The formation of functional synapses requires a proper dialogue between incoming axons and their future synaptic targets. As axons approach their target, they are instructed to slow down and remodel to form proper presynaptic terminals. Although significant progress has been made in the identification of the mechanisms that control axon guidance, little is known about the mechanisms that regulate the conversion of actively growing axon into a presynaptic terminal. We found that Wnt secreted proteins are retrograde signals that regulate the terminal arborization of axons and synaptic differentiation. Wnts released from postsynaptic neurons induce extensive remodelling on incoming axons. This remodelling is manifested by a decrease in axon extension with a concomitant increase in growth-cone size. This morphological change is correlated with changes in the dynamics and organization of microtubules. Studies of a vertebrate synapse and the Drosophila neuromuscular junction suggest that a conserved Wnt signalling pathway modulates presynaptic microtubules as axons remodel during synapse formation. In this paper I discuss the role of the Wnt–Dvl (Dishevelled protein)–GSK-3β (glycogen synthase kinase-3β) signalling pathway in axon remodelling during synapse formation in the central nervous system.
The formation of functional neuronal connections requires the correct navigation of axons to their synaptic targets, the formation of complex dendritic arbours and the assembly of functional synapses. During the journey to their targets, axons encounter a number of attractive and repulsive molecules that guide them to their appropriate targets [1,2]. Upon arrival at the target field, axons slow down their growth, branch and remodel to form appropriate terminals where the presynaptic apparatus is assembled. Transmembrane proteins such as neuroligin/neurexin and cadherins have been well recognized as molecules that regulate the assembly at both sides of the synapse . However, target-derived molecules such as Wnts and fibroblast growth factors are also emerging as key players in synapse formation [4–6]. These secreted factors function as retrograde signals to regulate the assembly of the presynaptic apparatus. Great emphasis has been given to identification of the molecules that initiate the recruitment of presynaptic molecules. However, the mechanisms that regulate the conversion of a growing axon into a presynaptic terminal remain poorly understood.
Members of the Wnt family of secreted proteins regulate the terminal arborization of axons and presynaptic differentiation [4,7]. Studies in the mouse cerebellum revealed a role for Wnt7a in the formation of the mossy-fibre–granule-cell synapse. Wnt7a is released by postsynaptic granule cells and acts on mossy fibre axons. Wnt7a has two distinct effects: it induces extensive axon remodelling and increases the clustering of synaptic proteins. The remodelling is manifested by decreased axon extension, increased spreading of the distal portions of the axon and growth cone, and a concomitant increased in axon branching . These changes in axon behaviour are different from those observed with attractive or repulsive axon guidance molecules, perhaps owing to the distinct function of Wnts. Although axons might encounter axon guidance molecules in transit to their targets, axons would encounter Wnt7a as they approached their final target. Based on these findings, we proposed that Wnt7a acts as a retrograde signal to regulate axon remodelling, a process that is intimately associated with the assembly of the presynaptic terminal.
Wnt-mediated remodelling is achieved through the activation of the canonical Wnt signalling pathway. Wnts can signal through three known pathways: the canonical, the planar cell polarity and the Ca2+ pathways . In the canonical pathway, binding of Wnts to their receptors Frizzled activates a downstream cascade that requires Dvl (Dishevelled protein), a cytoplasmic scaffold protein . Activated Dvl inhibits the serine/threonine kinase, GSK-3β (glycogen synthase kinase-3β), resulting in the accumulation of β-catenin and its translocation to the nucleus where the complex between β-catenin and TCF/LEF (T-cell factor/lymphoid enhancer factor) leads to the expression of target genes . Using drugs that inhibit the activity of GSK-3β, we found that Wnt proteins signal through GSK-3β to induce remodelling.
What is the mechanism by which Wnts regulate axon remodelling? The decreased axon extension suggests changes in the dynamics of microtubules, whereas increased spreading and membrane protrusion suggests actin involvement. Thus Wnt signalling is likely to affect these two major cytoskeletal components. In this review, however, I discuss the role of Wnt signalling on microtubules at remodelled axons during the initial stages of synapse formation.
