Deconstructing GSK-3: The Fine Regulation of Its Activity

3.1. Regulation by Phosphorylation

The first regulatory mechanism described of GSK-3 activity involved the phosphorylation of specific residues of GSK-3 by other kinases, and more recently through autophosphorylation [17, 28].

Four different regions and residues have been described in the GSK-3 molecule. The first one corresponds with a serine residue at positions 21 in GSK-3α and 9 in GSK-3β. It has been clearly established that phosphorylation of serine 21 or 9 correlates with the inhibition of its kinase activity [29–31]. Many protein kinases are capable of phosphorylating GSK-3 at this residue, such as Akt, ILK, PKA, and p90Rsk [32, 33], and many physiological situations of inhibition of GSK-3 correlate with serine phosphorylation, such as Insulin/IGF1, NGF, or estradiol treatments, not only in neurons [34].

Additionally, two other regulatory sites have been described. One is the threonine 43, present only in the isoform GSK-3β, which may be phosphorylated by Erk [35]. This phosphorylation correlated with GSK-3 inhibition. Second, serine 389 and threonine 390 present in GSK-3β have been shown to be phosphorylated by p38 MAPK [36]. In both cases, the data suggested that this phosphorylation may increase the capacity of Ser-9 to be phosphorylated rather than promote a direct inhibition (see Figure 1).

In contrast, tyrosine phosphorylation present in positions 279 in GSK-3α or 216 in GSK-3β, appears to correlate with an increase of its kinase activity [37]. Different candidates such as Pyk-2 and Fyn kinases have been reported to be able to phosphorylate GSK-3 in vitro on tyrosine. In addition, MEK1/2 has been showed to have this capacity only in fibroblasts [38, 39]. This data contrast with those reported in Dictyostelium discoideum where there is compelling evidence indicating that ZAK 1 is responsible for generating tyrosine phosphorylation in GSK-3 [14, 40]. However, no homologue of such kinase has been found in mammals.

More recently, an alternative hypothesis has been proposed for the regulation of GSK-3 tyrosine phosphorylation. This hypothesis suggests that in mammalian systems phosphotyrosine in GSK-3 corresponds to an intramolecular autophosphorylation event and may be regulated by Hsp90 [41]. Molecular dynamics and crystallographic studies clearly suggest that Tyr216 renders the kinase active through interactions with Arg220 and Arg223, stabilizing the activation loop and allowing full substrate accessibility [42, 43]. However, this hypothesis still lacks a cellular demonstration.

However, our data indicated that not all pharmacological inhibitors of GSK-3 decrease the level of phosphotyrosine. Therefore, lithium chloride inhibits GSK-3 activity, but this inhibition does not alter its pTyr content [44]. Moreover, in neuronal cells, tyrosine phosphorylation of residue 216 or 279 increased following exposure to LPA [37] and even upon exposure of neurons to β-amyloid or PrP [13, 45, 46] in a clear correlation with an increase in GSK-3 activity. In addition, in many neuronal cells, the pharmacological inhibition of tyrosine phosphatases with ortho-vanadate increases the basal level of GSK-3-pTyr [44]. Thus, considering all these data, in addition to this tantalizing autoregulatory system proposed, we hypothesized that some as-yet-unidentified tyrosine kinases and phosphatases may also regulate this kinase (see Figure 2).

3.2. Regulation by Protein Complex Association

One regulatory mechanism that is still not fully understood involves the interaction of the GSK-3 with structural proteins (scaffold proteins). It is well known that GSK-3 contributes to a multiprotein complex formed by axin and adenomatous polyposis coli (APC), among others (for review see, i.e., [47]). This protein complex is the core of canonical Wnt signalling (see below). Indeed, in the absence of ligand, GSK-3 is able to phosphorylate β-catenin for targeting it for proteasome degradation [48]. More recently, some data suggests that this complex may be specific for GSK-3β2 isoform [49], which opens the possibility of a deeper analysis of “specific functions of GSK-3 isoforms.”

Another system of protein-kinase interaction was denoted as GSK-3-binding protein (GBP or FRAT) [25, 50]. Three different FRATs have been cloned and characterized; however, their mechanism of action is not well understood. FRAT1 appears to act as an inhibitory system [51], whereas FRAT2 appears to preferentially increase GSK-3-mediated phosphorylation in some residues [52]. Surprisingly, recent data demonstrated that FRAT is dispensable because the triple FRAT-knockout mouse lacks any major defect in brain development [53]. All these data indicated that the precise role of FRAT in GSK-3 regulation is still to be defined.

Using the binding site on GSK-3 for FRAT/GBP, a GSK-3-interacting protein symbolized by GSKIP has been cloned and characterized. GSKIP can block phosphorylation of different substrates and functions as a negative regulator of GSK-3 beta [54].

