Signaling across the synapse: a role for Wnt and Dishevelled in presynaptic assembly and neurotransmitter release

During the formation of synaptic connections, axons remodel and begin to assemble the machinery required for neurotransmitter release upon arrival to their synaptic targets. An intrinsic genetic program is likely to regulate the synthesis of synaptic components and their targeting to pre- and postsynaptic sites. It is becoming clear that the cross communication between the pre- and postsynaptic terminals is essential for the coordinated assembly at both sides of the synapse. Although great emphasis has been given to the role of membrane-bound proteins such as neuroligin–neurexin (Scheiffele et al., 2000) and cadherins in this process, there is increasing evidence that secreted molecules such as Wnt, fibroblast growth factor (FGF), and TGFβ also play a crucial role in synaptic assembly and growth (Hall et al., 2000; Withers et al., 2000; Packard et al., 2002; McCabe et al., 2003; Umemori et al., 2004). However, little is known about the mechanisms by which these signals regulate synapse formation and to what extent changes in synaptic assembly translate into function.

Recent studies on neuroligin and neurexin have strengthened their role in synapse formation, as these molecules provide a local signal that stimulates synaptic assembly. Consistent with this notion, neuroligin and neurexin interact with components of the synaptic machinery (for review see Dean and Dresbach, 2006). In contrast, a distinct role for Wnts, FGFs, and thrombospondin has been proposed (Waites et al., 2005). In this model, these secreted signals could indirectly regulate synaptic formation by accelerating neuronal maturation through changes in the transcription and/or translation of synaptic components (Waites et al., 2005). Thus, synapses are formed through a sequence of events in which secreted factors stimulate neuronal maturation, thus, priming neurons for synapse formation followed by the focal action of membrane proteins that stimulate synaptic assembly. However, this model has not been fully tested and, hence, the precise role for secreted molecules in synapse formation remains to be established.

Wnt signaling plays a key role in diverse aspects of neuronal connectivity by regulating axon guidance, dendritic development, axon remodeling and synapse formation (Ciani and Salinas, 2005). In the cerebellum, Wnt7a is expressed in granule cells (GCs) at the time when mossy fiber (MF) axons, their presynaptic partners, reach the cerebellar cortex and make synaptic contact with GCs (Lucas and Salinas, 1997). Upon contact, MF terminals are extensively remodeled, resulting in the formation of complex and elaborate structures called glomerular rosettes (Hamori and Somogyi, 1983a). The extensive interdigitation of several GC dendrites into a single MF axon leads to a significant increase in the area of contact and is thought to contribute to some of the unusual functional properties observed at the MF-GC synapse (DiGregorio et al., 2002; Xu-Friedman and Regehr, 2003). These morphological changes are concurrent with the accumulation of presynaptic proteins and the formation of active zones. In Wnt7a-deficient mice, this extensive remodeling is affected, and the accumulation of the presynaptic protein synapsin I at glomerular rosettes is delayed (Hall et al., 2000). Conversely, gain of function of Wnt7a increases the remodeling and clustering of synapsin I in MF axons in vitro (Hall et al., 2000). Thus, Wnt7a functions as a retrograde GC signal that acts on MF axons to regulate presynaptic differentiation.

The signaling pathway activated by Wnt7a during synaptogenesis remains unknown. During early development, Wnt proteins signal through at least three pathways (Fig. 1 A). Signaling through Frizzled and, in some instances through Ryk, has been shown to require the cytoplasmic scaffold protein Dishevelled (Dvl; Cadigan and Liu, 2006). However, the discovery of new Wnt receptors (Mikels and Nusse, 2006) raises the question of whether Dvl is an absolute requirement for all Wnt-mediated functions. In the β-catenin pathway, Wnts activate Dvl to inhibit the serine/threonine kinase Gsk-3β, resulting in the elevation of β-catenin levels and its subsequent translocation to the nucleus to regulate transcription (Ciani and Salinas, 2005). In neurons, lithium, which is an inhibitor of GSK-3β, mimics the remodeling and presynaptic differentiation effects of Wnt7a suggesting that the canonical pathway might be involved (Hall et al., 2000). However, the poor specificity of lithium as a Gsk-3β inhibitor poses the question as to whether Wnt7a signals through this pathway to regulate synapse formation. Thus, the downstream events leading to synapse formation induced by Wnts remain poorly understood.

To address the mechanism by which Wnt7a regulates presynaptic differentiation and to determine the functional impact of deficit in Wnt signaling, we have examined the contribution of Dvl (Logan and Nusse, 2004). New studies have revealed that Dvl functions locally within a cellular compartment (Ciani et al., 2003; Itoh et al., 2005) and can bind to microtubules to increase their stability in both dividing cells and differentiated neurons (Krylova et al., 2000). In addition, Dvl has been shown to regulate dendritic morphology in hippocampal neurons (Rosso et al., 2005). Loss of function of Dvl1, which is one of the three mouse Dvl genes, results in abnormal behavior that is manifested by defects in social interactions (Lijam et al., 1997). However, the mechanism leading to this defect remains unexplored. Interestingly, although Dvl has been shown to be involved in synapse formation at the neuromuscular junction (Luo, 2002), the role for Dvl1 at central synapses is unknown.

