Introduction

It is currently believed that the glycogen synthase kinase 3 (GSK3) family of intracellular kinases are regulators of the physiology of developing and adult central nervous system (CNS) neurons. In fact, absence or misregulation of GSK3 functions has been observed in several CNS diseases such as Alzheimer’s disease (e.g., Ref. [1]), mood disorders and schizophrenia (e.g., Refs. [2, 3]), and anxiety or behavioral abnormalities associated with serotonin deficiency [4, 5]. The relevance of this family of kinases is demonstrated by emerging roles in neuronal physiology identified in recent years (see Refs. [6, 7] for review). For example, GSK3 kinases control gene transcription, axonal transport, and cytoskeletal dynamics in growth cones (see Ref. [8] for a recent review). Indeed, GSK3 kinases serve as key molecules in the coordination of cytoskeletal elements by controlling microtubule dynamics and assembly via regulation of several microtubule-binding proteins (MBPs). Thus, changes in GSK3 participation in the control of the neuronal cytoskeleton have been associated with neurodegeneration. In addition, several studies have implicated the modulation of the GSK3 activity and the upstream or downstream regulators in axonal regeneration after lesion (e.g., Refs. [9–13]). However, the role of GSK3 in neuronal shape maintenance (as a regulator of neuronal polarization) and plasticity is unclear due to different reports with contradictory findings. In this review, we would like to present an overview of the published roles of GSK3 beta (GSK3β) signaling in neuronal growth and regeneration in order to clarify these discrepancies. We will focus on the most relevant data in both neuron development and axon regeneration after injury. In the review, we will use the term GSK3β to refer directly to the kinase and GSK3 when the reported effects are not clearly associated specifically with GSK3β. Due to the large amount of information concerning GSK3 family of kinases and their various roles in different processes, we refer the reader to the reviews by Beurel et al. [14] and Lee and Kim [15] for the particular roles of GSK3 kinases in the immune system and in glucose metabolism, respectively, and to the review by Coen and Goedert [16] for the therapeutic potential of GSK3 inhibitors in several diseases.

GSK3β, a Fine-Tuned Kinase of the GSK3 Family in Neural Tissue

The GSK3 proteins are serine/threonine (Ser/Thr) kinases originally identified as key enzymes in glycogen metabolism [17, 18]. Two isoforms, GSK3α and GSK3β, are encoded by different genes with 95 % identity in their kinase domains and 85 % sequence homology [19, 20]. GSK3α and GSK3β are expressed in the central nervous system and participate in several non-redundant functions [21]. GSK3 kinases are implicated in multiple processes during neural development and in adult stages including neurogenesis, neuroprotection, migration, neuronal polarization, and axonal growth and maintenance, as well as intracellular transport. Phosphorylation of most GSK3 substrates requires prior proline phosphorylation by a priming kinase [e.g. Casein kinase I-II, cyclin-dependent kinase 5 (Cdk5) or dual-specificity tyrosine phosphorylation-regulated kinase (DYRK)], but unprimed substrates have also been reported for some proteins (e.g., Tau, phosphorylated at Ser396/404) [22]. GSK3β is the most commonly expressed isoform in the nervous system [23] and its spliced variant, GSK3β2, is the most enriched isoform in neurons of humans and rodents [24]. GSK3β expression is widespread throughout all regions of the developing and adult brain, although it is greatest in the hippocampus, thalamus, cortex, and Purkinje cells of the cerebellum in the adult (e.g., Ref. [25]). At the cellular level, GSK3β is present throughout the cell body and processes (dendrites and axon) of post-mitotic neurons (e.g., Ref. [25]).

