Discovery of novel imidazopyridine GSK-3β inhibitors supported by computational approaches

molecules

Article

Discovery of Novel Imidazopyridine GSK-3β

Inhibitors Supported by Computational Approaches

Rosa Buonﬁglio, Federica Prati , Martina Bischetti, Claudia Cavarischia, Guido Furlotti and

Rosella Ombrato *

Angelini Pharma S.p.A., Viale Amelia, 70, 00181 Rome, Italy; rosa.buonﬁglio@angelinipharma.com (R.B.);

federica.prati@angelinipharma.com (F.P.); martina.bischetti@angelinipharma.com (M.B.);

claudia.cavarischia@angelinipharma.com (C.C.); guido.furlotti@angelinipharma.com (G.F.)

*

Correspondence: rosella.ombrato@angelinipharma.com

ꢀ

ꢁꢂꢀꢃꢄꢅꢆꢇ

Academic Editor: Maria Cristina De Rosa

Received: 14 April 2020; Accepted: 2 May 2020; Published: 5 May 2020

ꢀ

ꢁꢂꢃꢄꢅꢆ

Abstract: The interest of research groups and pharmaceutical companies to discover novel GSK-3

β

inhibitors has increased over the years considering the involvement of this enzyme in many

pathophysiological processes and diseases. Along this line, we recently reported on 1H-indazole-

3-carboxamide (INDZ) derivatives 1–6, showing good GSK-3β inhibition activity. However, they

suﬀered from generally poor central nervous system (CNS) permeability. Here, we describe the

design, synthesis, and in vitro characterization of novel imidazo[1,5-a]pyridine-1-carboxamide (IMID

1

) and imidazo[1,5-a]pyridine-3-carboxamide (IMID 2) compounds (

In detail, structure-based approaches and ﬁne-tuning of physicochemical properties guided the design

of derivatives 18 resulting in ameliorated absorption, distribution, metabolism, and excretion

ADME) properties. A crystal structure of 16 in complex with GSK-3 enzyme (PDB entry 6Y9S)

conﬁrmed the in silico models. Despite the nanomolar inhibition activity, the new core compounds

showed a reduction in potency with respect to INDZ derivatives . In this context, Molecular

7–18) to overcome such liability.

7–

(

β

1–6

Dynamics (MD) and Quantum Mechanics (QM) based approaches along with NMR investigation

helped to rationalize the observed structure activity relationship (SAR). With these ﬁndings, the key

role of the acidic hydrogen of the central core for a tight interaction within the ATP pocket of the

enzyme reﬂecting in good GSK-3β aﬃnity was demonstrated.

Keywords: glycogen synthase kinase-3

β (GSK-3β); 1H-indazole-3-carboxamide core (INDZ);

imidazo[1,5-a]pyridine-1-carboxamide core (IMID 1); imidazo[1,5-a]pyridine-3-carboxamide core

(

IMID 2); CNS permeability; molecular docking; molecular dynamics (MD); quantum mechanics

QM); X-ray crystallography; NMR characterization

1

. Introduction

Glycogen synthase kinase-3 (GSK-3) is a ubiquitous, multifunctional and constitutively active

serine/threonine kinase, involved in numerous pathways and over one hundred diﬀerent putative

substrates [1].

In humans, GSK-3 enzyme is encoded by two independent genes that regulate the expression of

two diverse isoforms, namely GSK-3

kinase domain [2] and substantial functional overlap [3–5].

GSK-3 is a signiﬁcantly networked enzyme, which regulates a plethora of physiological functions

in peripheral tissues and the central nervous system (CNS), spanning from metabolism and cell cycle

regulation to brain development. Accordingly, balanced GSK-3 regulation and signaling is key to

human physiology but often compromised in diverse disease conditions such as diabetes, cancers,

and brain disorders [ ]. As for CNS-related pathologies, aberrant GSK-3 activity has been reported

α and GSK-3β, with high sequence homology (~98%) in their

β

6

,7

β

Molecules 2020, 25, 2163; doi:10.3390 /molecules25092163

www.mdpi.com /journal /molecules

Molecules 2020, 25, 2163

2 of 29

in neurological or neurodevelopmental diseases, such as neurodegenerative disorders (Alzheimer’s

and Parkinson’s diseases), bipolar disorder, depression, and Fragile X syndrome [ 11]. Moreover,

clinical, genetic, and pharmacological investigations suggested that GSK-3 inhibition may attenuate

signaling dysfunction in such diseases [12 13]. Indeed, the role of GSK-3 in mood disorder was

8–

β

,

β

highlighted by studies on the mechanism of actions of the established treatments lithium and valproate,

which were both found to inhibit GSK-3β [14,15].

Based on this evidence, intensive eﬀorts have been made in the search for novel GSK-3β inhibitors

as innovative therapeutic agents, which, however, have not been translated into clinical practice yet.

Therefore, GSK-3β inhibitors are still needed.

On these premises, a discovery program started at Angelini Pharma with a computer-aided protocol

based on structure-based virtual screening and similarity search, which led to the identiﬁcation of the

1

H-indazole-3-carboxamide moiety (INDZ) as a new scaﬀold for GSK-3

optimization campaigns followed to optimize GSK-3 inhibition potency, as well as ADMET properties,

and resulted in INDZ derivatives 1–6 [17–19] (Figure 1 and Table 1).

β inhibition [16]. Subsequent

β

Figure 1. Exploration of IMID 1 and IMID 2 derivatives 7–18 starting from previous INDZ 1–6 to

improve CNS penetration.

Despite few derivatives showing promising in vitro proﬁles, enhanced blood brain barrier (BBB)

penetration was still desirable to hinder CNS disorders. Indeed, the BBB is a dynamic and functional

structure which poses a vast challenge in the development of drugs acting on the CNS [20]. The major

route of drug permeation through the BBB is passive diﬀusion, which can be boosted by ﬁne-tuning

several physicochemical properties. A variety of approaches has been put forward to guide the design

of BBB penetrant compounds based upon these features [21–23].

On these bases, starting from the INDZ core, compound physicochemical property evaluation

and computer-aided methods were applied to support a scaﬀold hopping approach leading to the

identiﬁcation of novel imidazo[1,5-a]pyridine-1-carboxamide (IMID 1) and imidazo[1,5-a]pyridine-

3-carboxamide (IMID 2) derivatives 7–18 (Figure 1 and Table 1) as putative GSK-3β inhibitors

with improved in vitro CNS permeability. Speciﬁcally, structure-based approaches were applied

to predict ligand–protein interactions, which were conﬁrmed by the high-resolution co-crystal

structure of compound 16 with human GSK-3β enzyme (PDB entry 6Y9S). Additionally, computational

investigations based on Molecular Dynamics (MD) simulations and Quantum Mechanics (QM) helped

to retrospectively rationalize the structure activity relationships (SARs) of the in vitro tested compounds.

NMR-based studies were also carried out to further substantiate the observed experimental data.

Molecules 2020, 25, 2163

4 of 29

2

. Results and Discussion

2

.1. Chemistry

INDZ derivatives

pyridines

Speciﬁcally, IMID 1 derivatives

1

–

6

were synthesized as previously described [17

–

19], whereas imidazo[1,5-a]

7–

18 were prepared according to the linear synthetic routes depicted in Schemes 1 and 2.

10 and 12 were obtained in a four-step synthesis starting from the

7–

commercially available 7-bromoimidazo[1,5-a]pyridine (Scheme 1). 7-bromoimidazo[1,5-a]pyridine

underwent Vilsmeier–Haack reaction in the presence of DMF and phosphorus oxychloride to

provide the corresponding aldehyde (19). Subsequent silver nitrate-mediated oxidation led to

the carboxylic acid 20, which was then coupled to 1-[1-(2-methoxyethyl)piperidin-4-yl]methanamine

and 1-(oxan-4-yl)methanamine under standard carboxylic acid-amine coupling reaction to aﬀord

amides 21 and 22, respectively. Final Suzuki coupling of 21 and 22 with the desired boronic acids gave

the target compounds

the additional conversion of the 7-bromo-N-[(oxan-4-yl)methyl]imidazo[1,5-a]pyridine-1-carboxamide

to the corresponding boronic ester 23 by treatment with bis(pinacolato)diboron and subsequent

Suzuki coupling with (3-bromopyridin-4-yl)methanol (Scheme 1).

IMID 2 derivatives 13 16 and 18 (Scheme 2) were prepared starting from the commercially

available 6-bromoimidazo[1,5-a]pyridine-3-carboxylate 24. Bromide 24 underwent standard Suzuki

7–10 and 12. The alkoxy-substituted pyridinyl derivative 11 was synthesized by

2

–

coupling reaction with the desired boronic acids to give the corresponding intermediates 26

and 29. Subsequent ester hydrolysis with lithium hydroxide provided lithium carboxylate salts 30

,

27

,

3

1

, and 33, which were ﬁnally coupled to 1-[1-(2-methoxyethyl)piperidin-4-yl]methanamine and

1

-(oxan-4-yl)methanamine to provide the target compounds. Similarly to compound 11, synthesis of

the alkoxy-substituted pyridinyl derivative 17 required the additional conversion of bromide 24 to the

corresponding boronic ester 25, which then entered the same three-step synthetic route as described

above (Scheme 2).

O

H

OH

Br

POCl , DMF

AgNO , NaOH

3

N

H O, rt, 3h

N

toluene,

2

100ºC, 1.5h

19 (57%)

20 (61%)

R₁

NH

R₁

NH

O

R B(OH) or R Br (for 11)

NH R

2

1

HATU/DIPEA

CsF, Pd(dppf)Cl

2

X

R₂

N

DMF, rt, 24-72h

dioxane/H O, 90°C, on

N

2

*

N

7: R =

N

, R =

(19%)

2

1: X = Br, R =

(56%)

O

1

2

O

N

*

O

*

8

: R =

, R =

(16%)

B pin , KOAc,

22: X = Br, R =

O

(78%)

1

2

1

O

*

PdCl dppf,

*

2

O

9

: R =

, R =

(34%)

1

2

dioxane/H O

N

2

O

B

*

100°C, on

23: X =

, R =

O

(76%)

O

1

O

10: R =

, R =

(28%)

(48%)

(61%)

1

2

N

OH

*

1

1: R =

O

, R =

1

2

N

*

O

12: R =

, R =

1

2

F

Scheme 1. Synthesis of imidazo[1,5-a]pyridine-1-carboxamide derivatives 7–12.

Molecules 2020, 25, 2163

5 of 29

O

OEt R B(OH) or R Br (for 17)

2

OEt

X

^C^s^F^,^P^d⁽^d^p^p^f⁾^C^l2

R₂

LiOH.H O

2

N

dioxane/H O, 90°C, on

2

MeOH/H O, 50°C, on

2

*

2

4: X = Br

26: R =

(64%)

B pin , KOAc,

2

N

PdCl dppf,

O

2

7: R =

(93%)

(51%)

(90%)

2

dioxane/H O

N

2

OH

O

100°C, on

25: X = _B_O(100%)

*

28: R =

2

*

N

*

2

9: R = _F

2

F

^R1

NH

O

OLi

NH R₁

HATU/DIPEA

2

^R2

R₂

N

DMF, rt, 24-72h

*

3

0: R =

(100%)

13: R =

N

, R =

(39%)

2

1

2

O

N

*

O

*

O

1: R =

(95%)

14: R =

, R =

(46%)

2

1

2

N

OH

*

2: R =

(100%)

15: R =

O

, R =

(9%)

2

1

2

N

*

O

*

3: R =

(93%)

16: R =

O

, R =

(22%)

(34%)

(15%)

2

1

2

OH

N

F

*

1

7: R =

O

, R =

1

2

N

F

*

1

8: R =

O

, R =

1

2

F

Scheme 2. Synthesis of imidazo[1,5-a]pyridine-3-carboxamide derivatives 13–18.

2

.2. Derivatives 1–6 and Rationale for New Design

The INDZ core was previously identiﬁed as an attractive scaﬀold for the development of novel

GSK-3 inhibitors [16 19]. Through in silico and crystallographic studies [16 17 19], it was found to

bind in the ATP binding domain of GSK-3 enzyme between the N- and C-terminal lobes (Figure 2a).

β

–

, ,

β

The INDZ moiety is located at the adenine binding site and contacts the hinge residues via three

hydrogen bonds (H-bonds) referred to as: 1) “deep” engaging the hydrogen (Ha) at position N-1

of the core with Asp133 carbonyl group, 2) “central” involving the N-2 of the core and Val135 N-H

group, and 3) “outer” between the hydrogen (Hb) of the carboxamide group and Val135 carbonyl

(

the three interactions are represented in Figure 2a as red, green, and orange dashed lines, respectively).

In addition, R and R substituents appended to the central core are oriented towards Arg141 in

1

2

the external solvent accessible part of the binding pocket and the inner cavity in proximity of

Lys85, respectively.

Compounds

representative derivatives of the INDZ series and served as starting point for an optimization exercise

towards the development of novel compounds with improved CNS penetration. Compounds 1–6

1–6, published ahead as potent GSK-3β inhibitors [17–19], are reported herein as

showed outstanding GSK-3

the most potent of the set (IC₅₀= 4 and 9 nM, respectively). Interestingly, both compounds

β

inhibitory activity in the nanomolar range, with derivatives

2

and

4

2

being

and

4

carry an isopropoxypyridinyl group as R substituent, resulting from 3 to 5-fold more potent than the

2

corresponding pyridinyl analogues

in complex with GSK-3 (2.08 Å, PDB entry 6Y9R) supported the expected binding mode (Figure 2b).

Here, the piperidine chain was solvent exposed towards Arg141, the INDZ was anchored to the hinge

1 and 3, respectively. The high-resolution X-ray crystal structure of

2

β

Molecules 2020, 25, 2163

6 of 29

domain via the tridentate H-bond interaction, the pyridine formed a H-bond with the catalytic Lys85,

and the isopropoxyl group faced the ribose binding site. In proximity of the ligand, a network of

hydration sites was found involving residues like Ile62, Leu132, Thr138, Gln185, Asn186, and Asp200.

Interestingly, MD simulations (see below) highlighted that the isopropoxyl group of

involved in additional water-mediated interactions, justifying the boost in potency with respect to the

pyridinyl analogue . Replacement of the R pyridinyl group of with a (hydroxymethyl)pyridinyl (

or a di-F-phenyl ( ) also retained double-digit nanomolar inhibition activity. Finally, when comparing

the enzymatic potency of the pyridinyl compound against or the isopropoxypyridinyl derivative

against , it appeared that the two matching pairs resulted almost equipotent, suggesting that the

R substituents do not signiﬁcantly impact on the GSK-3 inhibitory activity. This is in line with the

observed INDZ binding mode, where the R group is solvent oriented.

