|Home | About | Journals | Submit | Contact Us | Français|
Recent evidence suggests that glycogen synthase kinase 3s (GSK 3s) and their upstream and downstream regulators have key roles in many fundamental processes during neurodevelopment. Disruption of GSK3 signaling adversely affects brain development and is associated with several neurodevelopmental disorders. Here, we discuss the mechanisms by which GSK3 activity is regulated in the nervous system and provide an overview of the recent advances in understanding how GSK3 signaling controls neurogenesis, neuronal polarization, and axon growth during brain development. These studies suggest GSK3 as a major signaling node that mediate multiple signaling molecules regulating neurodevelopment, such as DISC1, Par3/6, and Wnts.
Glycogen synthase kinase 3s are serine/threonine kinases that were originally identified as key regulatory enzymes in glucose metabolism1, 2. There are two isoforms, GSK3α and GSK3β, encoded by separate genes, which are overall 85% homologous to each other, with 95% identity in the kinase domains3. In rodents and humans GSK3β2, an alternative splice variant of GSK3β has been reported, which contains a 13 amino acid insertion in an external loop near the catalytic domain4. In contrast to the ubiquitously expressed GSK3β1, GSK3β2 is expressed specifically in the nervous system, with the highest levels found during development4. Interestingly, recent studies suggest that GSK3β2 plays a specific part in neuronal morphogenesis in vitro5, 6, and it remains to be determined whether different isoforms have specific functions in vivo during brain development. Since their discovery, GSK3s have been shown to mediate various signaling pathways, among which the growth factor and Wnt signaling pathways are the most studied7. Consistent with the roles of growth factors and Wnt proteins in the nervous system, especially during neurodevelopment, emerging evidence points to GSK3s as key regulators in multiple neurodevelopmental processes, including neurogenesis, neuronal migration, neuronal polarization and axon growth and guidance.
How do GSK3s regulate such a wide spectrum of developmental events? The answer may lie in the broad range of GSK3 substrates. Among GSK3 substrates are many transcription factors, such as cAMP response element-binding protein (CREB)8, nuclear factor of activated T-cells (NFAT)9, 10, neurogenin 211, Smad112, c-Jun13, and β-catenin14, all of which play important parts in the regulation of gene expression throughout neurodevelopment. GSK3s regulate these transcription factors by controlling their protein levels, DNA binding activities, and/or nuclear localization. In addition to gene expression, cell morphogenesis requires reorganization of the cytoskeleton, especially microtubules. GSK3s regulate the activity of several microtubule-associated proteins (MAPs)15, and so might control mitotic spindle reorganization during cell division, coordinated movement of the leading process and soma during neuronal migration, and directed growth cone advancement during axon growth and guidance, all of which require coordinated control of microtubule dynamics. Changes in GSK3 activity have been associated with many psychiatric and neurodegenerative diseases, such as Alzheimer’s disease, schizophrenia, and autism spectrum disorders, and it has become increasingly apparent that GSK3 might be a common therapeutic target for different classes of psychiatric drugs16, 17. Indeed, lithium, a direct inhibitor of GSK318, has been used in humans as a mood stabilizer for over 50 years19. The hypothesis that disturbances of brain development play a part in the etiology of these disorders is further supported by the fact that several genetic susceptibility factors for psychiatric disorders have key roles in neurodevelopment. Intriguingly, many of the genes associated with schizophrenia encode proteins involved in GSK3 signaling, such as Disrupted-In-Schizophrenia 1 (DISC1), Neuregulin 1, and Frizzled-320. Genes associated with autism spectrum disorders also encode proteins that are involved in the modulation of or that are affected by GSK3 activity, including phosphatase and tensin homolog deleted on chromosome 10 (PTEN)21, DISC122, serotonin23, tuberous sclerosis complex 1/2 (TSC1/2)24, and adenomatous polyposis coli (APC)25, 26. Therefore, a better understanding of the role of GSK3 in neurodevelopment could provide insight into the etiology of these disorders and possibly open up the potential of a new library of therapeutic targets.
In this Review, we provide an overview of the involvement of GSK3 signaling in neurodevelopment, with a particular emphasis on neurogenesis, neuronal polarization, and axon growth. We will also discuss the potential crosstalk between GSK3 signaling and other pathways that are implicated in these developmental steps.
GSK3 is an unusual kinase, in the sense that it has a high basal activity in resting cells and is inactivated by upstream regulators in response to stimuli. Several signaling pathways regulate GSK3 activity (Fig. 1). Activation of the phosphatidylinositol 3-kinase (PI3K) pathway downstream of receptor tyrosine kinase signaling is thought to result in inactivation of GSK3 through phosphorylation of the N-terminal serine residue (S9 in GSK3β and S21 in GSK3α). It is widely accepted that AKT is the major mediator of this serine phosphorylation and subsequent inactivation of GSK3 (Fig. 1)27, 28. However, mutant mice (GSK3α-S21A/GSK3β-S9A double knock-in mice) with GSK3s that cannot be phosphorylated at their N-termini developed normally with no overt phenotype in the nervous system (although insulin-induced activation of glycogen synthase was disrupted in the muscle)29. This suggests that alternative pathways might regulate GSK3 activity, especially in the nervous system. Indeed, a recent study30 showed that p38 MAPK regulates the activity of GSK3β, but not GSK3α, by inducing an inhibitory phosphorylation on S389 (Fig. 1). However, p38 MAPK-mediated phosphorylation is insufficient to provide a regulatory mechanism of GSK3 in the nervous system because both isoforms of GSK3 are required for normal development31.