We examined Dvl1^−/− mutant mice for possible defects in synapse formation and for the contribution of Dvl1 to Wnt7a function in this process. We found a functional interaction between Wnt7a and Dvl1 during the formation of the MF-GC synapse. In vivo, double Wnt7a^−/−; Dvl1^−/− mutant mice exhibit an enhanced defect in the localization of presynaptic proteins at MF terminals when compared with single mutant mice. Ultrastructural studies reveal that glomerular rosettes in the Wnt7a/Dvl1 mutant are simpler, yet synapses still form with normal active zones. Importantly, electrophysiological recordings of double mutant mice reveal a decreased frequency of mEPSCs without changes in amplitude, indicating a defect in neurotransmitter release. Furthermore, we show by gain and loss-of-function studies that Dvl is necessary to regulate the formation of presynaptic clusters and synaptic vesicle recycling sites. Moreover, the presence of Dvl protein at presynaptic sites and its ability to increase the clustering of Bassoon, which is a cytomatrix protein involved in synaptic assembly, are consistent with the notion that Wnt regulates synaptic assembly through Dvl. Our studies also raise the interesting possibility of a role for Wnt signaling in synaptic function.

To begin to dissect the mechanism by which Wnts regulate synapse formation, we first examined the distribution of Dvl in neurons and found that endogenous Dvl is present in synaptosomal fractions isolated from adult brains (Fig. 1 B). Interestingly, other components of the Wnt pathway, such as β-catenin and Gsk-3β, are also present in synaptosomes. The specificity of this fraction was confirmed by the presence of synaptic proteins such as PSD95 and synaptophysin (Fig. 1 B). To further test whether Dvl is present at synaptic sites, we examined the colocalization of Dvl with presynaptic markers in cultured hippocampal neurons. When expressed in hippocampal neurons, Dvl1-HA has a punctate distribution along the axon and colocalizes with endogenous synapsin I (Fig. 1 C). Importantly, endogenous Dvl is found in bright and small puncta, as well as in larger and diffuse areas along the axon (Fig. 1 D). Some of the bright Dvl puncta colocalizes with endogenous clusters of synapsin I and Bassoon (Fig. 1 D). To examine whether Dvl localizes to stable synaptic sites, Dvl1-HA and PSD95-GFP were expressed in 14 d in vitro (DIV) cultures. Some Dvl1 clusters colocalize with synapsin I in close apposition with PSD95-GFP (Fig. S1). Collectively, these findings suggest that Dvl is present presynaptically at central synapses.

To test the function of Dvl in synapse formation, we examined the consequences of Wnt signaling on the formation of presynaptic clusters during the initial stages of synaptic assembly. Several studies have suggested that GC release factors, such as Wnt7a and FGF22, modulate synapse formation between MFs and GCs (Hall et al., 2000; Umemori et al., 2004). We tested the effect of GC conditioned media (CM) on the clustering of synapsin I, which provides a readout of presynaptic differentiation. We found that CM from cerebellar GCs induces the formation of numerous and very large clusters of synapsin I in MF cultures (Fig. S2 A). Wnt7a, which is released by GCs, also induces clustering, but to a lesser extent than GC-secreted factors (Fig. S2 A). Wnt7b, which is highly related to Wnt7a and is also expressed by cerebellar GCs (not shown), increases clustering of synapsin I puncta as observed with Wnt7a (Fig. S2 A). Thus, both endogenous Wnt7a and Wnt7b could regulate synapse formation between MFs and GCs in the cerebellum. The effect of Wnt7b could explain the rescue of the phenotype of the Wnt7a mutant at postnatal day (P) 15 (Hall et al., 2000), as expression of Wnt7b increases in GCs at this stage (unpublished data). We tested whether inhibition of Wnt activity by addition of the secreted Wnt inhibitor Sfrp-1 (Finch et al., 1997; Rattner et al., 1997) was able to block the clustering activity of GC factors. Indeed, Sfrp-1 blocks most of the clustering activity of GCs (Fig. S2 B). These findings indicate that Wnts, possibly Wnt7a and Wnt7b, contribute significantly to the clustering activity of factors released by cerebellar GCs.

To further examine the role of Wnts in presynaptic differentiation, we tested the effect of Wnts on several synaptic proteins. As Wnt7b CM exhibits a more reliable level of activity than Wnt7a CM, we have hereon used Wnt7b for gain-of-function studies. In MF axons, Wnt7b induces an 85% increase in the number of clusters labeled with VAMP2 (24 ± 2.3 clusters per 100 μm), compared with controls (13.3 ± 2.0 clusters per 100 μm; P < 0.05; Fig. 2, A and C). Furthermore, the Wnt antagonist Sfrp-1 blocks Wnt7b activity, as the number of VAMP2 puncta (14.1 ± 1.0 clusters per 100 μm) is similar to control (Fig. 2, A and C). In addition, VAMP2 clusters are larger in cultures exposed to Wnt7b (80.4 ± 3.3 μm²), compared with control (51.3 ± 1.8 μm²) and Sfrp-1 (48.2 ± 2.8 μm²; P < 0.05). We then examined the effect of Wnt on the clustering of Bassoon. In 10 DIV hippocampal cultures, Wnt7b induces a 28% increase in the number of Bassoon puncta with 22.8 ± 1.9 clusters per 100 μm, compared with 17.7 ± 1.5 clusters per 100 μm in control (Fig. 2, B and D). Conversely, Sfrp1 reduces the Wnt effect to almost control levels (18.2 ± 1.9 clusters per 100 μm). Western blot analysis shows that Wnt7b does not affect the level of presynaptic proteins such as synapsin I, VAMP2, Munc18, or syntaxin in cultured MFs (Fig. 2 E). In addition, β-catenin levels are unaffected (Fig. 2 E), suggesting that Wnt might signal through a β-catenin–independent pathway. Alternatively, Wnt could induce small changes in β-catenin levels, which are not detectable after overnight treatment. Collectively, these findings suggest that Wnt increases the clustering of presynaptic markers without affecting the overall levels of presynaptic proteins.

