphB during synapse development. Upon binding EphrinB, EphB catalyzes the tyrosine phosphorylation of E5 which triggers E5 degradation. We identify Ube3A as the ubiquitin ligase that mediates E5 degradation, thus allowing synapse formation to proceed. As UBE3A is mutated in Angelman syndrome and duplicated in some forms of Autism Spectrum Disorders (ASDs), these findings suggest a possible mechanism by which the mutation of Ube3A might lead to cognitive dysfunction (Jiang et al., 1998; Kishino et al., 1997). Specifically, we provide evidence that in the absence of Ube3A, the level of E5 is elevated and propose that this may lead to the enhanced suppression of EphB-mediated excitatory synapse formation, thereby contributing to Angelman syndrome and, possibly, ASDs.
To identify mechanisms that restrict the ability of EphBs to promote an increase in excitatory synapse number, we searched for RhoA guanine nucleotide exchange factors (GEFs) that specifically activate RhoA signaling, are expressed in the same population of neurons that express EphB, are expressed at the same time during development as EphB, and interact with EphB. Structure-function studies of GEFs identified amino acid residues in the activation domain of Rho family GEFs that specifically identify the GEFs as activators of RhoA rather than Rac or Cdc42. Applying this criterion, fourteen GEFs were identified that specifically activate RhoA (Rossman et al., 2005). Of these GEFs we found by in situ hybridization that E5 has a similar expression pattern to EphB in the hippocampus (Fig 1A). These findings raised the possibility that E5 might mediate the effect of EphB on developing synapses.
We asked if E5 interacts physically with EphB. We transfected HEK293T (293) cells with plasmids encoding Myc-tagged E5, E1, or a vector control together with Flag-tagged EphB2 or EphA4 and asked if these proteins co-immunoprecipitate. Extracts were prepared from the transfected 293 cells and EphA4 or EphB2 immunoprecipitated with Flag antibodies. The immunoprecipitates were subjected to SDS polyacrylamide gel electrophoresis (SDS-PAGE) and blotted with anti-Myc antibody (α-Myc). We found that E5 co-immunoprecipitates with EphB2 but not with EphA4 (Fig 1B). The relatively weak E5 interaction with EphA4 is consistent with published experiments (Ogita et al., 2003). By contrast, E1 is co-immunoprecipitated by EphA4 but not EphB2 (Shamah et al., 2001). These findings suggest that E5 interacts preferentially with EphB2.
To extend this analysis we investigated whether EphB2 interacts with E5 in neurons. Neurons from embryonic day 16 (E16) mouse brains were lysed in RIPA buffer and the lysates incubated with affinity purified anti-C-terminal E5 (α-C-E5) or control (IgG) antibodies. The immunoprecipitates were then resolved by SDS-PAGE and immunoblotted with affinity purified anti-N-terminal E5 (α-N-E5) or EphB2 (α-EphB2) antibodies (Fig 1C). This analysis revealed that endogenous, neuronal EphB2 is immunoprecipitated by α-C-E5 but not IgG. Moreover, using lysates from brains of wild type or E5 knockout mice (E5−/−, see Fig S1), we find that α-C-E5 immunoprecipitates EphB2 only from brain lysates when E5 is present (Fig 1D). Taken together, these findings suggest that EphB interacts with Ephexin5 in neurons.
As an independent means of assessing if EphB and E5 interact with one another, we used immunofluorescence microscopy to determine if these two proteins co-localize in neurons. Cultured mouse hippocampal neurons were transfected with a plasmid expressing green fluorescent protein (GFP). The GFP-expressing neurons were imaged and quantified for the co-localization of EphB2 and E5 puncta by staining with α-C-E5 and α-EphB2. This analysis revealed that EphB2 and E5 co-localize along dendrites (Fig 1E). We find that 40% of EphB staining overlaps with α-C-E5 staining early during the development of excitatory synapses. After eight days in vitro (DIV) the overlap of EphB with E5 within neuronal dendrites decreases to below the level that would be detected by random chance. This change suggests that EphB interacts with E5 early during development, possibly to inhibit EphB synapse formation.