In the canonical Wnt pathway, Wnt-induced dissociation of GSK3 from its substrate β-catenin leads to the stabilization and activation of β-catenin. Upon Wnt stimulation, GSK3 is recruited to the membrane, where PPPSPXS motifs in the co-receptor low-density lipoprotein receptor-related protein (LRP) 5/6 associate with and inhibit GSK3 activity towards β-catenin (Fig. 1)32, 33. Regulation of GSK3 by LRP5/6 is achieved via protein-protein interaction, which sequesters GSK3 away from its substrate without affecting the kinase activity per se. Regulation of GSK3 activity towards a particular substrate rather than general inhibition of its kinase activity per se is also exemplified in a recent finding22, which suggested that by physically interacting with GSK3 DISC1 prevents GSK3 from phosphorylating β-catenin.
Both in response to Wnt34 and growth factors35 GSK3 activity is regulated by a conserved cell polarity pathway involving Cdc42-Par3/6-PKCς (Fig. 1). The regulation of GSK3 activity by the Par complex was first described in directed cell migration36. Specifically, localized activation of the small GTPase Cdc42 at the leading edge recruits the Par3/Par6/PKCς complex, which interacts with GSK3β, leading to its phosphorylation and inactivation36. Although GSK3 phosphorylation occurs as a consequence, PKCς is unable to directly phosphorylate GSK327 and a study using fibroblasts derived from GSK3α-S21A/GSK3β-S9A double knock-in mice37 demonstrated that GSK3 phosphorylation is not required for the Par3/6-mediated GSK3 inhibition. Instead, dishevelled (Dvl) and axin have been suggested to mediate GSK3 inhibition through this pathway by inducing a physical association between Dvl and PKCς during directed cell migration (Fig. 1). Interestingly, a recent study38 shows that Par3 directly binds PI3K and enhances its activity, providing an alternative explanation for the increased GSK3 phosphorylation downstream of Par3. Together, these lines of evidence suggest protein-protein interaction as a recurrent theme in the regulation of GSK3 activity, which in many cases does not require changes in the phosphorylation status of GSK3 (Box 1).
The most widely used method to establish GSK3 activity is to examine S9/S21 phosphorylation. However, results from GSK3α-S21A/GSK3β-S9A double knock-in mice29 raise the question of whether this serine phosphorylation of GSK3 is the major regulatory mechanism in the CNS. More importantly, it is becoming increasingly apparent that changes in GSK3 activity are not always accompanied by changes in their phosphorylation status. Thus, serine phosphorylation of GSK3 might not be used as the sole indicator of GSK3 activity.
The second approach is the in vitro kinase assay. Endogenous GSK3 can be immunoprecipitated from cell lysates to detect its activity towards its substrates. However, because the interaction of GSK3 with their regulators in cells might be transient and only a small percentage of total GSK3 in cells interacts with a particular regulator118, it is possible that the changes in GSK3 activity are not preserved in the in vitro kinase assay or are less likely to be detected.
The third approach is to examine the tyrosine phosphorylation in GSK3 (Y279 in GSK3α and Y216 in GSK3β), which facilitates its activity by promoting substrate accessibility119. Although there is evidence for signal-dependent regulation of tyrosine phosphorylation22, 120, 121, tyrosine phosphorylation of GSK3 is thought to occur through a post-translational and intramolecular autophosphorylation event119, 121, which in many cases is not dynamically regulated upon external stimuli.
The last approach is to examine changes in the phosphorylation status of known GSK3 substrates. Because GSK3 might not have equal access to all potential substrates in vivo owing to subcellular localization, this method allows us to detect more physiologically relevant changes in GSK3 activity. However, when this method is applied to assess GSK3 activation/inactivation in a pathway that does not have known GSK3 substrates, one should consider multiple potential substrates that can represent a diverse family of proteins, for instance, representing primed versus unprimed substrates with different subcellular localizations.
Recent evidence22, 39, 40 suggests that GSK3 signaling is essential for the coordination of progenitor proliferation and differentiation during brain development. Because the developmental process of mouse neocortex (Fig. 2) has been well-documented, we will use it as a model system to discuss the role of GSK3 signaling in neural development.
A recent study39 revealed a pivotal role of GSK3 signaling in the developing nervous system by selectively deleting both Gsk3α and Gsk3β in neural progenitors. The double knock-out mice had a significantly increased cortical surface area with a convoluted shape, due to an over-expansion of the neural progenitor cells. However, the cortex was thinner than that of control littermates, indicating a reduced number of neurons. Indeed, further analysis using various markers of progenitors, intermediate progenitor cells (IPCs) and post-mitotic neurons showed that deletion of GSK3s markedly enhanced the proliferation of progenitor cells while suppressing neuronal differentiation, indicating that spatiotemporal regulation of GSK3 activity is required for appropriate transition from the proliferative to the neurogenic phase during brain development.