To begin to address the mechanism by which Wnts regulate synaptogenesis, we examined the time course of Wnt action. We found that Wnt7b could induce clustering of VAMP2 within 15 min in MFs and hippocampal neurons without a visible effect on axon remodeling (Fig. S3, A–D). We then examined whether Dvl mimics the effect of Wnts on synaptic protein clustering. As explants of MFs cannot be easily transfected, we used hippocampal cultures. In EGFP-expressing control neurons, few synapsin I clusters are found along the axon (Fig. 2 F), which is similar to untransfected cells in these early cultures. In contrast, Dvl1-HA–expressing neurons exhibit a twofold increase in the number of synapsin I clusters (Fig. 2, F and H). Importantly, expression of Dvl also increases the number of Bassoon clusters by 50% (Fig. 2, G and I). Thus, Dvl1 mimics the effect of Wnts on the clustering of presynaptic markers, which is a readout of presynaptic differentiation.

We then examined the consequence of Dvl1 loss of function. MFs isolated from wild-type mice contain 28.3 ± 1.6 clusters of VAMP2 per 100 μm (Fig. 3, A and B). In contrast, MFs from Dvl1 mutant mice exhibit a 57% decrease in the number of VAMP2 puncta (16.0 ± 1.2 clusters per 100 μm; P < 0.05; Fig. 3, A and B). Furthermore, the average size of VAMP2 clusters in wild-type MFs is 53.9 ± 3.5 μm², whereas MFs from Dvl1 mutant have an average size of 37.3 ± 2.6 μm², demonstrating that the loss of Dvl1 function leads to a 44% decrease in the size of VAMP2 clusters (P < 0.01). Loss of function of both Wnt7a and Dvl1 in presynaptic differentiation also results in a significant decrease in the number of VAMP2 clusters to similar levels to Dvl1 mutant MFs (Fig. S3, E and F). This finding is consistent with the fact that isolated MFs from pontine nuclei have not yet been exposed to endogenous Wnt7a from GCs, as they were isolated before contacting GCs in the cerebellum. Interestingly, some presynaptic differentiation still occurs in Dvl1 mutant MFs. The expression of Dvl2 and Dvl3 in MFs (unpublished data) could contribute to the residual level of presynaptic differentiation in the Dvl1 mutant. Consistent with this view, we found that exposure of Dvl1 mutant MFs to Wnt7b does result in a small increase in the number of VAMP2 clusters, but does not fully rescue the Dvl1 phenotype to wild-type levels (Fig. 3, C and D). Collectively, these results demonstrate that Dvl1 is required for optimal Wnt signaling to mediate presynaptic differentiation.

The aforementioned results suggest that Wnt signaling through Dvl1 regulates presynaptic differentiation by increasing the assembly of presynaptic sites. To test this directly, we have examined the effect of Wnt7b on synaptic vesicle recycling in MFs. Synaptic vesicle recycling was determined by the uptake of an antibody against the intralumenal domain of synaptotagmin (Matteoli et al., 1992; Kraszewski et al., 1995). Upon depolarization with a high potassium buffer, the synaptotagmin antibody is detected in vesicles corresponding to the readily releasable and reserve pools of synaptic vesicles, which have undergone exocytosis and endocytosis (Sudhof et al., 1993; Kraszewski et al., 1995). Consistent with a role in presynaptic assembly, Wnt7b increases the number of recycling puncta labeled by anti-synaptotagmin antibody by ∼61% (31.4 ± 2.2 puncta per 100 μm) when compared with controls (18.9 ± 2.8 puncta per 100 μm; P < 0.01; Fig. 4, A and B). Importantly, the presence of Sfrp-1 significantly reduces the number of recycling vesicles when compared with Wnt7b CM (16.7 ± 3.1 puncta per 100 μm; Fig. 4, A and B). Wnt7b also induces a significant increase (57%) in the size of recycling clusters (85.4 ± 4.7 μm²) compared with control (49.1 ± 3.5 μm²) or Sfrp-1–treated cultures (53.2 ± 2.8 μm²; P < 0.05). Western blot analyses showed that Wnt7b does not affect the level of synaptotagmin protein (Fig. 4 E), indicating that the increased number of recycling puncta reflects a true increase in the recycling rate. Moreover, synaptic vesicle recycling measured by FM1-43 dye uptake also showed that Wnt7b increases recycling (Fig. 4, C and D). These results demonstrate that Wnt7b regulates the formation of functional presynaptic terminals.

To test the requirement of Dvl1 in synaptic vesicle recycling, MFs from Dvl1 mutant mice were examined. We used pontine explants from P0 animals when MFs have not yet been in contact with Wnt7a-expressing GCs. MFs from Dvl1 mutant mice exhibit a 37% decrease in the number of recycling puncta (25 ± 1.4 puncta per 100 μm) compared with wild type (34 ± 2.0 puncta per 100 μm; P < 0.05; Fig. 4, F and G). The recycling clusters are also significantly smaller in Dvl1 mutant MFs (22.9 ± 1.6 μm²) compared with wild type (30.8 ± 2.2 μm²; P < 0.05). These results indicate that Dvl1 is required in MF terminals for normal synaptic vesicle recycling.

