Many of this kind of GSK-3 inhibitors isolated from marine organisms were identified during the search for inhibitors for cyclin-dependent protein kinases (CDKs) with anti-tumor activity. The dual activity of these inhibitors (and others) toward GSK-3 and CDKs is a direct result of their structural similarity within the ATP-binding domain (about 86% sequence similarity).

The bis-indole indirubin isolated from the traditional Chinese medicine for treatment of myelocyte leukemia was initially characterized as a CDK inhibitor and then found to be a potent GSK-3 inhibitor (Hoessel et al., 1999; Leclerc et al., 2001). It inhibits both protein kinases within the nanomolar concentration range. The indirubin analogs that were synthesized and tested (collectively termed here “indirubins”) showed inhibitory activity toward both GSK-3 and CDKs (Leclerc et al., 2001; Meijer et al., 2003). The indirubin analog 6-bromoindirubin, isolated from a marine invertebrate, the mollusk known as “Tyrian purple,” showed a certain selectivity toward GSK-3 over CDKs (Meijer et al., 2003). Accordingly, a synthetic cell-permeable derivative, 6-bromoindirubin-3′-oxime (6BIO; Figure 1), was developed and was about 16-fold more selective for GSK-3 relative to CDKs (Meijer et al., 2003; Polychronopoulos et al., 2004). The biological activity of 6BIO has been evaluated in several neuronal systems. It reduced tau phosphorylation in cultured cortical neurons (Martin et al., 2009), and inhibited neurite outgrowth in cerebellar and embryonic or postnatal dorsal root ganglion (DRG) neurons (Kim et al., 2006; Alabed et al., 2011). This effect appeared to be dependent on the degree of GSK-3 inhibition, as a weak inhibition of GSK-3 promoted axon branching (Kim et al., 2006). 6BIO enhanced self renewal and pluripotency in human ES cells (Sato et al., 2004), via its ability to act as a Wnt mimetic (Sato et al., 2004). Additional studies with indirubins demonstrated their ability to provide neuroprotection against kainic acid, MPTP (1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine), and trophic deprivation (Hongisto et al., 2003; Jin et al., 2005; Wang et al., 2007; Magiatis et al., 2010). In vivo systemic administration of indirubin-3′-oxime to APP/Presenilin-1 transgenic mice, a established AD model, attenuated many AD symptoms including tau hyperphosphorylation, Aβ accumulation, inflammation, and spatial memory deficits (Ding et al., 2010). In contrast, treatment with indirubin-3′-oxime did not reduce tau phosphorylation in cultured neurons or in the rat brain (Selenica et al., 2007). These discrepancies could be due to the limited bioavailability of indirubin-3′-oxime, which, like other indirubins, has limited water solubility (Leclerc et al., 2001). Recent work reported the synthesis of additional indirubin derivatives with improved water solubility (Vougogiannopoulou et al., 2008). Their biological activity has not yet been reported.

Sponges (Porifera), as the best known source of bioactive marine natural products in metazoans, play a significant role in marine drug discovery and development. The alkaloids debromohymenialdisine (DBH) and hymenialdisine (HD), which contain both bromopyrrole and guanidine groups (Williams and Faulkner, 1996; Figure 1) were originally isolated from sponges belonging to the Agelasidae, Axinellidae, and Halichondridae. HD structure was established using X-ray crystallography (Cimino et al., 1982) and it is a potent protein kinase inhibitor targeting GSK-3β, CDKs, MEK1, CK1, and Chk1 (Tasdemir et al., 2002; Sharma and Tepe, 2004) among others. In addition, HD has the ability to inhibit GSK-3β in vivo and also blocks the in vivo phosphorylation of the microtubule-binding protein tau at sites which are hyperphosphorylated by GSK-3 in AD (Meijer et al., 2000). As observed with other GSK-3 inhibitors, HD, and derivatives act in competition with ATP (Wan et al., 2004). Recently, it has been shown that HD and DBH are stored in spherulous cells from Axinella sp. (Song et al., 2011), which can help to define the bioactive production strategy in terms of sponge aquaculture.