Various signaling pathways have been identified to regulate neural progenitor proliferation, including Wnt41, sonic hedgehog (Shh)42, fibroblast growth factor (FGF)43, and Notch signaling pathway44. Intriguingly, GSK3 is implicated in the regulation of each of these pathways. In the canonical Wnt pathway, inhibition of GSK3 is crucial for the stabilization and nuclear translocation of β-catenin that leads to the subsequent activation of the T-cell factor (TCF) 4-dependent gene transcription. As discussed above, although the regulatory mechanism is not clear, GSK3 is a well-known downstream effector of PI3K, which provides an explanation for the disruption of FGF signaling in GSK3-deleted brains39. The mechanism by which Shh signaling was elevated in Gsk3-deleted brains39 is unclear, but a study in Drosophila45 suggests that GSK3 controls the stability of key transcriptional effectors (Ci/Gli proteins) in the Hedgehog pathway in a similar way as it regulates β-catenin in the Wnt pathway. There are contradicting results of how GSK3 regulates Notch signaling. In one study46 Notch protein was stabilized by GSK3-mediated phosphorylation46, whereas in another study the transcriptional activity of Notch was enhanced by inhibition of GSK347. However, the fact that Notch signaling was hyperactivated in Gsk3-deficient brains39 suggests that GSK3 negatively regulates Notch signaling in neural progenitors. Together, these lines of evidence suggest that GSK3 might function as a node molecule of multiple signaling pathways and thereby coordinate the proliferation and differentiation of neural progenitors.
The role of GSK3 signaling in neurogenesis is further supported by two recent studies22, 40, in which DISC1 or Par3 — upstream regulators of GSK3 — were manipulated in neural progenitors. Ectopic expression of either DISC1 or Par3 increased progenitor proliferation and suppressed neuronal differentiation, reminiscent of Gsk3 knock-out mice. One study22 showed that DISC1 directly interacts with GSK3, which seemed to prevent GSK3 from phosphorylating β-catenin and that over-expression of DISC1 led to the activation of canonical Wnt signaling. The other study40 did not directly test the involvement of GSK3 but showed that over-expression of Par3 led to the activation of Notch signaling, which was also observed in Gsk3 knock-out mice39. As Wnt and Notch signaling independently promote the proliferation of progenitors in Gsk3 knock-out mice39, it could be suggested that DISC1 and Par3 regulate GSK3 activity in the Wnt and the Notch pathway, respectively, allowing pathway-specific control of GSK3. These lines of evidence suggest that inhibition of GSK3 signaling, by either deleting Gsk3s39 or over-expressing their negative regulators22, 40, promotes the proliferation of progenitor cells while suppressing their differentiation into neurons.
Kim et al.39 found that phosphorylation levels of c-Myc and β-catenin, targets of GSK3, were increased at later stages in development when progenitor proliferation is subsiding and neuronal differentiation predominates. Because both c-Myc48 and β-catenin are pro-proliferation factors and phosphorylation by GSK3 leads to their degradation, this finding suggests that activation of GSK3 promotes neuronal differentiation. Supporting this notion, Disc1 knockdown in neural progenitors caused premature neuronal differentiation at the expense of the size of the progenitor pool22. When GSK3β signaling was examined, the levels of phosphorylated β-catenin and tyrosine-phosphorylated GSK3β, which is associated with increased kinase activity (see Box 1), were increased22. Moreover, it was shown that pharmacological inhibition of GSK3 prevented the progenitor proliferation defects induced by Disc1 knockdown22. These results suggest that the premature differentiation observed in DISC1-depleted cells might have been due to activation of GSK3 signaling, although future studies of Disc1 knock-out mice are needed to confirm this conclusion. Depletion of Par3 also increases the neuronal population with a concomitant decrease in the number of progenitor cells40. Future studies are required to determine if this defect is also attributed to GSK3 activation.
As mentioned above, GSK3 has high basal activity in resting cells, and signaling is usually initiated by GSK3 inactivation in response to extracellular factors. However, lysophosphatidic acid (LPA) appears to increase GSK3 activity above basal levels, and that this activation is in part mediated by the small GTPase RhoA49. Consistent with the idea that GSK3 activation promotes neuronal differentiation, treatment of ex vivo cultured cortical hemispheres with LPA promoted terminal mitosis of neural progenitors, leading to their differentiation into neurons50. Moreover, a recent study51 showed that Lfc, a guanine nucleotide exchange factor of RhoA, promotes neuronal differentiation51. Knocking down Lfc, which inhibits Rho activation, increased the pool of radial glial progenitors and prevented neurogenesis, whereas genetic silencing of Tctex-1, a negative regulator of Lfc, had the opposite effect51. Although these results are consistent with the hypothesis that activation and inactivation of GSK3 promote neuronal differentiation and progenitor proliferation, respectively, it remains to be determined if these effects are indeed mediated by regulation of GSK3 activity.
