Small-Molecule Inhibitors of GSK-3: Structural Insights and Their Application to Alzheimer's Disease Models

The world health organization (WHO) estimated that 18 million people are struck by Alzheimer's disease (AD). The USA, France, Germany, and other countries launched major programmes targeting the identification of risk factors, the improvement of caretaking, and fundamental research aiming to postpone the onset of AD. The glycogen synthase kinase 3 (GSK-3) is implicated in multiple cellular processes and has been linked to the pathogenesis of several diseases including diabetes mellitus, cancer, and AD. Inhibition of GSK-3 leads to neuroprotective effects, decreased β-amyloid production, and a reduction in tau hyperphosphorylation, which are all associated with AD. Various classes of small molecule GSK-3 inhibitors have been published in patents and original publications. Herein, we present a comprehensive summary of small molecules reported to interact with GSK-3. We illustrate the interactions of the inhibitors with the active site. Furthermore, we refer to the biological characterisation in terms of activity and selectivity for GSK-3, elucidate in vivo studies and pre-/clinical trials.


Introduction
Protein kinases regulate diverse cellular functions and thus are frequently exploited in drug discovery programmes [1]. They regulate signal transduction processes by phosphorylating serine, threonine and tyrosine residues in key proteins. The signalling pathways involved contribute to the pathology in many diseases [2,3]. Glycogen synthase kinase 3 (GSK-3) was identified in the late 1970s and is a constitutively active, ubiquitous expressed serine/threonine kinase, which participates in a number of physiological processes ranging from glycogen metabolism to gene transcription [4]. Initially, the focus of pharmaceutical companies concerning GSK-3 was on diabetes mellitus, but since GSK-3 was linked to Alzheimer's disease (AD), the focus has moved from diabetes to AD. GSK-3 has been linked to all primary abnormalities associated with AD. GSK-3 interacts with different components of the plaque producing amyloid system, participates in phosphorylating the microtubule binding protein tau that contributes to the formation of neurofibrillary tangles, and has an influence on presenilin and other AD-associated proteins [4][5][6][7][8]. Two related isoforms of GSK-3 are present in mammalians, GSK-3α and β, which share 98% homology in their catalytic domains and have similar biochemical properties [9]. The isoforms are similar in their catalytic domains, yet differ significantly in their N-terminal regions [10]. Alois Alzheimer's first report of the neuropathological hallmarks of AD dates back to 1907 [11,12]. The histopathology of the AD brain is characterized by the presence of abnormal filamentous tau-protein inclusions in nerve cells and extracellular amyloid deposits [13,14]. Partially phosphorylated tau contains sequence motifs that support association with tubulin, which entails the stabilization of microtubules in AD-uneffected cells. The pathological hyperphosphorylation of these motifs destabilizes microtubules and consequently interferes with tubulin binding. The misfolding of hyperphosphorylated tau involves the E/Z-isomerisation of a phosphorylated Ser-Pro motive, which leads to the formation of insoluble neurofibrillary tangles (NFTs) and intraneuronal aggregates of paired helical filaments (PHFs) [9,15]. GSK-3 phosphorylates multiple sites on tau both in vitro and in vivo [9]. It exerts a central and crucial role in the pathogenesis of both familial and sporadic forms of AD. Early-onset forms of familial Alzheimer's disease (FAD) have been linked to mutations in amyloid precursor protein (APP), presenilin-1 (PS-1), and presenilin-2 (PS-2). These mutations adversely affect APP processing and result in the increased production of the 40-42 amino acid long β-Amyloid (Aβ) peptides, which are the major component of amyloid deposits. Several risk factors have been associated with sporadic Alzheimer's disease (SAD). The most prevalent factors are aging and the presence of specific ApoE isoforms, which have been implicated in Aβ clearance. Sporadic Alzheimer's disease can be caused by the activation of β-secretase, which results in enhanced formation of Aβ. Enhanced Aβ production or deficiency in Aβ clearance will result in the deposition of Aβ aggregates [4,16]. Recent work suggests that enhanced GSK-3 activity increases Aβ production [17]. Several studies support that GSK-3 inhibition leads to decreased Aβ production and a reduction in tau hyperphosphorylation [1].
A plethora of GSK-3 inhibitors has been described, and most of the biological effects were reported for in vitro and cellular studies [17]. These studies, the number of patent applications, and a successful phase II trial indicate that GSK-3 is a promising drug target for AD therapy, but the ultimate proof of concept has not been presented yet. GSK-3 is highly enriched in the brain, and several publications indicate that the GSK-3β isoform is a key kinase required for abnormal hyperphosphorylation of tau [18,19]. Spittaels et al. generated a double-transgenic mouse overexpressing human protein tau and constitutively active human GSK-3β and ascertained that this kinase is implicated in aberrant tau phosphorylation and in addition reduced tau binding capacity to microtubules [15,20]. The homology of the ATPbinding pocket in GSK-3α and GSK-3β presents an obstacle for the development of isoform selective inhibitors. All GSK-3 inhibitors developed until now are able to inhibit the two isoforms with similar potency, except Λ-OS1 (36), which showed a selectivity (up to 7 fold) for GSK-3α [8,21,22].
The structures of GSK-3β cocrystallized with several inhibitors have been solved by X-ray crystallography recently. These structures provide a remarkable possibility to design both novel and selective GSK-3 inhibitors. There are two fundamental options to inhibit GSK-3: non-ATP competitive inhibition and ATP competitive inhibition.
The non-ATP competitive inhibitors, for example, substrate competitive inhibitors, usually engage in a weakbinding interaction with the enzyme [23]. Non-ATP competitive inhibitors do not compete with the high intracellular ATP-concentration and thus offer a distinct pharmacological advantage. Moreover, the involvement of GSK-3 in several essential signalling pathways imposes a limit on the GSK-3 inhibition, complete inhibition will result in adverse events.

