Pyridinylimidazoles as GSK3β Inhibitors: The Impact of Tautomerism on Compound Activity via Water Networks

Abstract

Glycogen synthase kinase-3β (GSK3β) is involved in many pathological conditions and represents an attractive drug target. We previously reported dual GSK3β/p38α mitogen-activated protein kinase inhibitors and identified N-(4-(4-(4-fluorophenyl)-2-methyl-1H-imidazol-5-yl)pyridin-2-yl)cyclopropanecarboxamide (1) as a potent dual inhibitor of both target kinases. In this study, we aimed to design selective GSK3β inhibitors based on our pyridinylimidazole scaffold. Our efforts resulted in several novel and potent GSK3β inhibitors with IC₅₀ values in the low nanomolar range. 5-(2-(Cyclopropanecarboxamido)pyridin-4-yl)-4-cyclopropyl-1H-imidazole-2-carboxamide (6g) displayed very good kinase selectivity as well as metabolical stability and inhibited GSK3β activity in neuronal SH-SY5Y cells. Interestingly, we observed the importance of the 2-methylimidazole’s tautomeric state for the compound activity. Finally, we reveal how this crucial tautomerism effect is surmounted by imidazole-2-carboxamides, which are able to stabilize the binding via enhanced water network interactions, regardless of their tautomeric state.

Glycogen synthase kinase-3β (GSK3β) is a ubiquitously expressed serine/threonine kinase, which plays an important role in a variety of different cell signaling pathways. GSK3β plays a crucial role in almost every pathway leading to the hallmarks of Alzheimer’s disease^1,2 and is often referred to as a tau-kinase due to its capacity to modulate tau hyperphosphorylation. Overactivity of GSK3β has also been connected to an increased production of β-amyloids,³ neuroinflammation, and oxidative stress.⁴

GSK3β has also been associated with a plethora of other pathological conditions such as diabetes,⁵ cancer,⁶⁻⁸ schizophrenia,⁹ bipolar disorders,¹⁰ and osteoporosis.¹¹ Thus, GSK3β is considered to be an attractive drug target.

We recently reported a series of pyridinylimidazoles as dual GSK3β/p38α MAP kinase (MAPK) inhibitors and identified trisubstituted imidazole 1 as a potent balanced inhibitor of both target enzymes (Figure 1).¹² Furthermore, we observed that the removal of the para-fluorophenyl ring (2), which might be located in the hydrophobic region (HR) I of the ATP binding site, resulted in a significantly reduced GSK3β inhibition with a complete loss of activity against p38α MAPK. In this study, our aim was to further improve the activity of our pyridinylimidazole scaffold while shifting the selectivity toward GSK3β.

Figure 1.

Pyridinylimidazole-based lead compounds 1 and 2.

Recently, employing quantum mechanics (QM) in drug design and development has become increasingly popular. For instance, QM can be utilized to improve docking and scoring, determining protonation states and optimizing structures as well as ligand binding energies.^13,14 Also, the importance of water in drug design is gaining more and more emphasis.¹⁵ The effect of water network stabilization for ligand binding has been demonstrated e.g. by Klebe and co-workers.¹⁶ In addition, molecular dynamics (MD) simulations offer valuable insights into ligand binding interactions.¹⁷ By utilizing QM calculations with MD simulations, we disclosed the importance of tautomerism and water networks for the activity of our pyridinylimidazole compounds. In this case, the observed SAR could not have been clarified by simplified computational tools, such as docking, which has major caveats especially related to the solvent effects and dynamics of the system¹⁸ that were found determining for the activity differences among imidazoles.

Results and Discussion

Detailed descriptions of the synthetic sequences are reported in Schemes S1–S14 (Supporting Information, SI).

To address the alarming diffuse trend of increasing the inhibitor logP value in the lead optimization,¹⁹ we monitored the lipophilic ligand efficacy (LLE) of the synthesized compounds.²⁰ The LLE of our lead compounds 1 and 2 was already high with values of 5.10 and 4.40, respectively (Table 1).

Table 1. Activity and Physicochemical Parameters of 2-Methylimidazoles 3a–o.

^an = 2.

^bCalculated with Canvas (Schrödinger LLC).²²

^cAccording to QM Conformer & Tautomer Predictor of Maestro (Schrödinger, LLC, New York, NY, 2018) (seeSI and Table S3for details).

^dValues taken from Heider et al.¹²

^eDetermined by ELISA activity assay.²³

^fn.d. = not determined.

^gThe intramolecular H-bond to amide conformation excluded (seeTable S3, SI).

Initially, we examined the influence of different substituents reaching into the HR I of GSK3β. To this end, we synthesized a series of 2-methylpyridinylimidazoles with different cyclo-aliphatic and aromatic moieties attached to the imidazole-C4 position.

