An essential step during the development of hippocampal neurons is the polarised outgrowth of a single axon. Recently, it has been suggested that inhibition of glycogen synthase kinase-3β (GSK-3β) via Akt/PKB-dependent phosphorylation of Ser9, specifically at the tip of the presumptive axon, is required for selective axonal outgrowth. We now report that, by using neurons from double knock-in mice in which Ser9 and Ser21 of the two GSK-3β isoforms have been replaced by Ala, polarity develops independently of phosphorylation at these sites. Nevertheless, global inhibition of GSK-3β disturbs polarity development by leading to the formation of multiple axon-like processes in both control and knock-in neurons. This unpolarised outgrowth is accompanied by the symmetric delivery of membrane components to all neurites. Finally, the adenomatous polyposis coli (APC) protein accumulates at the tip of one neurite before and during axon elongation, but global inhibition of GSK-3β leads to APC protein accumulation in all neurites. We conclude that GSK-3β inhibition promotes the development of neuronal polarity, but that this is not mediated by Akt/PKB-dependent phosphorylation.
The establishment of axonal-dendritic polarity is a key event in the development of the nervous system. It has been extensively studied using isolated hippocampal neurons, which polarise to form a single axon from three to five preformed minor neurites (Dotti et al., 1988). A recent report suggested that polarised centrosomal-based activities at the immediate post-mitotic stage are necessary and sufficient to define the position of the first-formed neurite and its subsequent axonal fate (de Anda et al., 2005). However, since axonal growth does not begin until all neurites have formed, this leaves open the question of what distinguishes the first and subsequent neurites. An early manifestation of axon formation is a selective bulk-flow of organelles and trans-Golgi-network (TGN)-derived vesicles before elongation can be observed morphologically (Bradke and Dotti, 1997), and the development of a more dynamic actin cytoskeleton in the growth cone of the presumptive axon (Bradke and Dotti, 1999). However, the spatially and temporally restricted molecular events that orchestrate these changes are poorly understood.
The localised inhibition of the kinase glycogen synthase kinase-3 (GSK-3) isoform β (GSK-3β) has been reported to play a key role in the decision to form an axon (Jiang et al., 2005; Shi et al., 2004; Yoshimura et al., 2005). GSK-3 has many potential substrates including several microtubule-binding proteins, such as the adenomatous polyposis coli (APC) tumour suppressor protein, Tau, Crmp-2 and MAP1B. Alterations in GSK-3 activity could, therefore, have multiple effects on microtubule dynamics. Phosphorylation of Crmp-2 by GSK-3, for example, reduces its microtubule-binding activity (Yoshimura et al., 2005), whereas GSK-3-mediated phosphorylation of APC inhibits its ability to bind microtubules (Zumbrunn et al., 2001). Interestingly, in migrating astrocytes, the polarised elongation of microtubules is mediated by inhibition of GSK-3 leading to the association of APC with microtubule plus-ends (Etienne-Manneville and Hall, 2003).
Several groups have reported that inactivation of GSK-3β in neurons occurs specifically at the tip of the axon and that this is mediated through a PI 3-kinase and the Akt/PKB pathway, culminating in phosphorylation of GSK-3β at Ser9 (Jiang et al., 2005; Shi et al., 2004). By contrast, we show here that, by using neurons derived from knock-in mice whose two isoforms of GSK-3 (GSK-3α and GSK-3β) have been rendered non-phosphorylatable by Akt/PKB (McManus et al., 2005), GSK-3 inhibition, which leads to polarised axonal outgrowth, is regulated independently of phosphorylation at Ser9 or Ser21.