2 might be also

1

3

5)

2

6

1

3

2

4

β

1

Figure 2.

extracted from crystallographic studies discussed elsewhere [16,17,19]. The 2D structure of the INDZ

core with explicit Ha and Hb atoms is also showed; ( ) X-ray co-crystal structure of inhibitor in

complex with GSK-3 enzyme (PDB entry 6Y9R). The deep, central and outer H-bond interactions are

(a) Generic binding mode of the INDZ core in the ATP binding domain of GSK-3β enzyme

b

2

β

represented as red, green, and orange dashed lines, respectively. The most relevant residues of the

ATP binding site are showed in light blue (for clarity, light transparency is used for some amino acids).

Relevant water molecules around inhibitor

.9.4 [24].

2 are also displayed. Both ﬁgures were prepared with VMD

1

Alongside enzymatic potency evaluation, preliminary brain permeability needs also to be assessed

for compounds to be successfully advanced in the CNS therapeutic area. Indeed, BBB penetration

is still the bottleneck in brain drug development and is the single most important factor limiting the

future growth of neurotherapeutics. The MDCK-MDR1 cell line is considered a good surrogate of

the in vivo BBB model. Therefore, we report herein on compounds 2, 3, 4, and 6 that, being the most

potent derivatives of the dataset, were selected as INDZ representatives for experimental testing in

an in vitro MDCK-MDR1 permeability assay (Table 1) [25]. The permeability coeﬃcients (Papp) were

calculated by measuring the transport of the compounds from the apical (A) to the basolateral (B)

direction (Table 1 and Equation (1)). Eﬄux Ratio (ER) was also measured (see Equation (2)) to quantify

eﬄux transport across a cell monolayer expressing P-glycoproteins (P-gp), by assessing transport in

both directions (A-B and B-A) [26].

−

6

Interestingly, compounds

3

and

3.3 × 10 cm/s, respectively), whereas compounds

.5 10 cm/s). Moreover, high ER values were obtained for derivatives

and . The observed experimental data could be rationalized by analysis of the

physicochemical proﬁle. Indeed, several properties have consistently been found to be relevant for

6

showed moderate permeability (Papp = 15.8

and resulted only poorly permeable (Papp

and and lower ratios

×

10 and

−6

1

2

4

<

−

6

×

2

4

for compounds

3

6

Molecules 2020, 25, 2163

7 of 29

optimal brain exposure, such as lipophilicity, expressed by the calculated logarithm of the octanol/water

partition coeﬃcient (cLogP) or calculated logarithm of the octanol/water partition coeﬃcient at

physiological pH 7.4 (cLogD), topological polar surface area (TPSA), ionization state (pK

of H-bond donors (HBD) and acceptors (HBA), rotatable bond (RB) count, and molecular weight

MW). Up to now, many published works endeavor to provide guidance in terms of physicochemical

a

), number

(

property range for the design of CNS penetrant molecules [27]. Many of these studies have focused on

the proﬁle of marketed CNS drugs. As an example, the median values derived from an analysis of

marketed CNS drugs conducted at Ely Lilly deﬁne a CNS drug proﬁle as: cLogP = 3.1, cLogD = 1.9,

a

TPSA = 40.5 Å, most basic pK = 7.9, HBD = 1, HBA = 4, RB count = 4, and MW = 300. Obviously,

these numbers are median values, and there are many marketed drugs with properties at either end

of the range [27]. Overall, CNS drug property studies, despite providing diﬀerent values, came to

the common conclusion that CNS drugs are generally smaller, more lipophilic, with fewer HBDs,

and lower TPSA compared with oral non-CNS drugs [21]. At this regard, cLogD (pH = 7.4), TPSA,

the most basic pK

a

, HBD, and MW of compounds 1–6 were calculated here with ACD/Percepta 2017.2,

for permeability estimation. Overall, the higher Papp values of derivative

3

and

6

with respect to

2

and

4

6

4

could be ascribed to a more desirable CNS proﬁle, characterized by higher lipophilicity (in particular

), reduced H-bond potential (TPSA), and lower molecular size (Table 1). Moreover, derivatives

showed higher ER compared to and

2

3

6, a result in agreement with the observation that increasing

the numbers of HBA and MW would confer greater likelihood of P-gp eﬄux [28,29].

In the context of CNS drug design, the HBD count, an important H-bond capacity descriptor,

is a critical CNS diﬀerentiating parameter [21–23]. Indeed, it has been reported that increased

HBD count is associated with lower passive solubility and higher risk of interaction with eﬄux

transporters. Particularly, compounds with HBD > 2 are highly unlikely to be brain penetrant.

Therefore, we conceived a new scaﬀold hopping design strategy to reduce the HBD count of the INDZ

core (HBD = 2) and boost the BBB penetration of the current series. In particular, we replaced the

indazole N-Ha with a C-Ha group and shuﬄed the ring nitrogen atoms resulting in the IMID 1 and

IMID 2 cores (Table 1). These modiﬁcations were intended to reduce the acidity of the Ha atom so to

enhance permeability, while retaining the tridentate H-bond interaction with the hinge residues and

enzymatic potency.

We expected SAR to be reasonably transferrable from the INDZ to the IMID 1 and IMID 2 series;

thus, new derivatives

7–18 were obtained by decorating the IMID 1 and IMID 2 cores with the same R₁

and R substituents discussed above for INDZs 1–6.

2

.3. In Silico Investigation of Novel 7–18 Derivatives

Taking advantage of the crystal structure of 2 in complex with GSK-3β, we ran Glide Standard

Precision (SP) ligand docking [30

the binding mode and the stability of the new series compared to the related INDZ compounds

see Materials and Methods for details). It followed that the docked ligands accommodated the central

core in the ATP binding site similarly to the INDZ group, while major variability in the orientations

of R and R groups was observed. Such conformational variability was mainly ascribed to: (i) the

,31] and Molecular Dynamics (MD) simulations [32] to evaluate

(

1

2

inherent ﬂexibility of the R appended moieties, (ii) the open solvent-exposed binding site occupied by

1

the R groups, and (iii) the wide binding site ﬁlled by the R residues.

1

2

Figure 3 highlights the good overlap of INDZ-IMID 1-IMID 2 triads bearing the same R and

1

R groups.

2

Deep visual inspection and energetics considerations (from

−

8.95 to 7.72 kcal/mol docking

−

score) led to the selection of one pose per ligand for further studies (Table 2). For a comprehensive

understanding of the IMID 1 and IMID 2 binding mode, the RMSD of the triplets was ﬁrst calculated

based on the common substructure of the core scaﬀold and carboxamide portion (excluding hydrogen

atoms), while the more ﬂexible R and R groups were disregarded (Table 2). In the match pair analysis,

1

2

the indazole was considered as reference structure.

Molecules 2020, 25, 2163

8 of 29

Figure 3. Binding poses of INDZ, IMID 1, and IMID 2 derivatives, reported as yellow, cyan, and orange

ball and sticks representation, respectively, in complex with GSK-3 . Each panel represents a triad

bearing the same R and R groups, ( 13, ( 14, ( 15, ( 10 16, ( 11 17, ( 12 18.

β

a

)

1-

7-

b)

2

-8

-

c)

3-

9-

d)

4

-

e)

5-

-

f

)

6

-

1

2

The most relevant residues of the ATP binding site are also showed in grey. Deep, central, and outer

H-bonds are represented in red, green, and orange, respectively, while dark dashed lines refer to other

interactions involving Phe67, Lys85, and Gln185.

Table 2. Glide SP docking score of the selected poses of derivatives

poses within each triplet (INDZ vs. IMID 1 and IMID 2 analogues). RMSD based on the core and

carboxamide portion. ²RMSD based on the whole molecule. Ref. stands for reference compound.

1

–

18. RMSD of the docking

1

Glide SP Docking

Score (kcal/mol)

RMSD Core +

Carboxamide (Å)

RMSD (Å) ²

ID

Core

1

7

INDZ

IMID 1

IMID 2

−8.48

−8.58

−8.69

−8.77

−7.94

−8.44

Ref.

0.44

0.34

Ref.

1.24

0.76

1

3

2

8

INDZ

IMID 1

IMID 2

Ref.

0.45

0.43

Ref.

1.36

0.92

1

4

3

9

INDZ

IMID 1

IMID 2

−8.88

−8.28

−8.25

−8.44

−7.91

−7.72

−8.16

−8.51

−8.61

−7.96

−8.14

−8.95

Ref.

0.15

0.19

Ref.

0.70

0.39

1

5

4

INDZ

IMID 1

IMID 2

Ref.

0.18

0.15

Ref.

1.00

1.51

1

0

6

5

INDZ

IMID 1

IMID 2

Ref.

0.22

0.34

Ref.

0.72

0.28

1

7

6

INDZ

IMID 1

IMID 2

Ref.

0.26

0.67

Ref.

1.08

1.42

1

2

8

Molecules 2020, 25, 2163

9 of 29

The new cores showed little displacement, in the range 0.15–0.67 Å RMSD, indicating very

high similarity to each other in the binding site. For sake of completeness, RMSD of the common

substructure of the whole molecules is also reported in Table 2, showing more variability within each

triplet when R and R groups are considered.

1

2

In this work, MD was intended as a postprocessing tool to validate and reﬁne Glide SP docking

solutions. In addition, we aimed to evaluate the complex stability since co-crystal structures were

unavailable at that time. The basic assumption was that a bad docking pose would have generated

an unstable MD trajectory, during which the ligand could even leave the binding site. Conversely,

a meaningful docking pose would have displayed stable and speciﬁc interactions with the target,

showing a low RMSD over time, with respect to the starting conformation [33].

Based on the docking results, compounds seemed to reasonably occupy the ATP binding site of

GSK-3

see Materials and Methods). To deﬁne the optimum MD time scale, 500 ns simulations of

in complex with GSK-3 enzyme were initially run. After a deep analysis, we found that the ﬁrst

00 ns were representative of the overall 500 ns trajectories. In fact, the main structural events were

β

kinase. To conﬁrm this hypothesis, we ran MD simulations of all derivatives using Desmond

(

2, 8, and 14

β

1

generally preserved and the interaction fraction with the most relevant residues of the binding site were

comparable (Figure S1 in Supplementary Materials). Therefore, a 100 ns time scale was picked for all

compounds as a good compromise between structural sampling thoroughness and computational cost.

A single 100 ns MD simulation per compound was assumed to provide exhaustive conformational

sampling given that the same combination of the R1 and R2 groups (considered the most ﬂuctuating

portions of the compounds) was simulated three times within each triplet.

Both protein and ligand RMSDs from the initial structures were calculated for each MD trajectory

showing good structural stability of the whole systems (Figure 4). In detail, the former spanned 0 to

2

.66 Å and the latter 0 to 4.25 Å. Larger ligand RMSD values were observed in compounds bearing the

R piperidine group, which ﬂuctuated in the solvent-exposed pocket. For each complex, RMSD of

1

protein and ligand resulted to be quite comparable, meaning that the ligand thoroughly occupied the

initial binding site without diﬀusing away. Ligand RMSF was also evaluated to investigate the most

ﬂexible moieties of the inhibitors (Figure S2). In all systems, R and R groups showed the highest

1

2

RMSF values compared to the stable core scaﬀolds.

Figure 4. RMSD evolution of (

MD simulation using frame 0 as reference. The RMSD ranges are 0–2.66 Å in a and 0–4.25 Å in b

dark green-yellow-white color scale corresponds to 0 to 4.25Å values).

a) protein and (b) ligand with respect to the protein, for each 100 ns

(

The pattern of H-bond interaction involving the hinge residues was essentially maintained along

all the simulations. In detail, Val135 carbonyl interacted with carboxamide N-Hb via stable H-bond

Molecules 2020, 25, 2163

10 of 29

with comparable persistence between INDZ and corresponding IMID 1 and IMID 2 compounds along

the MD trajectories. Conversely, the interaction occurrence of the central H-bond within each triad was

variable, with INDZ derivatives contacting Val135 more persistently than IMID 1 and IMID 2 analogues

(

Table S1). The inner H-bond involving Asp133 was found in all systems although interaction distances

and angles changed from INDZ to IMID 1 and IMID 2 core scaﬀolds (see below). These results suggest

that INDZ derivatives are more robustly anchored to the hinge residues than IMID 1 and IMID 2

analogues. Nevertheless, none of the simulations run on the three cores showed ligands moving

away from the binding pocket. The amide carbonyl was exposed to water molecules through which it

interacted with Ile62 and Gln185 to a diﬀerent extent in all trajectories spanning from 10 up to 50%

persistence (Figure S3). Albeit not preserved during the entire simulations, this along with the outer

H-bond seemed to stabilize the amide group preventing its rotation. Diﬀerent results were obtained

for the analyzed R and R groups. In 1-7

-

13 and

2-8-14 systems, the highly ﬂexible piperidine group

1

2

ﬂuctuated in the open solvent-exposed pocket and contacted Gln72 and Pro136 via intermittent direct

or indirect interactions (Figure S3). A water-bridged H-bond involving the R oxanyl ring was mostly

1

observed with Pro136 in the remaining systems (Figure S3). Other solvent-mediated contacts were

found along all MD simulations between the R group and Tyr134, Glu137, and Arg141. Overall,

1

such indirect bonds randomly observed along the simulations reﬂected the high ﬂexibility of the R₁

groups and the open binding pocket in which they lied. On the other side, aromatic and heteroaromatic

rings variably interacted with the surrounding residues. In the 1-7-13 triad, the pyridine system

stably occupied the region around the positively charged Lys85 residue, but optimal geometry for

a stable contact was found to a variable extent in the three systems. Pi–pi interaction between the

heteroaromatic ring and Phe67 was detected along the trajectories moving the nitrogen atom away

from Lys85. The meta-pyridinyl substituent of 2, 8, and 14 derivatives was found to engage Asp200

from the highly conserved motif DFG via a water-mediated H-bond. Consequently, the pyridinyl

ring fairly rotated to ease the rearrangement of the 5-substituent preventing a stable contact of its

nitrogen atom with Lys85. De facto, this charged-assisted H-bond occurred no more than 30% of the

simulation time in

and 15, the unsubstituted pyridinyl ring preserved the contact with the positively charged Lys85

for over 70% MD simulations. Instead, rotation of pyridine ring in moved away the nitrogen atom

2 and 8 trajectories and was not found in the 14 system. Moving to derivatives

3

9

resulting in 50% occurrence of such an H-bond. The R group was also placed towards Phe67 in a

2

pi–pi stacking interaction. Next, compounds 10 and 16 showed none or weak interaction with Lys85,

while the isopropoxyl group conﬁrmed the anticipated interaction with Asp200 via a water molecule.