Small-Molecule Inhibitors of Glycogen Synthase Kinase 3
Several ATP competitive GSK-3 inhibitors from different structural classes are highlighted in this paper. The in vitro and in vivo data are summarized if available. It should be noted that the IC 50 values strongly depend on assay conditions and thus may vary 100 fold depending on ATP and enzyme concentration as well as incubation time. The interactions between the inhibitors and the ATP binding pocket are depicted.

Lithium
Chloride. Lithium chloride (LiCl) was the first GSK-3 inhibitor to be discovered. However, there are several other biological targets for lithium resulting in adverse events and a rather small therapeutic window. This effectively rules out the use of LiCl in the therapy of AD. The mechanism by which lithium inhibits GSK-3 is unknown, but two hypotheses were proposed: (a) lithium (Li + ) is a competitive inhibitor of GSK-3 with respect to magnesium (Mg 2+ ), but neither competitive to substrate nor to ATP (b) lithium inhibits potassium deprivation [25][26][27]. This paper focuses on small organic molecules that target specifically GSK-3, thus the activity of lithium salts will not be reviewed. Also covalent or irreversible inhibitors, like the TDZD NP-12, will be noted, but not further discussed as well as the FRATtide, a peptide derived from FRAT1, which binds to GSK-3 and blocks GSK-3 from interacting with axin [28,29].

Maleimide Derivatives.
Maleimide derivatives have been reported as scaffolds for ATP competitive GSK-3 inhibitors. Researchers at SmithKline Beecham Pharmaceuticals reported that 3-anilino-4-arylmaleimides 1-3 (Table 1) are potent and selective inhibitors of GSK-3 [30]. The compounds displayed GSK-3 inhibition (IC 50 ) at low nanomolar concentrations. The selectivity of compound 2 was evaluated using a panel of 25 kinases. The majority of the kinases showed less than 30% inhibition at an inhibitor concentration of 10 μM. The complex of 3 (named I-5) with GSK-3β ( Figure 1) elucidated the binding mode to the ATP pocket [31]. Herein, the maleimide nitrogen interacts with the carbonyl oxygen of Asp133 and the oxygen of compound 3 with the backbone nitrogen of Val135. Two additional interactions are observed between a carboxylate oxygen and Arg141 and with Gln185. Compound 4 (SB-216763) inhibited GSK-3α with an IC 50 of 34 nM (Table 2). Derivative 4 exhibited little or no inhibition of the 24 kinases tested in the panel [32]. Incubation of cerebellar granule neurones with this compound reduced the death rate in a concentration-dependent manner in response to either stimulus. The maximal neuroprotection was observed with 3 μM of SB-216763 [33]. A 60% reduction in GSK-3β activity levels was observed in the hippocampus, but not in the cortex of SB-216763 treated animals versus vehicle-treated rats [34]. The compounds 5-11 revealed good potencies against GSK-3. Concerning compounds 7 and 9-11, much higher potencies (IC 50 < 4 nM) were ascertained in comparison to compound 1 (Table 2) [36][37][38].
The compounds 10 and 11 demonstrated very good GSK-3 inhibition and improved selectivity against PKCβII, CDK2, and CDK4 and inhibit tau phosphorylation in a neuronal cell line [38]. Whereas compound 12 was analyzed by in silico docking, the binding mode of 13 (IC 50 = 0.6 nM) was determined by X-ray crystallographic analysis ( Figure 2) [40]. Similar to compound 3, the maleimide nitrogen of 13 interacts with the carbonyl oxygen of Asp133 and its oxygen with the backbone nitrogen of Val135. Furthermore, compound 13 forms a crucial hydrogen bond to Gln185. Maleimide 13 was screened against 317 kinases to provide data on kinase selectivity in order to predict potential safety issues. It was found that compound 13 inhibits only 36 kinases with >90% inhibition at 10 μM [40]. Compounds 14 and 15 are reversible GSK-3 inhibitors in the low micromolar or high nanomolar range and can be used as chemical precursors for the corresponding halomethylketones (HMKs). These HMKs are irreversible inhibitors, they alkylate Cys199, which is located in the ATP binding site of GSK-3 [41]. Further maleimide derivatives are listed in Table 3. All of them, compounds 16-27 showed IC 50 /K i values in the low nanomolar range [42][43][44][45][46][47][48]. The majority of these structures was synthesized by Johnson & Johnson Pharmaceutical R & D. The compounds 17 and 18 were evaluated against a broad panel of 55 protein kinases. 10 μM of derivative 17 or 18 inhibited GSK-3β kinase activity by 100% in the presence of 10 μM ATP. Compound 18 exhibited excellent selectivity for GSK-3β except for the moderate selectivity against PKCβII. Meanwhile, the bis-7-azaindolylmaleimide 17 exhibited high selectivity at GSK-3β against all kinases tested [43]. The activity of compound 23 was tested in a kinase panel containing 100 diverse protein kinases. An IC 50 of 3 nM and a 460 fold selectivity for GSK-3β over PKCβII was reported [46]. The X-ray structure analysis of cocrystallized compound 23 and GSK-3β is illustrated in Figure 3. The maleimide establishes key hydrogen bond contacts with the residues Asp133 and Val135. The compounds 25 and 26 were docked into the ATP binding site of GSK-3β (PDB code 1R0E). Similar to other maleimide derivatives, the maleimide nitrogen interacts with the carbonyl oxygen of Asp133 and the oxygen with the backbone NH of Val135. In addition to these main interactions, the hydroxymethyl group of compound 26 interacts with the side chain of Arg141 [48].  The benzo[e]isoindole-1,3-dione 28 (Table 4) displays an IC 50 of 304 nM in the presence of 100 μM ATP. It was evaluated in a panel of 22 representative kinases. 5 μM of compound 28 inhibited 91% of GSK-3β activity as well as 71% of CDK2/cyclin A and 53% of KDR (VEGFR2) [49]. The inhibition of GSK-3 leads to ectopic activation of the Wnt pathway during zebrafish development, resulting in a headless embryo. Thus, zebrafish embryos, which were treated with compound 28, had highly restricted brain defects that ranged from smaller eyes and forebrain to a complete loss of these structures [49]. Compound 29 displayed a lower IC 50   3 days of incubation [50]. The synthetic activity dedicated to this structural class has been strong and is ongoing [51][52][53][54][55].