GSK-3 is required for the establishment of polarised axonal growth
After plating, freshly dissociated hippocampal neurons initially form 3-5 buds or lamellipodia (stage 1; for definition of stages in the development of polarity of hippocampal neurons see Dotti et al., 1998). Highly dynamic `minor neurites' emerge from these buds within 24 hours (stage 2). By 48 hours, 70% of neurons have developed a clear, polarised phenotype with a single, long process (stage 3; Fig. 1A,C) that will become the axon. To examine the role of GSK-3 in the establishment of neuronal polarity, we treated hippocampal neurons with different small-molecule inhibitors. Addition of either of the ATP-competitive inhibitors SB216763 or SB415286 during a 4-48 hour time period induced the symmetric outgrowth (Fig. 1A,D) of several Tau-1-positive axon-like processes, similar to what has been reported by others (Jiang et al., 2005; Yoshimura et al., 2005). These inhibitors stabilise cytosolic β-catenin, demonstrating efficient GSK-3 inhibition (Fig. 1B). We have also used two structurally distinct inhibitors of GSK-3 that have been reported to be even more specific and potent towards GSK-3, CHIR 99021 and AR-A014418 (Cohen and Goedert, 2004). Their effect in compromising polarity was similar to the SB inhibitors (Fig. 1A,C). This `depolarizing' effect of GSK-3 inhibition was most effective at the transition from stage 2 to 3 (supplementary material Fig. S1). The supernumerary long neurites have clear axonal properties because they each contain the marker proteins Tau-1 and plasma membrane ganglioside sialidase (PMGS, Fig. 1D) (Da Silva et al., 2005).
Regulation of GSK-3 activity during polarity establishment
It has been reported that inhibition of GSK-3β associated with neuronal polarisation is regulated by phosphorylation at its serine residue 9 by the kinase Akt/PKB. However, localised accumulation of GSK-3β phosphorylated at Ser9 (GSK-3β-P at Ser9) was described only after the axon had formed raising the possibility that this is a consequence rather than a cause of polarisation (Jiang et al., 2005). To investigate this further, we made use of an antibody specific to phosphorylated GSK-3β (anti-GSK-3β-P antibody). The specificity of this antibody was first confirmed using neurons isolated from knock-in mice in which the Akt/PKB Ser phosphorylation sites of both GSK-3β alleles had been replaced with Ala (GSK-3β21A/21A/9A/9A, supplementary material Fig. S2). Using this antibody with wild-type neurons and normalising the immunofluorescence signal to the signal obtained from a purely cytosolic protein, green fluorescent protein (GFP) or a dye staining unspecifically all proteins, dichlorotriazinylaminofluorescein (DTAF), revealed no significant accumulation of either GSK-3β-P at Ser9 or total GSK-3β in neurites at stage 2, or in the axon at stage 3 (Fig. 2A,B).
To assess whether phosphorylation of GSK-3 is at all required for axon formation, we used hippocampal neurons isolated from double knock-in mice, in which the Akt/PKB serine phosphorylation sites of both alleles of both GSK-3β and GSK-3α had been replaced with alanine (GSK3α/β21A/21A/9A/9A), and from littermate mice (GSK3α/β+/+/+/+) (McManus et al., 2005). The levels of GSK-3α and GSK-3β protein were similar in the two types of neurons but, as expected, there was no detectable GSK-3 phosphorylation in neurons from the knock-in mouse using the anti-GSK-3 α and β-P antibodies (Fig. 3A). As shown in Fig. 3B,C, the development of neurons from the double knock-in neurons was indistinguishable from that of wild-type neurons. After 48 hours in culture, fully polarised neurons with only a single Tau-1-positive axon and minor neurites of equal length developed in both cases (Fig. 3B,C,H). During 48-70 hours in culture, the percentage of polarised stage-3 neurons increased similarly in both types of neurons (Fig. 3G). Importantly, an in vivo analysis of knock-in neurons showed axonal and dendritic projections indistinguishable from wild-type littermates (Fig. 3D). Thus, we conclude that Ser9 or Ser21 phosphorylation of GSK-3 isoforms is not required for the establishment of a polarised neuron in vivo or in vitro.
To determine whether GSK-3 inhibition can still induce multiple axon-like projections in knock-in mice, we made use of the GSK-3 inhibitors. After 70 hours in the presence of SB216763, about one third of wild-type neurons and knock-in neurons developed multiple, axon-like processes (Fig. 3E-G).