To determine if E5 activates RhoA, we transfected 293 cells with a control plasmid or a plasmid that drives the expression of Myc-tagged mouse E5. We prepared extracts from the transfected cells and incubated the extracts with a GST-fusion protein that includes the Rhotekin-Binding Domain (GST-RBD), a protein domain that selectively interacts with active (GTP-bound) but not inactive (GDP-bound) RhoA. Following SDS-PAGE of the proteins in the extract that bind to GST-RBD, RhoA binding to GST-RBD was measured by immunoblotting with α-RhoA antibodies. We found that cells expressing E5 exhibited higher levels of activated RhoA compared to cells transfected with a control plasmid, indicating that E5 activates RhoA (Fig 2A).
When a similar series of experiments were performed using a GST-fusion Pak-Binding Domain (GST-PBD) which specifically interacts with active forms of two other Rho GTPases, Rac1 and Cdc42, we found that E5 does not induce the binding of GST-PBD to Rac1 or Cdc42. In contrast, E1-expressing cells displayed enhanced binding of Rac1 and Cdc42 to GST-PBD. We conclude that E5 activates RhoA but not Rac1 or Cdc42 (Fig S2A).
To determine whether E5 activation of RhoA requires the GEF activity of E5, we generated a mutant form of E5 in which its GEF activity is impaired. To identify the residues required for Ephexin5 guanine nucleotide exchange activity we compared its Dbl-homology (DH) domain to the DH domain of other RhoA-specific GEFs (Snyder et al., 2002). We identified within the α5 helix of E5’s DH domain three amino acids that are conserved in other GEFs that, like E5, activate RhoA but not Rac1 and Cdc42 (Fig S2B). To generate a form of E5 predicted to be inactive as a GEF, we mutated these three conserved amino acids (L562, Q566, and R567) to alanine (E5-LQR). Using the GST-RBD pull down assay we found that although E5-WT and E5-LQR are expressed at similar levels, the E5-LQR mutant is significantly impaired relative to WT in its ability to activate RhoA (Fig 2B). As a control, we mutated other conserved residues within the α5 DH region to alanine (Q547, S548, R555, and L556). When we tested this mutant we observed no defect in RhoA activation, suggesting that the E5-LQR mutation specifically disrupts the GEF activity of E5 and that the inability of the LQR mutant to activate RhoA is not a general consequence of disrupting the α5 region of Ephexin5 (Fig S2C). Taken together, these findings indicate that E5 requires an intact conserved GEF domain to promote RhoA activity in 293 cells, suggesting that E5 functions as a RhoA GEF.
We next asked if E5 expression affects RhoA activity in the brain. We lysed P3 whole brains from wild type or E5−/− mice and performed a GST-RBD pull down assay. This analysis revealed a significant decrease in RhoA activation in brain extracts from E5−/− mice compared to wild type mice, suggesting that E5 is required to maintain wild type levels of RhoA activity in the brain (Fig 2C).
Our findings indicate that E5 interacts with EphB, a key regulator of excitatory synapse development. Thus, we asked whether E5 plays a role in the development of excitatory synapses. We generated two short hairpin RNA constructs that each knocks down E5 protein levels when expressed in 293 cells or cultured hippocampal neurons (Fig S3A–S3B). These shRNAs were introduced into cultured hippocampal neurons together with a plasmid that drives expression of green fluorescent protein (GFP) to allow detection of the transfected cells. We found by staining with α-N-E5 antibodies that the E5 shRNAs (E5-shRNA), but not scrambled hairpin control shRNAs (ctrl-shRNA), efficiently knocked down E5 expression in the transfected neurons (Fig S3C).
By staining with antibodies that recognize pre- and post- synaptic proteins or by visualizing dendritic spines in GFP transfected neurons we observed a significant increase in the number of excitatory synapses and dendritic spines that are present on the E5-shRNA-expressing neurons compared to neurons expressing ctrl-shRNAs (Fig 3A and 3B). By contrast, we failed to detect a significant change in dendritic spine length or width under these conditions (Fig S3D). These findings suggest that E5 functions to restrict spine/excitatory synapse number but has no significant effect on spine morphology. Consistent with these conclusions, we found that overexpression of E5 in hippocampal neurons leads to a decrease in the number of excitatory synapses that are present on the E5-overexpressing neurons (Fig 3C). This ability of E5 to negatively regulate excitatory synapse number requires its RhoA GEF activity, as overexpression of E5-LQR had no effect on synapse number (Fig 3D).