The mechanism of axon remodelling by Wnts
To examine in detail the effect of Wnts on axons, we performed time-lapse recordings. Our studies revealed that Wnt changes the behaviour of axons within the first 10–20 min upon exposure to Wnt. First, Wnt induces growth-cone pausing and subsequently induces extensive membrane protrusion at the growth cone. Increased membrane activity is manifested by a constant retrograde and anterograde movement of the plasma membrane at the distal portion of the axon and at the growth cone. Axon pausing is correlated with a decrease in the length of axons. Wnt increases the number of both dynamic and stable pools of microtubules along the axon shaft and particularly the number of stable microtubules at growth cones. Thus, Wnt-induced remodelling is correlated with changes in microtubule dynamics.
To begin to address the mechanisms by which Wnt proteins regulate axon remodelling, we have focused our attention on Dvl. We found that Dvl1 is highly present in neurons of the cortex, hippocampus, pons and cerebellum . Thus, Dvl could mediate the response to Wnt in mossy fibre axons. Indeed, we found that expression of Dvl in neurons induces remodelling as observed with Wnts, by increasing the size of growth cones and the diameter of axons while decreasing the length of axons. This extensive remodelling is also correlated with the presence of large numbers of dynamic and stable microtubules .
How does Dvl regulate microtubules? Endogenous Dvl is localized in a puncta distribution in the neuronal cell body and along neurites. Interestingly, Dvl is tightly associated to axonal microtubules, particularly stable microtubules . Importantly, this association to microtubules is correlated with the ability of Dvl to stabilize microtubules . To assess the requirement of Dvl in microtubule dynamics, we examined the consequence of Dvl1 deficiency. Neurons lacking Dvl1 lose their microtubules faster than controls, demonstrating that Dvl1 is essential for microtubule stability. This microtubule stability is controlled through a divergent Wnt canonical pathway. We found that expression of wild-type GSK-3β blocks Dvl function on microtubules. Conversely, inhibition of GSK-3β using pharmacological agents increases microtubule stability. Interestingly, Dvl can still stabilize microtubules when TCF-mediated transcription is blocked by a dominant negative TCF. Moreover, inhibition of transcription by actinomycin D does not affect the ability of Dvl to stabilize microtubules . These findings demonstrate that Wnt–Dvl-mediated microtubule stability is achieved through a canonical pathway that does not require transcriptional activity and diverges downstream of GSK-3β.
Another interesting feature of Dvl is that it acts locally to regulate microtubule stability. To develop an inducible expression system, Dvl protein was fused to ER (oestrogen receptor) to create a fusion protein (Dvl–ER) that becomes functional upon induction with β-oestradiol. We found that after a short period of induction, functional Dvl–ER was localized to the neuronal cell body, whereas a longer period of induction led to the distribution of Dvl–ER along the axon. Taking advantage of this differential distribution of Dvl, we asked whether Dvl acts locally. We found that after a shorter period of induction, microtubules were stabilized only in the cell body. A longer period of induction, in contrast, led to the stabilization of axonal microtubules . We found a 100% correlation between the localization of Dvl and the presence of stable microtubules, strongly suggesting that Dvl functions locally to regulate microtubule dynamics.
How does GSK-3β regulate microtubules? Our previous studies have shown that GSK-3β directly phosphorylates MAP-1B (microtubule-associated protein 1B), a microtubule-associated protein that is highly present in axons . Importantly, Wnt or expression of Dvl decreases MAP-1B phosphorylation by GSK-3β in axons [12,13]. This change is correlated with increased microtubule stability. Conversely, phosphorylation of MAP-1B by GSK-3β has been shown to maintain more dynamic microtubules . Based on these findings, we proposed a model in which Dvl that is present in microtubules inhibits GSK-3β (possibly a pool of GSK-3β present on microtubules) to regulate MAP-1B phosphorylation, resulting in changes in microtubule dynamics.
Wnt signalling also regulates the organization of microtubules, as Wnt induces unbundling and looping of microtubules at enlarged growth cones. Both populations of microtubules (dynamic and stable) are affected. Microtubule unbundling occurs at the distal portions of the axon and particularly at spread areas in the axon shaft. Looped microtubules occur at growth cones, probably contributing to the increased size of the growth cones. Expression of Dvl or inhibition of GSK-3β also induces looping and unbundling of microtubules . Thus the Wnt–Dvl–GSK-3β pathway regulates both the dynamics and organization of microtubules.
