Neurons are the masters of membrane specialization. Normally, each polarized neuron has one axon and many dendrites, but it is not clear how this neuronal polarity is formed and maintained. Two recent reports in Cell suggest that glycogen synthase kinase 3β (GSK3β) might be a key factor in translating extracellular cues into changes in cytoskeletal organization.

In culture, embryonic hippocampal neurons are initially nonpolarised, with immature neurites, and one of these neurites then grows out to become an axon. Jiang and colleagues found that GSK3β is present in all processes in both polarized and nonpolarized neurons. However, the distribution of phosphorylated GSK3β, which is inactive, is more interesting. This is mainly present in the tips of axons in polarized neurons.

Overexpression of a constitutively active GSK3β mutant inhibits axon formation without affecting the number of dendrites. By contrast, both the inhibition of GSK3β activity by pharmacological and peptide inhibitors and the suppression of GSK3β expression by short-hairpin RNAs result in the formation of multiple axon-like processes. Not only do these processes look like axons and express axonal markers, they can also support synaptic vesicle recycling — a basic feature of functional axons. Among the kinases that might phosphorylate GSK3β and regulate its activity, AKT is differentially localized in axons. Therefore, the authors went on to test whether it promotes axon formation by functioning as an upstream inhibitor of GSK3β. Overexpression of a constitutively active AKT mutant resulted in the formation of multiple axons, which could be partially reversed by co-transfection of GSK3β.

But how does GSK3β regulate axon formation? Axon specification is thought to involve elongation of an immature neurite, and microtubule assembly in the tip of the neurite is essential. In the second study, Yoshimura and colleagues identified a substrate of GSK3β, collapsin response mediator protein 2 (CRMP2), which is involved in regulating axon formation and might be the missing link between GSK3β activity and microtubule organization. GSK3β phosphorylates CRMP2, thereby reducing its ability to bind to tubulin, the building block of microtubules. Like dephosphorylated GSK3β, nonphosphorylated CRMP2 is predominantly present in the tips of axon growth cones, and its overexpression also results in the formation of multiple axons. Overexpression of nonphosphorylated CRMP2 can counteract the effect of GSK3β on axon growth, which indicates that GSK3β regulates neuronal polarity through CRMP2. Interestingly, neurotrophin 3 and brain-derived neurotrophic factor (BDNF) — factors that promote axon formation by activating the AKT-GSK3β pathway — decrease CRMP2 phosphorylation in neurons.

The two studies put together an interesting theory of how a particular neuronal process can be selected to form an axon. In nonpolarized neurons, all neurites contain the active form of GSK3β and phosphorylated CRMP2, which has low tubulin-binding activity. This blockade of microtubule assembly can be removed by extracellular cues that activate AKT, which, in turn, phosphorylates and inactivates GSK3β. Whether other pathways are involved and how they might fit into the picture remains to be seen.