To assess the effect of reducing E5 levels on the functional properties of excitatory synapses, we recorded miniature excitatory postsynaptic currents (mEPSCs) from cultured hippocampal neurons transfected with E5-shRNA or ctrl-shRNA. We observed an increase in the frequency and amplitude of mEPSCs on neurons expressing E5-shRNA compared to ctrl-shRNA (Fig 3E). This suggests that E5 acts postsynaptically to restrict excitatory synapse function. The increase in mEPSC frequency could be due to an increase in presynaptic vesicle release onto the transfected neuron or an increase in the number of excitatory synapses that are present on the transfected neuron. We favor the latter possibility since our transfection protocol selectively reduces E5 levels postsynaptically and also because the increase in synapse number is most consistent with the increase in co-staining of pre- and post-synaptic markers that we observe when the level of E5 is reduced. The possibility that E5 functions postsynaptically is further supported by immunofluorescence staining experiments demonstrating that E5 is enriched in dendrites relative to axons (Fig S1F).
As an independent means of assessing the importance of E5 in the control of excitatory synapse number, we cultured hippocampal neurons from E5−/− mice or their wild type littermates for 10 days in vitro and then, following transfection of a GFP-expressing plasmid into these neurons, quantified the number of excitatory synapses present on the transfected neuron at DIV14. We observed a three-fold increase in the number of synapses that are present on E5−/− neurons compared to E5+/− neurons (Fig 4A). Taken together with the E5-shRNA knockdown and E5 overexpression analyses, these findings suggest that E5 acts postsynaptically to reduce excitatory synapse number.
We next asked if E5 regulates synapse number in the context of an intact developing neuronal circuit using conditional E5 (E5fl/fl) animals (see Fig S1). Upon introduction of Cre recombinase into E5fl/fl cells, exons 4–8 of the E5 gene are excised resulting in a cell that no longer produces E5 protein (data not shown). Organotypic slices were prepared from the hippocampus of the E5fl/fl mice or their wild type littermates. Using the biolistic transfection method, a plasmid expressing Cre recombinase was introduced into a low percentage of neurons in the slices. We found that introduction of a Cre-expressing plasmid into E5fl/fl neurons in the hippocampal slice led to a significant increase in the density of dendritic spines present on the Cre-expressing neurons (Fig 4B). By contrast, expression of Cre in neurons of a wild type hippocampal slice has no effect on dendritic spine density. The length and width of dendritic spines analyzed in these experiments showed no significant difference between wild type and E5−/− neurons (Fig S4). Thus, elimination of E5 expression in neurons in the context of an intact neuronal circuit leads to an increase in the number of dendritic spines.
To assess the role of E5 in hippocampal circuit development in vivo, we performed acute slice physiology experiments in the CA1 region of the hippocampus from wild type or E5−/− mice. We find that relative to wild type neurons, in E5−/− CA1 pyramidal neurons there are more frequent excitatory events that have larger amplitude (Fig 4C). A possible explanation for these findings is that when E5 function is disrupted during in vivo development more excitatory synapses form resulting in more excitatory post-synaptic events. To test this possibility, we used array tomography to quantify the number of excitatory synapses that form in the CA1 stratum radiatum of wild type and E5−/− mice. We observed a ~2-fold increase in the number of excitatory synapses within the CA1 region of the E5−/− hippocampus compared to wild type mice (Fig 4D). Specifically, the number of juxtaposed synapsin and PSD-95 puncta was quantified and considered a measurement of the number of excitatory synapses that form within the CA1 region of the hippocampus in vivo. This analysis revealed a significant increase in the number of PSD-95 puncta but no change in the number of synapsin puncta density (Fig 4D). This suggests that the increase in excitatory synapse number in the stratum radiatum of E5−/− mice is likely due to the absence of E5 post-synaptically and that when E5 is present within dendrites it functions to negatively regulate synapse number in vivo. On the basis of these results, we conclude that a key function of E5 is to restrict excitatory synapse number during the development of neuronal circuits.