The conventional view is that GSK3 is constitutively active in resting cells. If this is the case, neural progenitors would undergo neuronal differentiation without any additional stimuli. However, a recent study52 showed that an extracellular cue, retinoic acid derived from forebrain meninges, is required for neuronal differentiation52. Interestingly, it has been shown that retinoic acid decreases S9 phosphorylation of GSK3β53 and increases GSK3 expression5, raising the intriguing possibility that retinoic acid signaling leads to the elevation of GSK3 activity above basal levels to control neurogenesis. Supporting this hypothesis, Par3 was significantly upregulated in neural progenitors in the absence of retinoic acid52. Collectively, these studies provide some indirect evidence that GSK3 activation might also be regulated by extracellular cues during neuronal differentiation. The most direct way to test if an increase in GSK3 activity promotes neuronal differentiation would be to over-express GSK3 in the developing brain. Unfortunately, the currently available GSK3 transgenic mice over-express GSK3β under the control of the Thy-1 promoter, which is specifically expressed in post-mitotic neurons54, rendering them unsuitable for examining the role of GSK3 in neurogenesis. A mouse model that over-expresses GSK3 specifically in neural progenitors is needed to test this idea.
Emerging evidence suggests that GSK3 regulates the stability of a wide range of proteins via activation of the ubiquitin-proteasome system (UPS)55, 56. Thus, one plausible mechanism by which GSK3 regulates neurogenesis is by controlling the levels of molecules involved in neurogenesis. Prominent candidates are transcriptional regulators, such as β-catenin in the Wnt pathway, Gli in the Shh pathway, and c-Myc in the FGF pathway39, 45, 48, all of which control progenitor proliferation through regulation of gene transcription. Supporting this hypothesis, during mitotic cell division GSK3-phosphorylated β-catenin, which is destined for degradation, is inherited by only one daughter cell57. Thus, it is possible that during asymmetrical division of radial glial cells, two daughter cells have different GSK3 activity owing to asymmetrical segregation of upstream GSK3 regulators (e.g. Par3). The daughter cell with lower GSK3 activity will accumulate β-catenin (and perhaps other pro-proliferative proteins) and thus remains a progenitor. Conversely, the daughter cell with higher GSK3 activity will degrade these proteins through activation of the UPS and differentiate into either a neuron or an IPC (Fig. 3).
In addition to proteins regulating gene transcription, GSK3 could also regulate other molecules that are important for controlling proliferation or differentiation. A recent study58 has shown that during asymmetrical division of radial glial progenitors, daughter cells inheriting the centrosome that contains the new centriole migrate away from the ventricular zone (VZ) and differentiate into neurons. By contrast, daughter cells inheriting the centrosome that contains the original, mature centriole retain the capacity to self-renew and remain in the VZ. Removal of Ninein, which is mainly localized in the mature centriole, disrupts the asymmetric segregation of the centrosome and promotes neuronal differentiation. Intriguingly, Ninein is a substrate of GSK359-61 and its level is regulated by the UPS62. Therefore, during asymmetrical division Ninein would be accumulated in the daughter cell with low GSK3 activity. As Ninein is a microtubule minus end-anchoring protein involved in the formation of astral microtubules (which anchor cells to their niches) daughter cells with Ninein in their centrosome are more likely to be anchored to the apical surface of VZ and inherit a progenitor fate. On the other hand, daughter cells with high GSK3 activity will have little Ninein in their centrosome, leading to their detachment from the VZ surface and becoming neurons or IPCs.
Another major class of GSK3 substrates consists of MAPs, which control several aspects of microtubule dynamics. APC is not only a component of the Wnt pathway that regulates β-catenin stability but also a MAP that belongs to a family of microtubule plus end-binding proteins (+TIPs)63. When GSK3 activity is inhibited, APC binds to microtubule plus end and there, it anchors the spindle microtubules to the kinetochore or the astral microtubules to the cell cortex, both of which are important for (symmetrical and asymmetrical) cell division. APC has a key role in asymmetrical stem cell division by anchoring one daughter cell in the specific niche so that it retains self-renewal capacity64. In the Wnt signaling pathway, APC interacts with GSK3 to degrade β-catenin, and, similar to Gsk3 deletion, loss of APC leads to stabilization of β-catenin and subsequent activation of TCF465. However, in contrast to Gsk3 knock-out, removal of Apc in progenitors results in reduced rather than increased proliferation, which is probably due to the role of APC in regulating microtubules during cell division.
In summary, GSK3 signaling is emerging as a key regulator of neurogenesis; through its role in the regulation of protein stability and microtubule assembly, it coordinates proliferation and differentiation (Fig. 3).