Neuronal GSK3β activity is under the control of numerous mechanisms and signaling pathways. In fact, GSK3β is regulated both positively and negatively by several kinases including protein kinase B (Akt), p38 mitogen-activated protein kinase (MAPK), and protein tyrosine phosphatase (PTPase; Fig. 1). Thus, whereas the phosphorylation of tyrosine 216 residue (Tyr216) leads to GSK3β activation, phosphorylation of Serine 9 residue (Ser9) reduces its activity [26, 27] (Fig. 1). In some cases, phosphorylation of Ser9 is the result of the combined action of two kinases (i.e., ERK1/2 and p90RSK [28]). In contrast to inhibitory regulation by Ser9 mainly associated in the literature with PKB/Akt activity [29], GSK3β activity can also be facilitated by auto-phosphorylation [30], by alterations in intracellular calcium levels [31] and by other tyrosine kinases like proline-rich tyrosine kinase 2 (Pyk2) through Tyr216 phosphorylation [32] (see Fig. 1 for a general overview of its regulation). In addition, the regulation of the GSK3β activity may also include protein–protein interaction (e.g., with Frat1) [33, 34].

Fig. 1
figure 1

A general overview of GSK3β regulation. GSK3β is subject to multiple regulatory mechanisms. Phosphorylation of Ser9 is probably the most important regulatory mechanism. Several kinases are capable of mediating this modification, including p70 S6 kinase, extracellular signal-regulated kinases (ERKs), p90Rsk (also called MAPKAP kinase-1), Akt (PKB), certain isoforms of protein kinase C (PKC), and cyclic AMP-dependent protein kinase (PKA). Stimulation of pGSK3β (Tyr216) could be mediated by alterations in intracellular calcium levels and a calcium-dependent tyrosine kinase, proline-rich tyrosine kinase 2 (PYK2), or Fyn, a member of the Src tyrosine family. pGSK3β (Tyr216) is also subject to the regulation of mitogen-activated protein kinase (MEK1/2), PTPase, and even autophosphorylation. Moreover, the protein phosphatases 1 and 2A (PP1 and PP2A) also regulate GSK3β phosphorylation levels. However, control of GSK3β activity is through its main targets (MAPs or transcription factors), finally leading to control of neurogenesis, neuronal polarization, and axonal outgrowth. See references in text

Full size image

GSK3β activity has been shown to be a key regulator in signaling pathways in neurons triggered by many extracellular molecules such as Wnt, neural growth factor (NGF), fibroblast growth factor (FGF), epidermal growth factor (EGF), and sonic hedgehog (Shh) [20, 35–37]. These signaling mechanisms have been largely characterized as regulators of proliferation, differentiation, and maturation of neural progenitors (e.g., Refs. [6, 38, 39]). Indeed, GSK3β exhibits crosstalk with several signaling pathways and is considered to have the potential to integrate these pathways into the control of neuronal physiology, particularly at the level of microtubular cytoskeleton dynamics.

GSK3β Activity in Neuronal Polarization and Plasticity

In developing neurons, the involvement of GSK3β in neuronal polarization coordinating mammalian target of rapamycin (mTOR), phosphoinositide 3-kinase (PI3K), Par3/6, protein kinase A (PKA)-LKB1, and Rho-GTPase signaling pathways has been reported (see also Ref. [7] for a recent review). Today, it is well known that neuron polarization begins with the rapid extension of a single neurite to produce an axon, while the remaining neurites differentiate into dendrites. This is followed by the selective localization of molecules that are responsible for differential growth and specific functions of the axon and dendrites [40]. During this process, polarized neurons develop axonal and dendritic trees by elongation and branching. In these processes, GSK3β activity plays a role at two levels: (i) controlling the phosphorylation of different proteins associated with the microtubular cytoskeleton and (ii) modulating local protein synthesis and degradation [41] Indeed, GSK3β substrates include MBPs such as adenomatous polyposis coli (APC), collapsin response mediator proteins 2 and 4 (CRMP2 and CRMP4), microtubule-associated proteins 1B (MAP1B) and Tau, and cytoplasmic linker proteins (CLIP-associate protein), as well as other transcription factors and kinases putatively involved in cytoskeleton modifications (Fig. 1). In fact, functions of GSK3 during polarization are not exclusive since it runs in parallel with other kinases localized in growing neurites also acting on cytoskeleton polymerization (e.g., microtubule affinity-regulating kinase (MARK) [42], serine/threonine kinase LKB1, and SAD kinases [43]).