We next considered the possibility that the ability of E5 to restrict excitatory synapse number might be controlled by EphB2 signaling. To test this idea, we asked whether reducing EphB2 signaling eliminates the increase in excitatory synapse number detected when E5 levels are knocked down by expression of E5-shRNA. To block EphB2 activation, we introduced into neurons a kinase dead version of EphB2 (EphB2-KD) which has been previously shown to block EphB2 signaling (Dalva et al., 2000). As described above, expression of E5-shRNA in neurons leads to a significant increase in the number of synapses that are present on the E5-shRNA-expressing neuron. However, this increase was reversed if the E5-shRNA was co-transfected with a plasmid that drives expression of EphB2-KD, but was not affected by co-transfection of a control plasmid (Fig 4E). These findings suggest that the increase in excitatory synapse number that occurs when E5 levels are reduced requires EphB signaling. Consistent with this conclusion, we find that if we overexpress wild type EphB2 in neurons more synapses are present on the EphB-expressing neuron. However, this effect is reduced if E5 is overexpressed in neurons together with EphB (Fig 4F). It is possible that the ability of overexpressed E5 to suppress the synapse-promoting effect of EphB2 reflects independent actions of these two signaling molecules. However, given that EphB2 and E5 interact with one another in neurons, the most likely interpretation of these results is that E5 functions directly to restrict the synapse-promoting effects of EphB2. If this were the case, we would predict that for EphB2 to positively regulate excitatory synapse development it would be necessary to inactivate and/or degrade E5.
We considered the possibility that since EphB2 is a tyrosine kinase it might inhibit the GEF activity or expression of the E5 protein by catalyzing the tyrosine phosphorylation of E5. In support of this possibility, stimulation of dissociated mouse hippocampal neurons with EB1 for 15 minutes led to an increase in the level of E5 tyrosine phosphorylation as detected by probing immunoprecipitated E5 with the pan-anti-phosphotyrosine antibody, 4G10 (Fig 5A).
We have previously shown that EphrinA1 stimulation of cultured neurons leads to the tyrosine phosphorylation of E1 at tyrosine 87 (Sahin et al., 2005). On the basis of this finding we hypothesized that exposure of neurons to EphrinB1 (EB1) might promote the phosphorylation of the analogous tyrosine residue (Y361) on E5 (Fig 5B) and that phosphorylation at this site might lead to E5 inactivation. To address this possibility, we overexpressed EphB2 in 293 cells together with wild type E5 or a mutant form of E5 in which Y361 is converted to a phenylalanine (E5-Y361F). Lysates were prepared from the transfected cells and after SDS-PAGE were immunoblotted with 4G10 (Fig 5C). We found that in the presence of EphB2, E5-WT, but not E5-Y361F, becomes tyrosine phosphorylated. These findings suggest that EphB2 catalyzes the tyrosine phosphorylation of E5 primarily at Y361.
To show definitively that E5 Y361 is tyrosine phosphorylated, we generated E5 phospho-Y361 antibodies (α-pY361). To demonstrate that these antibodies specifically recognizes the Y361-phosphorylated form of E5, we immunoblotted cell lysates prepared from 293 cells that express EphB2 and either E5-WT or E5-Y361F with α-pY361. This analysis demonstrated that the α-pY361 bind to wild type E5 but not E5-Y361F (Fig 5C). Furthermore, using α-pY361 we found that when wild type EphB2, but not a kinase dead or cytoplasmic truncated version of EphB2, is expressed in 293 cells together with E5, E5 becomes phosphorylated at Y361 (Fig S5A). In contrast, when EphA4 or EphA2 were expressed in 293 cells we detected little to no phosphorylation of E5 at Y361 (Fig S5B). These findings suggest that EphB2, but not EphAs, promote E5 Y361 phosphorylation (pY361).
We also found by immunoblotting with the α-pY361 that E5 is phosphorylated at Y361 in the hippocampus of wild type but not E5−/− mice (Fig S5C), and that EB1 stimulation of cultured hippocampal neurons leads to E5 Y361 phosphorylation (Fig 5D). By immunofluorescence microscopy we detect punctate α-pY361 staining along the dendrites of EB1-treated wild type neurons, but less staining in untreated neurons (Fig 5E). This result suggests that E5 becomes newly phosphorylated at Y361 upon exposure of hippocampal neurons to EB1.