Our previous study demonstrated that Wnt7a-null mutant mice exhibit a defect in synaptic differentiation that is manifested by a delayed accumulation of synapsin I at MF terminals between P10 and P12 (Hall et al., 2000). By P15, this defect is no longer evident (Hall et al., 2000). This could be caused by compensation by other factors, possibly by other Wnts like Wnt7b, which are expressed in the cerebellum. Therefore, we decided to examine the consequence of the loss of function of Dvl1 during synapse formation. Dvl1 is highly expressed in the cerebellum (Sussman et al., 1994; Krylova et al., 2000) and, thus, could mediate Wnt7a function. Importantly, MFs from Dvl1 mutants exhibit a defect in presynaptic differentiation. To test a possible genetic interaction between Wnt7a and Dvl1 in vivo, we also examined the cerebellum of double mutant mice. Interestingly, genetic studies in Drosophila melanogaster have shown that defects in hypomorphic wingless/Wnt mutants can be enhanced by mutations in dishevelled (Theisen et al., 1994). Therefore, we predict that a concomitant loss of Dvl1 and Wnt7a would enhance the Wnt7a phenotype.

We first analyzed Wnt7a, Dvl1 single mutant, and Wnt7a/Dvl1 double mutant mice for possible defects in the accumulation of synapsin I at P10. Indeed, Dvl1 mutants exhibit a defect in the accumulation of synapsin I at glomerular rosettes compared with wild type, as observed in the Wnt7a single mutant (Fig. 5 A). More importantly, double mutant mice exhibit a stronger defect than the single mutants at P10 (Fig. 5 A). Quantification of glomerular areas stained by synapsin I revealed a shift in the size distribution of synapsin I–stained rosettes in all the mutants (Fig. 5 B), which was made evident by the increase in the percentage of smaller (<40 μm²) with a concomitant decrease of larger ones (>20 μm²; P < 0.05 for single mutants, P < 0.01 for double mutants).

We then examined glomerular rosettes at P15, which is the peak of cerebellar synaptogenesis. We found that the defect in presynaptic protein accumulation recovers by P15 in the single Wnt7a and Dvl1 mutants (Fig. 5 C). In contrast, double mutant mice still exhibit a significant defect in the accumulation of synapsin I at glomerular rosettes (Fig. 5 C). Quantification shows that double mutant mice exhibit an increase of ∼50% in the number of smaller synapsin I–stained glomerular rosettes (<40 μm²) with a concomitant decrease in the number of larger ones (Fig. 5 D). Furthermore, other presynaptic proteins, such as synaptobrevin/VAMP2 (Fig. S4 A), SV2 (Fig. S4 C), and synaptophysin (Fig. S4 E), were also examined at P15. Although some defects are detected in large glomerular rosettes in single mutants (Fig. S4, B, D, and F), a clear defect is observed in the double mutant, as the number of small rosettes increases with a concomitant decrease in larger rosettes (Fig. S4, B, D, and F). These findings indicate that there is a genetic interaction between Wnt7a and Dvl1 at cerebellar synapses, as Dvl1 deficiency enhances the phenotype of Wnt7a in cerebellar glomerular rosettes.

The defect in the localization of synaptic proteins to the presynaptic MF terminals suggests that Wnt signaling could regulate the levels of these proteins. However, Western blot analyses of whole cerebellar homogenates from P15 double mutant mice reveal no differences in the levels of VAMP2, synaptophysin, and synapsin I (Fig. S4 G). Other proteins, such as calmodulin-associated serine/threonine kinase (CASK), synaptosomal-associated protein of 25 kD, and β-catenin, are also unaffected between wild-type and double mutant mice (Fig. S4 G). These results combined with those obtained in cultured MFs indicate that Wnt7a-Dvl1 signaling regulates the distribution rather than the levels of presynaptic proteins at MF terminals during synapse formation.

During postnatal development, a single MF makes contact with several GCs to form the glomerular rosette, a multisynaptic structure. Upon contact, GCs start to interdigitate into the MF terminal, resulting in increased perimeter and complexity of this terminal (Hamori and Somogyi, 1983b). These morphological changes are correlated with synaptic maturation. Therefore, a delay in the accumulation of presynaptic proteins observed in Wnt7a and Dvl1 mutants could be caused by a delay in the maturation of MFs or by a decrease in the size of the presynaptic terminal. To investigate this more closely, we analyzed the morphology of glomerular rosettes by EM at P15.

Our EM studies revealed that MF terminals have a simpler structure in the double Wnt7a/Dvl1 mutant cerebella. We measured complexity, which was calculated based on perimeter squared divided by area (Hamori and Somogyi, 1983b), as a parameter of the irregularity of the terminal. We found that the area and complexity of MF terminals (Table I) are not significantly different between wild-type (Fig. 6, A and A'), Wnt7a (Fig. 6, B and B') and Dvl1 (Fig. 6, C and C') single mutant mice (ranging from 32.30 ± 1.25 to 29.95 ± 1.09). In contrast, glomerular rosettes from double mutant mice exhibit a significant decrease in complexity (26.24 ± 0.63; P < 0.01; Fig. 6, D and D'). As the area of the terminals is not affected in double mutants, this finding indicates that the defect in complexity is caused by less interdigitation of GCs into the MF terminal. These results show that Wnt-Dvl signaling is required for the proper structural maturation of the MF-GC synapse in vivo. As the size of glomerular rosettes is unaffected in the mutants, the decrease in the area labeled with presynaptic proteins at glomerular rosettes (Fig. 5) is caused by a defect in the localization of presynaptic markers.