The potent activity of HD as a competitive kinase inhibitor aroused interest in synthesizing a pyrrole-azepin-8-one ring system bonded to a glycocyamidine ring scaffold (Nguyen and Tepe, 2009). Crystallographic data of the HD complex with CDK2, led to the rational development of new analogs with increased potency and selectivity over GSK-3, CDK5, and CDK1. Moreover, using different docking methods and molecular dynamics simulation, the structural determinants that govern target selectivity on HD derivatives has been studied, explaining the significant inhibitory effect on GSK-3 of the related HD metabolite dibromocantharelline (Zhang et al., 2011). The specific residue Cys199 near the binding site of GSK-3 provides new clues for the design of potent and selective inhibitors.

Meridianins are brominated 3-(2-aminopyrimidine)-indoles, which are isolated from the tunicate Aplidium meridianum, an ascidian collected near the South Georgia Islands. These compounds inhibit various protein kinases such as CDKs, GSK-3, PKA, and CK1 (Gompel et al., 2004; Figure 1). The ability of Meridianins to prevent cell proliferation and to induce apoptosis, demonstrated their ability to enter cells and to interfere with the activity of kinases important for cell division and cell death (Gompel et al., 2004). Different medicinal chemistry approaches were employed to prepare more selective GSK-3 inhibitors (Akue-Gedu et al., 2009). However, in all cases, they were most potent toward CDKs, and inhibition of GSK-3 was marginal (Echalier et al., 2008).

Among the first synthetic small molecule GSK-3 inhibitors reported were the purine analogs, the aminopyrimidines, developed by Chiron. The potent inhibitors CHIR98014 (CT98014), CHIR98023 (CT98023), CHIR99021 (CT99021) (collectively termed here the CHIRs) inhibit GSK-3 within the nanomolar concentration range (Ring et al., 2003; Figure 1). Systemic analysis that profiled protein kinase inhibitors also confirmed the high selectivity of CHIRs toward GSK-3 (Bain et al., 2003, 2007). A limited number of studies have tested CHIRs in neuronal systems (these compounds had been chiefly tested in diabetic models). They showed that CHIRs potently reduced tau phosphorylation in cultured neurons and the rat brain (Selenica et al., 2007). In addition, treatment with CHIRs inhibited neurite outgrowth in cerebellar and DRG neurons (Alabed et al., 2011) and blocked NMDA-mediated long-term depression (LTD) in hippocampus slices (Peineau et al., 2009), indicating an unexpected role of GSK-3 in LTD maintenance (Peineau et al., 2009). In agreement with the 6BIO studies, CHIRs enhanced self renewal and pluripotency in mouse ES cells mimicking the activation of Wnt signaling pathway (Ying et al., 2008; Li et al., 2011). Finally, CHIRs reduced neuronal death in a cerebral ischemia rat model (Kelly et al., 2004), and enhanced the levels of the survival motor neuron protein (SMN) in spinal muscular atrophy (SMA; Makhortova et al., 2011).