Staurosporine and Organometallic
Inhibitors. Staurosporine (31) is a natural product from the bacterium Streptomyces staurosporeus [56]. It exerts antimicrobial, hypotensive, and cytotoxic activity [57]. Staurosporine (31) is also a potent GSK-3β inhibitor with a reported IC 50 value of 15 nM (Table 5) [58]. The cocrystal structure of Staurosporine (31) with GSK-3β is elucidated in Figure 4. Again, the maleimide 31 interacts with the Asp133 carbonyl oxygen and the backbone nitrogen of Val135. These are the only direct hydrogen bonds observed between GSK-3β and this inhibitor [31]. The other polar interactions, for example, water-mediated interactions are not denoted. The organometallic ruthenium complex 32 is a remarkably potent inhibitor of GSK-3 ( Table 5). The IC 50 for GSK-3α is 3 nM and 10 nM for GSK-3β. The compound was evaluated against 10 kinases and displayed activity against GSK-3 and RSK1 (IC 50 = 100 nM) only. The authors observed a water-mediated contact between the carbonyl ligand and the carbamide of Gln185 and assumed that this contributes to the specificity of GSK-3 [59]. Compound 33 showed improved selectivity. A test panel of more than 50 kinases was not significantly inhibited at 100 nM. Compound 33 was reported to activate the Wnt signaling pathway in cell culture experiments at nanomolar concentrations. The treatment of zebrafish embryos in early development by compound 33 caused malformations, for example, a head structure lacking the eyes and a stunted tail [60]. The S-diastereomer of compound 34 is a weaker GSK-3β inhibitor than the Rdiastereomer: IC 50 of 0.22 nM at 100 μM. Compound 34 (Rdiastereomer, Table 5) exerts the highest reported potency against GSK-3β with an IC 50 of 40 pM. Unfortunately, the evaluation in a kinase panel was not reported yet [61]. The complex of derivative 34 with GSK-3β ( Figure 5    lacking the carbonyl revealed the contribution of this carbonyl to potency and selectivity [22,61,62]. [67][68][69][70][71][72][73][74]. Indirubins are likewise indole derivatives. They are related to the naturally occurring indigo dyes, and their pharmacological properties have long been known in traditional Chinese medicine, for example, in the treatment of leukemias [15]. Except for the dye indirubin 38, all other noted indole derivatives (37, 39-50) revealed an IC 50 below 100 nM. Compound 37, the indirubin-3 -monoxime, was reported to display an IC 50 of 22 nM on GSK-3 inhibition (Table 6). This unselective compound is a strong inhibitor of the closely related kinases CDK1 and CDK5 [58]. Compound 37 did not affect p-tau levels in neither cortex nor hippocampus of P12 rats despite the high concentration of the compound in the brain (13 μM) [34]. Three important interactions of compound 37 and GSK-3β are displayed in the X-ray in Figure 8 [31]. Maternal Wnt activityz is necessary for dorsal axis formation in Xenopus laevis embryos, whereas head formation requires the inhibition of zygotic Wnt activity. The Xenopus laevis embryos were treated with compound 39 (BIO) or LiCl, respectively; in order to challenge these two GSK-3 regulated events in vivo. LiCl treatment leads to a hyper dorsoanteriorization at the expense of trunk and tail when applied during early cleavage stage. 39 exerted the same effect on the embryos. Compound 39 inhibited GSK-3 with an IC 50 of 5 nM, but also CDK1, CDK2, and CDK5 in a nanomolar range [63,64]. The binding mode of 39 in the ATP pocket of GSK-3β has been determined by X-ray crystallographic analysis. The four major interactions between the inhibitor and GSK-3β are shown in Figure 9. The nitrogen interacts with the carbonyl oxygen of Asp133 and the oxygen with the backbone nitrogen of Val135. Moreover, interactions with the oxygens of Val135 and Pro136 as well as the van der Waals contact between the bromine and Leu132 were observed. Compounds 41 and 43 are potent GSK-3 inhibitors and very selective against CDK1/cyclin B and CDK5/p25 [64].