The simpler morphology of the glomerular rosettes and the delay in the accumulation of presynaptic proteins in double mutant mice led us to examine in detail for possible structural defects at the presynaptic terminal. Several parameters were measured, including the number of synaptic contacts per glomerular rosette, the number of docked and reserve pool vesicles, the width of synaptic cleft, the length of synaptic contact, and the thickness of the postsynaptic density. Docked vesicles were defined as those vesicles present within a distance of 50 nm of the active zone, whereas the reserve pool was defined as vesicles present up to 200 nm away from the active zone (Pozzo-Miller et al., 1999). Morphometric analyses showed no significant differences in the number of synaptic contacts per glomerular rosette, in the number of docked or reserve pool vesicles between single mutant, double mutant and wild-type animals (Table I). However, double mutant mice exhibit a mild, but significant, decrease in the width of the synaptic cleft (P < 0.01) and the postsynaptic density (P < 0.01; Table I). These results suggest that apart from a defect in complexity, active zones are structurally normal in single and double Wnt7a/Dvl1 mutant mice.

The changes in synaptic vesicle recycling in MFs observed in gain- and loss-of-function studies led us to test for possible functional defects at the MF-GC synapse. Acute cerebellar slices were used to analyze functional properties of the MF-GC synapse in the double mutant. We chose to perform electrophysiological recordings at P15, as this stage is the peak of cerebellar synaptogenesis. Importantly, P15 GCs are less heterogeneous than at earlier ages, thus, more consistent recordings can be obtained. Capacitance measurement during whole-cell voltage clamp recordings enables the measurement of estimated total membrane area (Cathala et al., 2003). We found that cerebellar GCs recorded from wild-type and double mutant mice had a similar membrane capacitance of 3.3 ± 0.3 pF (n = 10) and 2.9 ± 0.2 pF (n = 10), respectively. Thus, no significant differences in the size of cerebellar GCs between wild-type and double mutant mice were observed.

The relatively small size of cerebellar GCs means that there is little dendritic filtering of the conductance change that results from activation of postsynaptic glutamate receptors. Therefore, analysis of miniature excitatory postsynaptic currents (mEPSCs) recorded in the presence of tetrodotoxin can be used to estimate the frequency of spontaneous vesicle fusion at the MF terminal. In whole-cell recordings from wild-type GCs, mEPSCs were clearly detected in 6 out of 10 recordings, whereas in the double mutant mEPSCs were detected in 8 out of 10 recordings. The mean mEPSC frequency in wild-type GCs was 0.23 ± 0.07 Hz (Fig. 7, A, B, and G). In double mutant GCs, the mEPSC frequency was consistently lower than wild type, with an average mEPSC frequency of only 0.05 ± 0.02 Hz (Fig. 7, D, E, and G; P < 0.05; unpaired t test). Interestingly, single Wnt7a or Dvl1 mutants showed mEPSC frequencies similar to wild type (not depicted). There were no significant differences in the mEPSC peak amplitudes (Fig. 7, C, F, and H), 10–90% rise-time, or 50% decay recorded in the either of the single knockout strains or in double mutant mice (Table S1). Therefore, a similar number of postsynaptic glutamate receptors appear to be activated after vesicle release at the MF terminal. The kinetics of activation and deactivation of these receptors were not obviously affected in any of the mutants. These results show that deficiency in Wnt7a or Dvl1 alone does not affect synaptic function at this stage, but a deficiency in both Wnt7a and Dvl1 results in a significant decrease of mEPSC frequency, indicating a defect in the release of neurotransmitters.

Studies in the cerebellum and spinal cord have revealed a role for Wnt proteins as target-derived signals that regulate the terminal arborization of axons and presynaptic differentiation (Hall et al., 2000; Krylova et al., 2002). We show that Dvl1 is required for proper Wnt signaling during synapse formation. Our studies on Dvl1 have unraveled new aspects of Wnt function at central synapses. The findings presented in this study indicate that early in synaptogenesis Wnt signaling regulates the assembly of the presynaptic terminal by increasing the formation of presynaptic clusters, including Bassoon, and the number of SV recycling sites. In vivo, a deficiency of Wnt7a and Dvl1 does not affect the morphology of active zones in the cerebellum, although a defect in neurotransmitter release is observed. Based on these findings, we propose a model in which Wnt signaling could regulate two processes, the assembly of the presynaptic terminal during the early stages of synaptogenesis and, later, the release of neurotransmitters.

Dvl has been postulated to function as a scaffold protein for efficient Wnt signaling. Importantly, Dvl1 mutant mice exhibit a defect in social behavior (Lijam et al., 1997), demonstrating that Dvl1 deficiency affects the function of the nervous system. However, its role during synapse formation has not been explored. Our new analyses of Dvl1 mutant mice and gain-of-function studies reveal that Dvl1 is required for the proper localization of presynaptic proteins as MF terminals from Dvl1 mutant mice exhibit a decreased number of synaptic protein clusters and SV recycling sites. Importantly, the addition of Wnt to Dvl1-deficient MFs only partially restores presynaptic protein clustering. Moreover, the Wnt7a/Dvl1 double mutant exhibits an enhanced defect in the accumulation of presynaptic proteins at cerebellar glomerular rosettes in vivo when compared with single Dvl1 or Wnt7a mutant mice. These findings demonstrate a novel role for Dvl1 as a mediator of Wnt function during synaptogenesis in the cerebellum.