The activation and modulation of GSK3 yields numerous outcomes during neuronal polarization that may trigger both axon growth promotion and inhibition depending on the cellular model and particular experimental condition. For example, several studies have suggested that reduced GSK3β activity is required for initial axon formation and elongation (e.g., Refs. [44–48]) and that increased GSK3β activity leads to neurite retraction during polarization (e.g., Ref. [49]). Yoshimura et al. [44] reported that neurotrophic treatment of cultured neurons leads to axon elongation and branching by inhibiting GSK3β via PI3K/Akt, thereby reducing phosphorylation of CRMP2 and decreasing microtubule polymerization (Fig. 3). In fact, it has been reported that both GSK3α and GSK3β play roles in axon formation [48]. However, a second group of studies reported opposite results indicating that elevated activity of GSK3 leads to increased axon elongation. As an example of this, it has been described how high expression and phosphorylation of particular GSK3β targets (e.g., CMRP2 or β-adducin) increased neurite extension in cultured cortical neurons and SH-SY5Y neuroblastoma cells [50–52]. Similarly, it has been described how pharmacological and shRNA blockage of GSK3 activity induces neurite retraction [53, 54]. Although well developed, most of these results have been derived in cultured neurons of neural cells under different experimental conditions.

In an interesting study, Garrido et al. [48] determined a “critical period” of GSK3 activity in polarizing neurons of 24 h. Thus, inhibition of GSK3 activity after this period does not compromise the elongation process but increases axonal branching. In fact, two proteins, Katanin (a severing microtubular protein) and KIF2 (a microtubule depolymerase that modulates axon branching), are potential GSK3 substrates, suggesting that inactivation of GSK3 may regulate their functions to control axon branching [55]. These results defined a dual participation of GSK3 kinases during polarization at the level of initiation of axon growth from undifferentiated neurites and the branching modeling of the growing axon.

However, these challenging questions forced researchers to move forward with in vivo analysis of the function of GSK3 in mouse models. In particular, recent results obtained using conditional GSK3β expressing mice are of particular interest and reinforce some of the previous results. Indeed, functions of GSK3β during neurite extension and branching in newborn neurons in the hippocampal dentate gyrus have been reported [56]. GSK3β overexpression under the CaMK-II promoter in newborn granule cells of the dentate gyrus causes dramatic changes in neurite extension and synaptic connectivity (Fig. 2). These effects can be reversed by down-regulating GSK3β in the same cells [56]. In fact, GSK3β-overexpressing granule cells displayed a relevant dendritic tree remodeling in the adult hippocampus after maturation, mimicking the remodeling observed in Alzheimer´s disease. In this respect, general or prolonged inhibition of GSK3β might induce multiple points throughout the axon/neurite capable of supporting efficient microtubule formation, resulting in more branching instead of axonal elongation (e.g., Refs. [44, 46, 48]; Fig. 2). These effects have been reported in vivo by Llorens-Martin and co-workers [56] who found an enlarged highly branched proximal dendritic tree and a drastic decrease in distal branching of the granule cells in these mice as previously shown in vitro (see above).

Fig. 2
figure 2

GSK3β activity in neurite growth and branching during development and in the adult. GSK3β plays a key role in neuronal polarization and neurite outgrowth in developing neurons. The suppression of GSK3β activity in cultured neurons induces axon formation and elongation. Once the axon has been determined in developing neurons, a compartmentalization in the neuron takes place that is maintained by the differential spatiotemporal regulation of GSK3β. Moreover, Llorens-Martin et al. recently demonstrated that GSK3β overexpression in newly generated granule neurons induces relevant branching in proximal dendrites rather than elongation. Then further GSK3β inhibition leads to a morphological rescue phenotype. Thus, it appears that GSK3β activity above or below a baseline level of activity leads to alterations in the balance between elongation and branching