We asked if EB1 stimulation of E5 Y361 phosphorylation leads to a change in E5 activity or expression. To investigate this possibility we asked if EphB suppresses E5-dependent RhoA activation in a phosphorylation-dependent manner. We transfected 293 cells with E5 in the presence or absence of EphB2 and measured RhoA activity using the RBD pull down assay (Fig 5F). We found that E5-dependent RhoA activation was reduced in 293 cells expressing EphB2 and E5 compared to cells expressing E5 alone. These findings are consistent with the possibility that EphB2-mediated tyrosine phosphorylation of E5 either leads to a suppression of E5’s ability to activate RhoA, or alternatively might trigger a decrease in E5 protein expression resulting in a decrease in RhoA activation. We found this latter possibility to be the case (Fig 5F, E5 loading control). Furthermore, when we compared lysates from the brains of wild type or EphB2−/− mice, we observed that E5 phosphorylation at Y361 is decreased while the levels of E5 expression are increased in the lysates from EphB2−/− mice (Fig 5G). These data suggest that EphB2 functions to phosphorylate and degrade E5.
Consistent with the idea that E5 expression is destabilized in the presence of EphB, we observed that in the dendrites of cultured hippocampal neurons overexpressing EphB2, endogenous E5 expression levels are reduced compared to control transfected neurons or neurons transfected with a kinase dead version of EphB2 (Fig S6A and B). When neurons were exposed to EB1 compared to EA1 for 60 minutes, we found by immunoblotting of neuronal extracts, or immunofluorescence staining with α-N-E5, that exposure to EB1 leads to a decrease in E5 expression (Fig 6A). The lack of complete loss of E5 expression by western blot may be due to the fact that EB1 stimulation leads to dendritic and not somatic loss of E5 expression. Moreover, immunofluorescence staining revealed a loss of E5 puncta specifically within the dendrites of EB1-stimulated neurons, consistent with the possibility that EB1/EphB-mediated degradation of E5 relieves an inhibitory constraint that suppresses excitatory synapse formation on dendrites. In support of this idea, we find by immunoblotting of extracts from mouse hippocampi that endogenous E5 protein levels are highest at postnatal day 3 prior to the time of maximal synapse formation and then decrease as synapse formation peaks in the postnatal period (Fig S6C). Northern blotting revealed that this decrease in E5 protein is not due to a change in the level of E5 mRNA expression (Fig S6C). Given that E5 protein levels decrease dramatically during the time period P7-P21 when synapse formation is maximal, these findings suggest that E5 may need to be degraded prior to synapse formation.
We asked whether EphB-mediated degradation of E5 could be reconstituted in heterologous cells. When EphB and Myc-tagged E5 were co-expressed in 293 cells we observed a significant decrease in E5 protein expression in the presence of EphB2. The presence of EphB2 had no effect on the level of expression of a related GEF, E1 (Fig 6B). We asked whether EphB-mediated degradation of E5 depends upon Y361 phosphorylation. We found that in 293 cells overexpressing Myc-tagged E5, the co-expression of EphB2, but not EphB2-KD, resulted in a significant decrease in E5 levels (Fig 6C). This suggests that EphB tyrosine kinase activity is required for E5 degradation. The EphB-mediated reduction in E5 levels is dependent on Y361 phosphorylation, as EphB2 expression had no effect on the level of E5 Y361F expression (Fig 6D). This suggests that the phosphorylation of E5 at Y361 triggers E5 degradation.
We considered the possibility that the Y361 phosphorylation-dependent decrease in E5 protein levels might be due to EphB-dependent stimulation of E5 proteasomal degradation. Consistent with this possibility we found that addition of the proteasome inhibitor lactacystin to 293 cells leads to a reversal of the EphB-dependent decrease in E5 protein levels, as measured by an increase in total ubiquitinated E5 (Fig S6D). In addition, in neuronal cultures the EB1 induced decrease in E5 protein expression is blocked if the proteasome inhibitor lactacystin is added prior to EB1 addition (Fig 6E). Notably, in the presence of lactacystin, E5 is ubiquitinated, further supporting the idea that E5 is degraded by the proteasome.
To test whether E5 is ubiquitinated in the brain, we incubated wild type or E5−/− brain lysates with α-C-E5 and after immunoprecipitation and SDS-PAGE, probed with α-ubiquitin antibodies. This analysis detected the presence of ubiquitinated species in α-C-E5 immunoprecipitates prepared from wild type but not E5−/− brain lysates (Fig 6F). These findings indicate that E5 is ubiquitinated in the brain.