The arylindolemaleimides SB-216763 and SB-415286 are highly selective GSK-3 inhibitors developed by GlaxoSmithKline that inhibit GSK-3 within the low nanomolar concentration range (collectively termed here SBs; Coghlan et al., 2000; Figure 1). A Number of studies demonstrated the neuroprotective effects of SBs against variety of pro-apoptotic conditions including inhibition of the PI3 kinase/Akt survival pathway, trophic deprivation, Aβ toxicity, heat shock, ethanol, NMDA excitotoxicity, and polyglutamine toxicity caused by the Huntington’s disease protein (Bijur et al., 2000; Cross et al., 2001; Culbert et al., 2001; Carmichael et al., 2002; Facci et al., 2003; Hongisto et al., 2003, 2008; Takadera and Ohyashiki, 2004; Hu et al., 2009). In addition, SB-216763 was shown to inhibit axon growth in postnatal and embryonic DRG neurons (Owens et al., 2003; Alabed et al., 2011). On the other hand, different studies reported that SB-216763 induced the formation of multiple long axons in hippocampal, cerebellar granular (CG), and DRG neurons (Padilla et al., 1997; Jiang et al., 2005; Yoshimura et al., 2005; Seira et al., 2011). Furthermore, it improved axon regeneration in lesioned neurons (Seira et al., 2011). Apparently it appeared that axon fate is dependent on the degree of inhibition of GSK-3, namely, strong inhibition of GSK-3 with high concentration of inhibitor inhibited axon growth, while weak inhibition promoted axon branching (Kim et al., 2006). An alternative explanation for these discrepancies relies on different culture length (Jiang et al., 2005). In any event, it has been clearly demonstrated that GSK-3 regulates neurite polarity and neurite outgrowth. The therapeutic activity of SBs has been further tested in several in vivo models. SB-216763 reduced ischemic cerebral damage in mice subjected to middle cerebral artery occlusion (Valerio et al., 2011), and enhanced locomotor recovery after spinal cord injury (Padilla et al., 1997). Like CHIRs, treatment with SBs inhibited NMDA-induced LTD (Peineau et al., 2009). In an AD model of mice injected with Aβ peptide, SB-216763 reduced Aβ neurotoxic effects including reduction in tau phosphorylation, caspase-3, and the activity of the stress activated kinase JNK (c-Jun N-terminal kinase; Hu et al., 2009). Administration of SB-216763 to disrupted-in-schizophrenia-1 (DISC1) knockdown mice ameliorated schizophrenic symptoms such as depressive behavior and hyper-locomotion (Mao et al., 2009). In the postnatal rat model, administration of SB-216763 reduced tau phosphorylation in the hippocampus (Selenica et al., 2007). In one of these studies, however, SB-216763 produced neurodegenerative-like effects and behavior deficits in healthy mice (Hu et al., 2009). This demonstrates that over-inhibition of GSK-3 may result in conditions that prevent neurons from operating normally. Thus, GSK-3 inhibitors should be used preferentially in pathologies associated with elevated GSK-3 activity. Finally, additional maleimide derivatives were recently identified by phenotypic screen for defects in zebrafish embryogenesis (Zhang et al., 2011). These compounds were potent GSK-3 inhibitors (Zhang et al., 2011), but their biological activity remains to be tested in relevant systems.

The amino thiazole AR-A014418 developed by AstraZeneca was shown to be a selective inhibitor toward GSK-3 when compared to other protein kinases including CDKs (Bhat et al., 2003; Figure 1). In their initial study, Bhat et al. (2003 showed that AR-A014418 was neuroprotective in apoptotic conditions induced by inhibition of PI3 kinase and prevented neurodegeneration in hippocampal slices exposed to Aβ peptide. Consistent with studies using SBs and 6BIO (Owens et al., 2003; Kim et al., 2006; Alabed et al., 2011), treatment with AR-A014418 prevented axon elongation in hippocampal neuronal culture (Shi et al., 2004). AR-A014418 has also been examined in behavior models. It was shown to produce an anti-depressive like behavior in the mouse FST (Gould et al., 2004a; Silva et al., 2008) and, in a manic behavior model of amphetamine-induced hyperactivity, it reduced manic activity (Kalinichev and Dawson, 2011). These results suggest that inhibition of GSK-3 may be beneficial in both depressive and manic episodes. Finally, in JNPL3 transgenic mice overexpressing mutant human tau, AR-A014418 reduced levels of aggregated insoluble tau in the brainstem (Noble et al., 2005), and, in an ALS transgenic mouse model, AR-A014418 attenuated motor neuron death and improved cognition (Koh et al., 2007). Yet, unexpectedly, AR-A014418 showed no effect on tau phosphorylation in the cortex or hippocampus in the postnatal rat model (Selenica et al., 2007). A structurally closed related compound AZD-1080 entered into clinical trials phase I for AD in 2006, but unfortunately the development has been discontinued.¹

