Straightforward Access to a New Class of Dual DYRK1A/CLK1 Inhibitors Possessing a Simple Dihydroquinoline Core

The DYRK (Dual-specificity tyrosine phosphorylation-regulated kinase) family of protein kinases is involved in the pathogenesis of several neurodegenerative diseases. Among them, the DYRK1A protein kinase is thought to be implicated in Alzheimer’s disease (AD) and Down syndrome, and as such, has emerged as an appealing therapeutic target. DYRKs are a subset of the CMGC (CDK, MAPKK, GSK3 and CLK) group of kinases. Within this group of kinases, the CDC2-like kinases (CLKs), such as CLK1, are closely related to DYRKs and have also sparked great interest as potential therapeutic targets for AD. Based on inhibitors previously described in the literature (namely TG003 and INDY), we report in this work a new class of dihydroquinolines exhibiting inhibitory activities in the nanomolar range on hDYRK1A and hCLK1. Moreover, there is overwhelming evidence that oxidative stress plays an important role in AD. Pleasingly, the most potent dual kinase inhibitor 1p exhibited antioxidant and radical scavenging properties. Finally, drug-likeness and molecular docking studies of this new class of DYRK1A/CLK1 inhibitors are also discussed in this article.


Introduction
Alzheimer' s disease (AD) is a neurodegenerative disease, clinically characterized by a progressive deterioration of cognitive functions, leading to an inexorable decline in functional abilities. In addition, AD patients often present neuropsychiatric and behavioral symptoms such as depression, restlessness and hallucinations leading to a deterioration of their quality of life [1]. The pathogenesis of AD is multifactorial, characterized by specific brain anatomical and pathological abnormalities: a significant extent of oxidative stress associated with the presence of extracellular deposits of amyloid-β(Aβ) peptide and intraneuronal neurofibrillary tangles [2][3][4].
Dual-specificity tyrosine phosphorylation-regulated kinase 1A (DYRK1A) is a protein kinase that is abnormally expressed in many diseases such as Down syndrome or AD [5][6][7]. In the literature, it was reported that DYRK1A is able to phosphorylate both APP on Thr668 [8] and presenilin-1 on Thr354 [9]. When the latter is hyper-phosphorylated, the activity of γ-secretase, one of the proteases that cleave APP, is increased. This leads to an abnormally high production of β-amyloid peptides, the main components of amyloid plaques, that are responsible for neurodegeneration [10,11].
DYRK1A was also found to be involved in the hyperphosphorylation of Tau protein (Tubulin-Associated Unit) [12][13][14][15][16], which triggers Tau to dissociate from neuronal microtubules and, shortly afterwards, causes microtubules to disassemble while leading to the formation of neurofibrillary tangles and final neurodegeneration. Furthermore, the CDK-like kinase (CLK) family, composed of four isoforms: CLK1, CLK2, CLK3 and CLK4, is also an interesting target for treating AD [7,[17][18][19][20]. As for DYRK1A, they are dual-specificity kinases. In particular, CLK1 is likely involved in AD by phosphorylating the serine residue in serine/arginine-rich (SR) proteins [19]. SR proteins are a family of splicing factors involved in the alternative splicing of Tau.
As part of a research program seeking new and readily accessible classes of DYRK1A and/or CLK1 kinase inhibitors, our attention was drawn to INDY [(Z)-1-(3-Ethyl-5-hydroxy-2(3H)-benzothiazolylidene)-2-propanone)] and its derivatives such as TG003 (its synthetic precursor) and ProINDY (its prodrug), reported by Ogawa [30] and Hagiwara [31], respectively ( Figure 1). These DYRK1A/CLK1 inhibitors are all derived from 5-methoxy-2methylbenzothiazole and show significant structural analogy to the 3-acetyl 7-substituted 1,4-dihydroquinoline derivatives. It is worth mentioning that this new class of dual DYRK1A and CLK1 inhibitors based on a biooxidizable 1,4-dihydroquinoline scaffold may also undergo oxidation into quinolinium salts. That could lead to antioxidant properties going hand in hand with the attenuation of oxidative stress in brain tissues. In the present article, we report the synthesis of these biooxidizable 1,4-dihydroquinoline derivatives, a preliminary in vitro biological evaluation toward DYRK1A and CLK1 kinases, and an examination of their redox properties in various biological media.
Furthermore, the CDK-like kinase (CLK) family, composed of four isoforms: CLK1, CLK2, CLK3 and CLK4, is also an interesting target for treating AD [7,[17][18][19][20]. As for DYRK1A, they are dual-specificity kinases. In particular, CLK1 is likely involved in AD by phosphorylating the serine residue in serine/arginine-rich (SR) proteins [19]. SR proteins are a family of splicing factors involved in the alternative splicing of Tau.
As part of a research program seeking new and readily accessible classes of DYRK1A and/or CLK1 kinase inhibitors, our attention was drawn to INDY [(Z)-1-(3-Ethyl-5-hydroxy-2(3H)-benzothiazolylidene)-2-propanone)] and its derivatives such as TG003 (its synthetic precursor) and ProINDY (its prodrug), reported by Ogawa [30] and Hagiwara [31], respectively ( Figure 1). These DYRK1A/CLK1 inhibitors are all derived from 5-methoxy-2-methylbenzothiazole and show significant structural analogy to the 3-acetyl 7-substituted 1,4-dihydroquinoline derivatives. It is worth mentioning that this new class of dual DYRK1A and CLK1 inhibitors based on a biooxidizable 1,4-dihydroquinoline scaffold may also undergo oxidation into quinolinium salts. That could lead to antioxidant properties going hand in hand with the attenuation of oxidative stress in brain tissues. In the present article, we report the synthesis of these biooxidizable 1,4-dihydroquinoline derivatives, a preliminary in vitro biological evaluation toward DYRK1A and CLK1 kinases, and an examination of their redox properties in various biological media.