Full size image

These results make it clear that the differences between specific domains of the growing neurite and differentiating axon may arise because developing neurons are able to spatially regulate GSK3β activity in different cellular domains along growing neurites. This may also come about from a particular distribution of receptors and/or intracellular partners modulating local activity of GSK3β toward specific domains of neuronal cytoskeleton. Particularly relevant is a putative scenario in which second messengers may act locally regulating kinases involved in neuronal polarization. Two examples of this are worth citing: Barnes et al. [43] determined that cAMP levels act through phosphorylated LKB1 and GSK3β to promote axon initiation, whereas cGMP-mediated signaling suppresses axon formation via reciprocal down-regulation of cAMP/PKA-dependent phosphorylation of LKB1 and GSK-3β [43]. Second, Jiang et al. [46] demonstrated that phosphatase and tensin homolog (PTEN), a phosphatase that converts PIP3 to PIP2, showed low activity in the axon compared to dendrites in developing neurons. This low activity of PTEN attenuates PIP3-dependent local activation of Akt in neurites other than the growing axon, leading to GSK3β inhibition and promotion of axonal growth and enhancement of neuronal polarization (Fig. 2). In support of this notion, it has recently been described how SIRT-1, a deacetylase that is mainly located in the axon and growth cones of developing neurons, inhibits GSK3 activity by deacetylating and activating Akt, leading to axon regrowth [57]. These actions are particularly relevant when considering axon formation and growth under the effects of some long-range axonal wiring cues during neural development since most of them act through cAMP levels [32, 58], in contrast with local control of neurite/axon formation by the balanced control of cGMP/cAMP levels. This is in line with a current notion that suggests that the neuronal growth cone may also sense different concentrations of some extracellular factors that modulate different intracellular mechanisms including GSK3 activity. This new idea implies that a single cue is able to trigger different spatial neuronal responses depending on their extracellular concentration. An example can be found in the growth cone responses to secreted semaphorins (e.g., Sema3A). Sema3A acts through Neuropilin-1 and Plexins, impairing neurite extension by modulating GSK3β activity [59–61]. However, a recent study by Manns et al. [62] demonstrated that while mTOR inhibition did not influence GSK3β activation, GSK3β inhibition promoted mTOR activity in cultured dorsal root ganglion (DRG) neurons exposed to different concentrations of Sema3A. Furthermore, regarding GSK3β activity and its reported spatiotemporal regulation, what is clear is that a simple analysis of the overall phosphorylation status of the kinase in the neuron does not correlate fully with the observed response. For example, this has been demonstrated for Netrin-1, a chemoattractive molecule for commissural axons that is able to guide axons by small changes in local concentration [63]. In fact, when analyzed after treatment, Netrin-1 acts by increasing the phosphorylation of GSK3β at Ser9 and Thr216 and by modulating MAP1B phosphorylation [64], which further mediates chemoattraction in vitro as well as in vivo (e.g., Refs. [64, 65]). Parallel to the spatiotemporal regulation, the specific experimental conditions may lead to an increase in or blocking of axonal growth depending on the substrates involved and the extent of GSK3β inhibition [54, 66]. In fact, today, we cannot rule out the scenario that hypothesizes that a decrease in GSK3β activity after intense treatment or local influence might reduce GSK3β activity to baseline levels, which would then promote axonal growth or branching. Furthermore, although examples have been reported for specific cell types (e.g., DRG neurons, [62] and hippocampal neurons [32]), we still do not know whether the mechanisms that regulate GSK3β also occur in all subsets of developmental neurons in the same manner.