The group of paullone compounds, in particular kenpaullone and alsterpaullone, are widely used in various experimental settings as GSK-3 inhibitors (Figure 1). Paullones are fused tetracyclic compounds that inhibit both GSK-3 and CDKs within the nanomolar concentration range (Schultz et al., 1999; Leost et al., 2000). A structurally similar compound, 1-azakenpaullone, is a more selective GSK-3 inhibitor (Kunick et al., 2004). Its derivative cazpaullone (9-cyano-1-azapaullone; Figure 1) has been recently characterized as a selective GSK-3 inhibitor (Stukenbrock et al., 2008). Both alsterpaullone and kenpaullone prevented neuron cell death in response to variety of insults including trophic deprivation, thapsigargin treatment, and mitochondrial stress (Takadera and Ohyashiki, 2004; Mishra et al., 2007; Petit-Paitel et al., 2009; Skardelly et al., 2011). Alsterpaullone was shown to reduce tau phosphorylation in cultured neurons (Leost et al., 2000; Selenica et al., 2007), and to block NMDA-induced LTD in hippocampal slices (Peineau et al., 2008). Kenpaullone decreased Aβ production in cells overexpressing APP (Phiel et al., 2003) and promoted differentiation of precursor cells into dopamine neurons (Castelo-Branco et al., 2004). This last supported the use of GSK-3 inhibition in treatment of Parkinson’s disease (Castelo-Branco et al., 2004). Recent work implicated a role for GSK-3 in SMA (Makhortova et al., 2011). Apparently, a search for small molecules that elevate the expression levels of SMN identified GSK-3 inhibitors as potential agents. Treatment with alsterpaullone verified this observation and showed that alsterpaullone slowed down the degradation of SMN in SMA human fibroblasts (Makhortova et al., 2011). Furthermore, the death of motor neurons induced by depletion of SMN was rescued by alsterpaullone (Makhortova et al., 2011). This study is a first indication of a therapeutic potential of inhibition of GSK-3 in SMA (Makhortova et al., 2011). Reports of alsterpaullone or kenpaullone in in vivo systems are limited. One study demonstrated the ability of alsterpaullone to reduce tau phosphorylation in the rat brain (Selenica et al., 2007).

Additional compounds were recently described as dual CDK/GSK-3 inhibitors. These include the purine derivatives pyrazolo [3,4-b] quinoxalines (Ortega et al., 2002), the pyrazolo[3,4-b]pyridine ring system (Chioua et al., 2009), the 9-oxo-thiazolo [5,4-f] quinazoline-2-carbonitrile derivatives (Loge et al., 2008), and the thiazolo[5,4-f]quinazolin-9-ones (Testard et al., 2006). Aloisines (6-phenyl[5H]pyrrolo[2,3-b]pyrazines) are a different class of dual CDK/GSK-3 inhibitors that inhibit the two kinases in the sub-micromolar range (Mettey et al., 2003). Aloisine A (Figure 1) is the most potent of those analogs tested and showed anti-proliferative effects in differentiated postmitotic neurons (Mettey et al., 2003). A series of bisindolylmaleimides were identified as potent GSK-3 inhibitors and were shown to enhance mouse ES cells self renewal in the presence of the leukemia inhibitory factor (LIF; Bone et al., 2009). Recent work identified TWS119, a 4,6-disubstituted pyrrolopyrimidine (Figure 1), from a phenotypic cellular screen for compounds with the ability to induce neuronal differentiation of pluripotent mouse ES cells; the compound proved to be a GSK-3 inhibitor (Ding et al., 2003). This result contrasts with earlier studies that demonstrated maintenance of pluripotency by other GSK-3 inhibitors (Sato et al., 2004; Ying et al., 2008). These differences could be due to differences between mouse and human ES cells, culture conditions, or differences in the developmental stages derivation (Welham et al., 2007). Finally, a novel series of macrocyclic polyoxygenated bis-7-azaindolylmaleimides were shown to be highly selective toward GSK-3 (Kuo et al., 2003; Shen et al., 2004). Their biological activity remains to be further elucidated.