Replacing the aromatic ring with cycloalkyl moieties at the imidazole-C4 position (3j–m) resulted in a substantial loss of activity, leading to modest inhibitors of GSK3β in the micromolar range displaying complete inactivity against p38α MAPK.

Compounds with bulky moieties, such as 2-naphtyl (3h), were inactive, probably because of a steric clash in the HR I.

Replacement of the para-fluorophenyl ring with the 5-membered heteroaromatic rings thiophene (3d) or furan (3e) as well as other minor changes on the para-fluorophenyl ring, such as addition of a methyl (3o) or a second fluorine atom (3i), led to inhibitors with slightly increased IC₅₀ values compared to 1. Introduction of a pyrimidine (3g) at the imidazole-C4 led to a completely inactive derivative. The less lipophilic ortho- and meta-hydroxyphenyl derivatives (3a and 3b, respectively) turned out to be potent GSK3β inhibitors with sound LLE values, while the para-hydroxyphenyl compound (3c) displayed substantially diminished inhibition against both kinases. All potent GSK3β inhibitors bearing an aromatic ring in the HR I, however, remained potent inhibitors of p38α MAPK, except for 3e with 10-fold selectivity for GSK3β.

To elucidate the observed dramatic loss of activity of certain compounds, we investigated the influence of the R₁-substituent on the imidazole ring’s tautomeric state. To this end, we conducted QM calculations to assess the probability of different tautomeric states and conformations for the compounds (Figure 2, seeSIfor details). Indeed, QM results indicated that the active compounds generally prefer tautomer A or at least represent a reasonable population of this tautomeric state (Table S3, Supporting Information). For instance, the low nanomolar inhibitors 3a, 3b, 3d, and 3e display a clear preference for tautomer A (>71.5%). In turn, the less active compounds 3c and 3j–m display a clearly diminished population of tautomer A (<17%).

Figure 2.

(A) The imidazole ring has two potential tautomeric forms: in tautomer A, the nitrogen next to the R₁-group is unprotonated and can act as a H-bond acceptor, whereas in tautomer B it is protonated and can act as a H-bond donor. (B) The R₁-group influences the preferred tautomeric state and conformation. As an example, the lowest energy conformations (in solution) of the highly active compound 3d and of the poorly active compound 3l are shown here. Compound 3d prefers the active conformation with tautomer A, and 3l exists in the inactive conformation with tautomer B.

Obviously, the preference of a specific tautomeric state does not fully determine the compound activity. For example, the inactive compound 3g clearly prefers tautomer A (85.9%), but the pyrimidine group is suboptimal for the hydrophobic region (solvent preference). On the contrary, the highly lipophilic para-fluorophenyl substituent in compound 1 clearly increases potency, despite its preference for tautomer B. Interestingly, the inactive 2-methoxyphenyl derivative 3f appears only in a specific conformation as tautomer A, wherein the methoxy-group folds on top of the pyridinyl ring (Table S3, Supporting Information), which most likely impedes the binding. Overall, the tautomeric state preference partially, but not solely, determines the 2-methylimidazole activity.

Next, we attempted to improve the binding affinity of 1 via enhancing interactions at the solvent interface in the HR II. To this end, MD simulations (200 ns) demonstrated the potential suitability of compounds bearing N-(pyridin-2-yl)tri- or tetrazolepropanamide moieties (4a,b; Figures S2–S4, Supporting Information). Both displayed cation−π interactions with the Arg141 and improved solvent interactions combined with significantly lower log P values (Table 2). The simulation of 4a highlighted an identical binding mode for the triazole ring as observed in a crystal structure (PDB ID: 5K5N).²¹

Table 2. Activity and Physicochemical Parameters of N-(Pyridin-2-yl)tri- or tetrazolepropanamide Bearing 2-Methylimidazoles 4a–d.

^an = 2.

^bCalculated with Canvas (Schrödinger LLC).²²

Compounds 4a and 4b show a similar potency as the lead compound 1 but with enhanced LLE values. Removal of the para-fluorophenyl anchor resulted in compounds 4c and 4d, both displaying substantially reduced inhibitory potency. This clearly results from the negative log P values of these compounds, which seems to compromise their binding affinity (entropic penalty). Nevertheless, these compounds still exhibit mediocre target inhibition and fit nicely into the SAR of the series.

To overcome the highlighted tautomerism-related issues observed with the 2-methylimidazoles, we designed and synthesized a series of imidazole-2-carboxamides. Instead of the acceptor nitrogen of tautomer A, the 2-carboxamides could neglect the tautomeric state of the imidazole by presenting the amide oxygen toward the Lys85 region. This amino acid side chain has been successfully targeted by carbonyl groups; for example, Pfizer disclosed 6-amino-4-(pyrimidin-4-yl)pyridones interacting with Lys85,²⁴ while Bristol-Myers Squibb reported potent pyrrolopyridinones.²⁵ Moreover, we investigated ethyl esters as well as a hydroxyl moiety for their suitability to address the Lys85 residue.