The corresponding INDZ derivative 4 contacted Lys85 more persistently and the oxygen atom of the

meta substituent underwent also an indirect interaction with Gln185 (Figure S3). The indirect contact

with Asp200 occurred around 50% of the simulation time in all systems with the isopropoxyl group.

Compared to the interaction pattern observed in unsubstituted pyridinyl derivatives, this additional

bond could contribute to ameliorated aﬃnity for GSK-3β. In the 5-11-17 triad, the R hydroxymethyl

2

group contacted Gln185 through a direct or water-mediated H-bond (Figure S3). An indirect interaction

of the ortho substituent with Asn186 was also observed. In addition, Lys85 occasionally contacted

the pyridinyl nitrogen of

di-F-phenyl group towards Lys85 so promoting an electrostatic-like interaction (Figure S3).

In 18 systems, hydrophobic contacts were also observed where ligands engaged Ile62, Val70,

Ala83, Leu132, Leu188, and Tyr134 at diﬀerent extents. A timeline representation of the contacts

5. Finally, the 6-12-18 triad mostly orientated the halogen atoms of the

1–

(H-bonds, hydrophobic, ionic, water bridges) with protein is reported in Figure S4.

On the whole, ligand docking and MD simulations provided valuable structural information about

the new IMID 1 and IMID 2 molecules. The initial binding poses appeared to be stable with preserved

contacts across the MD trajectories, with subtle diversity of central and deep H-bonds. Thus, the newly

designed compounds were basically suggested as comparable to the previous INDZ analogues.

Molecules 2020, 25, 2163

11 of 29

2

.4. In Vitro Data of Novel IMID 1 and IMID 2 Derivatives

Driven by the promising results, compounds

7

–

18 were synthesized as reported in Chemistry

Section 2.1. Their biological activity was determined using an in vitro GSK-3

elsewhere [17 19]. It followed that 18 exhibited good inhibition activity mostly in the nanomolar scale

with an approximate 8- to 50-fold increase in IC₅₀with respect to (Table 1). In vitro MDCK-MDR1

permeability of IMID 1 and IMID 2 derivatives was also experimentally assessed (Table 1). Derivatives

15, and 18 resulted in improved permeability compared to their INDZ analogues. In this respect,

a decrease of TPSA was observed moving from INDZ to the corresponding analogues 12 (IMID

) and 18 (IMID 2), and from to 15. The SAR analysis highlighted the detrimental impact of the

positively charged piperidine ring of 14 triplet and derivative 13 on CNS permeability. In fact,

these compounds showed low Papp values. When the piperidine was replaced with an oxanyl ring,

while keeping the isopropoxypyridinyl group on the other side ( vs. and vs. 10), a low CNS

penetration was still observed, indicating a coexisting unfavorable contribution (high TPSA) of the R₂

moiety to permeability. The comparison of the 10 pyridinyl derivatives with the corresponding 12

di-F-phenyl molecules conﬁrmed an ameliorated BBB permeability when lipophilicity raises.

β assay described

,

7–

1–6

12,

6

1

3

2-8-

2

4

8

4

–

6–

In the new series of compounds, a general reduction of the ER values was also observed. This

was consistent with published guidelines according to which removal of H-bond groups or reduction

of H-bond strength can reduce P-gp eﬄux [34]. Moreover, it is known that compounds with the most

basic pK of less than 8.0 have a much higher likelihood of being identiﬁed as a non-P-gp substrate [34].

a

This unambiguously emerged from the measured ER of 8, 13, and 14 and by comparing derivatives 8

and 10 and 13 and 15.

Hence, the optimization of physicochemical properties and the introduction of focused chemical

modiﬁcations to contain the critical HBD count of IMID 1 and IMID 2 series enabled to successfully

overcome the indazole issues for CNS penetration.

2

.5. Retrospective In Silico Analyses and NMR Studies to Rationalize in Vitro Data

This far, ligand docking and MD simulations were described driving the INDZ optimization

campaign in a predictive way. To strengthen the results of the calculations, we generated a

high-resolution X-ray structure of 16 in complex with the protein (PDB entry 6Y9S). In line with our

prediction, 16 conﬁrmed to inhibit GSK-3β enzyme by binding within the ATP pocket. The tridentate

interaction with the hinge residues, the oxanyl ring towards Arg141, R group contacting Lys85, and a

2

pool of water molecules towards the open cavity and the back pocket were found. Compared to

the docked structure, a very good consensus was observed with core RMSD = 0.35 Å (Figure 5),

thus validating our models. The experimental binding mode also conﬁrmed the stable water-bridged

contact of the isopropoxyl group with Asp200 detected in the MD simulation (Figure S3).

Thus, we carried out other in silico analyses starting from MD simulations to elucidate

retrospectively the factors responsible for the unanticipated potency decrease of the new series.

Due to the highly similar inhibition activity, the following is not meant to provide a rank of IMID 1

and IMID 2 compounds. Our results are intended to compare the new derivatives with the INDZ

analogues instead, regardless of the tiny IC₅₀diﬀerences within each core series.

First, the average binding free energy (

born surface area (MM-GBSA) was calculated from each production trajectory [35

the INDZ molecules always resulted in better MM-GBSA scores compared to the corresponding IMID

and IMID 2 derivatives, in agreement with the in vitro potency (Table 3). In fact, energetics terms

turned out to be generally more favorable for the INDZ compounds within each triplet.

∆

G_bind) based on the molecular mechanics-generalized

,36]. Interestingly,

1

Molecules 2020, 25, 2163

12 of 29

Figure 5. Overlap of the docking poses (yellow) versus co-crystal structure (green) of 16 in the complex

with GSK-3β enzyme.

Table 3. MM-GBSA ∆G_bind average from each MD simulation (100 snapshots per compound).

Compound

Core

MM-GBSA ∆G_bind (kcal/mol)

1

7

INDZ

IMID 1

IMID 2

−63.69

−58.34

−56.96

−70.98

−65.23

−63.42

1

3

2

8

INDZ

IMID 1

IMID 2

1

4

3

9

INDZ

IMID 1

IMID 2

−53.88

−49.25

−50.19

−59.97

−57.98

−58.30

−55.87

−53.68

−51.97

−55.56

−51.02

−52.11

1

5

4

INDZ

IMID 1

IMID 2

1

0

6

5

INDZ

IMID 1

IMID 2

1

7

6

INDZ

IMID 1

IMID 2

1

2

8

In both the IMID series, we supposed the drop in potency to be attributable to the less acidic C-Ha

group of the central core. Together with other moieties, it tightly anchors the compounds in the binding

pocket, but the nature of the H-bond is weaker compared to the one involving conventional INDZ

N-Ha. The general characteristics of this type of interaction to take into account are: (i) the distance

between the hydrogen and the acceptor (A), (ii) the angle formed by the donor (D), the hydrogen itself

and A, and (iii) the strength [37

across the molecules.

,38]. To diﬀerent levels, these components were herein investigated

We examined how the binding mode changes as the core scaﬀolds are varied, in terms of distance

between each hinge binding atom of the ligands and the partner heavy atom of Asp133 and Val135

along the MD simulations. The analysis revealed that moving from INDZ to IMID 1 and IMID 2

analogues, an elongation of the deep H-bond occurred, while the outer and central interactions

remained approximately constant in length (red versus green and orange scatter plots in Figure 6).

The stretching of the deep H-bond was found in IMID 1 and IMID 2 compounds of all triplets regardless

Molecules 2020, 25, 2163

13 of 29

of the R and R moieties. This suggests a possible rotation of the ligand about the hinge binding

1

2

residues which might aﬀect the SAR and elucidate the potency reduction. As anticipated in the MD

discussion, persistence percentage of the central H-bond is higher in INDZ compounds compared to

IMID 1 and IMID 2 analogues. The optimally spaced deep H-bond of the INDZ system might reﬂect

on such high persistence. Conversely, the central H-bond is more scattered along the simulations of

IMID 1 and IMID 2 derivatives as a consequence of the elongated interaction of the C-Ha.

Figure 6. Distance of the deep (red), central (green) and outer (orange) H-bond interactions across

100 ns MD simulation of derivatives 1–18. The plots are ordered per INDZ-IMID 1-IMID 2 triplet.

Dashed lines represent the mean value of the H-bond distance.

In the MD trajectories, optimal D-H-A angles promoting strong interaction of the INDZ N-Ha

with the hinge portion were also observed, while the values lowered moving to IMID 1 and IMID 2

cores (Figure 7).

Figure 7. D-H-A angles (degrees) of the deep H-bond interaction across 100 ns simulation of 1–18

systems. The plots are ordered per INDZ-IMID 1-IMID 2 triplet and colored as yellow, cyan, and orange,

respectively. Dashed lines correspond to the mean value of the D-H-A angle.

Molecules 2020, 25, 2163

14 of 29

Based on the observed arrangements, we may ascribe the H-bond between IMID C-Ha and Asp133

carbonyl group to a weakened interaction compared to INDZ N-Ha, so substantiating the aﬃnity

decrease of IMID 1 and IMID 2 derivatives versus the GSK-3β enzyme.

Lastly, the H-bond acidity prediction was investigated via QM methods using Jaguar software [39].

The Vα(r) descriptor of the core Ha and the carboxamide Hb was calculated. This is an eﬀective

and useful predictor of H-bond strength deﬁned in Peter Kenny’s work and recently implemented in

Schrödinger suite [40]. Larger values correlate with stronger HBDs, while HBAs are characterized by

low numbers. Although Kenny’s analysis never really considered C-H donors, the principle of the

calculation is still well deﬁned. Here, all ligands from Glide SP docking were used as input for V

α(r)

calculation. (Table 4).

Table 4. Vα(r) values (kcal/mol) of the core Ha and carboxamide Hb.

Vα(r)

Compound

Core

Ha

Hb

1

7

INDZ

IMID1

IMID2

256.37

211.13

198.73

245.08

234.78

247.25

1

3

2

8

INDZ

IMID1

IMID2

258.51

212.36

198.57

248.31

237.47

235.57

1

4

3

9

INDZ

IMID1

IMID2

219.28

174.58

160.91

191.94

184.25

185.69

1

5

4

INDZ

IMID1

IMID2

219.30

173.32

159.89

192.82

183.04

184.80

1

0

1

6

5

INDZ

IMID1

IMID2

218.93

176.54

160.85

191.53

184.68

186.16

11

7

1

6

INDZ

IMID1

IMID2

218.71

175.11

160.20

192.92

183.15

185.42

12

8

1

The calculation indicated that the INDZ core has the strongest H-bond potential in all compounds,

with a very strong N-Ha donor. This illustrates the importance of this moiety at the hinge for

good aﬃnity, but it equally explains the reason why it is felt to be an issue for CNS penetration.

The calculations also showed that the C-Ha of the IMID 1 and IMID 2 cores are weaker H-bond donors,

with some variability between the two new series. In each triplet, there seemed to be a large V

diﬀerence when Ha was compared between INDZ derivatives and the corresponding IMID 1 and

IMID 2 analogues ( (r) > 40). On the other hand, the V (r) values between related IMID 1 and

α(r)

∆

Vα

α

IMID 2 compounds within the same triplet resulted in a minor gap. These in silico results appeared

to be in line with the experimental IC₅₀values, where potency reduction of one or more orders of

magnitude was observed moving from INDZ compounds to the new series, while IMID 1 and IMID

2

derivatives were quite comparably potent as GSK-3β inhibitors. About Hb, no big changes were

observed from INDZ to IMID 1 and IMID 2 scaﬀolds. This suggested that the acidic N-Ha/C-Ha group

was a signiﬁcant discriminant of the cores for inhibition activity, while the carboxamide N-Hb did not

make a real diﬀerence. With this, we proved V

α

(r) to be a valuable descriptor to explore SAR in terms

of H-bond acidity.

1

These observations were further substantiated using H-NMR characterization. By way of example,

we selected the triplet 2-8-14. In particular, H-bond donating abilities of the core N-Ha/C-Ha and

the carboxamide N-Hb of the three compounds were investigated by means of NMR spectroscopy.

Molecules 2020, 25, 2163

15 of 29

In NMR experiments, proton chemical shifts depend upon many factors, among which there are

H-bond interactions. In general, a proton involved in such interaction exhibits a decreased shielding

and therefore tends to move downﬁeld in the NMR spectrum.

1

First, the complete assignment of the H-NMR spectra of

2,

8

and 14 compounds was carried out in

1

13

1

13

CDCl by means of the simultaneous H- H COSY, H- H NOESY, H- C HSQC, and H- C HMBC

3

2

D NMR analyses (see Supplementary Materials), to unequivocally assign the Ha and Hb resonances.

1

Subsequently, the H-NMR spectra were acquired in both CDCl and DMSO-d as “not good”

3

6

and “good” H-bond acceptor solvents, respectively [41]. The variation of chemical shifts of Ha and Hb

protons in the two solvents was calculated as ∆δ

(

δ

− δ_C_D_C_l₃), using the TMS signal as a reference

DMSO

at 0 ppm (Table 5 and Figures S5–S7). The target concentration of the compounds was selected by a

serial dilution method (reaching a 0.3 mM ﬁnal concentration), until a stable chemical shift diﬀerence

between the two solvents was observed, and the solute–solute interactions were expected as low

as possible.

1

Table 5. Chemical shifts of Ha and Hb observed in the H-NMR spectra of

2,

8

and 14 derivatives (0.3

mM) in CDCl and DMSO-d .

3

6

Ha

Compound

Core

δ (ppm) in DMSO-d6

δ (ppm) in CDCl3

∆δ (ppm)

2

8

INDZ

IMID 1

IMID 2

13.7

8.5

7.6

10.3

8.0

7.5

3.4

0.5

0.1

1

4

Hb

Compound

Core

δ (ppm) in DMSO-d6

δ (ppm) in CDCl3

∆δ (ppm)

2

8

INDZ

IMID 1

IMID 2

8.4

8.1

8.6

7.2

7.3

7.5

1.2

0.8

1.1

1

4

The large chemical shift diﬀerence of 3.4 ppm observed for proton Ha in compound

2 going

from DMSO-d to CDCl suggested that this proton was involved in H-bond with the former solvent,

6

3

excluding the formation of an intramolecular H-bond. On the contrary, the decrease in ∆δ observed for

proton Ha of compounds 8 and 14 suggested a lower acidity of the IMID 1 and IMID 2 core scaﬀolds.

This is in agreement with the pharmacological results where potency reduces moving from the INDZ

compounds to the new series, whereas IMID 1 and IMID 2 derivatives are similarly double-digit

nanomolar GSK-3β inhibitors. On the other hand, no large diﬀerences were observed for the ∆δ of

Hb within the triplet (Table 5), suggesting comparable acidity of the carboxamide N-Hb in the three

diﬀerent scaﬀolds. Moreover, the sharper and better resolved triplet observed for the Hb signals

moving from CDCl to DMSO-d (Figures S5–S7) conﬁrmed the involvement of the protons in H-bond

3

6

interactions with the latter solvent, which prevented them from engaging in exchange processes with

the residual amounts of water in solution.