Indole Derivatives. Several pharmaceutical companies have reported and patented indole derivatives as GSK-3 inhibitors
In 2007, this class of indirubine derivatives was patented by Meijer et al. [75]. This patent lists in vivo activity for several compounds. The derivatives 44-46 feature an extended amino side chain at position R 1 . They were prepared to enhance selectivity and water solubility versus compound 39. These compounds were reported to inhibit β-catenin Ser33/37/Thr41 phosphorylation by GSK-3. Compounds 44-46 were also less cytotoxic than compound 39 in the MTS reduction assay of SH-SY5Y neuroblastoma cells [65]. The compounds 47-50 are potent, yet unselective GSK-3 inhibitors [66].  of the most potent GSK-3β inhibitors and competes with ATP for binding to GSK-3β (Table 7). Compound 55 was evaluated in a kinase panel with 25 kinases and exhibited high selectivity for GSK-3α/β, CDK1/ cyclin B, CDK2/cyclin A, CDK2/cyclin E, and CDK5/p35. All measured IC 50 values were in the nanomolar range. Alsterpaullone 55 was reported to inhibit the in vivo phosphorylation of tau at AD-specific sites by GSK-3β [76]. Paullone 55 formulated in 20% DMSO/25% Tween-80 and injected s.c. led to a reduction in 43 kDa tau phosphorylation in cortex after 2 h, though not in the 49 kDa isoform. Furthermore, the phosphorylation levels of the 43 kDa and 49 kDa isoforms in the hippocampus were significantly decreased [34]. Most of the paullone derivatives were patented by Meijer and Kunick in 2001 [81]. The binding mode of 55 has been determined by X-ray crystallographic analysis ( Figure 10) [31]. The interactions of alsterpaullone and GSK-3β include two hydrogen bonds with Val135 and one interaction between the nitro group and the side chain amino group of Lys85. The compounds 56 and 57 inhibited GSK-3 as well as CDK1/cyclin B and CDK5/p25 in the nanomolar range [78]. The exchange of nitrogen and carbon in compound 58 and 59 decreased the inhibitory activity, but provided selectivity to 58, which is void of CDK-inhibitory effects [79]. However, compound 58 was tested against three kinases (GSK-3β, CDK1/cyclin  Table 7, leads to the most potent inhibitor 64. Compound 64, with an IC 50 of 0.8 nM is likewise nonselective for GSK-3. It preferentially inhibited the CDK/GSK-3 family in the nanomolar range and VEGFR-2, VEGFR-3, and Src in the submicromolar range [80]. Table 8). The precursor was identified by a pharmacophore search of the in house database. The compounds 65-67 were profiled against a panel of 25 kinases including GSK-3β. An excellent selectivity was obtained against the majority of the kinases. However, a significant inhibition of CDK2/cyclin A was reported [82]. Compounds 68 and 69 displayed improved potency, but nevertheless inhibited CDK2/cyclin A. Only compound 70 showed an excellent GSK-3 potency and improved CDK-2/cyclin A selectivity [83]. The IC 50 comparison of compounds 71-73 revealed that the Ph-4-OH motif is the best. The most potent inhibitor 74 (IC 50 = 0.8 nM) was tested against a panel of kinases and showed a reduction in the overall selectivity profile [84,85]. Compounds 75-77 showed an excellent selectivity against CDK-2/cyclin A. In addition, compound 77 demonstrated an excellent overall selectivity profile against all kinases of the panel [85]. Unfortunately, no further studies were published concerning this series.

Pyrazolamide Derivatives. GlaxoSmithKline identified another class of GSK-3 inhibitors in 2003 (
Takeda Pharmaceutical disclosed compound 78 in 2009. It is a very potent inhibitor of GSK-3 with an IC 50 of 2.3 nM for GSK-3α and 2.0 nM for GSK-3β. It had no inhibitory effect on 23 kinases, and only a weak inhibition was detected for CDK1/cyclin B, CDK2/cyclin A, CDK5, and JNK1. The cold-water stress model (CWS) with male C57BL/6Njcl mice was used to evaluate the in vivo efficacy of compound 78. CWS transiently induces in vivo tau hyperphosphorylation as reported in previous studies. The compound displayed highly potent inhibition of in vivo phosphorylation in CWS mice and reduced sarkosyl insoluble tau in old homozygous JNPL3 mice. Furthermore, compound 78 inhibited tau phosphorylation at GSK-3-directed sites in rat primary neuronal cells and mouse brain tissue [86]. Several structures containing a pyrazole core have shown promising results in animal models of diabetes, for structural information see denoted references.
The pyrazolamides developed by Wyeth-Ayerst Research, now Pfizer, showed significant plasma glucose-lowering activity (16-42% reduction) in genetically obese, diabetic db/db mice [87,88]. Novartis published a series of compounds with ED 50 values for glucose reduction in ob/ob mice of 3.0 mg/kg/day [89]. In 2011, the Merck Research Laboratories disclosed a potent human glucagon receptor antagonist with good pharmacokinetic profiles in four preclinical species. One of these compounds showed excellent oral pharmacodynamic efficacy in rhesus monkeys and transgenic mice by blocking glucagon-induced hyperglycemia [90]. Tables 9 and 10. This series is characterized by a high number of patent applications, particularly by Vertex Pharmaceuticals . Compounds 79 and 81 inhibited GSK-3β in the nanomolar range. The in silico docking of compound 80 into the ATP binding pocket of GSK-3β suggested that 80 makes two hydrogen bonds with the hinge region and one interaction with the positively charged Arg141 [28,91]. Unfortunately, the activity of these compounds in a kinase panel was not disclosed. The compounds 82-90 are weak inhibitors of GSK-3, and there are no public data for in vivo activity nor selectivity. Docking of 90 into the PDB structure 1Q5K of GSK-3β suggested that one nitrogen and the secondary amine form hydrogen bonds to the hinge region at the Val135 NH and carbonyl, respectively. The following compounds 91-93 showed IC 50 values below 100 nM, but 91 and 92 inhibit Aurora A in the nanomolar range. Derivative 93 (GSK-3α IC 50 = 61 nM, GSK-3β IC 50 = 41 nM) inhibited Aurora A at micromolar concentration only. In vivo studies of this compound in rats showed 34% oral bioavailability and good exposure [92]. Compound 94, a lead compound for ERK2 inhibition, is a nonselective GSK-3 inhibitor and inhibited all 5 kinases in the panel, which included GSK-3 [93]. The complex of compound 94 with GSK-3β ( Figure 11) provided the binding mode in the ATP pocket of GSK-3. The secondary amine interacts with the carbonyl oxygen of Val135 and the nitrogen of compound 94 with the backbone NH of