Collectively, studies with ATP-competitive inhibitors verified that GSK-3 is essential for maintenance of normal neuronal activities, but also demonstrated an important role for GSK-3 in the etiological mechanisms of neurodegeneration and psychiatric disorders. Crystallographic data identified specific interactions within the ATP-binding pocket and with additional residues located at the surface of the C-lobe. These include Asp 133, Arg 141 Gln185, Asp200, and Arg220 and the conserved salt bridge of Lys 85/Glu 97 (Figure 3A). However, the limited specificity of ATP-competitive inhibitors is a serious drawback as demonstrated in series of studies that profiled many protein kinase inhibitors (Davies et al., 2000; Bain et al., 2003, 2007). It is possible that the high toxicity of this type of drugs prevented entering into the clinical phase or resulted in a failure in clinical treatments. It may be possible to improve specificity by exploiting unique features within the ATP-binding fold of GSK-3, but this strategy requires further validation. As a last caveat, the dual inhibition of GSK-3 and CDKs may be not disadvantageous. CDK5 which is largely restricted to neurons, is essential for regulation of many neuronal functions (Jessberger et al., 2009), and its aberrant activity has been implicated in variety of neurodegenerative conditions (Cruz and Tsai, 2004). CDK5 also serves as the “priming kinase” that phosphorylates prior its phosphorylation by GSK-3 (Li et al., 2006; Plattner et al., 2006). Thus, inhibition of both GSK-3 and CDK5 may be of therapeutic benefit, and the use of such dual activity inhibitors should not be discouraged until further evaluation in preclinical models.

As we have highlighted above, there are many compounds with different scaffolds able to inhibit GSK-3 in an ATP-competitive manner; these compounds may have adverse secondary effects if used in chronic treatment. The human kinome has more than 500 protein kinases that share a high degree of homology in the catalytic site, and in particular, within the ATP-binding pocket. Achieving kinase selectivity is one of the main challenges in the search and design of protein kinases inhibitors (Eglen and Reisine, 2009). ATP non-competitive GSK-3 inhibitors are likely to be more selective than those that inhibit ATP binding, since they should bind to unique regions within the kinase providing a more subtle modulation of kinase activity than simply blocking ATP entrance. This point is of utmost importance for GSK-3 modulation as a therapeutic approach, because only the aberrant GSK-3 activity should be inhibited.

There are different chemical families of organic compounds reported in the literature that do not compete with ATP in their GSK-3 inhibition and different binding modes to the enzyme have been described. The first reported family was the small heterocyclic thiadiazolidinones (TDZD) family (Martinez et al., 2002; Figure 2A). Although their mechanism of action has not yet been experimentally confirmed, a possible role has been postulated for an interaction with cysteine 199, a key residue located in the active site of GSK-3 (Mazanetz and Fischer, 2007). TDZD did not show inhibition over various kinases including PKA, PKC, CK-2, and CDK1/cyclin B. Treatment of primary culture neuronal cells with TDZD reduced tau phosphorylation, and treatment with diverse TDZDs such as NP00111 (Figure 2A), NP031112, NP03115, produced neuroprotective and antidepressant activities in several animals models (Luna-Medina et al., 2007; Rosa et al., 2008a,b).