A Brief Overview of the Molecular Factors that Impair Axon Regrowth After Adult CNS Lesion

In the adult mammalian central nervous system (CNS), axons have a limited capacity for regrowth after lesion (e.g., Refs. [67–69]). Several intrinsic and extrinsic neural processes converge in the absence of axonal regeneration including: (i) low potential to recapitulate the intrinsic developmental program in adult CNS neurons, (ii) low renewal of damaged neurons with endogenous stem cells, which impairs putative autonomous cell therapy in most CNS regions, and (iii) the overexpression of numerous inhibitory molecules by reactive cells after adult CNS lesions. These reactive cells form the meningoglial scar, a dense meshwork of astrocytes, oligodendrocyte precursors, microglia, and infiltrated meningeal cells (e.g., Refs. [70–72]). Indeed, this scar is considered both a physical and a biochemical barrier to axon regeneration and functional recovery, although some reports suggest that it may have beneficial effects shortly after the lesion (see Ref. [70] for review). Today, it is generally accepted that the molecular environment around the meningoglial scar is, in combination with the loss of intrinsic growth capacity of the injured axons, a determinant of the ability or failure of adult axons to regenerate (e.g., Refs. [73–77]). As indicated, the scar produces inhibitory and nonpermissive factors such as semaphorins, ephrins, tenascins, chondroitin sulfate proteoglycans (CSPGs), and myelin-associated inhibitors (MAIs) [71]. In this review we will focus on semaphorins, CSPGs, and MAIs as examples of inhibitory signals that may modulate GSK3 and especially GSK3β activity.

Semaphorins and Their Receptors

Although initially characterized as repulsive axonal guidance cues, semaphorins are now regarded as relevant contributors to morphogenesis and homeostasis for a wide range of tissue types (e.g., see Ref. [78] for a detailed description of semaphorins and their receptors). Semaphorin-mediated long- and short-range repulsive, and attractive, guidance has a profound influence on neuronal physiology (see Refs. [79, 80] for review). In fact, in recent years, several members of the superfamily of semaphorins have been associated with neural lesions, and some of these directly with inhibitory processes after injury. For example, Sema3A, Sema3C, Sema6B, and Sema6C are overexpressed after telencephalic lesions in several reactive as well as nonreactive cells [59, 81–84]. Sema6C and Sema3A lead to GSK3β-dependent axonal growth cone collapse [82, 85]. Moreover, it has also been reported that three other semaphorins, Sema4D, Sema7A, and Sema6A, are overexpressed after spinal cord injury [86–88] and may act though the same signaling pathways as Sema6C and Sema3A (GSK3β activity modulation) leading to inhibitory processes.

Chondroitin Sulfate Proteoglycans and Their Receptors

Proteoglycans (PGs) are highly glycosylated proteins synthesized by reactive astrocytes [89, 90], meningeal cells [91, 92], NG2 cells [93], and macrophages after lesion [94]. They are made up of a core protein and long unbranched glycosaminoglycan (GAG) chains such as chondroitin sulfate (CS) and heparan sulfate (HS). CSPGs are involved in several neuronal processes like migration, axon guidance, and the promotion or inhibition of neuritic growth (e.g., see Ref. [95]) during CNS development. However, CSPGs are also overexpressed in many types of CNS lesion (e.g., Refs. [96–101]). The modes of action induced by CSPGs have not yet been fully determined. Most of their inhibitory effect depends on GAG chains (e.g., Ref. [102]), although for particular CSPGs, such as NG2, the protein core also displays inhibitory properties [103]. Two proteins, the transmembrane protein tyrosine phosphatase 3 (PTPσ) [104–107] and the leukocyte common antigen-related phosphatase (LAR), are CSPG receptors that may modulate GSK3 activity [108]. In fact, from a molecular point of view CSPGs initially diminish Akt and GSK3β-Ser9 phosphorylation in cell lines and cultured neurons [10].