Val135. Two additional interactions are present between OH and Asp200 and between the carbonyl and the primary amine of Lys85. The arylimidazoles 95 (CHIR 99021) and 96 (CHIR 98014) are very effective ATP competitive inhibitors of murine and rat GSK-3 (IC 50 ≤ 10 nM). Both compounds exhibited 500 to 10000 fold selectivity for GSK-3 versus 20 other kinases tested. These GSK-3 inhibitors rapidly lower blood glucose levels in diabetic rodent models and enhance glucose transport as well as GS activation in insulin-resistant oxidative skeletal muscle from  (Table 10). Compound 100 (GSK-3β IC 50 = 5 nM) displayed excellent selectivity against 25 kinases, including CDK2/cyclin A, which was inhibited with an IC 50    (EC 50 = 0.39 μM) [95]. Compound 101 was also profiled by cross-screening against a variety of kinases and showed an excellent overall selectivity against all kinases tested including CDK2. The docking of 101 into the ATP binding site of GSK-3β provided a likely interaction. One nitrogen and NH 2 of aminopyrimidine are anchored to the carbonyl moiety and NH of Val135 via hydrogen bond interactions. The 3pyridine moiety is located close to Lys85 of the conserved salt  bridge (Lys85/Glu97). The EC 50 value of compound 101 in the glycogen accumulation assay in L6 cells was 3.2 μM and thus 9 fold higher than for compound 100 [96]. Tables 11-13 list GSK-3 inhibitors featuring the oxadiazole moiety. The depicted 1,2,5oxadiazoles, 103-109, revealed IC 50 values from 0.1 μM to more than 1.1 μM. Compounds 104-106 were screened for inhibitory activity against a panel of 32 kinases at 100 μM ATP concentration. All compounds gave at least 100 fold selectivity for GSK-3 compared to CDK2. However, 10 μM of compound 104 inhibited other kinases like MSK1 and DYRK1A, the latter activity is of interest for neurodegenerative diseases. The above mentioned compounds 104-106 displayed both sufficient cell penetration and suitable water solubility [123]. The dioxolane 110 was identified as GSK-3β inhibitor (IC 50 = 65 nM; Table 12) by high throughput screening. The X-ray analysis of the GSK-3β cocrystallized compound confirmed the interaction with the ATP binding site ( Figure 12). One oxygen and a neighbouring hydrogen atom of the benzodioxole establish hydrogen bonds with the amide NH hydrogen and carbonyl oxygen of Val135 in the hinge region. The nitrogen atoms of the oxadiazole engage in an unique hydrogen bond relay network between Lys85-Glu97-Asp200 via two water molecules (interactions not shown). Further interactions are denoted in Figure 12. After an extensive derivatisation, for example, 111-114, compound 114 ( Figure 13) exerts a 28 fold increased activity (GSK-3β IC 50 = 2.3 nM) compared to its homologue 110 (Table 12). The selectivity of 114 was evaluated in a panel of more than 20 kinases to reveal more than 1000 fold selectivity against CDK1, CDK2, and CDK5. In addition, rat cassette dosing experiments of compound 114 were performed, which revealed low oral bioavailability. The cocrystal Table 11: Examples of 1,2,5-oxadiazole inhibitors with biological activity against GSK-3, selectivity, X-ray, and reference. structure of 114 bound to GSK-3β was not fully characterized due to the cleavage of the S-C bond in the X-ray beam ( Figure 13) [125]. The cocrystal structure indicated that the nitrogen of the benzimidazole forms a hydrogen bond with the backbone NH of the hinge region at Val135, and one nitrogen of the oxadiazole ring makes a hydrogen bond with the NH of Asp200. The hydrogen atom on the carbon of the benzimidazole made an additional hydrogen bond with the carbonyl oxygen of Val135 [125]. The compounds 115 and 116 are enantiomers just as compounds 117 and 118. The S-isomers 116 and 118 were found to be eutomers. It was reported that the S-isomers possessed good oral absorption in nonfasted Crl: CD(SD)IGS rats with a bioavailability of 72.8% for derivative 116 and 65.5% for derivative 118. Furthermore, the compounds exhibited favourable blood-brain barrier (BBB) permeability. Compounds 116 and 118 were tested for inhibitory activity against more than 20 kinases and displayed no significant activity. This indicates that these compounds are highly selective GSK-3 inhibitors. The binding mode of 116 in the ATP pocket of GSK-3β has been determined by X-ray crystallographic analysis ( Figure 14) [126]. One of the nitrogen atoms of the oxadiazole was revealed to interact via a hydrogen bond with the side chain of Lys85 and the other with Asp200. The oxygen atom and one carbon hydrogen of the benzofuran ring interact with the main chain of Val135. Another interaction was observed between one carbon hydrogen and the carbonyl of Asp133. Derivatives 116 and 118 were tested in vivo using the CWS model in mice. Here, tau phosphorylation was induced at several GSK-3β-directed sites such as Ser199, Thr205, Thr231, and Ser396 in mice. As reported, tau phosphorylation was significantly reduced by 35% for compound 116 and 38% for compound 118 [126]. Compound 116 was further investigated and revealed significantly decreased hippocampal tau phosphorylation as well as suppression of tau pathology without affecting amyloid β pathology [127]. A multistage virtual screening of 289903 molecules resulted in 59 hits of potential GSK-3β inhibitors, for example,     (Table 14) are nonselective GSK-3 inhibitors and showed moderate activity against CDKs [130]. This stands in contrast to the thiazolylurea 130 (AR-A014418), which strongly inhibited GSK-3 (IC 50 = 104 nM) but not any other kinase in the panel. AR-A014418 inhibited tau phosphorylation in transfected 3T3-fibroblasts in a dose-dependent fashion exhibiting an IC 50 of 2.7 μM. The neuronal loss was reduced  in the organotypic culture (N2A cells) and compound 130 by itself did not affect neuronal viability. The cocrystal structure analysis of AR-A014418 and GSK-3β revealed that this compound binds to the hinge region via three hydrogen bond interactions (Figure 15) [131]. Furthermore, compound 130 significantly reduced insoluble tau levels in the brainstem of JNPL3 mice when compared with vehicle treated animals [132]. Surprisingly, another research group reported that AR-A014418 showed no effect on the phosphorylation levels of neither 43 kDa nor 49 kDa tau in the cortex or hippocampus of postnatal model rats [34]. Novel compounds based on the scaffold of AR-A014418 were synthesized recently. They showed improved in vitro activity and reduced toxicity in the wildtype zebrafish embryo assay [133].