With these others positive data, safety regulatory studies of one TDZD candidate was done with the aim to determinate the human dose. Very relevant for the translation from the bench to the treatment of AD patients are the results obtained in the chronic oral treatment for 3 months performed with TDZD compound named NP-12, to the double transgenic APP × tau rodent model (Ribe et al., 2005). Treatment with NP-12 improved cognitive performance in the Morris water maze test, while all histopathological features related to AD pathology such as beta-amyloid plaque load, tau hyperphosphorylation, gliosis, and neurons death were significantly reduced (Sereno et al., 2009). Of interest is the observation regarding the increase in insulin growth factor-1 (IGF-1) in mice brains of both on wild type and APP × PS1 mice, after an acute oral treatment with NP-12 for 5 days (Bolos et al., 2011). IGF-1 is a potent neurotrophic peptide with therapeutic value for many neurodegenerative diseases including AD pathology (Torres-Aleman, 2007). These increased IGF-1 levels in mice would mean that there is not only a direct action of GSK-3 inhibitors on the brain, but, there is also an effect on some peripheral signaling pathways (as activation of the IGF-1 transporter megalin), that could be modified with this innovative treatment improving the AD pathology.

Much of the work done in the hit-to-lead process, and lead optimization of TDZD heterocyclic family has been reported recently (Martinez et al., 2008). Two molecules have been well characterized. The first is TDZD-8, which is one of the pharmacological tools frequently used to study the biological and pathological roles of GSK-3 in cellular and animal models (Beaulieu et al., 2004; Cuzzocrea et al., 2006; Lipina et al., 2011). The second molecule is NP031112, also termed NP-12 or tideglusib. It is a brain permeable small molecule currently used in clinical trials phase II for AD and progressive supranuclear palsy (PSP).

Data from the phase IIa trial of tideglusib were recently reported and indicated a trend in improved cognitive abilities of the mild to moderate AD patients treated for 24 weeks (del Ser, 2010). The phase IIb trial for AD² is ongoing with more than 20 centers involved. FDA and EMEA have approved the orphan drug status for the development of tideglusib in the rare tauopathy PSP³. The TAUROS study is ongoing and results are expected to be finalized by the end of 2011.

The second family of compounds known as ATP non-competitive GSK-3 inhibitors is the halomethylketone (HMK) derivatives (Conde et al., 2003) which have been recently described as the first irreversible inhibitors of this enzyme (Perez et al., 2009a; Figure 2A). In this case, inactivation of the enzyme is due to the formation of an irreversible covalent sulfur–carbon bond between the key cysteine 199 located at the entrance to the ATP site of GSK-3 and the HMK moiety (Perez et al., 2011).

HMKs are cell-permeable compounds. They are able to decrease tau hyperphosphorylation on primary neurons cell culture after 2 h of treatment (Perez et al., 2009a). They are rather selective in a wide protein kinase panel and its off-target activity were determined on different CNS receptors binding assays without any significant positive data (Perez et al., 2009a). There are a couple of HMKs commercially available from different commercial sources that confirms the importance of this series of compounds as pharmacological tools for the study of GSK-3 physiology and pathology in different cell models (Yasuda et al., 2009).

Manzamines are complex β-carboline alkaloids isolated from Indo-Pacific sponges and characterized as having an intricate and novel polycyclic system (Hu et al., 2003a). Following a discovery program of GSK-3 inhibitors from marine sources, it was found that manzamine A (Figure 2B) inhibits human GSK-3β in vitro more than 70% at 25 μM (Rao et al., 2006). In order to identify the pharmacophore responsible for this new enzymatic inhibition, the potential GSK-3 inhibition of carboline and ircinal A, which can be considered the chemical precursors of manzamine A, were tested. Both moieties are inactive in their ability to bind to GSK-3, indicating the entire manzamine molecule is responsible for this activity. To further assess the potential of manzamine A in the treatment of AD, its ability to inhibit several different kinases (GSK-3β, GSK-3α, CDK1, PKA, MAPK, and CDK5) and decrease the hyperphosphorylation of tau protein mediated by GSK-3 in human neuroblastoma cell cultures was investigated (Hamann et al., 2007). Manzamine A specifically inhibits GSK-3β and CDK5 (the two key players in the hyperphosphorylation of tau protein in AD) with IC₅₀s of 10 and 1.5 μM respectively; it is ineffective toward others kinases tested (Hamann et al., 2007). Kinetic studies indicated an ATP non-competitive inhibition regarding GSK-3 (Hamann et al., 2007) while susbstrate competitive inhibition has been recently proved experimentally (Palomo et al., 2011).