Ultrastructural analyses revealed that double mutant mice exhibit defects in the complexity of glomerular rosettes. Interestingly, the structure of the active zones is normal, with no apparent defects in the number of synaptic vesicles, despite a defect in presynaptic protein accumulation at the MF terminal. One possible interpretation for this finding is that double mutant mice might represent a partial loss of function rather than a complete loss of function of Wnt signaling. Consistent with this view, we found that Dvl2 and Dvl3 as well as Wnt7b are expressed in the cerebellum during the period of synaptogenesis (unpublished data). Thus, the presence of other Wnts or Dvl genes expressed in the cerebellum might partially compensate for the loss of Wnt7a and Dvl1 function.

How does Wnt-Dvl signaling regulate synapse formation? Wnt proteins could regulate synapse formation by at least two mechanisms. Wnt could accelerate the maturation of neurons by increasing the expression of synaptic proteins and those involved in synaptic assembly. Alternatively, Wnts could regulate synaptic assembly by stimulating the localization or recruitment of synaptic components to future synaptic sites. In the first model, Wnts could regulate the transcription and/or translation of synaptic proteins. However, we found no evidence for this model, as the level of several synaptic proteins do not change in the double mutant cerebellum or in neurons exposed to Wnt. Interestingly, other secreted factors implicated in synapse formation, such as FGF22 and thrombospondin also regulate clustering without affecting the levels of presynaptic proteins (Umemori et al., 2004; Christopherson et al., 2005). In addition, we show that Dvl colocalizes with endogenous Bassoon and increases the clustering of Bassoon, which is a protein implicated in synaptic assembly. We therefore favor the view that Wnts are not just priming factors as proposed by Waites et al. (2005), but that Wnt-Dvl signaling regulates synapse formation by accelerating presynaptic assembly.

Wnts regulate presynaptic remodeling (Hall et al., 2000; Krylova et al., 2002) suggesting the possibility that Wnt-mediated effect on presynaptic differentiation could be a consequence of changes in axon morphology. However, short-term exposure of MFs to Wnt still induces presynaptic protein clustering without affecting branching and detectable changes in axon remodeling (Fig. S3, A–D). Moreover, our studies in hippocampal neurons demonstrate that Wnt7b does not induce any detectable changes in axon morphology, yet presynaptic differentiation is induced. These findings suggest that Wnt signaling regulates presynaptic assembly independently of significant changes in morphology.

Does Wnt-Dvl signaling play a role at the postsynaptic terminal? Although our studies have been focused on the presynaptic terminal, Dvl1 is also present in developing dendrites and regulates dendritic morphogenesis in hippocampal neurons (Rosso et al., 2005). Interestingly, we found a decrease in the width of the PSD at the cerebellar synapse in double mutant mice. However, no obvious changes in the amplitude of mEPSCs were observed in double mutant mice, suggesting that a similar number of postsynaptic glutamate receptors were activated after neurotransmitter release. However, further studies are needed to establish whether Wnt signaling plays a postsynaptic role.

Dvl was originally identified as a cytoplasmic protein that mediates Wnt signaling to the nucleus (Noordermeer et al., 1994; Siegfried et al., 1994). Recently Dvl has been shown to locally regulate signaling events within different cellular compartments (Ciani et al., 2003; Itoh et al., 2005). In neurons, we found that Dvl1 exhibits a punctate distribution along the axon as observed with many presynaptic proteins and, indeed, a pool of Dvl1 colocalizes with synapsin I and Bassoon clusters. Importantly, a fraction of Dvl clusters are in close apposition with the postsynaptic protein PSD95, suggesting that Dvl is present at both orphan and synaptic sites. Consistent with its synaptic localization in cultured neurons, endogenous Dvl1 is found in synaptosomal fractions from adult brain. These findings suggest that Wnt signaling could signal locally by activating synaptic Dvl.

Although active zones appear normal in the Wnt7a/Dvl1 double mutant, electrophysiological recordings at the MF-GC synapse revealed a functional defect. Double mutant mice exhibit a significant decrease in the frequency of mEPSCs without affecting the amplitude and duration of mEPSCs suggesting that Wnt-Dvl signaling regulates the rate of neurotransmitter release without affecting the postsynaptic terminal.

Our studies are consistent with the notion that Wnt-Dvl signaling regulates synaptic assembly and function. Although the decrease in neurotransmitter release in the double mutant could be a consequence of an early defect in synapse formation, the fact that active zones are normal suggests that another mechanism could be involved. We propose that a partial loss of Wnt signaling permits the formation of structurally normal active zones but compromises the release of neurotransmitter by a yet unknown mechanism. Therefore, Wnt signaling could play a role in the modulation of synaptic activity in addition to synaptogenesis. Future studies will reveal whether the coordinated action of target-released molecules and membrane-bound proteins plays a role not only in synapse formation but also in synaptic modulation.