Miscellaneous
Heterocyclic Derivatives with GSK-3 Activity. Table 15 lists GSK-3 inhibitors featuring a benzimidazole core. There are neither in vivo assays nor selectivity data published for these potential metal chelators 131-134  moderately active GSK-3β inhibitors, but showed activity for CDKs (Table 15(a)) [135].
There is no report for an in vitro IC 50 of the synthetic xanthine propentofylline (PPF, compound 138, Table 15(a)), but studies in the Tg mouse model of AD indicated that  PPF exerts a dual effect: reduction of both pathological amyloidogenesis and tau phosphorylation while reducing the ratio of activated versus inactivated GSK-3β [136].
The pyrazolone 139 was identified as potential scaffold in a compound screen for novel GSK-3 inhibitors. Derivatisation has increased the potency of the derivatives (K i from 1.49 mM to 0.8 nM (Table 15(b)). The compounds 140, 141, and 143 showed selectivity in a kinase panel of 14 kinases. A cocrystallization was realized with compound 142 ( Figure 17) and provided the interactions of the pyrazole moiety with the GSK-3 backbone aminoacids Asp133 and Val135. In addition, the methoxy substituents make hydrogen-bonding contacts with both Asp200 and Lys85. These interactions and the hydrophobic contacts of the phenyl rings are thought to be responsible for the potency and selectivity [137]. A recent docking study of the pyrazolone 144 confirmed the interactions of compound 142 with GSK-3β. It is an active (IC 50 of 34 nM) and selective (kinase panel of 40 kinases) structural analogue to the previously described maleimides 9-27 (Tables 2 and 3). Furthermore, compound 144 was evaluated in a model of oxidative stress induced by homocysteic acid and displayed full neuroprotective activity at 1 μM. Additionally, it was able to reduce locomotor activity in the chlordiazepoxide/amphetamineinduced hyperactivity model in vivo [138]. The compounds 145-147 are moderate GSK-3 inhibitors and were not tested against other kinases to evaluate their selectivity (Table 15(b)). Compound 145 increased the glycogen content in the liver of Sprague-Dawley rats in a dose-dependent manner [139].
Compound 1 is a HTS hit, which was cocrystallized with GSK-3β and solved by X-ray crystallography ( Figure 18). The overlap of the inhibitor Compound 1 with a pyridone derivative revealed interactions with the catalytic triad, especially Lys85, and the protein backbone aminoacid Val135 of GSK-3β [140]. This leads to the latest ATP competitive GSK-3 inhibitors, the pyridones 148-150 (Table 15( 16.1 nM against GSK-3β. It was selective in a broad panel of kinases, including CK2 and CLK1. Compared to compound 148, compound 150 exhibited an improved in vitro human liver microsome intrinsic clearance value of 16.3 mL/min/kg. The in vivo CNS penetration assay demonstrated a good free brain to free plasma ratio and the assessment of potential genetic toxicology hazards was negative for compound 150 [140]. Last but not least, there were several GSK-3 inhibitors isolated from marine organisms. These alkaloids have the potential to provide new scaffolds comparable to the established indirubines. Manzamines, meridianins, hymenialdisine and dibromocantharelline inhibit GSK-3β in the μM range and display promising selectivity and in vivo results (Table 15(c); compounds 151-154) [141][142][143][144].

Activity and Selectivity Profiling
A plethora of GSK-3 inhibitors was discovered in recent years, and most of these displayed good-to-excellent inhibition of this kinase. However, selectivity and safety against other kinases remains to be a challenge. The structural analysis of the ATP competitive inhibitors may guide the development of more selective GSK-3 inhibitors. All ATP competitive inhibitors establish hydrogen bonds with the backbone atoms of Asp133 and Val135. Contact to these aminoacids is a key to enhance affinity to GSK-3, but it does not provide selectivity over other kinases. Moreover, Pro136 appears in several complexes to strengthen the interaction of the inhibitor with the backbone (Figure 19). One region of GSK-3 may offer privileged access to enhanced activity and selectivity: it is the region characterized by the aminoacids Lys85, Glu97, and Asp200. Lys85 was observed to form a saltbridge with Glu97 and simultaneously with Asp200, which is expected to be less significant and potent as the one with Glu97 [145]. The interaction of an inhibitor with this region holds potential to increase activity and selectivity for GSK-3. This interaction can be mediated via water molecules, as Ac tiv ity and sel ect ivit y Ac tiv ity and sel ect ivit y Ac tiv ity Ac tiv ity Acti vity and sele ctiv ity Acti vity and sele ctiv ity Figure 19: Schematic view in the ATP binding pocket of GSK-3β; important areas for activity and selectivity are denoted by PDB code 1I09 [146]. observed for AR-A014418 (130), or be established by direct contact. We observed in our dataset that a direct interaction with this region may cause a loss of selectivity (data not shown). There is a salt-bridge formed by Glu137 and Arg141 in the entrance area of the ATP pocket of GSK-3β. This region seems to contribute to the activity and selectivity of several inhibitors.