Overall, preliminary NMR data suggested, as expected, that the INDZ core has the strongest

H-bond potential playing a crucial role in the aﬃnity for GSK-3β binding pocket. Moreover, Vα(r) was

in accordance with ∆δ, as a signiﬁcant acidity drop of Ha was observed moving from INDZ to IMID 1

and IMID 2 in both calculated and experimental parameters.

The proton spectra of the 2, 8 and 14 were also acquired in deuterated DMSO at ﬁve diﬀerent

temperatures (298, 303, 308, 313, and 318 K). The temperature coeﬃcients (T_c_o_e_ﬀ) were calculated for

N-Hb protons by linear ﬁtting of the chemical shift as function of temperature curves (as exempliﬁed

in Figure 8 for compound 14).

Molecules 2020, 25, 2163

16 of 29

Figure 8. Temperature coeﬃcient calculated for N-Hb of compound 14 (IMID 2).

In general, amide proton chemical shifts show a strong temperature dependence and this is

mainly due to the occurrence of H-bonds [42]. Chemical shifts depend on the inverse third power of

the distance between amide proton and H-bond acceptor; the average H-bond length increases with

temperature because of the increased thermal motions [43]. Therefore, the deshielding eﬀect induced

by an acceptor decreases at higher temperature and amide protons result in being shifted upﬁeld.

In line with this general trend, N-Ha and N-Hb chemical shifts followed this behavior and showed a

linear dependence with temperature, allowing the determination of T_c_o_e_ﬀvalues from the slope of the

curves (Table 6).

Table 6. Temperature coeﬃcients (T_c_o_e_ﬀ) obtained for N-Ha proton of compound

of compounds 2, 8, and 14.

2 and N-Hb protons

T_c_o_e_ﬀ(ppb/K)

Hydrogen

INDZ (2)

IMID 1 (8)

IMID 2 (14)

Ha

Hb

−4.6

−6.8

-

−6.3

−6.6

Hb atoms have shown similar T_c_o_e_ﬀ, more negative than that of the Ha proton in the INDZ core.

Indeed, for weaker H-bonds, structural perturbations are more pronounced and N-Hb chemical shifts

are more temperature dependent (more negative). In previous studies, T_c_o_e_ﬀvalues in the range

of -12.3 to

intermolecular H-bonds [44], while a mean value equal to

solvent-exposed amide protons of the protein backbone [42]. In our case, the obtained values reﬂected

the presence of weak intermolecular H-bonds in DMSO-d ; the more negative T values for the

−

0.5 ppb/K were reported for -OH groups which were solvent exposed and involved in

−

7.1 2.5 ppb/K was determined for the

±

6

coeﬀ

Hb atoms suggested that, by increasing the temperature, intermolecular solute-solvent H-bonds are

broken more easily for N-Hb than for N-Ha within the INDZ core, in agreement with the conclusions

drawn in light of the ∆δ values.

From the ∆δ downﬁeld shifts reported above, it was possible to estimate the H-bond energies

E_H_B. The most universal approach is the empirical estimation of the H-bond strength by use of the

Equation (3) (see Materials and Methods) [45]. The equation is only valid for weak and moderate

values of the H-bond energy. The energy varies within a wide range from weak (1–4 kcal/mol) to

medium (4–15 kcal/mol) to strong H-bonds (15–40 kcal/mol).

Based on the calculated values (Table 7), Ha of compound

2 showed the highest energy, while

comparable and lower values were observed concerning Hb, regardless of the central core. Accordingly,

Molecules 2020, 25, 2163

17 of 29

we can assert that Ha of INDZ and Hb of all core scaﬀolds are involved in intermolecular H-bonds

that can be classiﬁed as weak/moderate to highly weak, respectively. Again, this outcome enables to

rationalize SAR conﬁrming the key role of the INDZ N-Ha for the high inhibition activity.

Table 7. Energy of the H-bonds E_H_B(kcal/mol) estimated using Equation (1) for compounds 2-8-14.

E_H_B± 0.2 (kcal/mol)

Hydrogen

INDZ (2)

IMID 1 (8)

IMID 2 (14)

Ha

Hb

3.7

1.7

-

1.2

-

1.5

3

. Materials and Methods

3

.1. Chemistry

Reagents were purchased from Sigma-Aldrich (Madrid, Spain) and Fluorochem (Hadﬁeld, UK)

and used as received.

Reaction progress was monitored by TLC or LC-MS. For TLC analysis, Merck silica gel 60 F254

(

0.04–0.063 mm) plates with detection by UV (214 or 254 nm) were used. LC-MS analyses were performed

on the Shimadzu LCMS-2020 Single Quadrupole Liquid Chromatograph Mass Spectrometer (Duisburg,

Germany) and column Acquity UPLC C18 (2.1 × 50 mm, 1.8 µm) (Etten-Leur, The Netherlands).

Final compound purity (always >95%) was determined by HPLC-UV, by comparing the area

of the target peak and total peak areas of minor impurities. HPLC-UV analyses were performed

with a pump/autosampler (Waters 2695, Alliance model), a UV photo diode array detector (Waters

2996 model), and a Waters system data management (Empower 2) (Etten-Leur, The Netherlands).

The column used was generally X-Bridge C18 (150 × 4.6 mm, 5 µm).

UPLC-QTOF analyses for HRMS determination of ﬁnal compounds were carried out by means of

a SYNAPT MS-ACQUITY UPLC system (Waters). The system was operated in positive ion mode in

the “V-Optics” conﬁguration. Leucine-enkephalin (200 pg/µL) was employed as the lock mass in order

to provide authenticated exact mass measurement in MS and MS/MS modes within 5 ppm RMS mass

accuracy. The column was an Acquity BEH C18 (50 × 2.1 mm, 1.7 µm).

1

13

Nuclear Magnetic Resonance Spectroscopy ( H-NMR and C-NMR) were obtained using Bruker

Advance systems (Milan, Italy), operating at 300, 400 and 500 MHz. Spectra were acquired at 300 K, using

DMSO-d , CD OD, and CDCl as solvents. All resonance bands were referenced to tetramethylsilane

6

3

1

(

internal standard). For H-NMR spectroscopy: (s) = singlet; (d) = doublet; (t) = triplet; (br) = broad;

dd) = double doublet; (dt) = double triplet; (ddd) = double double doublet; (dtd) = double triple

doublet; (m) = multiplet; J= coupling constant; and δ = chemical shift (in ppm).

Compounds were puriﬁed with one of the following techniques: (i) ﬂash chromatography

on silica gel (Grace Reveleris ﬂash chromatography system with a 40

0 mL/min), with an appropriate gradient of mixtures of DCM/MeOH or Cyclohexane/EtOAc as

eluents; (ii) preparative HPLC-MS system (sample was dissolved at 50 mg/mL in DMSO/CH CN in

µM silica cartridge, ﬂow =

6

3

1

:1 ratio), using an appropriate gradient of CH CN and H O + 0.1% NH HCO ; ﬂow =40 mL/min.

3 2 4 3

The preparative HPLC-MS system consisted of a Waters 2767 Sample manager, a Waters 2478 dual

absorbance detector, and a Waters Micromass ZQ single quadrupole mass spectrometer with an

electrospray ionization (ESI) source. The column used was a X-Bridge Prep C18 5 m OBD with

λ

µ

1

9 × 10 mm (Waters) precolumn. Fraction collection was available from the system software MassLynx

v. 4.1 (Etten-Leur, The Netherlands). Detection wavelength was set to 230 nm and temperature to

2

◦

5 C.

Molecules 2020, 25, 2163

18 of 29

General Procedures (A)–(D):

General amide coupling procedure (A) for the synthesis of compounds 21, 22 and 13–18.

To a solution of carboxylic acid/lithium carboxylate salt (1.0 eq.) in DMF (0.07 M) HATU (1.2 eq.),

DIPEA (4.5 eq.) and the proper amine (1.0–1.2 eq.) were added. The solution was stirred at rt until

full conversion was observed by HPLC-MS (24–72 h). The solvent was evaporated under reduced

pressure, the residue was taken up with EtOAc and washed with 1N aqueous NaOH solution (

×

2) and

brine. The combined organic phases were dried over anhydrous MgSO , ﬁltered and concentrated in

4

vacuo. The crude was puriﬁed by ﬂash chromatography (SiO , DCM/MeOH for 21 and 22) or reverse

2

phase chromatography (C18, NH HCO /ACN for 13–18) to give the target compound.

4

3

General boronate synthesis procedure (B) for the synthesis of compounds 23 and 25.

A mixture of bromide (1.0 eq.), bis(pinacolato)diboron (2.0 eq.) and potassium acetate (2.0 eq.) in

1

,4-dioxane (0.12 M) was placed in a Schlenk tube. After degassing with N for 10 min, Pd(dppf)Cl

2

◦

(0.15 eq.) was added and the mixture was heated at 100 C overnight. Conversion was checked by

HPLC-MS. The mixture was diluted with EtOAc/MeOH, the insoluble solids were removed by ﬁltration

through Celite and the solvent was evaporated under reduced pressure. The crude was puriﬁed by

ﬂash chromatography (SiO , DCM/MeOH) to give the target compound.

2

General Suzuki coupling procedure (C) for the synthesis of compounds 26–29 and 7–12.

A mixture of bromide (1.0 eq.), boronic acid/ester (1.0 eq.) and cesium ﬂuoride (2.0 eq.) in

1

,4-dioxane and water (4:1, 0.1M) was placed in a Schlenk tube. After degassing with N for 10 min,

2

◦

Pd(dppf)Cl (0.1 eq.) was added and the mixture was heated at 90 C overnight. Conversion was

2

checked by HPLC-MS. The mixture was diluted with EtOAc/MeOH, the insoluble solids were removed

by ﬁltration through Celite, and the solvent was evaporated under reduced pressure. The crude was

puriﬁed by ﬂash chromatography (SiO , Cyclohexane/EtOAc for 26

,

27 and 29; DCM/MeOH for 28) or

2

reverse phase chromatography (C18, NH HCO /ACN for 7–12) to give the target compound.

4

3

General ester hydrolysis procedure (D) for the synthesis of compounds 30–33.

To a solution of ethyl carboxylate ester (1.0 eq.) in MeOH and water (1:1, 0.03 M), lithium

◦

hydroxide hydrate (2.0 eq.) was added. Conversion was checked by HPLC-MS. After stirring at 50 C

overnight, the reaction mixture was concentrated to dryness under reduced pressure to give the target

compound that was used in the next step without further puriﬁcation.

7–Bromoimidazo[1,5-a]pyridine-1-carbaldehyde (19). To a solution of 7-bromoimidazo[1,5-a]pyridine

(

500 mg, 2.54 mmol) in toluene (2.5 mL), DMF (0.22 mL, 2.87 mmol) and phosphorus oxychloride

0.33 mL, 3.55 mmol) were added. After stirring at 100 C for 1.5 h, the reaction mixture was carefully

◦

quenched with ice and 1N aqueous NaOH solution. The aqueous phase was extracted with DCM

3), the combined organic layers were dried over anhydrous MgSO , ﬁltered and evaporated under

(

×

4

reduced pressure. The crude was puriﬁed by ﬂash chromatography (SiO , Cyclohexane/EtOAc),

aﬀording 327 mg of the title compound. Yield: 57%. LC-MS (ESI) m/z: 224.9 [M + H] .

2

+

7

1

(

-Bromoimidazo[1,5-a]pyridine-1-carboxylic acid (20). A mixture of 7-bromoimidazo[1,5-a]pyridine-

-carbaldehyde 19 (519 mg, 2.306 mmol), silver nitrate (509 mg, 3.00 mmol), and sodium hydroxide

277 mg, 6.92 mmol) in water (4 mL) was stirred at rt for 3 h. The reaction was then diluted with water

and the black solid was ﬁltered oﬀ and washed with water. The ﬁltrate was made acidic with 1N

aqueous HCl solution and the resulting precipitate was ﬁltered and dried. The crude (375 mg) was

+

used in the next step without further puriﬁcation. Yield: 61%. LC-MS (ESI) m/z: 240.9 [M + H] .

Molecules 2020, 25, 2163

19 of 29

7

-Bromo-N-{[1-(2-methoxyethyl)piperidin-4-yl]methyl}imidazo[1,5-a]pyridine-1-carboxamide (21). The title

compound was obtained according to general procedure (A) using 7-bromoimidazo[1,5-a]pyridine-

-carboxylic acid 20 (0.73 mmol) and 1-[1-(2-methoxyethyl)piperidin-4-yl]methanamine (0.73 mmol).

1

+

1

Yield = 56% (200 mg). LC-MS (ESI) m/z: 395.1 [M + H] . H-NMR (400 MHz, CD OD)

δ

8.35-8.32

3

(

m, 1H), 8.30 (s, 1H), 8.28 (dd, 1H, J = 7.4, 0.9 Hz), 6.94 (dd, 1H, J = 7.4, 1.9 Hz), 3.72–3.68 (m, 2H), 3.57

d, 2H, J = 10.3 Hz), 3.40 (s, 3H), 3.38 (d, 2H, J = 6.6 Hz), 3.30-3.26 (m, 1H), 3.06–2.93 (m, 3H), 2.07–1.90

m, 3H), 1.66–1.53 ppm (m, 2H).

7

-Bromo-N-[(oxan-4-yl)methyl]imidazo[1,5-a]pyridine-1-carboxamide (22). The title compound was obtained

according to general procedure (A) using 7-bromoimidazo[1,5-a]pyridine-1-carboxylic acid 20

(

[

1

(

0.73 mmol) and 1-(oxan-4-yl)methanamine (0.73 mmol). Yield = 78% (192 mg). LC-MS (ESI) m/z: 338.0

+

1

M + H] . H-NMR (400 MHz, CD OD)

δ

8.33 (dt, 1H, J = 1.8, 0.7 Hz), 8.27 (s, 1H), 8.25 (dd, 1H, J = 7.4,

.0 Hz), 6.91 (dd, 1H, J = 7.4, 2.0 Hz), 3.95 (dd, 2H, J = 11.4, 2.6 Hz), 3.40 (td, 2H, J = 11.9, 2.1 Hz), 3.30

d, 2H, J = 6.9 Hz), 1.94-1.82 (m, 1H), 1.74-1.65 (m, 2H), 1.42-1.29 ppm (m, 2H).

3

N-((Tetrahydro-2H-pyran-4-yl)methyl)-7-(4,4,5,5-tetramethyl-1,3,2-dioxaborolan-2-yl)imidazo[1,5-a]pyridine-1-

carboxamide (23). The title compound was obtained according to general procedure (B) using 7-bromo-

N-[(oxan-4-yl)methyl]imidazo[1,5-a]pyridine-1-carboxamide 22 (1.36 mmol). Yield = 76% (340 mg).