3.3. Regulation by Priming/Substrate Specificity

As previously mentioned, the specificity of many kinases is governed by a consensus sequence of aminoacids sequence. However, as almost general rule, GSK-3 substrates do not require a very specific sequence, but a previous (primed) phosphorylation residue modified by a priming kinase located four aminoacids, C-terminal, to the Ser or Thr residue to be modified by GSK-3. The crystal structure of human GSK-3β provides a model for the binding of prephosphorylated substrates to the kinase (PDB ID are 1I09 [22] and 1H8F [21]). According to it, primed Ser/Thr is recognized by a positively charged binding pocket formed by residues Arg96, Arg180, and Lys205 that facilitates the binding of the phosphate group of primed substrates.

Some “priming kinases” have been identified, such as cdk-5 [55–57], PAR-1 [58], casein kinase I [59], PK-C [60], or PK-A [57]. However, it is not clear so far whether a second set of “nonprimed” substrates may define a different group of functions [61]. In addition, different glycogen synthase kinase-3 isoforms appear to exhibit distinct substrate preference in the brain [62].

3.4. Regulation by Subcellular Localization

In developmental brain, the presence of GSK-3 was high at E18 and peaked on P8, decreasing after that period [63]. In addition, this report showed that the developmental profile of GSK-3α and GSK-3β is different, having β downregulated after birth which suggested a differential role in neuronal development. However, the putative differential role of each isoform has been explored in few reports, that is, [7]. It is important to indicate that a portion of GSK-3, mostly β, has been reported to be associated in the growth cone. This GSK-3 pool appears to respond rapidly, being modified by phosphorylation and/or relocated in the growth cone by external signals such as Semaphorins [64] or NGF [65].

GSK-3 activity is also dependent on its subcellular localization; some data illustrated the presence of GSK-3α and β in many neuronal compartments and in primary neurons, either in axon, dendrite, or in nucleus [66, 67]. In addition, GSK-3 has been found in the cytoplasm, nucleus and the mitochondria [18]. Considering the list of GSK-3 substrates reported, it is evident that most of its activity should occur in the cytoplasm and in the nucleus, while we have less information about GSK-3 potential targets in the mitochondria. Recent data suggested that proteins such as Mcl-1 [68] and hexokinase [69] may be regulated by GSK-3 activity. It has been suggested that nuclear GSK-3 may be involved in phosphorylation of many transcription factors such as cyclin D1, β-catenin, HSF-1, NFAT, and cAMP-response element-binding protein, among others (Table 1), for review see [28, 70, 71]. Also, it has been proposed that GSK-3 in the nucleus may have a role in alternative splicing [72]. In addition, proapoptotic stimuli induce nuclear accumulation of GSK-3β [73]; however, this hypothesis has been not established in other neuronal death paradigms (D. Simon, unpublished observations).

Further insight into GSK-3 regulation has been gained very recently by revealing an essential role of multivesicular endosomes in the Wnt signalling pathway. A combination of protease protection assays, detergent permeabilization, and cryoimmunoelectron microscopy demonstrated that Wnt activation of the Frizzled and LRP6 receptors triggers sequestration of GSK-3 into these membrane-bounded organelles, leading to decreased GSK-3 levels in the cytosol [74]. This process seems to require β-catenin, forming a feed-forward loop by facilitating GSK-3 sequestration.

3.5. Regulation by Proteolytic Cleavage

A new mechanism of GSK-3 regulation has been recently proposed. This regulation involves the removal by calpain of a fragment from the N-terminal region of GSK-3, including the regulatory serines 9/21. After removal of that fragment, GSK-3 becomes activated [75]. The same study showed that both isoforms α and β are cleaved by calpain, although with different susceptibility. Moreover, GSK-3 truncation has been observed in human and mouse postmortem brain tissue [76]. It is noteworthy to consider that a similar mechanism has been described for β-catenin in hippocampal neurons, where after NMDA-receptor-dependent activation, calpain induced the cleavage of β-catenin at the N terminus, generating stable and truncated forms which maintain its transcriptional capacity [77]. Likewise, GSK-3 truncation is mediated by extracellular calcium and can be inhibited by memantine [76], an NMDA antagonist used for the treatment of Alzheimer's disease. Interestingly, GSK-3β has also been recently shown to be cleaved at the N-terminus (and subsequently activated) by matrix metalloproteinase-2 (MMP-2) in cardiomyoblasts [78].

4.1. Inhibitory Pathways

Among these pathways, the signalling triggered by Insulin or IGF-1 [19, 79, 80] and NGF/BDNF/NT3 [81, 82] has similar features. These tyrosine kinase receptors initiated cytoplasm signals in which the inhibition of GSK-3 activity is a common feature. It is generally accepted that the kinase implicated in this inhibition is PKB/Akt [29–31], even though kinases such as PKA or ILK have also been implicated [32, 33]. In all cases, phosphorylation on serine 21 and 9 (α and β, resp.) represents the inhibition of the GSK3 kinase activity, as previously mentioned.