Another important interaction between GSK-3 and an organometallic inhibitor was reported by Bregman et al. They observed a water-mediated contact between the carbonyl ligand of the inhibitor and the carboxylate of Gln185 [59]. This water molecule was also found in other GSK-3β inhibitor complexes and seems to be responsible for an increased activity and especially for an improved selectivity towards GSK-3.
Docking studies with our inhibitors confirmed this interaction and further explained the selectivity of these inhibitors (data not shown). Feng et al. observed that the ruthenium-coordinated CO ligand of their inhibitor interacts with the flexible glycine-rich loop formed by Ile62, Gly63, Phe67, and Val70. This pocket, which is not shown in Figure 18, seems to be crucial for potency and selectivity [22]. Figure 18 illustrates in a simple scheme how to enhance activity and selectivity of inhibitors. The interaction with at least two of the three areas deems necessary to provide active and selective inhibitors.

Examples of In Vivo Tests
Over the last decade, several animal models have been developed to study tauopathies and other neurodegenerative disorders in vivo. Despite some obvious advantages of the diverse AD invertebrate systems, the vertebrate animal models of AD are generally favoured (for detailed invertebrate reviews, see [147][148][149]). Vertebrate models are evolutionary and morphologically closer related to humans, which makes the direct translation of experimental results easier and more reliable. A conditionally GSK-3β overexpressing mouse was reported by Lucas et al. in 2001. They demonstrated that GSK-3β overexpressed in vivo results in neurodegeneration and proposed that these mice can be used to study some aspects of AD [150]. In 2005, Perez et al. and Ribe et al. developed and characterized a double-transgenic mouse line based on overexpression of human mutant APP and tau [151,152]. They treated this transgenic mouse model with NP12, a non-ATP competitive GSK-3β inhibitor, and observed lower levels of tau phosphorylation, decreased amyloid deposition and prevention of memory deficits [18]. The APP-V7171 × Tau-P301L mice with combined amyloid and tau pathology and the GSK-3β × Tau-P301L mice with tauopathy only were reported by Terwel et al. in 2008. These models offer the possibility to explore molecular signals that act upstream and downstream of, or in parallel with GSK-3 isozymes [153]. Two transgenic mice models were developed to study the interaction between APP or Aβ and tau: the tripletransgenic model (3 × Tg-AD) harboring PS1 M146V , APP Swe and tau P301L transgenes and the TAPP mice [154,155].
The transgenic mice have been the major species used for modelling AD and frontotemporal dementia (FTD). JNPL3 mice are well-characterized transgenic mice that express human 4R0N tau with a FTDP-17 (P301L) mutation [156]. In particular, the levels of sarkosyl insoluble tau in JNPL3 mice increase in an age-dependent manner and comigrate with insoluble tau from AD and FTDP-17 brains. Thus, treatment with GSK-3 inhibitors should result in a significant reduction of sarkosyl insoluble tau, see, for example, compound 78 [86,132]. Besides JNPL3 mice, the transgenic zebrafish larvae have advanced as an AD model system. It combines many of the advantages of invertebrate and vertebrate models. The latest model is a Gal4/UAS-based vector system that efficiently generates transgenic zebrafish overexpressing high levels of human Tau-P301L or other diseaseassociated proteins. This tau-transgenic fish model could be an effective in vivo screening tool to identify quickly promising GSK-3 inhibitors and eliminate compounds without reasonable in vivo activity early in the screening process [157,158].
GSK-3 inhibitors have also specific effects on early wildtype zebrafish development when treatment occurs between 4 and 24 hpf. Thus this animal model can be used to test the efficacy of GSK-3 inhibitors in vivo [133,157]. Furthermore, there are two in vivo model systems with transiently induced tau hyperphosphorylation that were used to evaluate the activity of GSK-3 inhibitors. One of them is the cold-water stress (CWS) model, which causes a rapid and reversible enhancement of tau phosphorylation in the mouse brain at several GSK-3β directed sites such as pSer199, pThr205, pThr231, and pSer396 [125,159]. The CWS-model proves the significant reduction of tau phosphorylation in LiCl-or compound 78-treated mice [86,160]. The in vivo activities of several other GSK-3 inhibitors (e.g., compound 55 and 96) were demonstrated in the postnatal rat model [34]. This in vivo model takes advantage of the well-characterized GSK-3β expression level in the early and later life cycle of rats [161,162]. The Wayne State University is currently in phase IV with lithium carbonate against bipolar disorder whereas the University of Sao Paulo is in phase II with lithium carbonate against GSK-3 for AD and cognitive impairment. Lithium against AD was submitted by the National Institute of Neurological Disorder and Stroke and is in phase II, but there was no update since March 2008. Two compounds from Noscira, NP-12 (TDZD) and NP-103, are in the pre-/clinical trials. NP-12 (NP031112, tideglusib) is currently in a phase IIb clinical trial for AD. At the moment, NP-103 and CG-301338 from CrystalGenomics and an GSK-3β inhibitor from Takeda are in preclinical trials. The development status of XD-4241 from Cambrex, SB-415286 from GlaxoSmith-Kline, a GSK-3β inhibitor from Amphora, SAR-502250 from Sanofi-Aventis, CEP-16805 from Cephalon, and an GSK-3β inhibitor from Lundbeck are not reported. The failure or progress of these compounds into preclinical and clinical trials will stimulate or discourage further research.