Treatment of SH-SY5Y cells in culture with manzamine A at different concentrations (5, 15, and 50 μM) resulted in a decrease in tau phosphorylation at the GSK-3 epitope Ser 396 (Hamann et al., 2007). as quantified by a specific ELISA sandwich methodology. Cell survival was determined in parallel by measuring LDH release. Manzamine A constitutes a promising scaffold from which more potent and selective GSK-3 inhibitors could be designed as potential therapeutic agents for the treatment of diseases mediated by GSK-3 such as the AD (Wahba and Hamann, 2011). Recently a potential binding site of manzamine A with GSK-3 was identified, and this will provide new directions in substrate competitive drug design (Peng et al., 2011).

Very recently extracts and compounds obtained from the marine organism Ircinia sp., and, more particularly, the furanoterpenoids isolated from the Mediterranean sponges Ircinia dendroides, Ircinia variabilis, and Ircinia oros, have been claimed as inhibitors of GSK-3 (Alonso et al., 2005). Fractionation and purification of active components from these extracts, guided by a GSK-3 inhibition assay, resulted in the isolation of furanosesquiterpenoids as new GSK-3 inhibitors with potential use as therapeutic agents. Palinurin and one unknown metabolite called tricantin were mainly isolated (Figure 2B). Kinetic analyses of isolated compounds were performed and showed that tricantin inhibits recombinant human GSK-3β with an IC₅₀ value of 7.5 μM, whereas palinurin exhibited an IC₅₀ value of 4.5 μM. They are cell-permeable inhibitors and described as ATP non-competitive GSK-3 inhibitor able to reduce tau phosphorylation in cell cultures (Alonso and Martinez, 2006). Total synthesis of palinurin has been recently described providing a new source for this lead compound (Perez et al., 2009b). Different studies have been performed to determine structure activity relationships among a series of congeneric molecules and the bioactive conformation has been proposed (Ermondi et al., 2011). However, the binding mode to GSK-3 remains unknown.

Peptides have also been described as potential protein kinase inhibitors (Eldar-Finkelman and Eisenstein, 2009). The use of peptides, which copy natural motifs that specifically influence kinase activity and/or its intracellular interactions with associated partners, may be a promising approach for selective inhibition of protein kinases. Although substrate–competitive inhibitors have often been overlooked due their relative weak inhibition, they provide numerous advantages over the ATP-competitive inhibitors, mainly in selectivity. This is due to the fact that substrate recognition and types of interactions vary considerably among kinases, whereas ATP-binding domains are structurally conserved. In the case of GSK-3, the relatively weak affinities of substrate–competitive inhibitors, and, in general, ATP non-competitive inhibitors, may be advantageous. GSK-3 is essential for many aspects of neuronal function; hence, drastic inhibition may result in deleterious effects (as has been observed with some of GSK-3 inhibitors). Furthermore, pathological GSK-3 over-expression does not exceed two to threefold over normal levels. Thus, moderate-to-weak inhibition of GSK-3 (about 50% inhibition) is actually a desired approach for treatment of disorders associated with elevated activity of GSK-3. Finally because substrate–competitive inhibitors are more selective than ATP-competitive molecules, substrate–competitive inhibitors may be a favorable choice for clinical use.