All mutant mice were maintained on a C57BL/6 background. Wnt7a;Dvl1 double mutant mice were obtained from crosses of heterozygous Wnt7a (A. McMahon, Harvard University, Cambridge, MA) and Dvl1 mutant mice (T. Wynshaw-Boris, University of California, San Diego, La Jolla, CA). Genotypes were determined by three-primer PCR using tail DNA. For Wnt7a, the primers used were forward, 5′-TTCTCTTCGTGGTAGCTCTGGGTG-3′, reverse, 5`-CTCCTTCCCGAAGACAGTACGCTCT-3′, and the forward Neo primer 5`-AGGCCTACCCGCTTCCATTGCTCA-3′. For Dvl1, the primers used were forward 5`-CGCCGCCGATCCCCTCTC-3′, reverse, 5`-TCT GCCCAATTCCACCTGCTTCTT-3′, and the Neo primer 5`-AGGCCT- ACCCGCTTCCATTGCTCA-3′.

Basilar pontine nuclei from wild-type, Dvl1 mutant, or Wnt7a^−/−;Dvl1^−/− mutant mice were dissected from newborn or P1 pups, cut into 100-μm² explants, and cultured in 16-well LabTeck chamber slides, as previously described (Hall et al., 2000). Explants were cultured for 2 d and then treated either overnight or for 15 min with control, Wnt7b CM, or Wnt7b CM plus Sfrp-1. Untreated mutant MFs were fixed after 3 d. Hippocampal cultures were prepared from E18 rats as previously described (Dotti et al., 1988). For transfections, a Nucleofector Amaxa system was used according to the manufacturer's protocol (Amaxa Biosystems). In brief, 3 × 10⁵ neurons were electroporated, plated at 2 × 10⁴ neurons per coverslip, and allowed to grow for 9 d before fixation. Constructs for transfections were EGFP, Dvl1-HA, and PSD95-GFP.

CM from cerebellar GCs were prepared as described in the previous section and used to treat pontine explants. Control, Wnt7a-HA, and Wnt7b-HA CM were obtained from stably transfected Rat1b cells. Sfrp-1 CM in Fig. S2 was prepared from QT6 cells transiently transfected with myc-tagged Sfrp-1 construct (J. Nathans, Johns Hopkins University, Baltimore, MD) and used to block GC CM (Hall et al., 2000). Neuronal culture medium (N2/B27) was added to cell lines and conditioned overnight. The levels of Wnt7b and Sfrp-1 proteins were assessed by Western blot. For other blocking experiments, 2.5 μg/ml of purified Sfrp-1 (R&D Systems) was preincubated with Wnt7b CM for 20 min at room temperature before addition to neurons.

The intralumenal antisynaptotagmin antibody uptake assay was performed as previously described (a gift from P. de Camilli, Yale University, New Haven, CT; Matteoli et al., 1992) on MFs at day 3. In brief, pontine explants were incubated for 5 min at 37°C with warm high potassium depolarization buffer; Krebs-Ringer solution (125 mM NaCl, 25 mM Hepes, pH 7.4, 5 mM KCl, 2 mM CaCl₂, 1.2 mM MgSO₄, 1.2 mM KH₂PO₄, and 6 mM glucose) supplemented with 110 mM KCl and 10 μM APV and CNQX and contained either 10 μg/ml of rabbit antiintralumenal synaptotagmin antibody or 10 μM of fixable FM1-43 (Invitrogen). After incubation, cultures were washed three times in warm Krebs-Ringer buffer, and then fixed and processed for immunofluorescence or directly mounted for FM1-43. Images were captured with a microscope (model BX60; Olympus) or with a motorized microscope (Carl Zeiss MicroImaging, Inc.) using a charge-coupled device camera (Orca ER; Hamamatsu). The number, size, and intensity of synaptotagmin puncta larger than 0.16 μm² were quantified using MetaMorph (Molecular Devices). As FM1-43 also labels nonsynaptic clusters, we used a more stringent threshold to select larger and brighter puncta, typically >0.20 μm². Axons were traced manually and the number of puncta per 100-μm neurite length and area (μm²) were quantified. Each experiment was performed in triplicate.

Hippocampal neurons and pontine explants were fixed with 4% paraformaldehyde, 4% sucrose in PBS (pontine explants have 0.5% EM grade glutaraldehyde added), and then permeabilized with 0.02% Triton X-100, followed by blocking with 5% BSA and incubation with primary antibodies for at least 1 h. When endogenous proteins were analyzed, neurons were fixed with methanol for 10 min at −20°C. Primary antibodies were against Dvl (Krylova et al., 2000), synaptobrevin/VAMP2 (Synaptic Systems), rabbit polyclonal intralumenal synaptotagmin (a gift from P. de Camilli), synapsin I (BD Biosciences), SV2 (Developmental Studies Hybridoma Bank), GAP-43 (Abcam), Bassoon (Bioquote Limited), and HA (Boehringer). The secondary antibodies Alexa Fluor 488 and 594 were obtained from Invitrogen. Each experiment was performed in triplicate and at least 15 images were taken per condition. The images were analyzed using MetaMorph. A threshold was set to capture clusters that were clearly distinguishable and did not merge with one another. The number, size, and intensity of the puncta were measured and Analysis of variance (ANOVA) statistical tests were performed.

Cerebella from wild-type and mutant animals were fixed in 4% PFA, embedded in paraffin, and processed as previously described (Hall et al., 2000). Antibodies against synapsin I, synaptophysin (CHEMICON International, Inc.), VAMP2, and SV2 were used. Localization of the proteins was visualized by the VECTASTAIN ABC peroxidase system (Vector Laboratories). Three independently bred animals were used for each of the four genotypes. An area of 2,700 glomerular rosettes per animal was measured. ANOVA statistical tests were performed.