Studies in cultured hippocampal and cortical neurons have identified a number of molecules involved in neuronal polarization. These proteins, which include PI3K, AKT, GSK3, the small GTPases Rac1 and Cdc42, and the Par polarity complex66-76, might act in distinct neuronal populations at specific stages of embryonic development, but it is also possible that they form a network that controls neural development in a coordinated manner. In this regard, GSK3 is an appealing candidate ‘coordinator’ because it can respond to and integrate upstream signals, and downstream, it can phosphorylate various substrates involved in a wide array of cellular activities77. Below we discuss the evidence for a role of GSK3 signaling in neuronal polarization and explore the potential crosstalk of GSK3 signaling with other pathways involved in polarization (Fig. 4).
Several lines of evidence point to GSK3 as an essential regulator of neuronal polarity. Studies in hippocampal neurons show that the inactive form of GSK3β (phosphorylated at S9) is located at the tip of each neurite before polarization, but when neurons begin to polarize and one of these neurites develops into the axon, phospho-GSK3β becomes concentrated at the axonal tip, suggesting that maintaining local inactivation of GSK3β at the nascent axon is crucial for polarization69, 73, 76. Indeed, global inhibition of GSK3 by small-molecule inhibitors or knocking down Gsk3β induces the formation of multiple axons, whereas over-expressing GSK3β-S9A prevents axon formation69. Furthermore, local inhibition of GSK3 can convert dendritic processes into axons in already polarized neurons69, suggesting that localized inactivation of GSK3 is required for both the establishment and maintenance of neuronal polarity69, 76. To date, it is not clear how GSK3 inactivity becomes limited to one neurite during neuronal polarization.
Neuronal polarization requires major reorganization of the growth cone cytoskeleton. It has been proposed that a loose actin network at the tip of the future axon enables microtubules to selectively engorge into the axonal growth cone, forming the platform for subsequent axon elongation78-82. A recent study83 showed that microtubule stabilization in one neurite precedes axon formation, and that changing microtubule dynamics is sufficient to alter axon and dendrite specification83, pointing to the notion that regulation of local microtubule stability and dynamics has an instructive role in neuronal polarization84.
Several substrates of GSK3 are involved in neuronal polarization by regulating microtubule dynamics; these include collapsin response mediator protein 2 (CRMP2), APC, Tau, and microtubule-associated protein (MAP) 1b. Binding of CRMP2 to tubulin dimers, which is abolished by GSK3β phosphorylation76, promotes microtubule polymerization85. CRMP2 is enriched in the nascent axon and over-expression of CRMP2 is sufficient to induce the formation of multiple axons86. In hippocampal neurons, the microtubule +TIP protein APC localizes at the tip of immature neurites, and later becomes enriched in the growth cone of the nascent axon87. Similar to CRMP2, microtubule-binding activity of APC is abrogated by GSK3β phosphorylation88. Thus, GSK3β inactivation in the nascent axon promotes the association between APC and microtubule plus ends and thereby stabilizes the growing ends of axonal microtubules to support axon extension. As with CRMP2 and APC, Tau phosphorylation by GSK3β abolishes its binding to microtubules and thus impairs microtubule assembly89. GSK3 phosphorylation of MAP1b maintains microtubules in a dynamic state90. However, the lack of polarity defect in Tau/Map1b double knock-out mice suggest that they mainly function to regulate axon growth rather than neuronal polarization91.
Local activation of PI3K and accumulation of phosphatidylinositol-3,4,5-trisphosphate (PIP3) at the tip of the nascent axon is thought to act as a landmark for the induction of neuronal polarization. Several studies69, 73, 76 have proposed a model of neuronal polarization in which local activation of the PI3K/AKT pathway leads to the inactivation of GSK3β by inducing S9 phosphorylation. Supporting this model, the myristoylated, constitutively active form of Akt (Myr-AKT), but not the wild-type, induced the formation of multiple axons, which was partially blocked by over-expression of GSK3β-S9A. Moreover, ectopic expression of PTEN (a phosphatase that counteracts the actions of PI3K) prevented axon formation, an effect that was reversed by a GSK3 inhibitor69. Conversely, knocking down PTEN induced the formation of multiple axons, which was prevented by expression of GSK3β-S9A, leading the authors to conclude that GSK3β acts downstream of PI3K/AKT69. However, neurons from GSK3α-S21A/GSK3β-S9A double knock-in mice develop normal polarity in vivo and in vitro92. Importantly, these neurons still form multiple axons when treated with various GSK3 inhibitors. These results confirm that regulation of GSK3 activity is crucial for neuronal polarization, but suggest that there might be alternative mechanisms for GSK3 inactivation. Similarly, the finding that expression of Myr-AKT induces multiple axons suggests an alternative pathway downstream of AKT for regulating polarization.
The Par3/Par6/PKCς complex is required for axon-dendrite specification of hippocampal neurons. Before polarization, Par3 and Par6 are localized at the tips of all processes but later on they become selectively enriched in the nascent axon and developing growth cone74, 93, which is consistent with their roles as negative regulators of GSK3. Localized inhibition of GSK3 at the nascent axon is also required for the polarized transport and proper localization of Par373, suggesting a feedback regulation between Par3 and GSK3.