The specific requirement of GSK-3 for pre-phosphorylated substrates supported the rational for the use of synthetic phosphorylated peptides as substrate–competitive inhibitors (Plotkin et al., 2003). The peptide L803-mts (11 residues) is a cell-permeable phosphorylated peptide derived from the GSK-3 substrate heat shock factor-1 (HSF-1; Figure 2C) that is very selective and inhibits cellular activity of GSK-3 within the low micromolar concentration range (Plotkin et al., 2003). L803-mts showed biological activity in diabetic models consisting with the paradigm of GSK-3 acting as a negative regulator of insulin signaling (Kaidanovich-Beilin and Eldar-Finkelman, 2006; Rao et al., 2007). Recent work further demonstrated the therapeutic activity of L803-mts in CNS models. Like other GSK-3 inhibitors, L803-mts promoted axon formation and elongation in hippocampal neurons (Kim et al., 2006). It was also shown to provide neuroprotection effects in neuron cultured cells exposed to 6-hydroxydopamine-induced cell death (Chen et al., 2004). In vivo treatment with L803-mts increased β-catenin levels in the mouse hippocampus and produced anti-depressive like activity in the FST (Kaidanovich-Beilin et al., 2004). Administration of L803-mts in a traumatic brain injury (TBI) model reversed depressive behavior in the injured animals (Shapira et al., 2007). Finally, L803-mts conferred low toxicity in neurons as compared with other GSK-3 inhibitors (Kim and Thayer, 2009).

Understanding of the mode of interaction of GSK-3 with its substrates is necessary for effective design and development of substrate–competitive inhibitors. A combined approach of mutagenesis and computational protein–protein docking analyses identified a novel substrate-binding site within the catalytic core of GSK-3β formed by Phe 67, Gln 89, Phe93, and Asn 95 (Ilouz et al., 2006; Licht-Murava et al., 2011), and Asp 181 as an additional binding site for the N-terminal pseudosubstrate (Ilouz et al., 2008; Figure 3B). These residues are spatially located near the ATP-binding site and the phosphate binding pocket that interacts with the phospho-serine moiety of the substrate (Dajani et al., 2001; ter Haar et al., 2001; Figure 3B). From studies of the binding mode of L803-mts with GSK-3 it was found that the inhibitor- and substrate-binding sites are not identical. Both substrate and L803-mts interact with the phosphate binding pocket, but the substrate interacts with the cavity bordered by Gln 89 and Asn 95, whereas L803-mts mainly interacts with Phe 93 and with a hydrophobic surface located away from the ATP-binding site (Licht-Murava et al., 2011; Figure 3B). This clarified our understanding of the different binding modes of substrates and inhibitors. Whereas the substrate requires appropriate alignment with the catalytic domain to allow catalysis, the inhibitor does not require an exact positioning within the catalytic cleft. Contacts other than those used by the substrate may be crucial for converting a substrate to an inhibitor. Based on this idea, new L803-mts variants were synthesized, and some were 3- to 10-fold better inhibitors than L803-mts (Licht-Murava et al., 2011). Taken together, substrate–competitive inhibitors of GSK-3 hold tremendous promise as potential therapeutics. An extensive understanding of molecular recognition of GSK-3 with its substrates and inhibitors should provide the basis for rational design and optimization of efficient and high affinity substrate–competitive inhibitors.

A different example for peptide inhibitors are two peptides derived from Frat (FRATide-39 residues) or Axin (Axin GID-25 residues) that compete with Axin binding to GSK-3, which in turn, leads to activation of Wnt signaling pathway (Hedgepeth et al., 1999; Thomas et al., 1999). Treatment with Axin GID induced multiple axon formation in hippocampal neurons (Jiang et al., 2005), and its exogenous expression in cerebellar granule neurons protected against trophic deprivation-induced cell death (Hongisto et al., 2003). These results could suggest that GSK-3 inhibition-mediated neuroprotection or axon formation involves activation of Wnt signaling and or an increase in Axin non-bound GSK-3 pool (Hongisto et al., 2003). However, from the therapeutic point of view, more relevant are small peptides.