MAIs and Their Receptors

It is well known that MAIs are highly regulated after injury by glial cells, oligodendrocytes, and, in the case of MAIs (Nogo-A), lesioned neurons (e.g., Refs. [11, 109–111]). Three MAIs have been studied in detail: Nogo-A, the myelin-associated glycoprotein (MAG), and the GPI–linked oligodendrocyte-myelin glycoprotein (OMgp) [112–114]. These molecules impair axon regrowth mainly by acting on a common neuronal receptor complex, the Nogo receptor complex. This receptor is formed by the GPI-anchored protein Nogo receptor 1 (NgR1) and three putative co-receptors: (p75 (NTR), TAJ/TROY and Lingo-1) [112–114]. In addition, new ligands have recently been reported as acting through the Nogo receptor complex (e.g., Refs. [115–118]) and new receptors for MAIs have also been described [119–125]. The participation of these ligands and receptors and their associated intracellular signaling mechanisms in axon regeneration is under constant evaluation (e.g., Refs. [72, 126]). Most studies have reported MAI and receptor functions in the CNS, although emerging roles in PNS regeneration have also been described [127]. The effects of MAIs in GSK3 activity are discussed below, since dual effects have been described.

Modulation of GSK3β Activity by Lesion-Associated Molecules Through Akt/GSK3β and Akt/mTOR Pathways

From several studies we know that PI3K/Akt signaling is one of the most studied targets of inhibitory molecules after neural lesions. Two pathways can be activated by PI3K/Akt: i) the Akt/GSK3β pathway and ii) the Akt/mTOR pathway. The modulation of these pathways by inhibitory molecules has been reported in several studies. Concerning Akt/GSK3β, some studies reported that secreted semaphorins activated GSK3 (e.g., Sema3F [128] or Sema3A [85, 129, 130]). Uchida et al. report that the sequential phosphorylation of CRMP2 by Cdk5 and GSK3β mediates growth cone collapse in response to Sema3A in mouse DRG neurons. Interestingly, they also report the formation of a complex involving GSK3β, Axin-1, and β-catenin which may play an important role in Sema3A signaling involving the endocytic pathway [85].

With respect to CSPG and the Akt/GSK3β pathway, Fisher et al. [108] found a reduction in Akt phosphorylation after CSPG treatment in cultured cerebellar granule neurons (CGNs). Then, reduced Akt activity correlated with enhanced GSK3β activity. In addition, using pharmacological treatment, Gao et al. [131] also showed that treatment with Amphotericin B, a polyene fungal antibiotic [132], induces an increase in the regenerative process through Akt activation and GSK3β inhibition after CSPG stimulation in neurons. Interestingly, a recent study reported that NgR1 and 3 are also functional receptors for CSPGs, since they interact with CSPG GAG chains and might trigger a signaling pathway in which the Rho family is involved [123]. Hence, there may be collaboration between the NgR family and LAR family of receptors since an additive effect on the number and length of regenerating axons following CNS injury was observed in NgR1-3/RPTPσ -/- knockout mice [123]. Based on this evidence we may suppose that CSPGs affect GSK3 and particularly GSK3β activity both through known [e.g., NgR1, NgR3, integrins, paired immunoglobuline-like receptor B (PirB), PTPσ, LAR] and unknown receptors.

In the Akt/mTOR mechanism, PTEN has been shown to attenuate the regrowth of injured CNS axons by suppressing mTOR [133]. Indeed, PTEN deletion in injured mature retinal ganglion neurons (RGCs) increases axonal regeneration by over-activating Akt and inhibiting GSK3β activity [133, 134]. In line with these results, Liu and co-workers also show that PTEN deletion leads to a regenerative response of corticospinal neurons in the spinal cord [12]. The marked increase in compensatory sprouting of intact corticospinal tract (CST) axons and regenerative growth of injured CST axons after PTEN deletion suggests that these two forms of regrowth have similar underlying mechanisms [12]. Unfortunately, however, the chronic loss of PTEN triggers unwanted effects, since long-term knock-out or knock-down in particular tissues may lead to neoplasias, such as glial tumors, which promote astrocyte hypertrophy and proliferation (see Ref. [135]).