The second well-documented pathway is Wnt/Wingless signalling [83, 84]. This signalling has been widely studied, and it has been shown to be essential in early embryonic patterning, cell fate, cellular polarity, and cell movement in both vertebrates and invertebrates [47, 85]. In many if not all cell systems, the canonical Wnt pathway is formed by a set of phylogenetically conserved proteins including the Wnt receptor frizzled (fz), and a coreceptor LRP5/6; Dishevelled (Dsh), and a scaffolding protein that activates a complex formed by Axin/APC/GSK3-β/β-catenin [47, 85–88].

In this pathway, in the absence of ligand, GSK-3β phosphorylates β-catenin, among other proteins, this phosphorylation constituting part of a degradation signal for β-catenin. However, in the presence of Wnt, the receptor complex triggers a signal in which Dsh inhibits the activity of GSK-3β by a mechanism not completely understood, so far. This system appears to be specific for GSK-3β as no counterpart has been described for GSK-3α to date; however, some GSK-3 activity appears to be necessary for Wnt signalling [89, 90]. More recently, a bioinformatics-based screen for proteins whose stability may be controlled by GSK-3 [91] has led to the identification of a number of multiple Wnt signalling target proteins, suggesting that this pathway controls a broad range of cellular activities apart form β-catenin-mediated transcriptional activation. Furthermore, GSK-3-mediated Wnt signalling seems to regulate the turnover of many cellular proteins [74, 91], indicating that GSK-3 phosphorylation-dependent protein degradation may be a widespread cellular mechanism to regulate a variety of cellular processes in response to extracellular signals [71].

Functional segregation of the insulin/growth factor and Wnt roles requires either that there be no exchange between the subsets of the cellular GSK-3β pool committed to each role, or that the recruitment of GSK-3β to the axin-APC complex can reverse or override inhibitory Ser9 phosphorylation present in a recruited GSK-3β molecule. Phosphatases capable of removing extant Ser9 phosphorylation are certainly known to be associated with the axin-APC complex [92, 93]. Alternatively, the very substantial enhancement in activity towards β-catenin afforded by the axin “scaffolding” may simply allow a primed β-catenin substrate to outcompete a pSer9-GSK-3β N-terminal peptide for access to the substrate-binding site [26].

Estrogens regulate many physiological processes and fulfil a wide range of functions during development and differentiation in mammals of both sexes. The actions of estrogens are mediated by estrogen receptors and have been classified as either “genomic actions” or “nongenomic, rapid actions.” The genomic actions are based on the capacity of the estrogen receptors (ERs) to modulate transcriptional activity either directly or through coactivators or corepressors, that is, [94, 95].

More recently, it has been shown that in addition to its direct transcriptional activity, estrogen receptors activate a set of cytoplasm signals in a similar manner to some growth factors. Hence, it has been reported that estradiol acts synergistically with IGF-1 in the brain or in neurons, activating the PI3K/Akt pathway [34, 96, 97]. We described that the addition of estradiol increases the serine phosphorylation of GSK-3. This inhibitory phosphorylation is time- and concentration dependent, and an antagonist of estradiol prevents this event. The kinase responsible is sensitive to the inhibition of the PI3K pathway, and for this reason, it seems that the best candidate would be Akt [98, 99]. A more detailed analysis of these new signals will give us clear evidence whether this pathway is completely convergent with those using PI3K/Akt/GSK3, as previously mentioned.

4.2. Activation Pathways

In neurons, LPA has been shown to induce neurite retraction and the rounding up of neuroblastoma cell lines [100]. In some primary neurons, it also promotes growth cone collapse and neurite retraction [37, 101]. This bioactive lipid acts as a growth factor through specific seven transmembrane domain receptors, denoted as lpa 1–4 [102, 103]. We described that GSK3 activity was increased after LPA treatment in diverse neuronal cells of different species in correlation with the neurite retraction process [104, 105]. This activation correlated with an increase in GSK3-PTyr and may be downstream Gα₁₂ or Gα₁₃ [101, 105, 106]. The previous inhibition of GSK3 activity prevents, at least in part, the growth cone collapse response. Similarly, it has been reported that three different GSK-3 antagonists (LiCl, SB-216763, and SB-415286) can inhibit the growth cone collapse response induced by Sema 3A [64].

However, the exact mechanism of how this activation of GSK3 occurs is not known, so far. Many reports indicate that in Dictyostelium discoideum GSK-3 activity may increase in response to cAMP binding to a heptahelical G-protein-coupled receptor. In this system, a tyrosine kinase and a tyrosine phosphatase have been described as regulators of GSK-3 activity [14, 40], but similar kinase and phosphatase have not been found in mammals.

Furthermore, it has been reported that Reelin and Netrin increased GSK-3 activity, similar to what LPA did. This Netrin or Reelin-dependent GSK-3 activation seems not to be a particular characteristic of the cell line or neuron used but rather a more general physiological process [107–109]. Indeed, even in situations where the final balance is an inhibition of GSK-3 kinase activity, such as following the addition of IGF1/Insulin or after estradiol addition, a transient activation of GSK-3 could be observed [34, 38]. All these data suggest that the upregulation and downregulation of this kinase is more complex than might initially have been considered.