In contrast to methylimidazole 2, imidazole-2-carboxamide 6a showed a >35-fold improvement in potency (LLE 7.49) (Table 3). In the case of compound 1, the introduction of a carboxamide moiety at the imidazole-C2 position (6h) did not substantially improve the inhibitory activity. Installation of an ethyl ester (5a and 5c) yielded mediocre inhibitors highlighting the importance of the amide function. Introduction of a hydroxy moiety at the imidazole-C2 methyl group resulted in 6i, showing a 2-fold reduction in GSK3β inhibition and no shift in the IC₅₀ value of p38α MAPK. In most cases, imidazole-2-carboxamides were better inhibitors of GSK3β than their corresponding 2-methylimidazole counterparts (e.g., 6c vs 3f, 6e vs 3o). Only the already potent 2-hydroxyphenyl 3a (vs 6b) displayed no improvement in activity.

Table 3. Inhibition Data and Physicochemical Parameters of Ethyl Imidazole-2-carboxylates 5 and Imidazole-2-carboxamides 6–8.

^an = 2.

^bCalculated with Canvas (Schrödinger LLC).²²

^cDetermined by ELISA activity assay.²³

^dn = 1.

The most striking differences existed in cycloalkyl substituted compounds 6f and 6g, exhibiting dramatically improved potency against GSK3β along with higher LLE values compared to the corresponding 2-methylimidazole derivatives 3m and 3j, which displayed only mediocre activities and preferred tautomer B. Moreover, all three compounds showed significant selectivity over p38α MAPK, and compound 6g was among the best from this series (GSK3β, IC₅₀: 0.003 μM; p38α MAPK, IC₅₀ >10 μM; LLE 7.64).

To further investigate these dramatic activity differences, we first confirmed that the amide group replacing the methyl group on the imidazole-C2 position had no influence on the tautomeric state preference (Table S3). As an example, the cyclopropyl-substituted compounds 3j and 6g display an analogous population of the tautomeric state A, namely 6.0% and 6.7% for 3j and 6g, respectively. Nevertheless, the potency of these two inhibitors is dramatically different, with the imidazole-2-carboxamide derivative 6g showing a 3 orders of magnitude higher activity than its methyl counterpart (3j).

To gain a deeper insight into the compound binding and the activity differences between 2-methylimidazoles and imidazole-2-carboxamides, we conducted a total of 8 μs MD simulations for the selected compounds bound to GSK3β in their preferred tautomeric state: 3j and 6g in tautomeric state B and 3a and 6b in tautomeric state A. The initial 1 μs MD simulations suggested unstable binding only for 3j, where its lipophilic cyclopropyl group is exposed to water in the HR I (Figure 3 and Figures S5 and S9, Supporting Information). Whereas with 6g the amide stabilizes a water network near Asp200 and the cyclopropyl group is shielded from the solvent, allowing the stable binding of tautomer B (Figure 3andFigures S5 and S9, Supporting Information). With tautomer A preferring 3a and 6b, the 2-hydroxyphenyl was shielded from solvent regardless of the methyl or amide group substituent on the imidazole ring (Figure 3 and Figures S7 and S9, Supporting Information). These observations were confirmed in unbiased simulations, conducted using another crystal structure as the starting configuration (Figures S6, S8 and S10, Supporting Information). Based on these data, the energetically favorable tautomer B of 3j does not support the suggested stabilizing interactions with the dynamic water network, which leads to water exposed HR I, whereas the preferred tautomer A of 3a is capable to shield the HR I from water via direct or water mediated interactions to Lys85 and maintain a stable binding (Figures S5–S8, Supporting Information). Thus, QM calculations with the MD simulations provide a potential explanation for the observed activity differences.

Figure 3.

Representative snapshots from MD simulations of compounds 3j (A), 3a (B), 6g (C), and 6b (D). (E) Compound 3j appears in a shifted binding orientation compared to 6g, whereas both 2-hydroxyphenyl derivatives (F) 3a and 6b display similar binding orientation in the simulations. The shift in the binding orientation of compound 3j occurs due to direct H-bond interaction from the imidazole to Asp200. This interaction, with the increased solvent exposure of the lipophilic cyclopropyl group (seeFigures S5–S6, Supporting Information), explains the 3 orders of magnitude difference in activity between 3j and 6g. The protein surface is illustrated in transparent light blue color and hydrogen bonds with yellow dashed lines in A–D.

Selected compounds (1, 3a, 3i, 6a, and 6h) were further tested for their GSK3β affinity in a previously reported ESI-QTOF assay (Tables S1 and S2, Supporting Information).²⁶ Using this completely different assay system, the potency trend of these GSK3β inhibitors obtained in the ADP-Glo activity assay was confirmed.