Summary
The moderate inhibition of GSK-3 by selective inhibitors with excellent pharmacokinetic properties and excellent blood-brain barrier permeation holds high potential for the treatment of AD. The failure of the first potent GSK-3 inhibitor by Astra-Zeneca indicates the complexity of this target and the therapeutic window of GSK-3 inhibition in adult mammals. Information on the failure is rather limited, it may be due to potential toxicity of the chemical scaffold: P450-mediated metabolism of thiazoles has resulted in hepatotoxicity and thus failure of phase III candidates previously.
Biological characterization has advanced GSK-3 as a potential drug target, and the inhibition of this protein kinase by small molecules resulted in significant inhibition of tau phosphorylation. We have described a wide range of molecules that inhibit GSK-3 and discussed their properties. We focused on several inhibitors interacting with the ATP binding pocket of GSK-3β. The water molecule interactions are not incorporated in the figures; however, they may play a crucial role in the hydrogen bond network between the inhibitor and the aminoacids. The aminoacids, which are responsible for strong interaction with the enzyme, have been identified. Noteworthy are Asp133, Val135, Glu137, Arg141, Gln185, Asp200, and Arg220, which constitute important aminoacids for interactions with the binding pocket of GSK-3. The conserved salt bridge Lys85/Glu97 materializes as an interesting interaction partner. We summarized crucial biological findings and provided an overview on the in vivo effects of some inhibitors. Additional in vivo assays can be retrieved via the references denoted in the tables next to the structures. We summarized the GSK-3 inhibitors, which are in pre-/clinical trials with the therapeutic indication of AD. Yet, a review will be biased, and we request your pardon or input, if your favourite inhibitor or in vivo studies are not adequately referred to.

Outlook
GSK-3 is an intriguing enzyme, which plays important roles in the pathogenesis of several diseases, for example, diabetes, cancer, and AD. The literature is immense and quite often provides conflicting statements and observations for this kinase. For example, a few studies observed that Pin1 (peptidyl-prolyl cis-trans isomerase) knockout mice display tau hyperphosphorylation and that this enzyme might have an inhibitive role in phosphorylating tau and GSK-3β, thus protecting against AD [163]. Whereas another study ascertained that the GSK-3 inhibitor BIO (compound 39) may be useful in regenerative medicine, by reversibly maintaining human embryonic stem cells in an undifferentiated state [164]. A plethora of GSK-3 inhibitors has been published since it has been linked to AD. Naturally, the vast majority thereof was reported by pharmaceutical companies. Many potent inhibitors with good selectivity have been disclosed so far. The challenge of medicinal chemists will be to develop inhibitors, which translate their potent enzymatic inhibition into cellular settings and finally humans, who will tolerate a moderate GSK-3 inhibition only. A mild GSK-3 inhibition (∼35%) is needed because such an inhibition level provides sufficient insulin sensitization without elevation of β-catenin levels. Thus a moderate inhibition minimizes significant mechanism-based toxicities, ranging from hypoglycemia to tumorigenesis [165,166]. A further indication of a moderate GSK-3 inhibition is the application of lithium for the treatment of bipolar disorder. The GSK-3 inhibitor lithium is estimated to inhibit approximately 25% of total GSK-3 activity. It was used for the treatment of bipolar disorder since the 1950s without association of hypoglycemia, increased levels of tumorigenesis, or deaths from cancer [167]. Nevertheless, the established in vivo studies have to be accompanied by the investigation of pleiotropic activity and the determination of a safe therapeutic window for chronic GSK-3 inhibition in humans. The X-ray analysis of cocrystallized structures revealed how the inhibitors interact with the ATP-binding pocket and provide information about essential interaction partners to improve potency and selectivity. The comparison of the IC 50 values will be much easier, if only one stringent GSK-3 in vitro assay would be utilized, which employs a defined final ATP and inhibitor concentration as well as a standardized incubation time. Currently, Alon et al. determined that GSK-3β is responsible for the phosphorylation of the embryonic tau isoforms in birds, which harbors the GSK-3β gene only. In consideration of their and former studies, Alon et al. assume that GSK-3α and β have distinct roles in phosphorylating tau in adult and embryonic brain in nonvertebrates. Furthermore, they raise the hypothesis that specific inhibition of GSK-3α may be useful for therapeutic intervention in AD [168]. This supposition is enhanced by former siRNA experiments [169]. But all GSK-3 inhibitors developed until now inhibit the two isoforms of GSK-3 equipotently, except Λ-OS1 (36) (∼7 fold more selective for GSK-3α). Still most of the GSK-3 inhibitors fail in model animals despite of their good inhibitory activity in cell free assays. This is frequently due to the lack of selectivity, insufficient cell permeation, and poor blood-brain barrier permeability. Appropriate animal models were developed in transparent zebrafish to study preclinical efficacy, metabolic stability and toxicity, but their potential is not fully exploited. These in vivo tests are fast (3 days), relatively inexpensive and suitable for larger screening efforts in 96-well format. Furthermore, it has been observed that resistance arises during the therapy. The problem is that most of the kinase inhibitors are ATP competitive type I inhibitors. A new generation of type II inhibitors, which binds to the ATP site and extends into an allosteric site, may provide a solution [15,170,171]. Such novel type II inhibitors must be active, selective and permeate the human blood-brain barrier, which bears further limitations for drug development. However, several pharmaceutical companies continue to develop ATP competitive GSK-3 inhibitors [172,173]. Just a few pharmaceutical companies have started pre-/clinical trials addressing the druggability of GSK-3 inhibition. Noscira launched the phase IIb trial ARGO of Tideglusib (NP-12) to treat mildto-moderate AD patients in April 2011. The trial period will be 65 weeks with 2 dosage regimes of Tideglusib (500 and 1000 mg/day oral suspension, ClinicalTrials.gov identifier: NCT01350362). The ongoing clinical trials may lead to a paradigm shift, if the GSK-3 inhibitors display efficacy and safety.