During proteasome-dependent degradation of proteins, specificity is conferred by E3 ligases or E2 conjugating enzymes that recognize the substrate to be degraded. The E3 ligase binds to the substrate and catalyzes the addition of polyubiquitin side chains to the substrate thereby promoting degradation via the proteasome (Hershko and Ciechanover, 1998). We considered several E3 ligases that have recently been implicated in synapse development as candidates that catalyze E5 degradation. One of these E3 ligases, Cbl-b, has previously been implicated in the degradation of EphAs and EphBs (Fasen et al., 2008; Sharfe et al., 2003). A second E3 ligase, Ube3A, has been shown to regulate synapse number. To determine if Ube3A and/or Cbl-b catalyze E5 degradation we first asked if either of these E3 ligases interacts with and degrades E5 in 293 cells. When these E3 ligases were epitope-tagged and expressed in 293 cells together with E5 we found that E5 co-immunoprecipitates with Ube3A but not with Cbl-b (Fig 7A). The co-immunoprecipitation of Ube3A with E5 was specific in that Ube3A was not co-immunoprecipitated with two other neuronal proteins, E1 or the transcription factor MEF2. In a previous study we have shown that Ube3A binds to substrates via a Ube3A binding domain (hereafter referred to as UBD (Greer et al., 2010). Using protein sequence alignment programs, ClustalW and ModBase, we identified a UBD in E5, providing further support for the idea that E5 might be a substrate of Ube3A (Fig S7A). Consistent with this hypothesis, we found that the level of E5 expression is reduced in 293 cells co-transfected with Ube3A compared to cells co-transfected with Cbl-b (Fig S7B).
We asked if EB1/EphB-mediated E5 degradation in neurons is catalyzed by Ube3A. To inhibit Ube3A activity we introduced into neurons a dominant interfering form of Ube3A (dnUbe3A) that contains a mutation in the ubiquitin ligase domain rendering Ube3A inactive. We have previously shown that even though dnUbe3A is catalytically inactive it still binds to E2 ligases and to its substrates and functions in a dominant negative manner to block the ability of wild type Ube3A to ubiquitinate its substrates (Greer et al., 2010). We found that when introduced into 293 cells dnUbe3A binds to E5 (Fig 7A). We also found by immunofluorescence microscopy that when overexpressed in neurons, dnUbe3Ablocks EB1/EphB stimulation of E5 degradation (Fig 7B). EB1/EphB stimulation of E5 degradation was also attenuated when Ube3A expression was knocked down by a shRNA that specifically targets the Ube3A mRNA (Fig 7B (Greer et al., 2010)). Notably, the presence of the dnUbe3A did not affect E5 expression in neurons in the absence of EphrinB stimulation, suggesting that EphrinB stimulation of E5 Y361 phosphorylation may be required for Ube3A-mediated degradation of E5 (Fig S7C).
To determine if Ube3A-dependent degradation of E5 might be relevant to the etiology of Angelman syndrome we asked if the absence of Ube3A in a mouse model of Angelman syndrome affects the level of E5 expression in the brain. We compared the level of E5 protein expression in the brains of wild type mice to that expressed in the brains of mice in which the maternally inherited Ube3A was disrupted (Ube3Am−/p+). Because the paternally inherited copy of Ube3A is silenced in the brain due to imprinting, the level of Ube3A expression in Ube3Am−/p+ neurons is very low. We found that the level of E5 expression in the brains of Ube3Am−/p+ mice was significantly higher than that detected in the brains of wild type mice (Fig 7C). Moreover, the level of ubiquitinated E5 in brains of Ube3Am−/p+ mice was significantly reduced compared to the brains of litter mate controls (Fig 7D). In addition we found that when neurons from wild type and Ube3Am−/p+ brains were cultured and then treated with EB1 the level of E5 protein was reduced upon EB1 treatment in wild type but not in Ube3Am−/p+ neurons (Fig 7E). Taken together, these findings suggest that in response to EB treatment E5 is tyrosine phosphorylated by an EphB-dependent mechanism, and that this leads to E5 degradation by a Ube3A-dependent mechanism. If E5 degradation is disrupted due to a loss of Ube3A as occurs in Angelman syndrome the result is an increase in E5 expression and a disruption of the proper control of excitatory synapse number during brain development.