+

1

LC-MS (ESI) m/z: 386.2 [M + H] . H-NMR (400 MHz, CD OD)

δ

8.62 (s, 1H), 8.31 (s, 1H), 8.27 (dd, 1H,

J = 7.0, 1.2 Hz), 6.99 (dd, 1H, J = 7.0, 1.0 Hz), 3.95 (dd, 2H, J = 11.4, 2.7 Hz), 3.41 (td, 2H, J = 11.9, 2.1 Hz),

.34-3.31 (m, 2H), 1.96–1.84 (m, 1H), 1.71 (dd, 2H, J = 13.1, 1.8 Hz), 1.43–1.28 (m, 2H), 1.36 ppm (s, 12H).

3

Ethyl 6-(4,4,5,5-tetramethyl-1,3,2-dioxaborolan-2-yl)imidazo[1,5-a]pyridine-3-carboxylate (25). The title

compound was obtained according to general procedure (B) using 6-bromoimidazo[1,5-a]pyridine-3-

+

1

carboxylate 24 (11.0 mmol). Yield = 100% (4.08 g). LC-MS (ESI) m/z: 317.2 [M + H] . H-NMR

400 MHz, CDCl₃) 9.70 (dd, 1H, J = 2.0, 1.0 Hz), 7.61–7.57 (m, 2H), 7.36–7.30 (m, 1H), 4.51 (q, 2H,

(

δ

J = 7.2 Hz), 1.49 (t, 3H, J = 7.1 Hz), 1.36 ppm (s, 12H, J = 3.3 Hz).

Ethyl 6-(pyridin-3-yl)imidazo[1,5-a]pyridine-3-carboxylate (26). The title compound was obtained according

to general procedure (C) using 6-bromoimidazo[1,5-a]pyridine-3-carboxylate 24 (0.55 mmol) and

+

pyridin-3-ylboronic acid (0.55 mmol). Yield = 64% (94 mg). LC-MS (ESI) m/z: 268.1 [M + H] .

Ethyl 6-{5-[(propan-2-yl)oxy]pyridin-3-yl}imidazo[1,5-a]pyridine-3-carboxylate (27). The title compound

was obtained according to general procedure (C) using 6-bromoimidazo[1,5-a]pyridine-3-carboxylate

2

4

(0.45 mmol) and {5-[(propan-2-yl)oxy]pyridin-3-yl}boronic acid (0.45 mmol). Yield = 93% (140 mg).

+

LC-MS (ESI) m/z: 326.1 [M + H] .

Ethyl 6-[4-(hydroxymethyl)pyridin-3-yl]imidazo[1,5-a]pyridine-3-carboxylate (28). The title compound was

obtained according to general procedure (C) using ethyl 6-(4,4,5,5-tetramethyl-1,3,2-dioxaborolan-2-yl)

imidazo[1,5-a]pyridine-3-carboxylate 25 (0.68 mmol) and (3-bromopyridin-4-yl)methanol (0.68 mmol).

+

Yield = 51% (105 mg). LC-MS (ESI) m/z: 298.1 [M + H] .

Ethyl 6-(2,3-diﬂuorophenyl)imidazo[1,5-a]pyridine-3-carboxylate (29). The title compound was obtained

according to general procedure (C) using 6-bromoimidazo[1,5-a]pyridine-3-carboxylate 24 (1.16 mmol)

and (2,3-diﬂuorophenyl)boronic acid (1.16 mmol). Yield = 90% (318 mg). LC-MS (ESI) m/z: 303.1

+

[

M + H] .

Lithium 6-(pyridin-3-yl)imidazo[1,5-a]pyridine-3-carboxylate (30). The title compound was obtained

according to general procedure (D) using ethyl 6-(pyridin-3-yl)imidazo[1,5-a]pyridine-3-carboxylate

+

1

26 (0.35 mmol). Yield = 100% (85 mg). LC-MS (ESI) m/z: 240.1 [M + H] . H-NMR (400 MHz, CD OD)

3

Molecules 2020, 25, 2163

20 of 29

δ

9.66 (s, 1H), 8.91 (d, 1H, J = 1.8 Hz), 8.63 (dd, 1H, J = 4.9, 1.5 Hz), 8.22 (ddd, 1H, J = 8.0, 2.2, 1.7 Hz,),

7

.95 (dd, 1H, J = 9.4, 0.9 Hz), 7.74 (s, 1H), 7.61 (dd, 1H, J = 7.7, 4.7 Hz), 7.54 ppm (dd, 1H, J = 9.4, 1.5 Hz).

Lithium 6-{5-[(propan-2-yl)oxy]pyridin-3-yl}imidazo[1,5-a]pyridine-3-carboxylate (31). The title compound

was obtained according to general procedure (D) using ethyl 6-{5-[(propan-2-yl)oxy]pyridin-3-yl}

imidazo[1,5-a]pyridine-3-carboxylate 27 (0.43 mmol). Yield = 95% (123 mg). LC-MS (ESI) m/z: 298.1

M + H]⁺. H-NMR (400 MHz, CD OD)

1

δ

9.63 (s, 1H), 8.44 (d, 1H, J = 1.6 Hz), 8.28 (d, 1H, J = 2.5 Hz),

[

3

7.93 (d, 1H, J = 9.1 Hz), 7.74 (s, 1H), 7.71–7.69 (m, 1H), 7.52 (dd, 1H, J = 9.4, 1.2 Hz), 4.87–4.78 (m, 1H),

1

.40 ppm (d, J = 6.0 Hz, 6H).

Lithium 6-(4-(hydroxymethyl)pyridin-3-yl)imidazo[1,5-a]pyridine-3-carboxylate (32). The title compound

was obtained according to general procedure (D) using ethyl 6-[4-(hydroxymethyl)pyridin-3-yl]

imidazo[1,5-a]pyridine-3-carboxylate 28 (0.35 mmol). Yield = 100% (97 mg). LC-MS (ESI) m/z: 270.1

+

[

M + H] .

Lithium 6-(2,3-diﬂuorophenyl)imidazo[1,5-a]pyridine-3-carboxylate (33). The title compound was obtained

according to general procedure (D) using ethyl 6-(2,3-diﬂuorophenyl)imidazo[1,5-a]pyridine-3-

+

carboxylate 29 (1.05 mmol). Yield = 93% (273 mg). LC-MS (ESI) m/z: 275.1 [M + H] .

N-((1-(2-Methoxyethyl)piperidin-4-yl)methyl)-7-(pyridin-3-yl)imidazo[1,5-a]pyridine-1-carboxamide (7). The

title compound was obtained according to general procedure (C) using 7-bromo-N- {[1-(2-methoxyethyl)

piperidin-4-yl]methyl}imidazo[1,5-a]pyridine-1-carboxamide 21 (0.25 mmol) and pyridin-3-ylboronic

+

acid (0.25 mmol). Yield = 19% (20 mg). HRMS (ESI): calculated 394.2237 [M + H] , found 394.2241

M + H]⁺. H-NMR (500 MHz, DMSO-d )

1

δ

8.98 (br s, 1H), 8.64–8.59 (m, 2H), 8.49 (br s, 1H), 8.38 (br s,

[

6

1

H), 8.17 (br d, 1H, J = 7.3 Hz), 8.09 (br s, 1H), 7.54 (br dd, 1H, J = 5.0, 6.9 Hz), 7.31 (br d, 1H, J = 6.4 Hz),

3

.41–3.39 (m, 1H), 3.22 (s, 1H), 3.20–3.18 (m, 2H), 2.84 (br d, 2H, J = 10.5 Hz), 2.42 (br t, 2H, J = 5.5 Hz),

13

1

.89 (br t, 2H, J = 11.0 Hz), 1.62 (br d, 2H, J = 12.4 Hz), 1.55 (br s, 1H), 1.21–1.14 ppm (m, 2H). C-NMR

(

125 MHz, DMSO-d₆) 163.2, 149.8, 147.8, 134.4, 133.8, 132.4, 131.3, 128.5, 125.4, 125.3, 124.5, 116.1,

δ

1

13.2, 70.6, 58.4, 57.9, 54.0, 44.2, 36.5, 30.3 ppm.

7-(5-Isopropoxypyridin-3-yl)-N-((1-(2-methoxyethyl)piperidin-4-yl)methyl)imidazo[1,5-a]pyridine-1-carboxamide

(8

). The title compound was obtained according to general procedure (C) using 7-bromo-N-{[1-(2-

methoxyethyl)piperidin-4-yl]methyl}imidazo[1,5-a]pyridine-1-carboxamide 21 (0.25 mmol) and

5-[(propan-2-yl)oxy]pyridin-3-yl}boronic acid (0.25 mmol). Yield = 16% (19 mg). HRMS (ESI):

{

+

+ 1

calculated 452.2656 [M + H] , found 452.2652 [M + H] . H-NMR (500 MHz, CDCl )

δ

8.54 (s, 1H),

.49 (d, 1H, J = 1.8 Hz), 8.30 (d, 1H, J = 2.7 Hz), 8.07 (dd, 1H, J = 0.9, 7.3 Hz), 8.03 (s, 1H), 7.47 (t, 1H,

J = 2.2 Hz), 7.29 (t, 1H, J = 6.0 Hz), 7.02 (dd, 1H, J = 1.8, 7.3 Hz), 4.73–4.65 (m, 1H), 3.52 (t, 2H, J =

3

8

5

2

6

.7 Hz), 3.39 (t, 2H, J = 6.4 Hz), 3.34 (s, 3H), 3.00 (br d, 2H, J = 11.0 Hz), 2.58 (t, 2H, J = 5.4 Hz), 2.02 (t,

H, J = 10.8 Hz), 1.80 (br d, 2H, J = 12.5 Hz), 1.71–1.63 (m, 1H), 1.52-1.42 (m, 2H), 1.39 ppm (d, 6H, J =

13

.0 Hz). C-NMR (125 MHz, CDCl₃)

δ 163.4, 154.3, 139.8, 139.2, 134.4, 132.7, 131.9, 126.6, 125.7, 122.9,

1

20.2, 117.6, 113.8, 70.9, 70.3, 58.9, 58.2, 54.0, 44.4, 36.3, 29.8, 21.9 ppm.

7-(Pyridin-3-yl)-N-((tetrahydro-2H-pyran-4-yl)methyl)imidazo[1,5-a]pyridine-1-carboxamide (9). The title

compound was obtained according to general procedure (C) using 7-bromo-N-[(oxan-4-yl)methyl]

imidazo[1,5-a]pyridine-1-carboxamide 22 (0.25 mmol) and pyridin-3-ylboronic acid (0.25 mmol). Yield

+

1

=

34% (28 mg). HRMS (ESI): calculated 337.1659 [M + H] , found 337.1662 [M + H] . H-NMR

400 MHz, MeOD) 8.95 (dd, J = 2.4, 0.8 Hz, 1H), 8.58 (dd, J = 4.9, 1.5 Hz, 1H), 8.48–8.46 (m, 2H), 8.34

s, 1H), 8.23–8.15 (m, 1H), 7.57–7.50 (m, 1H), 7.23 (dd, J = 7.7, 1.5 Hz, 1H), 3.96–3.88 (m, 2H), 3.42 (td,

(

δ

J = 11.8, 2.1 Hz, 2H), 3.36–3.34 (m, 2H), 1.98–1.85 (m, 1H), 1.77–1.68 (m, 2H), 1.44–1.28 ppm (m, 2H).

Molecules 2020, 25, 2163

21 of 29

7-(5-Isopropoxypyridin-3-yl)-N-((tetrahydro-2H-pyran-4-yl)methyl)imidazo[1,5-a]pyridine-1-carboxamide (10).

The title compound was obtained according to general procedure (C) using 7-bromo-N-[(oxan-4-yl)methyl]

imidazo[1,5-a]pyridine-1-carboxamide 22 (0.25 mmol) and {5-[(propan-2-yl)oxy]pyridin-3-yl}boronic

+

acid (0.25 mmol). Yield = 28% (30 mg). HRMS (ESI): calculated 395.2078 [M + H] , found 395.2077

+

1

[

(

2

M + H] . H-NMR (400 MHz, MeOD)

d, J = 2.7 Hz, 1H), 7.75–7.73 (m, 1H), 7.24–7.20 (m, 1H), 4.83–4.81 (m, 1H), 3.96 (dd, J = 11.3, 2.7 Hz,

H), 3.42 (td, J = 11.9, 2.1 Hz, 2H), 3.35 (d, J = 6.9 Hz, 2H), 1.91–2.11 (m, 1H), 1.72 (d, J = 12.9 Hz, 2H),

δ 8.49 (d, J = 1.9 Hz, 1H), 8.46 – 8.43 (m, 2H), 8.34 (s, 1H), 8.23

1

.40 (d, J = 6.0 Hz, 6H), 1.42–1.31 ppm (m, 2H).

7-(4-(Hydroxymethyl)pyridin-3-yl)-N-((tetrahydro-2H-pyran-4-yl)methyl)imidazo [1,5-a]pyridine-1-carboxamide

(11). The title compound was obtained according to general procedure (C) using N-((tetrahydro-2H-

pyran-4-yl)methyl)-7-(4,4,5,5-tetramethyl-1,3,2-dioxaborolan-2-yl)imidazo[1,5-a]pyridine-1-carboxamide

1

2

3

(0.35 mmol) and (3-bromopyridin-4-yl)methanol (0.35 mmol). Yield = 48% (12 mg). H-NMR

(

400 MHz, MeOD)

δ 8.90 (d, J = 5.2 Hz, 1H), 8.79 (bs, 1H), 8.74 (dd, J = 7.2, 1.1 Hz, 1H), 8.67 (bs, 1H),

8.44 (bs, 1H), 8.04 (d, J = 5.1 Hz, 1H), 7.23 (dd, J = 7.2, 1.7 Hz, 1H), 4.99 (s, 2H), 4.29–4.24 (m, 2H), 3.72

(

td, J = 11.9, 2.1 Hz, 2H), 3.65 (d, J = 4.1 Hz, 2H), 2.28–2.15 (m, 1H), 2.05–1.98 (m, 2H), 1.72–1.62 ppm

m, 2H).

7

-(2,3-Diﬂuorophenyl)-N-((tetrahydro-2H-pyran-4-yl)methyl)imidazo[1,5-a]pyridine-1-carboxamide (12). The

title compound was obtained according to general procedure (C) using 7-bromo-N-[(oxan-4-yl)methyl]

imidazo[1,5-a]pyridine-1-carboxamide 22 (0.18 mmol) and (2,3-diﬂuorophenyl)boronic acid (0.18 mmol).