In vivo remodelling of axons at synaptic terminals
What is the consequence of loss of function of Wnt signalling in vivo? To address the in vivo role of Wnt, we examined the Wnt7a mouse mutant for possible defects at the mossy-fibre–granule-cell synapse. Wnt7a is highly expressed in cerebellar granule cells at the time when mossy fibres begin to contact granule-cell dendrites . During early postnatal development, mossy fibres originated outside of the cerebellum arrive at the cerebellar cortex and form synapses with granule cells (Figure 1). Upon contact, mossy fibres are extensively remodelled, a process that is characterized by axon spreading, growth-cone enlargement and the formation of very irregular mossy-fibre terminals . As development progresses, several granule-cell dendrites interdigitate into a single mossy fibre axon, creating an irregular-shaped presynaptic terminal that contains several active zones . Studies of granule-cell-deficient mutants have suggested that granule-cell factors regulate mossy-fibre remodelling . Several pieces of evidence suggest that Wnt7a from granule cells is mediating this remodelling. First, Wnt7a induces the remodelling of mossy fibre axons in culture  similarly to those observed in vivo. Secondly, secreted Wnt inhibitors such as SFRP-1 (secreted frizzled related protein 1) block the remodelling activity of granule-cell factors when added to culture mossy fibres . Finally, the Wnt7a mutant mice exhibit a defect in the remodelling of the mossy fibres in the cerebellum . These results demonstrate that Wnt signalling is required for the proper remodelling of presynaptic terminals in the cerebellum.
Proposed model for the role of Wnt proteins in axon remodelling during synapse formation in the mouse cerebellum
During remodelling, presynaptic proteins accumulate at mossy fibres, resulting in the formation of a multisynaptic structure called the glomerular rosette. Analyses of the Wnt7a mutant also revealed a defect in the accumulation of several presynaptic proteins, suggesting that Wnt7a is required for both remodelling of the terminal and presynaptic differentiation. Clustering of presynaptic proteins occurs along the axon, but particularly at spread areas in the axon shaft. Interestingly, axon spreading occurs where unbundled microtubules are present (F. Lucas and P.C. Salinas, unpublished work). These spread areas might constitute immature en passant synapses . These findings raise the interesting possibility that changes in microtubule dynamics and organization drive axon remodelling, which in turn contributes to the accumulation of presynaptic molecules to future synaptic sites.
Parallels between the vertebrate and the Drosophila synapse
What is the evidence that the Wnt–GSK-3β signalling pathway functions in vivo to remodel axons during synapse formation? Recent studies at the Drosophila NMJ (neuromuscular junction) have demonstrated that Wg (Wingless protein)/Wnt regulates synaptogenesis through Futsch, the MAP-1B homologue . wg mutants exhibit defects in the size of the presynaptic terminal and in the structure of active zones. Importantly, these defects are associated with changes in the microtubules at presynaptic boutons . In wild-type synapses, microtubules are decorated with the MAP Futsch, a protein previously shown to control the growth of the NMJ . Importantly, recent genetic studies have shown that Sgg, a GSK-3β homologue, negatively controls the growth of the synapse through changes in Futsch and microtubule dynamics . These studies provide clear genetic evidence that the Wnt–GSK-3β–MAP-1B pathway is essential for synapse formation and growth in the fly.
Does a similar pathway operate at central synapses in the vertebrate? We have provided very strong evidence that Wnts remodel axons through changes in microtubules and that Wnt signalling regulates the remodelling of axon terminals during the initial stages of synapse formation at mossy-fibre–granule-cell synapses. However, the role of microtubules at synapses has been questioned as microtubules appear not to be present at active zones or in most synaptic boutons of the vertebrate central synapse . It is possible that extensive remodelling mediated through microtubules is characteristic of large synaptic terminals such the glomerular rosettes, or perhaps this is an early event in synapse formation. Changes in microtubules could contribute to the morphological conversion of an actively growing axon into a proper synaptic bouton. Or do microtubules play a more active role in synaptic assembly? The strong parallel between the Drosophila NMJ and the mossy-fibre–granule-cell synapse raises the interesting notion that the Wnt–Dvl–GSK-3β pathway might play a general role in the formation, and possibly the growth, of vertebrate synapses.
Cell Architecture: from Structure to Function: A Focus Topic at BioScience2005, held at SECC Glasgow, U.K., 17–21 July 2005. Edited by S. Cockroft (University College London, U.K.), Y. Goda (University College London, U.K.), R. Insall (Birmingham, U.K.) and M. Wakelam (Birmingham, U.K.).
This work was funded by the Biotechnology and Biological Sciences Research Council and the Wellcome Trust.