GSK-3β accumulates in the Golgi region and is involved in the regulation of polarised traffic
Since we (Fig. 2B) and others (Jiang et al., 2005) could not find any specific accumulation of GSK-3β in the axonal tip of stage-3 neurons, or in a single neurite of stage-2 neurons, we analysed its intracellular localisation. At the light-microscopy level, GSK-3β is enriched in the Golgi-centrosome region of hippocampal neurons (Fig. 4A). Even after extraction of live cells with Triton X-100 (Fig. 4B) or saponin (data not shown) GSK-3β remains associated with the Golgi region, suggesting that it is membrane bound (supplementary material Fig. S3B). This is supported by biochemical experiments, in which GSK-3β was found in the same fraction as a Golgi marker in a continuous sucrose gradient (supplementary material Fig. S3A) and in the membrane pellet (Fig. 4C). This prominent localisation of GSK-3β raises the possibility that GSK-3 is involved in regulating polarised traffic to the axon. Indeed, GSK-3 inhibition leads to symmetric outgrowth from all neurites at a similar rate (22 μm/24 hours) (Fig. 4D), which is around three times slower than normal axonal outgrowth (69 μm/24 hours), but three times faster than that of minor neurites (8 μm/24 hours) in control neurons. The sum of overall neurite length, therefore, remains unchanged (supplementary material Fig. S3C).
To determine whether GSK-3 controls polarised, axonal membrane transport, we analysed the level of membrane trafficking along a selected neurite segment by using phase-contrast microscopy (supplementary material Fig. S4 and Movies 1-4) as described by Bradke and Dotti (Bradke and Dotti, 1997). After GSK-3 inhibition, membrane traffic - which is normally directed preferentially to axons - becomes symmetrically distributed towards all neurites (Fig. 4E). Moreover, new membrane carriers exiting from the TGN (which are concentrated in the axon of control cells) were homogeneously distributed in all processes after treatment with a GSK-3 inhibitor (supplementary material Fig. S3D). This was not due to a dispersal of the Golgi itself, which stayed intact in treated neurons (data not shown). The overall velocity of anterograde and retrograde membrane transport was not changed by GSK-3 inhibition - as can be seen by the overlapping peaks of membrane carrier velocities in the frequency-distribution diagrams in Fig. 4F - but only the polarity of the transport, which means the number of carriers travelling towards the axon (Fig. 4E) with respect to minor neurites.
Finally, ablation experiments (Goslin and Banker, 1989) suggest that activities at the axonal tip are not required for the maintenance of axonal outgrowth. Thus, when axons of stage-2+ or stage-3 neurons are cut such that the axonal stump remains longer than the remaining neurites, 81% of the axons re-grow (Fig. 4G).
APC accumulation at neurite tips is promoted by GSK-3 inhibition
The ability of APC to bind and stabilise MTs is influenced by GSK-3 activity (Zumbrunn et al., 2001) and APC has been localised to axon tips in polarised neurons (Zhou et al., 2004) and before axon formation (Votin et al., 2005). To examine the importance of APC in regulating polarity, we analysed its distribution in stage-2 neurons before any morphological signs of axon elongation could be detected. As shown in Fig. 5A, APC accumulation could be detected specifically in one of the minor neurites in stage 2 neurons (quantification: Fig. 5E). To see whether this early accumulation correlates with the future axon, we localised APC during the course of axon formation. In late stage-2+ neurons, APC strongly accumulates in presumptive axonal tips, which can be identified by a large growth cone (Bradke and Dotti, 1997). In 70% of stage-2+ neurons, the largest growth cone correlated with the highest APC accumulation in the tip (Fig. 5B). Moreover, in fully polarised (stage 3) neurons, APC was found at the axonal tip in 100% of neurons (Fig. 5C,E). The average normalised intensity in the axonal tip is five times higher than the average intensity of all the minor neurite tips and 2.5 times higher than the mostly intensely stained of the minor neurites (Fig. 5E). APC can also be found along the length of neurites but is strongly accumulated in the axonal tip, as demonstrated by measuring the profile of APC staining intensity along the axon and neurites (Fig. 5D).