+

+ 1

Yield = 61% (42 mg). HRMS (ESI): calculated 372.1518 [M + H] , found 372.1527 [M + H] . H-NMR

500 MHz, DMSO-d₆) 8.56 (d, 1H, J = 7.2 Hz), 8.51 (s, 1H), 8.30 (s, 1H), 8.16 (br t, 1H, J = 6.2 Hz),

.52–7.47 (m, 2H), 7.35–7.37 (m, 1H), 7.10 (br d, 1H, J = 7.2 Hz), 3.83 (br dd, 2H, J = 2.4, 11.2 Hz), 3.25

t, 2H, J = 10.0 Hz), 3.19 (t, 2H, J = 6.6 Hz), 1.86–1.77 (m, 1H), 1.57 (d, 2H, J = 10.0 Hz), 1.27–1.16 ppm

m, 2H). ¹³C-NMR (125 MHz, DMSO-d₆,)

163.2, 150.7, 147.8, 130.9, 129.2, 128.8, 128.7, 125.9, 125.8,

25.6, 124.9, 118.9, 117.8, 114.6, 67.2, 44.3, 35.6, 31.0 ppm.

(

δ

7

(

δ

1

N-((1-(2-Methoxyethyl)piperidin-4-yl)methyl)-6-(pyridin-3-yl)imidazo[1,5-a]pyridine-3-carboxamide (13). The title

compound was obtained according to general procedure (A) using lithium 6-(pyridin-3-yl)imidazo

[

1,5-a]pyridine-3-carboxylate 30 (0.40 mmol) and 1-[1-(2-methoxyethyl)piperidin-4-yl]methanamine

+

(

0.40 mmol). Yield = 39% (63 mg). HRMS (ESI): calculated 394.2237 [M + H] , found 394.2229 [M + H] .

1

H-NMR (500 MHz, DMSO-d₆,)

J = 1.9, 7.9 Hz), 7.91 (d, 1H, J = 9.3 Hz), 7.64 (s, 1H), 7.55 (dd, 1H, J = 4.9, 7.8 Hz), 7.45 (dd, 1H, J =

.4, 9.3 Hz), 3.40 (t, 2H, J = 5.9 Hz), 3.22–3.19 (m, 5H), 2.83 (br d, 2H, J = 11.3 Hz), 2.42 (t, 2H, J =

.0 Hz), 1.89 (br t, 2H, J = 10.9 Hz), 1.63 (br d, 2H, J = 12.5 Hz), 1.59–1.56 (m, 1H), 1.22–1.14 ppm (m,

δ 9.70 (s, 1H), 8.92 (d, 1H, J = 1.8 Hz), 8.66–8.62 (m, 2H), 8.12 (td, 1H,

1

6

2

13

H). C-NMR (125 MHz, DMSO-d₆)

δ 159.7, 149.7, 148.0, 134.8, 132.9, 132.5, 130.5, 125.2, 124.5, 122.6,

1

21.9, 121.0, 119.2, 70.6, 58.4, 57.9, 54.0, 44.4, 36.3, 30.3 ppm.

6

-(5-Isopropoxypyridin-3-yl)-N-((1-(2-methoxyethyl)piperidin-4-yl)methyl)imidazo[1,5-a]pyridine-3-carboxamide

14). The title compound was obtained according to general procedure (A) using lithium 6-{5-[(propan-

-yl)oxy]pyridin-3-yl}imidazo[1,5-a]pyridine-3-carboxylate 31 (0.30 mmol) and 1-[1-(2-methoxyethyl)

piperidin-4-yl]methanamine (0.30 mmol). Yield = 46% (60 mg). HRMS (ESI): calculated 452.2656

(

2

M + H]⁺, found 452.2647 [M + H] . H-NMR (500 MHz, CDCl )

+

1

δ

9.75 (s, 1H), 8.45 (d, 1H, J = 1.5 Hz),

.31 (d, 1H, J = 2.6 Hz), 7.66 (d, 1H, J = 9.2 Hz), 7.53-7.50 (m, 1H), 7.49 (s, 1H), 7.42 (t, 1H, J = 2.1 Hz),

.21 (dd, 1H, J = 1.1, 9.3 Hz), 4.72 – 4.67 (m, 1H), 3.51 (t, 2H, J = 5.6 Hz), 3.38 (t, 2H, J = 6.5 Hz), 3.34

[

8

7

3

(

s, 3H), 2.99 (br d, 2H, J = 11.4 Hz), 2.57 (t, 2H, J =5.7 Hz), 2.01 (br t, 2H, J = 11.0 Hz), 1.79 (br d, 2H,

J = 12.7 Hz), 1.69–1.63 (m, 1H), 1.48–1.42 (m, 1H), 1.40 ppm (d, 1H, J = 6.1 Hz). ¹³C-NMR (125 MHz,

CDCl₃) 159.8, 154.2, 140.2, 138.5, 133.6, 132.7, 130.3, 125.6, 123.3, 121.8, 121.0, 120.6, 118.3, 70.8, 70.0,

8.9, 58.0, 53.8, 44.5, 36.1, 29.6, 21.9 ppm.

δ

5

Molecules 2020, 25, 2163

22 of 29

6

-(Pyridin-3-yl)-N-((tetrahydro-2H-pyran-4-yl)methyl)imidazo[1,5-a]pyridine-3-carboxamide (15). The title

compound was obtained according to general procedure (A) using lithium 6-(pyridin-3-yl)imidazo

1,5-a]pyridine-3-carboxylate 30 (1.35 mmol) and 1-(oxan-4-yl)methanamine (1.35 mmol). Yield = 9%

[

+

+ 1

(

41 mg). HRMS (ESI): calculated 337.1659 [M + H] , found 337.1657 [M + H] . H-NMR (400 MHz,

MeOD) 9.72 (s, 1H), 8.90–8.89 (m, 1H), 8.60 (dd, J = 4.9, 1.5 Hz, 1H), 8.18 (ddd, J = 8.0, 2.3, 1.6 Hz,

H), 7.85 (dd, J = 9.4, 1.0 Hz, 1H), 7.61–7.57 (m, 2H), 7.41 (dd, J = 9.4, 1.6 Hz, 1H), 3.96 (dd, J = 11.4,

.7 Hz, 2H), 3.42 (td, J = 11.9, 2.1 Hz, 2H), 3.35 (d, J = 6.9 Hz, 2H), 1.99–1.87 (m, 1H), 1.75–1.71 (m, 2H),

.45–1.33 ppm (m, 2H).

δ

1

2

1

6

-(5-Isopropoxypyridin-3-yl)-N-((tetrahydro-2H-pyran-4-yl)methyl)imidazo[1,5-a]pyridine-3-carboxamide (16).

The title compound was obtained according to general procedure (A) using lithium 6-{5-[(propan-2-yl)

oxy]pyridin-3-yl}imidazo[1,5-a]pyridine-3-carboxylate 31 (0.43 mmol) and 1-(oxan-4-yl)methanamine

+

(

0.43 mmol). Yield = 22% (37 mg). HRMS (ESI): calculated 395.2078 [M + H] , found 395.2075 [M + H] .

1

H-NMR (400 MHz, MeOD)

J = 9.4, 1.0 Hz, 1H), 7.65–7.63 (m, 1H), 7.56 (d, J = 0.7 Hz, 1H), 7.35 (dd, J = 9.4, 1.5 Hz, 1H), 4.83–4.79

m, 1H), 3.95 (dd, J = 11.4, 2.8 Hz, 2H), 3.44–3.38 (m, 2H), 3.33 (d, J = 6.9 Hz, 2H), 1.97–1.88 (m, 1H),

.74–1.70 (m, 2H), 1.42–1.32 ppm (m, 8H).

δ 9.66 (s, 1H), 8.39 (d, J = 1.5 Hz, 1H), 8.23 (d, J = 2.5 Hz, 1H), 7.80 (dd,

(

1

6-(4-(Hydroxymethyl)pyridin-3-yl)-N-((tetrahydro-2H-pyran-4-yl)methyl)imidazo[1,5-a]pyridine-3-carboxamide

(17). The title compound was obtained according to general procedure (A) using lithium 6-(4-

(

hydroxymethyl)pyridin-3-yl)imidazo[1,5-a]pyridine-3-carboxylate 32 (0.35 mmol) and 1-(oxan-4-yl)

+

methanamine (0.42 mmol). Yield = 45% (34 mg). HRMS (ESI): calculated 367.1764 [M + H] , found

+

1

3

7

2

67.1763 [M + H] . H-NMR (400 MHz, MeOD) δ 9.43 (bs, 1H), 8.61 (d, J = 5.2 Hz, 1H), 8.49 (s, 1H),

.81 (dd, J = 9.3, 1.0 Hz, 1H), 7.74 (d, J = 5.2 Hz, 1H), 7.62 (s, 1H), 7.11 (dd, J = 9.3, 1.5 Hz, 1H), 4.66 (s,

H), 3.95 (dd, J = 11.4, 2.7 Hz, 2H), 3.41 (td, J = 11.9, 2.1 Hz, 2H), 3.34 (d, J = 4.8 Hz, 2H), 1.97–1.85 (m,

1

H), 1.74–1.68 (m, 2H), 1.42–1.30 ppm (m, 2H).

6

-(2,3-Diﬂuorophenyl)-N-((tetrahydro-2H-pyran-4-yl)methyl)imidazo[1,5-a]pyridine-3-carboxamide (18). The

title compound was obtained according to general procedure (A) using lithium 6-(2,3-diﬂuorophenyl)

imidazo[1,5-a]pyridine-3-carboxylate 33 (1.05 mmol) and 1-(oxan-4-yl)methanamine (1.05 mmol). Yield

+

1

=

15% (59 mg). HRMS (ESI): calculated 372.1518 [M + H] , found 372.1520 [M + H] . H-NMR

500 MHz, DMSO-d₆) 9.66 (s, 1H), 8.69 (t, 1H, J = 10.0Hz), 7.90 (d, 1H, J = 9.3 Hz), 7.66 (s, 1H),

.56–7.50 (m, 1H) 7.49–7.46 (m, 1H), 7.39–7.35 (m, 1H), 7.30 (d, 1H, J = 10.0Hz), 3.84 (br dd, 2H, J = 2.4,

(

7

δ

1

1.3 Hz), 3.25 (t, 2H, J = 10.0Hz), 3.21 (t, 2H, J = 10.0Hz), 1.89–1.81 (m, 1H), 1.58 (br d, 2H, J = 15.0Hz),

1

13

.25–1.17 ppm (m, 2H). H C-NMR (125 MHz, DMSO-d )

δ

159.8, 150.7, 147.8, 132.3, 130.5, 127.3,

6

26.2, 125.9, 124.6, 123.0, 121.8, 121.1, 118.8, 117.9, 67.2, 44.5, 35.5, 31.0 ppm.

3

.2. X-ray Crystallography

3

.2.1. Crystals of GSK-3β in Complex with Ligand 2

Prior to crystallization trials, the protein (protein buﬀer: 20 mM MOPS pH 7.0, 93.5 mM

NaCl,10% Glycerin, 1 mM DTT, 1xPIC) was diluted to 4.0 mg/mL and supplemented with 2 mM of the

1

H-indazole-3-carboxamide inhibitor

2

. Crystals of GSK-3

β

/Compound

2

were grown in the following

◦

conditions: 12–20% v/v PEG 8000, 0.13 M NaCl, 0.10 M TrisAcetate pH 7.5–8.0, 5 mM TCEP at 4 C.

Crystals appeared within 3 to 7 days in hanging drop vapor diﬀusion setups and were cryo-protected

with glycerol, and vitriﬁed in liquid nitrogen. Data collection and model statistics of the crystal

structure of GSK-3

diﬀraction data have been collected from complex crystals of human GSK-3

the SWISS LIGHT SOURCE (SLS, Villigen, Switzerland) using cryogenic conditions. The crystals

belong to space group C 2 2 2 . Data were processed to 2.08 Å resolution using the programs XDS and

β

in complex with ligand

2

are shown in the Supplementary Materials. The X-ray

β

with the ligand at

2

1

XSCALE [46]. The phase information necessary to determine and analyze the structure of GSK-3

β

Molecules 2020, 25, 2163

23 of 29

in complex with ligand 2 was obtained by molecular replacement using a previously solved GSK-3β

structure as a search model. CORINA was used for ligand parameterization [47]. Subsequent model

building and reﬁnement was performed with the software packages CCP4 and COOT [48 49]. For the

,

measure of the free R-factor, a measure to cross-validate correctness of the ﬁnal model, about 5.9%

of measured reﬂections were excluded from the reﬁnement procedure. Several rounds of manual

model building in COOT and bulk solvent correction, positional, B-factor, and TLS reﬁnement using

REFMAC yielded the ﬁnal model [48,49]. The atomic coordinates and structure factors have been

deposited in the Protein Data Bank (PDB entry 6Y9R).

3

.2.2. Crystals of GSK-3β in Complex with Ligand 16

Prior to crystallization trials, the protein (protein buﬀer: 20 mM MOPS pH 7.0, 93.5 mM

NaCl,10% Glycerin, 1mM DTT, 1xPIC) was diluted to 4.0 mg/mL and supplemented with 1 mM of the

imidazo[1,5-a]pyridine-3-carboxamide inhibitor 16. Crystals of GSK-3

β

/compound 16 were grown

◦

in the following condition: 20–28% PEG 8000, 0.13 M NaCl, 0.10 M TrisAcetate pH 7.75–8.25 at 4 C.

Crystals appeared after 2 to 5 days in hanging drop vapor diﬀusion setups, were also cryo-protected

with glycerol, and vitriﬁed in liquid nitrogen. Data collection and model statistics of the crystal

structure of GSK-3

diﬀraction data have been collected from complex crystals of human GSK-3

SWISS LIGHT SOURCE (SLS, Villigen, Switzerland) using cryogenic conditions. The crystals belong to

space group P 2 . Data were processed to 2.03 Å resolution using the programs XDS and XSCALE [46].

β

in complex with ligand 16 are shown in the Supplementary Materials. The X-ray

β

with the ligand 16 at the

1

The phase information necessary to determine and analyze the structure of GSK-3

ligand 16 was obtained by molecular replacement using a previously solved GSK-3

search model. CORINA was used for ligand parameterization [47]. Subsequent model building and

reﬁnement were performed with the software packages CCP4 and COOT [48 49]. For the measure of

β

in complex with

β

structure as a

,

the free R-factor, a measure to cross-validate correctness of the ﬁnal model, about 3.3% of measured

reﬂections were excluded from the reﬁnement procedure. Several rounds of manual model building in

COOT and bulk solvent correction, positional, B-factor, and TLS reﬁnement using REFMAC yielded

the ﬁnal model [48

,49 ]. The atomic coordinates and structure factors have been deposited in the Protein

Data Bank, www.pdb.org (PDB entry 6Y9S).