Similar to the aforementioned mechanism that controls directed cell migration involving the Par3/6-GSK3 pathway37, in hippocampal neurons Dvl becomes enriched at the tip of the axon in response to Wnt5a and induces axon formation via activation of PKCς in the Par3/Par6/PKCς complex94. In these neurons, Dvl over-expression leads to the production of multiple axons but did not affect GSK3β-S9 phosphorylation, and Dvl was still able to induce multiple axon formation in the presence of GSK3β-S9A94. Based on these results, authors concluded that this pathway might not involve GSK3β, but might employ microtubule affinity-regulating kinase (MARK) 2/Par1 as an alternative downstream effector to control microtubule dynamics94. In this model, activation of PKCς in response to Wnt5a/Dvl leads to the inhibitory phosphorylation of MARK2, which in turn causes dephosphorylation of MAPs such as Tau (Fig. 4). Consistent with this, a previous observation showed that knocking down Mark2 decreased Tau phosphorylation and induced the formation of multiple axons, whereas ectopic expression of MARK2 increased Tau phosphorylation and blocked axon growth68. On the other hand, based on the remarkable similarity of the players (Wnt5a, Dvl and PKCς) involved in neuronal polarization94 and cell migration37, it could be suggested that GSK3 inactivation has a role downstream of Wnt/Dvl/PKCς to control neuronal polarization as well. Interestingly, in Drosophila, MARK acts as a priming kinase, triggering subsequent Tau phosphorylation by GSK3β95, suggesting that inactivation of MARK2 and GSK3β at the axonal tip together reduce Tau phosphorylation downstream of Par3/Par6/PKCς complex. Further complicating the situation are the findings that GSK3β can both directly activate96 and inhibit97 MARK2 via phosphorylation. The exact roles of MARK2 and GSK3β and their interplay during neuronal polarization remain to be fully defined (Fig. 4).
The SAD kinases, SAD-A and SAD-B, which are expressed in the developing nervous system, contain a kinase domain related to that of C. elegans PAR-1 and its vertebrate orthologs, MARK1–MARK470. In the brains of Sad-A, Sad-B double knock-out (Sad-null) mice, segregation of neuronal subtypes to sublayers within the cortical plate was disordered and major axonal tracts were missing, suggesting that SAD kinases are required for neural development in vivo. When cultured in vitro, hippocampal neurons from Sad-null mice develop processes of equivalent lengths, each of which contains both dephosphorylated Tau (Tau1), which is often used as an axonal marker, and MAP2, which is normally enriched in dendrites70. In Sad-null brains, Tau1 was found not only in the axon-rich intermediate zone but also in the cortical plate, where dendrites elaborate70, suggesting that SAD kinases play a pivotal role in neuronal polarization.
Similarly, neurons lacking Lkb1, the ortholog of Par4 in C. elegans, develop neurites of equivalent length containing both Tau1 and MAP267, 72. In the cortices of Lkb1 knock-out mice, levels of phosphorylated SAD-A and SAD-B were greatly reduced, and biochemical experiments revealed that LKB1 phosphorylated SAD kinases67. Phosphorylated forms of LKB1 and SAD kinases were concentrated in axons, and LKB1-mediated SAD phosphorylation correlated with Tau phosphorylation67, 72. Thus, once LKB1 is activated in axons, it activates SAD kinases, which in turn phosphorylate MAPs such as Tau. Although it seems somewhat contradictory that SAD kinases are active in axons, which seems to have low levels of phosphorylated Tau98, these studies66, 67, 70, 72 provide convincing evidence that LKB1 and SAD kinases are required for neuronal polarization in vivo. Interestingly, when Lkb1 was knocked down by in utero electroporation, S9-phosphorylated GSK3β was not detectable at the tip of any neurites in cortical neurons despite the uniform distribution of GSK3β66, suggesting that GSK3β could be a downstream target of LKB1 (Fig. 4). To date, it remains unclear if the defects in neuronal polarization observed in Lkb1-depleted neurons involve GSK3 signaling. It is, however, interesting to note that a study in Xenopus showed that LKB1/XEEK1 regulates GSK3 activity and physically associates with GSK3β and aPKC in vivo99. Future studies should examine whether LKB1 regulates axonal specification by inhibiting GSK3β in addition to activating SAD kinases.
Mutations in Tsc1 and Tsc2 cause tuberous sclerosis, a disease characterized by tumor predisposition and neurological abnormalities including epilepsy, mental retardation, and autism100. TSC1/TSC2 act as a heterodimer that negatively regulates mammalian target of rapamycin (mTOR) signaling101. Upon activation of the PI3K/AKT pathway, AKT induces inhibitory phosphorylation of TSC2, by which its GTPase activating protein (GAP) activity towards Rheb is reduced, subsequently leading to the activation of mTOR kinase (Fig. 4). mTOR in turn phosphorylates translational regulators, such as ribosomal S6 kinase (S6K) and eukaryotic translation initiation factor 4E-binding protein (4EBP), and thereby induces protein synthesis102-104. It has been suggested that inactivation of the TSC1/TSC2 complex and subsequent mTOR activation in a single neurite regulates axon formation by inducing the translation of polarity proteins such as SAD kinases105. Supporting this notion, phosphorylated forms of S6K and 4EBP were enriched in the axon during polarization, suggesting local activation of mTOR signaling105, 106. Together with the aforementioned elevation of SAD kinase activity by LKB167, the control of SAD protein levels by TSC/mTOR signaling might play a role in axon specification. Because GSK3α-S21A/GSK3β-S9A double knock-in mice show normal neuronal polarization92, AKT phosphorylation of TSC2 has been suggested as an alternative mechanism to convey PI3K/AKT signaling.