SDS-PAGE and immunoblotting were performed from pontine explants or from cerebella using standard methods. Pontine samples were analyzed for cytoplasmic synaptotagmin (Synaptic Systems), VAMP2, synapsin I, Munc18, and β-catenin (BD Biosciences). Gsk-3β (BD Biosciences) was used as a loading control, whereas P15 cerebellar homogenates were analyzed for CASK (CHEMICON International, Inc.), synaptophysin (CHEMICON International, Inc.), VAMP2, synaptosomal-associated protein of 25 kD (Synaptic Systems), synapsin I, β-catenin, and tubulin (Abcam).

Synaptosomes were prepared as previously described (Cohen et al., 1977). In brief, adult mouse brains were homogenized in buffer A (0.32 M sucrose, 4 mM Hepes, pH 7.4, plus protease and phosphatase inhibitors) and centrifuged at 800 g for 10 min. All procedures were performed at 4°C. The supernatant was clarified by centrifugation at 9,000 g for 15 min, and the pellet containing myelin and synaptosomal and mitochondrial structures was resuspended in buffer A and layered on top of a discontinuous gradient containing 0.8/1.0/1.2 M sucrose in 4 mM Hepes, pH 7.4, plus protease inhibitors, and centrifuged at 82,500 g for 90 min. Synaptic membranes were taken from a 1–1.2-M interface and resuspended in buffer B (0.32 M sucrose, 4 mM Hepes, pH 7.4, and 150 mM NaCl plus inhibitors). Proteins were quantified by Lowry assay.

Small pieces from the internal granular layer of cerebellar lobes V, VI, and VIII of P15 animals were dissected and fixed in 3% EM grade glutaraldehyde in 0.1 M phosphate buffer, pH 7.3, at 4°C for at least 5 d. Samples from three wild-type, 3 Wnt7a^−/−, 3 Dvl1^−/−, and 5 Wnt7a^−/−; Dvl1^−/− mice were analyzed. Tissues were processed at the EM Unit of King's College London. Photographs of whole glomerular rosettes were taken at 20,000×, whereas photographs of synapses were taken at 72,000×. Photographs were scanned and analyzed. Glomerular complexity was estimated using the formula perimeter²/area (Hamori and Somogyi, 1983b). Docked vesicles were considered as vesicles present within 50 nm (one vesicle diameter) of the active zone, whereas vesicles located up to 200 nm away from the docked vesicles were considered as the reserve pool (Pozzo-Miller et al., 1999). ANOVA statistical tests were performed.

Parasagittal slices (250 μm thick) were cut from P14–P15 cerebellar vermis of mice. The slicing solution contained the following: 125 mM NaCl, 2.5 mM KCl, 1 mM CaCl₂, 2 mM MgCl₂, 26 mM NaHCO₃, 1.25 mM NaH₂PO₄, 25 mM glucose, 0.02–0.08 mM d-2-amino-5-phosphonopentanoic acid (d-AP5), pH 7.4, when bubbled with 95% O₂ and 5% CO₂. Patch-clamp recordings were made at room temperature (22–23°C), with an amplifier (Axopatch 700B; Axon Instruments). 50 mM picrotoxin and 1 mM strychnine was included in the recording solution to block GABA_A and glycine receptors, respectively, and in the presence 0.5 mM TTX to isolate mEPSCs. Patch pipette “internal” solution contained the following: 140 mM CsCl, 4 mM NaCl, 0.5 mM CaCl₂, 10 mM Hepes, 5 mM EGTA, and 2 mM Mg-ATP (adjusted to pH 7.3 with CsOH). Data analysis was performed using custom written software (Istvan Mody and Thorsten Hodapp, University of California, Las Angeles, Los Angeles, CA). Spontaneous mEPSCs were filtered at 2 kHz and digitized at 10 kHz. As all data were normally distributed (Shapiro-Wilk test), statistical differences between groups were tested using the t test and considered significant at P < 0.05 (STATISTICA 5.1; StatSoft).

Fig. S1 shows that Dvl1 colocalizes with synapsin I at synaptic sites. Fig. S2 shows that Wnt7a and Wnt7b have similar effects on synaptic vesicle clustering. Fig. S3 shows that short-term exposure to Wnt7b is sufficient to induce presynaptic assembly in mossy fibers and hippocampal neurons and Wnt7a^−/−; Dvl1^−/− double mutant (DKO) mossy fibers exhibit a defect in presynaptic clustering. Fig. S4 shows that Wnt7a;Dvl1 double mutant exhibits a defect in the accumulation of several presynaptic proteins at cerebellar glomerular rosettes at P15. Table S1 shows the properties of spontaneous EPSCs in GCs from wild-type and Wnt7a^−/−; Dvl1^−/− P15 animals.

We would like to thank to Drs. Any McMahon, Daniel Sussman, and Tony Wynshaw-Boris for generously providing the Wnt-7a and Dvl1 mutant mice, Eckard Gunderfinger for the Bassoon-EGFP construct, and Jeremy Nathans for Sfrp-1 construct, Pietro de Camilli for antibodies, and John Pacy for EM processing. We also thank Angela Brennan for helping with EM analyses and members of our laboratory for useful discussion and comments on the manuscript.

The Wellcome Trust, the Biotechnology and Biological Sciences Research Council, and the Medical Research Council (to J. Herreros) supported this work.

We declare that there are no commercial affiliations or conflict of interests.