If APC is involved in axon formation, then inhibition of GSK-3 should affect APC localisation. Indeed, the addition of the GSK-3 inhibitor SB216763 dramatically changes the distribution of APC. Concomitant with the induction of multiple axon-like extensions, we observe the accumulation of APC at the tips of all the long, axon-like processes after treatment with inhibitor (Fig. 5F-H). This effect is independent of GSK-3 phosphorylation, because the accumulation of APC at the axon tip can also be seen in neurons from GSK3α/β21A/21A/9A/9A knock-in mice (Fig. 5I). We did not observe a similar selective axonal accumulation for other plus-end microtubule-binding proteins, including EB1 (Fig. 5J), CLIP170 or CLIP115 (data not shown). When axons are severed, as in Fig. 4G, APC accumulated again in the axonal tip during re-growth in the majority of neurons (>80%).
We show here that, GSK-3 inhibition in hippocampal neurons leads to the symmetric outgrowth of multiple axon-like processes (Fig. 1A,C,D). The axon-like nature of these processes was evident by the accumulation of several marker proteins such as dephosphorylated Tau (Tau-1) and PMGS (Da Silva et al., 2005), (Fig. 1D). Underlying this symmetric neurite outgrowth is a symmetric trafficking of membrane carriers (Fig. 4E, supplementary material Fig. S3D). However, since the net outgrowth of all neurites stays unchanged, we suggest that GSK-3 activity does not regulate the levels of membrane trafficking but rather its directionality.
Although the mechanism linking GSK-3 inhibition to axonal outgrowth is not clear, the protein APC seems to play an important role. This plus-end microtubule-binding protein shows polarised accumulation at the tip of one neurite even in stage-2 neurons and preceding fast axonal outgrowth (Fig. 5A,E) (Votin et al., 2005). That this early accumulation of APC depends on GSK-3 inhibition is suggested by the observation that global GSK-3 inhibition leads to the appearance of APC at the tips of all neurites (Fig. 5F-H). A role for APC in axon outgrowth has been reported by others (Shi et al., 2004) and, furthermore, a dominant-negative mutant of APC prevents axon outgrowth. This suggests that APC is, at least in part, responsible for the GSK-3 effect. The localised accumulation of APC in the presumptive axon could act by conferring early asymmetry on microtubule dynamics, because APC binds and stabilises microtubules and can promote their capture by membrane-associated complexes (Zumbrunn et al., 2001). One consequence of GSK-3 inhibition could be a change in microtubule-mediated membrane trafficking, an event known to be crucial during the course of polarisation (Bradke and Dotti, 1997; Martinez-Arca et al., 2001; Tang, 2001). In fact, we have found that inhibition of GSK-3 activity results in equal flow of membrane traffic (Fig. 4E) and the equal distribution of TGN-derived vesicles in all neurites (supplementary Fig. S3D), rather than polarised flow to a single axon. Interestingly, a large pool of GSK-3 is localised in the centrosome/Golgi area (Fig. 4A, supplementary Fig. S3A), but whether this pool of GSK-3 facilitates polarised trafficking or whether polarised trafficking is a consequence of polarity establishment is not clear.
GSK-3 activity can be regulated in many ways. In the Wnt signalling pathway, GSK-3 inhibition is essential for stabilisation of β-catenin and is thought to occur through protein-protein interactions. Interestingly, it was shown that, Wnt signalling is involved in the regulation of the anterior posterior polarity of neurons in C. elegans (Hilliard and Bargmann, 2006). However, in contrast to our observations, disturbance of Wnt signalling in C. elegans leads to an inversion of the polarity axis rather than loss of polarity.