Notably, a direct interaction between TSC/mTOR signaling and GSK3 has been reported in the regulation of cell growth. Inoki et al.107 showed that inhibition of GSK3 is sufficient to activate mTOR-dependent protein translation through regulation of TSC2 phosphorylation. It is thus plausible that suppression of GSK3 activity in the axonal growth cone has a role in the activation of local translational machinery for the synthesis of key proteins for axonogenesis. Interestingly, mTOR activation in the axon leads to local translation of Crmp2 and Tau, both of which are GSK3 substrates106. It will be of particular interest to investigate whether the suppression of GSK3 activity at the tip of the axon controls neuronal development by regulating local translation of polarity genes, including its own substrates, in addition to directly controlling microtubule dynamics by regulating MAP phosphorylation.
Local inhibition of GSK3 in the axon indicates higher GSK3 activity in dendrites. Does active GSK3 in dendrites play any role in polarization? Local protein degradation by activation of the UPS is an appealing mechanism to control asymmetrical accumulation of polarity proteins. Inhibition of the UPS with MG132 or lactacystin before or after the establishment of neuronal polarity leads to the formation of multiple axons108, suggesting that UPS activation is required for the establishment and maintenance of neuronal polarity. Interestingly, AKT, Par3 and aPKC are ubiquitinated at early stages of polarization before they become accumulated at axonal tips108. UPS inhibition restores the presence of AKT in all neurites, which is accompanied by multiple axon formation, suggesting that AKT is a target of the UPS for local degradation in dendrites108. UPS inhibition does not seem to alter the protein level of GSK3β, but an increase in GSK3β-S9 phosphorylation was observed108, suggesting changes in its activity. A recent study55 indicates that GSK3β regulates proteolytic degradation of a much broader range of proteins than was previously appreciated55. It will be interesting to investigate whether GSK3 activation in dendrites is involved in regulation of local protein degradation that contributes to redistribution of key molecules during polarization.
It is well accepted that coordinated regulation of local axon assembly at the growth cone and gene transcription in the neuronal soma is required for efficient axon growth109. This section will discuss evidence for GSK3 signaling in those processes.
Local inhibition of GSK3 signaling is essential for promoting microtubule assembly at the growth cone. However, in seemingly discrepant studies, inhibition of GSK3 blocked axon growth. To reconcile this controversy, Kim et al31 suggested a model in which inhibition of GSK3 can both enhance and prevent axon growth depending on the substrates involved. An interesting feature of GSK3 is that, many substrates of GSK3s require phosphorylation by a distinct kinase — an event known as priming — before they can be phosphorylated by GSK3s (primed substrates), although GSK3s can also phosphorylate some substrates without priming (unprimed substrates). Among GSK3 substrates that regulate microtubule assembly at the growth cone, APC and CRMP2 are primed substrates, and GSK3 phosphorylation abrogates their microtubule-binding affinity76, 88. Consistent with local inhibition of GSK3 in the distal axon, dephosphorylated forms of CRMP2 and APC are enriched in the growth cone, promoting axon formation and mediating neurotrophin-induced axon growth76, 110. Inhibition of GSK3 activity specifically towards primed (but not unprimed) substrates (using a GSK3β-R96A construct)28, 111 results in reduced CRMP2 phosphorylation and increased axon outgrowth31. By contrast, MAP1b is an unprimed substrate, which can be phosphorylated by GSK3 on S1260 and T1265 without priming90. Phosphorylation of MAP1b at these sites renders microtubules more dynamic, so that they can efficiently probe the intracellular space and respond to extracellular signals, activities that are essential for axon growth15, 112. Therefore, to ensure efficient axon growth, GSK3 activity should be precisely controlled, so that its activity towards one subset of substrates is specifically blocked while its activity towards others is preserved (Fig. 5). Indeed, in growing axons phosphorylated MAP1b is enriched at the distal ends of axons90, where dephosphorylated forms of APC and CRMP2 are concentrated.
These findings have led to a working model in which preferential suppression of GSK3 activity towards primed substrates promotes axon growth, whereas inhibition towards unprimed substrates blocks axon growth (Fig. 5)31. Identification of additional GSK3 substrates enriched in the growth cone will help to test this hypothesis and better understand how GSK3 signaling controls local axon assembly. Moreover, further studies are required to elucidate the molecular mechanisms by which GSK3 activity towards different substrates can be differentially regulated.