Moreover, inhibitors 3a and 6g were tested for their metabolic stability by incubation with human liver microsomes (HLM) over a period of 4 h (Tables S4 and S5, Supporting Information). Both compounds displayed excellent metabolic stability in this assay.

Further pharmacological profiling of the potent GSK3β inhibitor 6g included the evaluation of its ability to inhibit relevant CYP isoforms (Table 4). At a test concentration of 10 μM, imidazole-2-carboxamide 6g shows a clean CYP inhibition profile. Only low inhibition of CYP1A2 was observed.

Table 4. Inhibition of CYP450 Isoenzymes.

	% inhibition of CYP isoform @ 10 μM
Cpd	1A2	2C9	2C19	2D6	3A4
6g	25.5	0.8	–2.0	–2.8	–2.3

To assess the overall kinome selectivity, the most promising inhibitor 6g was screened against a representative panel of 68 diverse kinases (Table S6, Supporting Information), including the target kinase GSK3β and all members of the MAP kinases. At a concentration of 0.5 μM (>160-fold its IC₅₀ value on the target kinase) only CDK2, CDK9, JNK3, MLK2, and VEGFR2 were substantially inhibited by 6g, suggesting an acceptable kinome selectivity for this compound.

To confirm the biological activity of imidazole-2-carboxamide 6g, we tested it in a cell-based GSK3β assay. At the tested concentration of 1 μM, 6g inhibits GSK3β activity, in terms of inactive phospho-GSK3α/β (Ser21/9) increase, after 1 h of treatment in neuronal SH-SY5Y cells (Figure 4).

Figure 4.

Inhibition of GSK3β activity in neuronal SH-SY5Y cells. Cells were incubated with compound 6g [1 μM] for 1 h. At the end of incubation, the phosphorylation of GSK3α/β (Ser21/9) (inactive GSK3α/β form) was determined by Western blotting. Data are expressed as the ratio between phospho-GSK3α/β and total GSK3β levels normalized against β-actin and reported as mean ± SD of at least three independent experiments (***p < 0.001 versus untreated cells; t test).

Since CDK2, CDK9, and VEGFR2 are off-targets of 6g, we also determined its cytotoxic profile on different cell lines after 48 h of incubation. A margin of safety is given concerning cytotoxic side effects. In the case of tested nontumorigenic cells, the concentration to cause a 50% decrease in cell viability is in the low micromolar range, which corresponds to >1000-fold the IC₅₀ value of the GSK3β kinase activity assay (Figure S11, Supporting Information).

In the case of the selected tumorigenic cells (Figure S12, Supporting Information), compound 6g shows antiproliferative activity in a human breast cancer cell line (Figure S12B, Supporting Information), at 0.1 μM and less at 0.01 μM. This may be a relevant result, as GSK3β is a target in the treatment of human breast cancer.²⁷ Breast cancer patients with overexpression of GSK3β presented poor prognosis, and GSK3β inhibition suppressed the viability and proliferation of breast cancer cells in vitro.²⁸

In summary, we synthesized a diverse set of 33 novel di- and trisubstituted pyridinylimidazoles. The most potent GSK3β inhibitors 6f, 6g, and 6j were selective over p38α MAPK and had reasonable (CNS) druglike log P values. Imidazole 6g was metabolically stable in HLM, displayed a very good selectivity profile, and showed no affinity toward pharmacologically relevant CYP isoenzymes. Importantly, the LLE values of the series illustrate that the series’ potency is not driven by molecular obesity.²⁹

The SAR of the synthesized compounds was explained with QM calculations and MD simulations. The series represent an interesting example of the influence of the 2-methylimidazole tautomerism on the compound activity. The effect of tautomerism was indirectly surmounted by introducing the water-network stabilizing 2-carboxamide, thus exemplifying the importance to consider that subtle molecular differences may have significant influence on the dynamics of the system.

Glossary

Abbreviations

GSK

glycogen synthase kinase

HR

hydrophobic region

MAPK

MAP kinase

LLE

lipophilic ligand efficacy

MD

molecular dynamics

QM

quantum mechanics

Supporting Information Available

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acsmedchemlett.9b00177.

• Experimental details for preparation of the compounds; ESI-QTOF assay; cell-based GSK3β assay; molecular modeling, QM calculations and MD simulations; metabolic stability in HLM; kinome selectivity screening; inhibition of CYP450 isoenzymes; cell toxicity data (PDF)

Author Contributions

^# F.H. and T.P. contributed equally to this work.

The authors declare no competing financial interest.

Notes

QM output conformations and MD movies and full raw trajectories are freely available at http://dx.doi.org/10.5281/zenodo.3362889.