MAIs and GSK3 Activity After Lesion: a Scenario Requiring Clarification

Some published data have directly associated GSK3β activity with MAIs. However, as observed above for neuronal polarization, there are some discrepancies in the literature concerning MAIs and GSK3 activity. In fact, several studies have shown increased GSK3β activity by stimulation with MAG or myelin extracts [10, 11, 136], whereas others found decreased GSK3β activity after stimulation with myelin extracts and the Nogo-P4 inhibitory peptide (residues 31–55 of the NogoA protein) [137] or OMgp [9] (Fig. 3). As an example of the first group of studies, Dill and co-workers demonstrated that the acute stimulation of PC12 cells and CGNs by MAG increased Akt and GSK3β-Ser9 phosphorylation [10]. Similar inhibition of GSK3 was described by Shen et al. [136] in N2a cells (see also Fig. 3 as example of the results published in Ref. 9). More relevantly, the published data attribute to GSK3β the main role of inhibiting axon regrowth after long-term exposure to inhibitory cues (see above), which may conflict with MAIs activities (in some studies). However, we should consider both the temporal and the GSK3 activity after lesion. For example, for some cultured cells, two different phases of kinase activation after MAIs exposure have been reported [11]. Thus, in contrast to GSK3β, ERK1/2 activity increased shortly after stimuli with myelin (Fig. 3). In a second phase, ERK1/2 decreased and GSK3β increased in parallel. As indicated above, ERK1/2 activity has been reported to inhibit GSK3β in hepatocellular carcinoma cells [28]. In fact, these data suggest an acute effect of ERK1/2 after stimulation with inhibitory molecules delaying the activation of GSK3β [11]. ERK1/2 activation by MAIs is mediated by fast EGF receptor (EGFR) transactivation [138]. Taking into account that, as indicated above, fast inhibition of GSK3 induces axon formation and growth (see above), it is reasonable to consider that ERK1/2 and GSK3 may act in a coordinated way to overcome from the very outset MAIs-mediated axon growth inhibition. Unfortunately, these neuronal responses are not followed up by fast regeneration and the lesioned axons remain collapsed. However, chronic inhibition of GSK3 activity by meningoglial scar molecules, such as CSPG, or secreted semaphorins impairing axonal growth may contribute to the axonal sprouting observed in chronic lesions [139, 140].

Fig. 3
figure 3

Proposed roles of GSK3β in axonal outgrowth inhibition. Alabed et al. demonstrate a scenario in which activation of the Nogo receptor complex or other receptors by myelin debris or other inhibitory molecules present at the injury site (e.g., CSPGs) triggers a downstream signaling pathway that leads to the phosphorylation of the GSK3β Ser9 epitope (1) and RhoA/ROCK pathway activation (2). This Ser9 phosphorylation results in changes in GSK3β activity, which switches from baseline activity to inactive status (1). The decreased activity may also direct CRMP4 dephosphorylation (3) and binding with RhoA (4). All these effects lead to modifications of the Actin or Microtubular cytoskeleton in the axon (5) and axonal outgrowth inhibition. Grey arrow represents the pathway direction without inhibitory stimulus that may lead to increased CRMP4 phosphorylation and axonal outgrowth stimulation. In these effects, the use of CRMP4-RhoA interfering peptide (C4RIP) (6) may promote increased neurite length, and the use of inhibitors (e.g., CT99021, SB216763, and SB415286), or inhibitory constructs like GSK3βs9A against GSK3β may lead to an inhibitory axonal outgrowth effect (7). Moreover, we and others (see Refs. [11, 138]) indicate that ERK1/2 is activated through a transactivational mechanism together with the RhoA pathway after inhibitory stimulus (8). In contrast current data suggest that ERK1/2 activity leads to an initial decrease in GSK3β activity that becomes stimulated in a second phase (9). Therefore, GSK3β may act directly on the microtubule binding proteins (e.g., MAP1B and Tau) (10) and through CRMP2 (11), contributing to microtubule destabilization. In parallel, RhoA (12) may facilitate CRMP2 phosphorylation by activation through GSK3β also acting on the F-Actin

Full size image

GSK3β Inhibition as a Strategy to Overcome Axon Regeneration: Lessons from In Vivo and In Vitro Models