Previous studies have revealed a role for EphrinB/EphB signaling in the development of excitatory synapses (Klein, 2009). However, the regulatory constraints that temper EphB-dependent synapse development so that excitatory synapses form at the right time and place, and in the correct number were not known. In this study we identify a RhoA GEF, E5, which functions to restrict EphB-dependent excitatory synapse development. E5 interacts with EphB prior to EphrinB binding, and by activating RhoA serves to inhibit synapse development. The binding of EphrinB to EphB as synapses form triggers the phosphorylation and degradation of E5 by a Ube3A-dependent mechanism. The reduction in E5 expression may allow EphB to promote excitatory synapse development by activating Rac and other proteins at the synapse.
The findings that E5 functions to restrict excitatory synapse number suggests that, even though EphBs promote excitatory synapse development, there are constraints on the activity of EphB so that synapse number is effectively controlled. There are several steps in the process of synapse development where E5 may function to restrict synapse number. One possibility is that E5 functions early in development as a barrier to excitatory synapse formation by activating RhoA and restricting the motility or growth of dendritic filopodia that are the sites of contact by the presynaptic neuron. For example, by inhibiting dendritic filopodia formation or motility, E5 may decrease the number of contacts the filopodia make with the presynaptic neuron, thus resulting in the formation of fewer synapses. An alternative possibility is that E5 functions to restrict synapse number later in development perhaps to counterbalance the positive effects of EphB on Rac that promote dendritic spine development. An additional possibility is that E5 functions after excitatory synapse development as a regulator of synapse elimination.
Our analyses of E5 function are most consistent with the possibility that E5 functions early in the process of synapse development. First, we find that E5 is expressed, active, and bound to EphB prior to synapse formation. Second, the interaction of EphrinB with EphB, a process that is thought to be an early step in excitatory synapse development, triggers the degradation of E5. Third, our preliminary time-lapse imaging studies suggest that E5 is localized to newly formed filopodia prior to synapse development where it appears to restrict filopodia motility and growth (Margolis et al. unpublished). Thus, E5 might function as an initial barrier to synapse formation until it is degraded upon EphrinB binding to EphB.
It is possible that through its interaction with EphB, E5 marks the sites where synapses will form, and that the degradation of E5 is a critical early step in excitatory synapse development. While the mechanisms by which E5 is degraded are not fully understood, our studies suggest that the phosphorylation of the N-terminus of E5 at Y361 triggers the Ube3A-mediated proteasomal degradation of E5. One possibility is that prior to pY361 the N- and C-terminal portions of E5 interact, thereby protecting E5 from degradation. The phosphorylation of E5 at Y361 may relieve this inhibitory constraint allowing for E5 ubiquitination and degradation. A similar mechanism has been shown to regulate the activation of the Rac GEF Vav, (Aghazadeh et al., 2000)). During EphrinA/EphA signaling it has been proposed that Vav-mediated endocytosis of the EphrinA/EphA complex may allow the conversion of the initial adhesive interaction between EphrinA and EphA-expressing cells into a repulsive interaction that results in growth cone collapse and axon repulsion. It is possible that E5 has a related function during EphB signaling at synapses. Typically the EB/EphB interaction is thought to be repulsive. This has been documented in studies of EphB’s role in the process of axon guidance (Egea and Klein, 2007; Flanagan and Vanderhaeghen, 1998). However, during synapse development the EphrinB/EphB interaction is thought to result in synapse formation, a process that requires an interaction between the developing pre- and post-synaptic specialization. One possibility is that when EphrinB and EphB mediate the interaction between the incoming axon and the developing dendrite, the interaction is facilitated by the degradation of E5 by Ube3A. Since E5 is a RhoA GEF, its presence might initially lead to repulsion between the incoming axon and the dendrite. However, the EphB-dependent degradation of E5 might convert this initial repulsive interaction into an attractive one.
The finding that Ube3A is the ubiquitin ligase that controls EphB-mediated E5 degradation is of interest given the role of Ube3A in human cognitive disorders such as Angelman syndrome and autism. The absence of Ube3A function in Angelman syndrome would be predicted to result in an increase in E5 protein expression, and thus a decrease in EphB-dependent synapse formation. Consistent with this possibility, we find in a mouse model for Angelman syndrome that the level of E5 protein expression is elevated and that in response to EphrinB treatment E5 is not degraded. Likewise, several studies have indicated that synapse development and function is disrupted in these mice (Jiang et al., 1998; Yashiro et al., 2009).