In the insulin pathway, GSK-3 is inactivated by phosphorylation of Ser9 (GSK-3β) or Ser21 (GSK-3α) (Doble and Woodgett, 2003). Recently, there have been two reports that GSK-3β-P at Ser9 accumulates in the axon of stage-3 hippocampal neurons (Jiang et al., 2005; Shi et al., 2004) or dorsal root ganglion neurons (Zhou et al., 2004). The authors concluded that this spatial accumulation was promoted by the local axon-specific activation of Akt/PKB (Jiang et al., 2005) or ILK (Zhou et al., 2004). Our data do not support this model: First, we could not detect any localised accumulation of total or phosphorylated GSK-3 either in the axonal tip or in neurites of stage-2 neurons (Fig. 2A,B). These differences could be owing to the experimental conditions or to the different way we have evaluated specific axonal accumulation. Second, and more importantly, we have found that neurons isolated from double knock-in mice, in which the Akt/PKB Ser phosphorylation sites of GSK-3α and GSK-3β had been replaced with Ala (McManus et al., 2005), develop normal polarity indistinguishable from wild type neurons (Fig. 3). Finally, phosphorylation of GSK-3β at Ser9 is also thought to be involved at later stages in neuronal development during axon outgrowth and guidance in peripheral neurons (Eickholt et al., 2002; Zhou et al., 2004). Although we have not directly investigated this for hippocampal neurons, the lack of any obvious neuronal defects in the knock-in mice suggests that phosphorylation is not a major mechanism of regulating GSK-3 during neuronal development in the CNS.
We conclude that, GSK-3 inhibition is an important signal in the establishment of neuronal polarity and GSK-3 mediates its effects, in part at least, through the polarised accumulation of APC. GSK-3 inhibition regulating these processes is not, however, initiated by phosphorylation through Akt/PKB.
Materials and Methods
Antibodies used were against APC (I. Näthke), EB1, GSK-3β (BD), Tau-1 (Chemicon), GSK-3α (Upstate), GSK-3β phosphorylated on Ser9, GSK-3α phosphorylated on Ser21 (Cell Signalling), α-tubulin (YOL1/34, Serotec), PMGS (J. Abad-Rodriguez), GFP (Molecular Probes), synaptobrevin (Synaptic Systems) and MAP-2 (Peninsula Laboratories). Inhibitors: SB216763 and SB415286 (Tocris), used at a final concentration of 20 μM or 40 μM. AR-A014418 and CHIR99021 were kindly provided by P. Cohen (U. Dundee, UK) and used at a concentration of 10 μM and 2 μM, respectively.
Hippocampal or cortical neurons from embryonic stage (E)18 or E19 embryos were prepared according to (Banker and Goslin, 1988). In brief, hippocampi were dissected and cells dissociated by trypsin (15 minutes, 37°C) and mechanical trituration using a fire polished Pasteur pipette. Neurons were plated immediately in dishes containing CO2 and temperature-equilibrated Dulbecco's modified Eagle's medium (DMEM) with 10% horse serum at a density of 2500 cells per cm2. After 4 hours incubation neurons were sufficiently attached and coverslips were placed inverted, separated by paraffin dots, onto astrocyte feeder-cell layers equilibrated in a serum-free complete medium (Brewer and Cotman, 1989). Hippocampal neurons were transfected before plating using the rat neuron nucleofector kit (Amaxa) according to the manufacturer's instructions.
Neurons were fixed for 15 minutes at room temperature in 4% paraformaldehyde (PFA), 4% sucrose, 2 mM MgCl2, 3 mM EGTA-PBS, or for 3 minutes in MeOH at -20°C, washed in PBS, quenched in 50 mM NH4Cl-PBS and permeabilised in 0.1% Triton X-100 in PBS. After blocking with 20% normal goat serum, the first antibody was added and incubated overnight at 4°C in 2% normal goat serum in PBS. The second antibody was added for 1 hour at room temperature and cells were embedded in mounting medium (Dako). To detect cytosolic proteins, 0.001% DTAF (Molecular Probes) was used. Fluorescence was monitored using conventional fluorescence microscopy or confocal microscopy (Biorad Radiance) and images were analysed by Image J.