The first experiments aimed to enhance axon regeneration by inhibiting GSK3β activity were carried out using LiCl as inhibitor [141–144]. LiCl seems to be a nonspecific GSK3β inhibitor [145] and it may affect other kinases and signaling pathways after treatment (see also Ref. [146]). In this respect, other highly selective GSK3β inhibitors, such as SB-415286 and SB-216763, have been developed to avoid these indirect effects [147–149]. Moreover, other groups of molecules, such as aminopyrimidines (purine analogues) like CT99021, seem to inhibit GSK3β in a nanomolar concentration range [150, 151]. Another approach was to combine treatment with LiCl and chondroitinase ABC [152] to induce ganglion cell axon regeneration in the retina and the rubrospinal tract after lesion [10, 152]. Therefore, the question concerning treatment with LiCl is whether these results are dependent on the effect of LiCl on GSK3β or on other kinases and/or factors that promote cell survival (see also Refs. [152–155]; Fig. 3). To determine the functional effects of GSK3β inhibition after in vivo spinal cord lesion, Dill et al. [10] compared systemic treatment of lesioned mice with LiCl or SB-415286 induced sprouting and axonal growth of the CST and raphe spinal axons, specifically in the dorsal transection and contusion models, which correlated with an improvement in behavioral recovery. Using in vitro models, our previous studies in organotypic slice co-cultures revealed that GSK3β remains active for several days post-lesion (Fig. 3) [59]. Treatment of the axotomized slices with GSK3β inhibitors (SB-415286 or SB-216763) promoted highly significant axonal growth and sprouting across the lesion [11]. Taken together, these data point to GSK3β activity as a therapeutic target to improve axon regeneration. However, our studies developed using NgR1 (NgR1 o/o) knockout slices also support the role of other inhibitory molecules, such as semaphorins and CSPGs and their receptors, in GSK3β activation in lesioned neurons (e.g., Refs. [59, 108, 156], see above for details; Fig. 3). In contrast to other reports, in our experiments MAIs and Sema3A stimulation enhanced Tau phosphorylation mediated by GSK3β, leading to cytoskeleton instability, neurite retraction, and growth cone collapse (e.g., Refs. [157, 158]; Fig. 3).

Recently, some groups have developed transgenic mice in an attempt to improve axon regeneration after injury and to better understand the role of GSK3 kinases in this scenario. However, deletion of the GSK3β gene in mice is lethal [159, 160]. Only GSK3β heterozygous mice survive, but they show a wide range of neurological defects [161, 162]. In contrast, it seems that mice lacking GSK3α are viable and relatively normal [163]. Thus, new studies using the recently developed conditional mice (see above) and the GSK3α knockouts will determine the potential of GSK3 and particularly GSK3β as targets after spinal cord lesion and its associated muscle dystrophy.

Perspectives

A complete understanding of the role of GSK3β after neural lesions requires additional experiments to be fully revealed, since discrepancies exist in the literature. However, recent biochemical results support histological data indicating that, after lesion, axons might try to regenerate shortly after lesion inhibiting GSK3β through increased ERK1/2 activity. However, due to the inhibitory environment, there is a progressive increase in GSK3β activity after injury that affects neurite outgrowth, but this might be involved in the axonal sprouting/branching observed in chronic lesions. However, due to the different experimental models, it would be of great interest to study the status and participation of the regulatory pathways of GSK3β activity after spinal cord lesion in newly generated mice with controlled expression of GSK3β in neurons. In fact, it seems reasonable to consider the possibility that some of the discrepancies observed between the different in vitro and in vivo studies may be associated with different experimental conditions or different lesion models, respectively. However, we also cannot rule out the presence of differentially regulated GSK3β in different domains of the developing and regenerating neuron that may affect the observed results. Indeed, we feel that analysis of the phosphorylation status of this kinase as a whole does not help to reveal its specific role in particular conditions. In this scenario, mice that may avoid unwanted effects of prolonged deletion of PI3K/Akt/mTOR or GSK3β are essential in helping unravel these challenging questions.