In addition to stimulating the tips of axons for cytoskeletal regulation, GSK3 has been implicated in the transcriptional control of axon growth via regulation of β-catenin and NFAT transcription factors. In the canonical Wnt pathway, Wnt3a induces axon growth from developing sensory neurons via accumulation of β-catenin and subsequent activation of TCF4 (Fig. 6)113, providing evidence for GSK3/β-catenin signaling in axon growth. GSK3 could also play a part in the transcriptional control of neurotrophin-induced axon growth by phosphorylating CREB and NFAT proteins. CREB is a well-established transcription factor downstream of neurotrophins, and Creb null mice display impaired axon growth from sensory and sympathetic ganglion neurons114. Neurotrophins (and other stimuli) lead to the activation of CREB kinases, which phosphorylate CREB on the transcriptional regulatory site S133, thereby recruiting the transcriptional coactivator CREB-binding protein (CBP) and so promoting the assembly of basal transcriptional complex115. In developing neurons, the DNA binding activity of CREB is tightly controlled by extracellular stimuli via nitric oxide-dependent signaling, as opposed to the classical model in which CREB is constitutively bound to promoter sequences116. Given that CREB phosphorylation by GSK3 abrogates its DNA binding activity8, it is plausible that GSK3 and nitric oxide signaling work in concert to control the duration and intensity of CREB-dependent transcription (Fig. 6). Neurotrophins and netrins also induce transcription of genes that are essential for axon growth by triggering Ca2+/calcineurin-dependent nuclear translocation of NFAT, and mice lacking calcineurin/NFAT signaling display profound defects in sensory axon projection and commissural axon growth117. Because NFAT phosphorylation by GSK3 inhibits its DNA binding activity10 and is required for nuclear export9, GSK3 is likely to have a role in NFAT-mediated gene transcription during axon growth (Fig. 6). Given that there are many other transcription factors modulated by GSK3, additional evidence for the transcriptional control of axon growth by GSK3 signaling awaits to be determined.
An algorithm that integrates consensus sequence motifs with contextual information has predicted GSK3 to be one of the kinases with a high number of substrates77. Although further studies are required to validate the physiological relevance of this prediction, the large number of putative substrates suggests that GSK3 could affect a broad range of cellular activities (Fig. 7). Indeed, recent studies implicate GSK3 in many fundamental processes during neural development, and a number of neurological disorders are associated with deficits in GSK3 signaling. It is thus important to understand GSK3 signaling, which requires, first, a re-evaluation of the involvement of GSK3 based on more reliable methods to detect GSK activation/inactivation, and second, the identification of additional in vivo GSK3 substrates.
The emerging picture is that GSK3 activity is controlled mainly by protein protein–interactions. Considering its many substrates, regulation by a complex and dynamic interaction network serves as an attractive model because it allows both pathway-specific control of GSK3 activity via signal-dependent formation of protein complexes and precise regulation of GSK activity towards particular substrates. It remains to be determined whether multiprotein complexes, as exemplified in the Wnt signaling, are a widespread mechanism for controlling the specificity of GSK3 actions. A distinguishable characteristic of GSK3 is the requirement of a priming event, which involves activation of another kinase. Thus, in combination with the signaling that directly regulates GSK3 activity, signaling that regulates the phosphorylation status of its substrates allows GSK signaling to achieve further specificity.
With respect to GSK3 regulation, studies so far have focused on the inhibitory mechanism. However, it is becoming increasingly apparent that activation of GSK3 comprises an important aspect of the regulatory scheme, suggesting that a delicate balance between activation and inactivation is required to shape GSK3 signaling. The data so far support the view that GSK3 activity is subjected to multiple layers of sophisticated regulation to ensure that GSK3 encounters its substrates in the right place at the right time. It will be of great interest to investigate how the complexity of GSK3 signaling is orchestrated in vivo to exert exquisite control over a plethora of cellular processes during neural development.
The authors were supported by grants from the Whitehall Foundation (to F.Q.Z), the Basil O’Connor Starter Scholar Research Award from the March of Dimes foundation (to F.Q.Z), and the NIH (1R01NS064288 to F.Q.Z). We apologize to authors whose work we were unable to include owing to space constraints.
Eun-Mi Hur received her Ph.D. in cell biology from Pohang University of Science and Technology, South Korea. After completing her Ph.D. under the supervision of Dr. Kyong-Tai Kim on receptor-mediated signal transduction and crosstalk, she joined the laboratory of Feng-Quan Zhou at Johns Hopkins University School of Medicine, where she studies the role of GSK3 signaling in neural development and axon regeneration. She is the recipient of the postdoctoral fellowship from the Christopher and Dana Reeve Foundation.
Feng-Quan Zhou is an Assistant Professor in the Department of Orthopaedic Surgery and The Soloman H. Snyder Department of Neuroscience at Johns Hopkins University School of Medicine. He received his Ph.D. in Anatomy and Cell Biology from State University of New York at Buffalo (mentored by Christopher Cohan), and postdoctoral training at University of North Carolina, Chapel Hill (mentored by William Snider). His current research focuses on the role of GSK3 signaling in neural development and the molecular mechanisms regulating axon regeneration. He is the recipient of the Basil O’Connor Starter Scholar Research Award.