The recent finding that the Ube3A gene lies within a region of chromosome 15 that is sometimes duplicated in autism raises the possibility that altered levels of Ephexin5 and the resulting defects in excitatory synapse restriction might also be a mechanism relevant to the etiology of autism (Glessner et al., 2009). If this is the case, a possible therapy for treating autism might be to reduce the level of Ube3A activity, and thus increase the level of Ephexin5 expression. It is important to consider that in addition to Ephexin5, Ube3A regulates the abundance of other synaptic proteins. Nevertheless, the ultimate effect of the aberrant expression of Ephexin5 and other Ube3A substrates on synapse development and function will require further study. It seems likely that such studies will provide further understanding of the development of human cognitive function and new insights into how this process goes awry in disorders such as Angelman syndrome and autism.
Details of DNA constructs can be found in supplementary section.
An E5 targeting vector was electroporated into 129 J1 ES cells, and positive clones were identified by Southern hybridization with two separate probes (See supplementary section).
Details of antibodies can be found in supplementary section.
Ube3a knockout mice were previously described (Greer et al., 2010). EphB2 knockout mice were previously described (Kayser et al., 2008). 293T cells were cultured in DMEM and transfected using the calcium phosphate method. Organotypic slice cultures were prepared from P6 mouse brains and biolistically transfected. Acute slices were prepared from P12-14 mice. Dissociated neurons were cultured in Neurobasal Medium supplemented with B27 and transfected using the Lipofectamine method. For details on cell culture, transfections and Ephrin stimulations see supplementary section.
Whole rat or mouse brains or cultured cells were collected and homogenized in RIPA buffer. For immunoprecipitations, lysed cells were centrifuged and supernatants were incubated with appropriate antibody for 2 hours at 4°C, followed by addition of Protein-A or Protein-G beads (Santa Cruz Biotechnology) for 1 hour, and washed three times with ice-cold RIPA buffer. For the α-PY361 detection experiment in 293T cells, samples were boiled in SDS buffer to disrupt the E5/EphB2 interaction and diluted 1:5 in 1.25X RIPA buffer prior to immunoprecipitation of E5-Myc. RBD and PBD pulldown assays were conducted according to the manufacture’s suggestions (Upstate Cell Signaling Solutions). For details see supplementary section.
To generate probes for in situ hybridization, mouse E5 and EphB2 cDNA were subcloned into pBluescript II SK (+). Bluescript plasmids containing E5 or EphB2 cDNA were linearized using the restriction enzyme BssHII. Sense and antisense probes were generated using DIG RNA labeling mix (Roche) according to manufacturer’s instructions. Full-length DIG-labeled probes were subjected to alkaline hydrolysis as described in supplementary section.
Neurons were paraformaldehyde fixed in PBS. For measuring synapse density, fixed neurons were incubated with α-PSD-95 and α-Synapsin antibodies followed by α-Cy3 and α-Cy5 antibodies to visualize the primary antibodies. For protein co-localization experiments fixed neurons were similarly treated using α-EphB2 antibodies and α-N-E5 antibodies or α-pY361-E5. For over-expression studies fixed neurons were incubated using α-Myc or α-Flag antibodies to visualize overexpressed E5-Myc or EphB2-Flag protein in the context of the GFP-labeled neurons. For details see supplementary section.
Images were acquired on a Zeiss LSM5 Pascal confocal microscope and spine and synapse analysis was performed as previously described (see supplementary section).
Dissociated hippocampal neurons from Ube3A knockout and wild-type mice were prepared as previously described (Greer et al., 2010).
Electrophysiology was performed using standard methods (see supplementary section).
We thank M. Thompson, Y. Zhou, and H. Ye for assistance in generating mice; E. Griffith, J. Zieg, S. Cohen, I. Spiegel, M. Andzelm, and the Greenberg lab for critical discussions. This work was supported by National Institute of Neurological Disorders and Stroke grant RO1 5R01NS045500 (M.E.G); NRSA Training grant 5T32AG00222-15 (S.S.M.); Edward R. and Anne G. Lefler postdoctoral fellowship (S.S.M.).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.