3

.3. MDCK-MDR1 Cell Line Assay

MDCK-MDR1 cells were obtained from the NIH (Rockville, MD, USA). Cells were seeded onto

Millipore Multiscreen Transwell plates at 3.4

assay incubations were carried out at 37 C in an atmosphere of 5% CO2 with a relative humidity of

×

105 cells/cm2 and cultured in DMEM. Cell culture and

◦

9

5%. The monolayers were prepared by rinsing both apical and basolateral surfaces twice with Hanks

◦

Balanced Salt Solution (HBSS) at the desired pH warmed to 37 C. Cells were then incubated with HBSS

at the desired pH in both apical and basolateral compartments for 40 min to stabilize physiological

parameters. The dosing solutions were prepared by diluting test compound with assay buﬀer to give a

ﬁnal test compound concentration of 10 µM (ﬁnal DMSO concentration of 1% v/v). Analytical standards

were prepared from test compound DMSO dilutions and transferred to buﬀer, maintaining a 1% v/v

DMSO concentration. Assay buﬀer was composed of supplemented HBSS pH 7.4. For assessment of

A–B permeability, HBSS was removed from the apical compartment and replaced with test compound

dosing solution. The apical compartment insert was then placed into a companion plate containing

fresh buﬀer (containing 1% v/v DMSO). For assessment of B–A permeability, HBSS was removed from

the companion plate and replaced with test compound dosing solution. Fresh buﬀer (containing 1%

v/v DMSO) was added to the apical compartment insert, which was then placed into the companion

plate. At 60 min, the apical compartment inserts and the companion plates were separated and apical

and basolateral samples diluted for analysis. Test compound permeability was assessed in duplicate.

Two control compounds were screened alongside the test compounds, propranolol (highly permeable),

and prazosin (a substrate for P-gp). Test and control compounds were quantiﬁed by LC-MS/MS

Molecules 2020, 25, 2163

24 of 29

cassette analysis using an 8-point calibration with appropriate dilution of the samples. Cyprotex

generic analytical conditions were used. The starting concentration (C ) was determined from the

0

dosing solution and the experimental recovery calculated from C and both apical and basolateral

0

compartment concentrations. The permeability coeﬃcient (Papp) for each compound was calculated

from the following Equation (1):

!

dQ/dt

C × A

0

Papp =

,

(1)

where dQ/dt is the rate of permeation of the drug across the cells, C is the donor compartment

0

concentration at time zero, and A is the area of the cell monolayer. C was obtained from analysis of

0

the dosing solution. For bi-directional experiments, ER was calculated from mean A–B and B–A data.

This was derived from Equation (2):

!

^Papp(B–A)

ER =

.

(2)

^Papp(A–B)

ER value greater than two indicates that drug eﬄux is occurring.

3

.4. Ligand Docking

The Schrödinger suite was used for all calculations. Ligands were converted to 3D structures

and prepared with LigPrep tool using default setting [50]. Overall, the most likely tautomers and

protonation states were considered for each ligand. Next, a conformational search was carried out

using ConfGen resulting in the generation of 10 conformers per ligand [51]. This represented the input

dataset for Glide SP docking [30

,31,52]. The coordinates of the crystal structure of GSK-3β in complex

with (PDB entry 6Y9R) were used. The protein was prepared using Maestro Protein Preparation

2

Wizard [53]. Hydrogen atoms were added, and water molecules were removed from the protein

structure. The default setting was used for receptor grid preparation. The binding site was speciﬁed

3

as a box of 10

×

10

×

10 Å centered on the centroid of the co-crystallized inhibitor. In addition, we

treated thiol and hydroxyl groups of Tyr134, Thr138, and Cys199 as rotatable (ﬂexible) groups. Three

H-bond constraints were deﬁned being the carbonyl oxygen of the hinge residues Asp133 and Val135,

and the hydrogen of the N-H group of Val135. In Glide SP docking, we required that at least one

such H-bond crucial for compound activity was formed. We also used the enhanced sampling option

to enhance the sampling of conformational space by adding variations on the input structure to the

conformational search and included aromatic ligand hydrogen atoms as donors. Visual inspection of

the docking results was carried out enabling the picking of one pose per ligand as a starting point for

further studies.

A rigid-body superposition of INDZ vs. IMID 1 and IMID 2 analogues within each triplets was

performed using Maestro software [54], and RMSD was computed (Table 2). The common substructure

was deﬁned considering all heavy atoms with the exception of the ones of the bicyclic system that

are diﬀerent moving from INDZ to IMID 1 and IMID 2 analogues. RMSD was computed without

changing structure positions.

Figures 3 and 5 were prepared with PyMOL 1.8.6.2 [55]. Crystal structure of 16 in complex of

GSK-3

β (Figure 5) was prepared using a default setting of the Maestro Protein Preparation Wizard

tool [53].

3

.5. MD Simulations

The selected binding poses of

simulation using the graphics processing unit (GPU) tool of the Desmond MD package as supplied

with the Schrödinger 2019-1 suite [32 56]. The Desmond System Builder was used to create a fully

1–18 within the ATP binding site of GSK-3β were processed in MD

,

solvated system with simple point charge (SPC) waters and a cubic solvation box with a 10 Å buﬀer

+

around the original complex. The system was electrostatically neutralized by adding Na ions in

deﬁned positions by using the ion placement option. Default parameters of OPLS-3e force ﬁeld were

Molecules 2020, 25, 2163

25 of 29

used. The prepared systems included approximately 58,900 atoms. The MD simulation protocol

followed an extended relaxation schedule consisting of the following steps:

2

1

.

100 ps, constant–volume ensemble (NVT) (T = 10 K); 50 kcal/mol/Å restraints on the solute

heavy atoms;

2

3

.

120 ps, NVT (T = 10 K); 50 kcal/mol/Å restraints on the solute heavy atoms;

2

120 ps, constant–pressure ensemble (NPT) (T = 10K, P = 1 atm); 50 kcal/mol/Å restraints on the

solute heavy atoms;

2

4

5

6

.

120 ps, NPT (T = 300 K, P = 1 atm); 50 kcal/mol/Å restraints on the solute heavy atoms;

2400 ps, NPT (T = 300 K, P = 1 atm); unrestrained;

a. Derivatives 2, 8 and 14: 500 ns, NPT (T = 300 K, P = 1 atm); unrestrained;

b. Remaining compounds: 100 ns, NPT (T = 300 K, P = 1 atm); unrestrained.

From 6a and 6b MD production step, 5000 and 1000 frames were respectively stored. The analysis

of the MD trajectory was basically performed with the Schrödinger suite. RMSD and RMSF of both

proteins and ligands, ligand–protein contacts along the simulations were calculated with the Simulation

Interaction Diagram (SID) tool implemented in the Desmond MD package. MD simulations were

visual inspected with VMD 1.9.4 [24]. By default, the deep H-bond involving C-Ha and Asp133 is not

detected with the SID tool. Therefore, distances and angles of such interaction in IMID 1 and IMID 2

simulations were monitored and calculated via VMD 1.9.4.

For the comparison of 500 and 100 ns MD simulations of 2, 8, and 14, raw data from the SID

output were processed using bash command lines and Knime workﬂows [57]. In Figure S1, the

interactions were categorized based on the deﬁnition of the Schrödinger tool. The stacked bar charts

were normalized over the course of the trajectory. Values over 1.0 were possible as some protein

residues may make multiple contacts of same subtype with the ligand. Figure S1 was realized with

Tibco Spotﬁre 7.11.1 (Palo Alto, CA, USA).

The percentages of interaction of the outer and central H-bond across each MD trajectory (Table

S1) and of other contacts discussed in the Results and Discussion section were basically extracted from

the SID outcomes.

The distances of the outer, central, and deep H-bonds along all the simulations were calculated

with VMD 1.9.4, combined with Knime workﬂow and plotted in Figure 6 with Tibco Spotﬁre 7.11.1

(

Palo Alto, CA, USA). Figure 7 with the angles of the deep H-bond interactions was created with the

same software.

By way of illustration, snapshots of some MD trajectories were extracted using VMD 1.9.4 and

reported in Figure S3. For the timeline representation of Figure S4, the interaction counts were basically

extracted from the SID results. In order table sto count the deep H-bond interactions in IMID 1 and

IMID 2 simulations, the SID geometric criteria for ligand-protein H-bond were applied to the distance

and angle values collected with VMD 1.9.4. Knime workﬂow, and bash command lines were used to

combine all these results in the timeline representation.

The average binding free energy (

∆G_bind) based on MM-GBSA was calculated using the

thermal_mmgbsa.py script on 100 frames extracted from each production trajectory [35,36] (Table 3).

3

.6. QM Studies

The predictions of hydrogen bond acidity are based on the work of Kenny [40]. The V

descriptor represents the value of the electrostatic potential at a distance r along the X-H vector from

the hydrogen atom. The 18 SP docking poses were used for the V (r) calculation as implemented in

Jaguar software [39]. B3LYP/6-31G** level of theory was used with r = 0.55 Å.

α

(r)

α

Molecules 2020, 25, 2163

26 of 29

3

.7. NMR Study

NMR experiments were performed in deuterated DMSO and CDCl solutions and recorded with

3

1

13

a Bruker Avance spectrometer operating at 500 MHz for H and 125 MHz for C, equipped with a

mm inverse broad-band probe and z-axis gradients.

The complete NMR spectra assignment of the three compounds was performed in CDCl solutions

5

3

at 298 K. COSY experiments were acquired with a spectral window of 15

at 4.5 ppm) using 2048 256 data points, 16 transients, with a relaxation delays of 2 s. NOESY

experiments were acquired with a spectral window of 15 15 ppm (carrier frequency at 4.5 ppm) using

096 256 data points, 32 transients, with a relaxation delay of 2 s and a mixing time of 600 ms. H- C

HSQC experiments were acquired with a spectral window of 15 ppm 200 ppm (carrier frequencies

at 6.5 and 100 ppm) using 2048 256 data points, 16 transients with relaxation delays of 2 s. H- C

HMBC experiments were acquired with a spectral window of 15 ppm 220 ppm (carrier frequency at

.5 and 110 ppm) using 4096 512 data points, 32 transients with relaxation delays of 2 s. All data

×

15 ppm (carrier frequency

×

1

13

4

×

1

13

×

6

×

were processed with TopSpin using a sine squared window function for each dimension of the 2D data.

The variations of chemical shift (∆δ) for protons of interest were calculated from the diﬀerence of

1

the chemical shift in DMSO-d6 and in CDCl , detected in the H-NMR spectra at 298 K.

3

1

For T_c_o_e_ﬀdetermination, H-NMR spectra of the three compounds were performed in deuterated

DMSO at ﬁve diﬀerent temperatures: 298 K, 303 K, 308 K, 313 K, and 318 K.

The empirical estimation of the H-bonding strength was performed using the following equation,

1

based on the ∆δ values obtained from H-NMR spectra:

E_H_B(∆δ) = ∆δ + (0.4 ± 0.2) kcal/mol.

(3)

4

. Conclusions

In this study, novel derivatives around the lead series core scaﬀold were investigated by means of

ligand docking and MD simulations. The novel IMID derivatives turned out to be GSK-3 inhibitors

β

in the nanomolar scale with a potency drop compared to the previous INDZ series but improved CNS

permeability. Interestingly, the in silico-derived models of the binding mode generated during the

design process well suited the crystallographic data (PDB entry 6Y9S). In order to rationalize SAR,

computational tools were also employed retrospectively. In particular, MD and QM-based calculations

shed light on the diﬀerences of the H-bond interactions between the core scaﬀolds and the hinge

residues moving from INDZ to IMID 1 and IMID 2 derivatives. The pharmacological proﬁle of IMID

1

and IMID 2 derivatives was associated to the lower acidity of the C-Ha versus N-Ha group of the

core scaﬀolds. NMR-based studies were also carried out to substantiate this hypothesis. Overall,

our ﬁndings provided adequate evidence that low C-Ha acidity and consequent unfavorable geometric

rearrangement aﬀected H-bond donating abilities so reducing the aﬃnity for GSK-3

β enzyme. This

represented signiﬁcant structural activity insight into our GSK-3 inhibitor design and demonstrated

β

the pivotal role of computational approaches properly supported by experimental proofs in the diﬀerent

steps of the drug discovery program. In the future, prospective application of the aforementioned in

silico tools may lead to further optimized classes of GSK-3β inhibitors.

Supplementary Materials: The following are available online, Figure S1: Ligand–Protein interactions monitored

throughout 100 ns (a) and 500 ns (b) simulations, Figure S2: Ligand RMSF from each MD simulation, Figure

S3: representative snapshots of MD simulations, Figure S4: Timeline representation of the interactions along

1

00 ns simulations for each compound, Figures S5–S7: H-NMR spectrum (500 MHz) of compounds

2

,

8

, and 14

in CDCl and DMSO-d6 at 298 K. Table S1: Percentage of interaction of the outer and central H-bond across each

MD trajectory, Table S2: Data collection and model statistics of crystal structure of

3

2 in complex with GSK-3β

enzyme (PDB entry 6Y9R), Table S3: Data collection and model statistics of crystal structure of 16 in complex with

GSK-3β enzyme (PDB entry 6Y9S), HNMR spectra of ﬁnal compounds, 2D NMR of compounds 2, 8, and 14.

Molecules 2020, 25, 2163

27 of 29

Author Contributions: Conceptualization, G.F. and R.O.; Data curation, R.B., F.P. and M.B.; Investigation, R.B. and

M.B.; Methodology, R.B., F.P., M.B. and C.C.; Project administration, R.O.; Supervision, G.F. and R.O.; Visualization,

R.B. and F.P.; Writing—original draft, R.B., F.P. and M.B.; Writing—review and editing, C.C., G.F. and R.O.

All authors have read and agreed to the published version of the manuscript.

Funding: This research received no external funding.

Acknowledgments: We would like to thank Sara Preciado (Enantia) for her support to the chemistry activities.

Conﬂicts of Interest: The authors declare no conﬂict of interest.

References

1

2

.

Sutherland, C. What Are the bona ﬁde GSK3 Substrates? Int. J. Alzheimers Dis. 2011, 2011, 505–607. [CrossRef ]

Woodgett, J.R. Molecular cloning and expression of glycogen synthase kinase-3/factor A. EMBO J. 1990, 9,

2

431–2438. [CrossRef ] [PubMed ]

3

.

Doble, B.W.; Patel, S.; Wood, G.A.; Kockeritz, L.K.; Woodgett, J.R. Functional redundancy of GSK-3alpha

and GSK-3beta in Wnt/beta-catenin signaling shown by using an allelic series of embryonic stem cell lines.