For immunofluorescence analyses of embryonic brain sections, embryonic brains were fixed for 4-6 hours in 4% PFA and equilibrated in 30% sucrose. 50-μm to 100-μm coronal sections were cut with a vibratome. Staining was performed as above with the difference that the blocking was done overnight at 4°C on a shaker, in PBS containing 0.5% Triton X-100 and 3% normal goat serum, and the first antibody was incubated overnight in 0.3% Triton X-100 at 4°C.
Analysis of membrane traffic was performed on a heated stage in a closed aluminium chamber. Images were captured in 1-second to 2-second intervals for 100 seconds. Membrane traffic was analysed by measuring the number of membrane carriers flowing through a proximally (immediately at the neurite shaft) and a distally (approximately half way along the nerurite length) located neurite segment of 10 μm. Values for the mean were calculated and sorted in descending order. In control neurons, axonal traffic (which is always higher than in neurites) is represented in the first column. The speed of membrane carriers was measured using Image J by following single vesicles for 4-20 seconds. For each condition (Fig. 4F), 50 vesicles were tracked.
Axons were cut by rapidly drawing a microinjection needle over the glass surface perpendicular to the axon length using a microinjection device (Eppendorf).
Evaluation of protein accumulation at the axonal tip
To quantify the accumulation of proteins at neurite or axon tips, neurons were either nucleofected with GFP or cytoplasmic proteins visualised with DTAF (Schindelholz and Reber, 1999). Neurons were immunolabelled for the molecule of interest and images were taken for the respective molecule (GSK-3, APC) and in parallel for GFP or DTAF. Using Image J, the ratio of the two images was calculated, the background subtracted and the mean ratio intensities in the neurite tips measured. Owing to the different architecture and volume of the different neurites, GFP or DTAF often accumulate non-specifically in the axon, and thus our measurements reflect the true accumulation of the protein of interest. As a control, we used the ratio between p38 and DTAF. p38 is a soluble protein (Fig. 4C) and is not located in axonal tips. The ratio intensities were sorted according to their strength and then normalised. The graphs represent sorted means of normalised ratio intensities at neurite tips. Since every neuron measured contained at least three neurites, only the first three neurites were shown. The slope of the three values represents the strength of accumulation. No accumulation represents a slope of zero, 100% accumulation in one neurite represents a slope of 1.
Separation of cytosolic and membrane proteins
Neurons were harvested in hypo-osmotic buffer (25 mM Tris, 2 mM EDTA, 1 mM DDT, protease inhibitors; pH 7.0) and cells disrupted by dispensing them through a 22G needle. Nuclei were removed by centrifugation for 5 minutes at 4°C at 600 g and membranes were harvested by centrifugation at 100,000 g. Cytosolic proteins remained in the supernatant, the membrane pellet was solubilised in 0.1% SDS.
Neurons were lysed in 50 mM Hepes pH 7.4, 150 mM NaCl, 10% glycerol, 1%Triton X-100, protease inhibitors. Proteins were separated by SDS-PAGE, transferred to nitrocellulose membranes and western blotted. Blots were incubated overnight with the first antibody in 5% milk-TBST and then with horseradish peroxidase (HRP)-coupled secondary antibody and detected by the ECL reagent (Amersham).
We are very grateful to Carlos Dotti for important suggestions and comments on the manuscript and technical help from Etienne Cassin and Bianca Hellias. We also thank Dario Alessi for his help and comments on this work. We thank David Becker (UCL, London, UK) for supporting the performance of some experiments, and Inke Näthke (University of Dundee, UK), Philip Cohen (MRC Unit, Dundee, UK) and Jose Abad-Rodriguez (Fondazione Ottolenghi, Turin, Italy) for reagents. This work was supported by a programme grant from Cancer Research UK and by a DFG fellowship (A.G.).
Supplementary material available online at http://jcs.biologists.org/cgi/content/full/119/19/3927/DC1
↵‡ Present address: Cavalieri Ottolenghi Scientific Institute, University of Turin, Regione Gonzole 10, 10043 Orbassano, Turin, Italy
↵§ Present address: Memorial Sloan-Kettering Cancer Center, 1275 York Avenue, New York, NY 10021, USA
- Accepted July 5, 2006.
- © The Company of Biologists Limited 2006