Dev. Cell 2007, 12, 957–971. [CrossRef ] [PubMed ]

4

.

Hoeﬂich, K.P.; Luo, J.; Rubie, E.A.; Tsao, M.S.; Jin, O.; Woodgett, J.R. Requirement for glycogen synthase

kinase-3beta in cell survival and NF-kappaB activation. Nature 2000, 406, 86–90. [CrossRef ] [PubMed ]

5

Patel, P.; Woodgett, J.R. Glycogen Synthase Kinase 3: A Kinase for All Pathways? Curr. Top. Dev. Biol. 2017

23, 277–302. [CrossRef ] [PubMed ]

Cohen, P.; Frame, S. The renaissance of GSK3. Nat. Rev. Mol. Cell Biol. 2001, 2, 769–776. [CrossRef ]

Hur, E.M.; Zhou, F.Q. GSK3 signaling in neural development. Nat. Rev. Neurosci. 2010, 11, 539–551.

CrossRef ]

,

1

6

7

.

[

8

9

1

.

Dandekar, M.P.; Valvassori, S.S.; Dal-Pont, G.C.; Quevedo, J. Glycogen Synthase Kinase-3beta as a Putative

Therapeutic Target for Bipolar Disorder. Curr. Drug. Metab. 2018, 19, 663–673. [CrossRef ]

Jope, R.S.; Roh, M.S. Glycogen synthase kinase-3 (GSK3) in psychiatric diseases and therapeutic interventions.

Curr. Drug Targets 2006, 7, 1421–1434. [CrossRef ]

0. Lei, P.; Ayton, S.; Bush, A.I.; Adlard, P.A. GSK-3 in Neurodegenerative Diseases. Int. J. Alzheimers Dis. 2011

011, 189–246. [CrossRef ]

1. Mines, M.A.; Jope, R.S. Glycogen synthase kinase-3: A promising therapeutic target for fragile x syndrome.

Front. Mol. Neurosci. 2011, 4, 35. [CrossRef ] [PubMed ]

,

2

2. Eldar-Finkelman, H. Glycogen synthase kinase 3: An emerging therapeutic target. Trends Mol. Med. 2002, 8,

1

26–132. [CrossRef ]

3. Meijer, L.; Flajolet, M.; Greengard, P. Pharmacological inhibitors of glycogen synthase kinase 3. Trends

Pharmacol. Sci. 2004, 25, 471–480. [CrossRef ] [PubMed ]

4. Klein, P.S.; Melton, D.A. A molecular mechanism for the eﬀect of lithium on development. Proc. Natl. Acad.

Sci. USA 1996, 93, 8455–8459. [CrossRef ]

5. Yuan, P.; Chen, G.; Manji, H.K. Lithium activates the c-Jun NH2-terminal kinases in vitro and in the CNS

in vivo. J. Neurochem. 1999, 73, 2299–2309. [CrossRef ]

6. Ombrato, R.; Cazzolla, N.; Mancini, F.; Mangano, G. Structure-Based Discovery of 1H-Indazole-3-

carboxamides as a Novel Structural Class of Human GSK-3 Inhibitors. J. Chem. Inf. Model. 2015, 55,

2

540–2551. [CrossRef ]

1

7. Furlotti, G.; Alisi, M.A.; Cazzolla, N.; Dragone, P.; Durando, L.; Magaro, G.; Mancini, F.; Mangano, G.;

Ombrato, R.; Vitiello, M.; et al. Hit Optimization of 5-Substituted-N-(piperidin-4-ylmethyl)-1H-indazole-

3-carboxamides: Potent Glycogen Synthase Kinase-3 (GSK-3) Inhibitors with in Vivo Activity in Model of

Mood Disorders. J. Med. Chem. 2015, 58, 8920–8937. [CrossRef ]

1

8. Furlotti, G.; Cavarischia, C.; Buonﬁglio, R.; Ombrato, R.; Iacoangeli, T. 1H-Indazole-3-Carboxamide

Compounds as Glycogen Synthase Kinase 3 beta Inhibitors. Patent WO2019215075, 6 May 2019.

9. Prati, F.; Buonﬁglio, R.; Furlotti, G.; Cavarischia, C.; Mangano, G.; Picollo, R.; Oggianu, L.; Di Matteo, A.;

Olivieri, S.; Bovi, G.; et al. Optimization of Indazole-Based GSK-3 Inhibitors with Mitigated hERG Issue and

In Vivo Activity in a Mood Disorder Model. ACS Med. Chem. Lett. 2020. [CrossRef ]

2

0. Zeiadeh, I.; Najjar, A.; Karaman, R. Strategies for Enhancing the Permeation of CNS-Active Drugs through

the Blood-Brain Barrier: A Review. Molecules 2018, 23, 1289. [CrossRef ]

Molecules 2020, 25, 2163

28 of 29

2

1. Ghose, A.K.; Herbertz, T.; Hudkins, R.L.; Dorsey, B.D.; Mallamo, J.P. Knowledge-Based, Central Nervous

System (CNS) Lead Selection and Lead Optimization for CNS Drug Discovery. ACS Chem. Neurosci. 2012, 3,

0–68. [CrossRef ]

2. Hitchcock, S.A.; Pennington, L.D. Structure-brain exposure relationships. J. Med. Chem. 2006, 49, 7559–7583.

CrossRef ] [PubMed ]

3. Rankovic, Z. CNS drug design: Balancing physicochemical properties for optimal brain exposure.

J. Med. Chem. 2015, 58, 2584–2608. [CrossRef ] [PubMed ]

4. Humphrey, W.; Dalke, A.; Schulten, K. VMD: Visual molecular dynamics. J. Mol. Graph. 1996, 14, 33–38.

5

2

[

[CrossRef ]

5. Horio, M.; Chin, K.V.; Currier, S.J.; Goldenberg, S.; Williams, C.; Pastan, I.; Gottesman, M.M.; Handler, J.

Transepithelial transport of drugs by the multidrug transporter in cultured Madin-Darby canine kidney cell

epithelia. J. Biol. Chem. 1989, 264, 14880–14884.

2

6. Saaby, L.; Brodin, B. A Critical View on In Vitro Analysis of P-glycoprotein (P-gp) Transport Kinetics.

J. Pharm. Sci. 2017, 106, 2257–2264. [CrossRef ]

7. Rankovic, Z. CNS Physicochemical Property Space Shaped by a Diverse Set of Molecules with Experimentally

Determined Exposure in the Mouse Brain. J. Med. Chem. 2017, 60, 5943–5954. [CrossRef ]

8. Didziapetris, R.; Japertas, P.; Avdeef, A.; Petrauskas, A. Classiﬁcation analysis of P-glycoprotein substrate

speciﬁcity. J. Drug Target. 2003, 11, 391–406. [CrossRef ]

9. Hochman, J.; Mei, Q.; Yamazaki, M.; Tang, C.; Prueksaritanont, T.; Bock, M.; Ha, S.; Lin, J. Role of Mechanistic

Transport Studies in Lead Optimization. In Optimizing the “Drug-Like” Properties of Leads in Drug Discovery;

Borchardt, R.T., Kerns, E., Hageman, M., Thakker, D., Stevens, J., Eds.; Springer: New York, NY, USA, 2006;

pp. 25–47.

3

0. Friesner, R.A.; Banks, J.L.; Murphy, R.B.; Halgren, T.A.; Klicic, J.J.; Mainz, D.T.; Repasky, M.P.; Knoll, E.H.;

Shelley, M.; Perry, J.K.; et al. Glide: A new approach for rapid, accurate docking and scoring. 1. Method and

assessment of docking accuracy. J. Med. Chem. 2004, 47, 1739–1749. [CrossRef ]

1. Halgren, T.A.; Murphy, R.B.; Friesner, R.A.; Beard, H.S.; Frye, L.L.; Pollard, W.T.; Banks, J.L. Glide: A new

approach for rapid, accurate docking and scoring. 2. Enrichment factors in database screening. J. Med. Chem.

2

004, 47, 1750–1759. [CrossRef ]

2. Bowers, K.J.; Chow, E.; Xu, H.; Dror, R.O.; Eastwood, M.P.; Gregersen, B.A.; Klepeis, J.L.; Kolossvary, I.;

Moraes, M.A.; Sacerdoti, F.D.; et al. Scalable algorithms for molecular dynamics simulations on commodity

clusters. In Proceedings of the 2006 ACM/IEEE conference on Supercomputing; Association for Computing

Machinery: Tampa, FL, USA; p. 84-es.

3

3. De Vivo, M.; Masetti, M.; Bottegoni, G.; Cavalli, A. Role of Molecular Dynamics and Related Methods in

Drug Discovery. J. Med. Chem. 2016, 59, 4035–4061. [CrossRef ]

4. Desai, P.V.; Raub, T.J.; Blanco, M.-J. How hydrogen bonds impact P-glycoprotein transport and permeability.

Bioorg. Med. Chem. Lett. 2012, 22, 6540–6548. [CrossRef ] [PubMed ]

5. Li, J.; Abel, R.; Zhu, K.; Cao, Y.; Zhao, S.; Friesner, R.A. The VSGB 2.0 model: A next generation energy model

for high resolution protein structure modeling. Proteins 2011, 79, 2794–2812. [CrossRef ] [PubMed ]

6. Shivakumar, D.; Williams, J.; Wu, Y.; Damm, W.; Shelley, J.; Sherman, W. Prediction of Absolute Solvation

Free Energies using Molecular Dynamics Free Energy Perturbation and the OPLS Force Field. J. Chem.

Theory Comput. 2010, 6, 1509–1519. [CrossRef ]

3

7. Jeﬀrey, G.A. An Introduction to Hydrogen Bonding; Oxford University Press: New York, NY, USA, 1997.

8. Huggins, M.L. The Hydrogen Bond (Pimentel, George, C.; McClellan, Aubrey, L.). J. Chem. Educ. 1960, 37,

A754. [CrossRef ]

3

9. Bochevarov, A.D.; Harder, E.; Hughes, T.F.; Greenwood, J.R.; Braden, D.A.; Philipp, D.M.; Rinaldo, D.;

Halls, M.D.; Zhang, J.; Friesner, R.A. Jaguar: A high-performance quantum chemistry software program

with strengths in life and materials sciences. Int. J. Quantum Chem. 2013, 113, 2110–2142. [CrossRef ]

0. Kenny, P.W. Hydrogen bonding, electrostatic potential, and molecular design. J. Chem. Inf. Model. 2009, 49,

4

1

234–1244. [CrossRef ] [PubMed ]

1. Türkmen, Y.E. Investigation of the hydrogen bond donating ability of 1,8-naphthalenediol byNMR

spectroscopy and its use as a hydrogen bonding catalyst. Turk. J. Chem. 2018, 42, 1398–1407. [CrossRef ]

2. Cierpicki, T.; Otlewski, J. Amide proton temperature coeﬃcients as hydrogen bond indicators in proteins.

J. Biomol. NMR 2001, 21, 249–261. [CrossRef ]

Molecules 2020, 25, 2163

29 of 29

4

3. Tilton, R.F., Jr.; Dewan, J.C.; Petsko, G.A. Eﬀects of temperature on protein structure and dynamics: X-ray

crystallographic studies of the protein ribonuclease-A at nine diﬀerent temperatures from 98 to 320 K.

Biochemistry 1992, 31, 2469–2481. [CrossRef ]

4

4. Kontogianni, V.G.; Charisiadis, P.; Primikyri, A.; Pappas, C.G.; Exarchou, V.; Tzakos, A.G.; Gerothanassis, I.P.

Hydrogen bonding probes of phenol -OH groups. Org. Biomol. Chem. 2013, 11, 1013–1025. [CrossRef ]

5. Afonin, A.V.; Vashchenko, A.V.; Sigalov, M.V. Estimating the energy of intramolecular hydrogen bonds from

(1)H-NMR and QTAIM calculations. Org. Biomol. Chem. 2016 , 14 , 11199–11211. [CrossRef ] [PubMed ]

4

6. Kabsch, W. XDS. Acta Cryst. Section D 2010, 66, 125–132. [CrossRef ] [PubMed ]

4

7. Sadowski, J.; Gasteiger, J.; Klebe, G. Comparison of Automatic Three-Dimensional Model Builders Using 639

X-Ray Structures. J. Chem. Inf. Comput. Sci. 1994, 34, 1000–1008. [CrossRef ]

4

8. Emsley, P.; Cowtan, K. Coot: Model-building tools for molecular graphics. Acta Crystallogr. D Biol. Crystallogr.

2

004, 60, 2126–2132. [CrossRef ]

9. Murshudov, G.N.; Vagin, A.A.; Dodson, E.J. Reﬁnement of macromolecular structures by the maximum-

likelihood method. Acta Crystallogr. D Biol. Crystallogr. 1997, 53, 240–255. [CrossRef ]

5

0. Schrödinger Release 2019-1; LigPrep, Schrödinger, LLC: New York, NY, USA, 2019.

1. Schrödinger Release 2019-1; ConfGen, Schrödinger, LLC: New York, NY, USA, 2019.

2. Schrödinger Release 2019-1; Glide, Schrödinger, LLC: New York, NY, USA, 2019.

3. Schrödinger Release 2019-1; Protein Preparation Wizard; Epik, Schrödinger, LLC: New York, NY, USA, 2016;

Impact, Schrödinger, LLC: New York, NY, USA, 2016; Prime, Schrödinger, LLC: New York, NY, USA, 2019.

4. Schrödinger Release 2019-1; Maestro, Schrödinger, LLC: New York, NY, USA, 2019.

5

5. The PyMOL Molecular Graphics System, Version 1.8; Schrödinger, LLC: New York, NY, USA, 2019.

6. Schrödinger Release 2019-1; Desmond Molecular Dynamics System, D.E. Shaw Research: New York, NY, USA,

2

019; Maestro-Desmond Interoperability Tools, Schrödinger: New York, NY, USA, 2019.

5

7. Berthold, M.R.; Cebron, N.; Dill, F.; Gabriel, T.R.; Kötter, T.; Meinl, T.; Ohl, P.; Sieb, C.; Thiel, K.; Wiswedel, B.

KNIME: The Konstanz Information Miner; Springer: Berlin/Heidelberg, Germany, 2008; pp. 319–326.

Sample Availability: Not available.

©

2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access

article distributed under the terms and conditions of the Creative Commons Attribution

(CC BY) license (http://creativecommons.org /licenses /by /4.0 /).

Article Doi

Discovery of novel imidazopyridine GSK-3β inhibitors supported by computational approaches

DOI: 10.3390/molecules25092163

Source and publish data:

Authors:

Article abstract of DOI:10.3390/molecules25092163

Full text of DOI:10.3390/molecules25092163

Products guided by the article

R&D Labs maybe for 865156-47-4

Relevant to this article

Hot Product