Synthesis of the Studied Dual hDYRK1A and hCLK1 Inhibitors
Apart from commercial quinoline 6a, 3-acetyl quinoline derivatives 6b-h were obtained by reacting the corresponding o-nitrobenzaldehyde derivatives 3b-h with 4,4-dimethoxybutan-2-one 4 in the presence of tin chloride as a reducing reagent (Scheme 1). This domino nitro reduction-Friedländer cyclisation, previously reported in the literature for the preparation of quinoline-3-carboxylic acid esters [32], provided access to the required 3-acetyl quinoline derivatives 6b-h in a one-step procedure from readily available and stable starting materials.
We then focused on the preparation of compounds 6i-m to investigate the influence of the electron withdrawing group (EWG) at C-3 position of the quinoline ring on the

Synthesis of the Studied Dual hDYRK1A and hCLK1 Inhibitors
Apart from commercial quinoline 6a, 3-acetyl quinoline derivatives 6b-h were obtained by reacting the corresponding o-nitrobenzaldehyde derivatives 3b-h with 4,4dimethoxybutan-2-one 4 in the presence of tin chloride as a reducing reagent (Scheme 1). This domino nitro reduction-Friedländer cyclisation, previously reported in the literature for the preparation of quinoline-3-carboxylic acid esters [32], provided access to the required 3-acetyl quinoline derivatives 6b-h in a one-step procedure from readily available and stable starting materials. a classical quaternization reaction with alkyl halides to afford the desired quinolinium salts 2a-p, which were then successfully regioselectively reduced with BNAH to the corresponding 1,4-dihydroquinolines 1a-p. With the target compounds 1a-p in hand, readily obtained in a few steps, and in most cases, in a good overall yield, we embarked on an initial biological evaluation of these dihydroquinoline scaffolds as potential new DYRK1A and CLK1 inhibitors (Table 1) We then focused on the preparation of compounds 6i-m to investigate the influence of the electron withdrawing group (EWG) at C-3 position of the quinoline ring on the inhibitory activity of the DYRK1A and CLK1 kinases. To this end, the 3-cyano-7-methoxy quinoline 6i previously reported in our group [33] was reacted with chlorotrimethylsilane and anhydrous methanol [34] to provide methyl ester 6j. Quinoline ester 6k was obtained by reacting thionyl chloride on the reported quinoline 3-carboxylic acid 5 [35] in ethanol. Finally 6l,m were prepared from quinoline derivative 5 using T3P ® and CDI amide coupling reagents, respectively. Subsequently, quinoline derivatives 6a-m were subjected to a classical quaternization reaction with alkyl halides to afford the desired quinolinium salts 2a-p, which were then successfully regioselectively reduced with BNAH to the corresponding 1,4-dihydroquinolines 1a-p. With the target compounds 1a-p in hand, readily obtained in a few steps, and in most cases, in a good overall yield, we embarked on an initial biological evaluation of these dihydroquinoline scaffolds as potential new DYRK1A and CLK1 inhibitors (Table 1).

In Vitro Evaluation on hDYRK1A and hCLK1 Kinases
We initially focused on the influence of the substitution patterns at the benzene ring by comparing the hDYRK1A and hCLK1 inhibitory activities of a set of dihydroquinolines 1a-h differently substituted at C-6 and C-7 (EWG = COMe and R 3 = Me). To our delight, compound 1h displayed interesting three-digit nanomolar hDYRK1A and hCLK1 inhibitory activities (Entry 8) comparable to those of INDY and TG003, whereas compounds 1ad showed no inhibition on either of the two kinases (Entries 1−4). This first series revealed that a methoxy group at C-7 is essential while an additional substituent at C-6 appears to be deleterious, resulting in a complete loss of the inhibitory activity (Entries 6,7). The influence of the EWG at C-3 position, which is crucial to ensure good stability of the 1,4dihydroquinoline, was then assessed by replacing the acetyl group in 1h by a variety of EWGs (nitrile, amides, esters). The resulting compounds 1i-m turned out to be inactive toward both hDYRK1A and hCLK1 kinases (Entries 9−13), with the sole exception of the ester derivatives 1j,k that exhibited micromolar hCLK1 inhibitory activities (Entries 10,11). After demonstrating that both an acetyl group at C-3 position and a methoxy group at C-7 position are required, we turned our attention to the role of the alkyl group on the nitrogen dihydroquinoline scaffold. Compared to N-methyl dihydroquinoline 1h, N-benzyl and N-phenethyl dihydroquinolines 1n,o exhibited lower hDYRK1A inhibitory activities, while displaying comparable three-digit nanomolar hCLK1 inhibitory activities (Entries 14,15). Gratifyingly, N-propyl dihydroquinoline 1p turned out to be 2.5-fold more potent than N-methyl dihydroquinoline 1h toward hDYRK1A and 8-fold more potent than N-methyl dihydroquinoline 1h toward hCLK1 (Entries 8,16). To complete this SAR analysis, we sought to determine whether or not the enamine function is a crucial element for this new class of dual DYRK1A/CLK1 inhibitors. To tackle this issue, quinolinium salts 2h and quinoline 6h were also evaluated for inhibition of hDYRK1A/hCLK1. None of these synthetic intermediates showed inhibitory activities on hDYRK1A/hCLK1 kinases (IC50 > 10 mM). These observations led us to conclude that, in addition to the acetyl group at C-3

In Vitro Evaluation on hDYRK1A and hCLK1 Kinases
We initially focused on the influence of the substitution patterns at the benzene ring by comparing the hDYRK1A and hCLK1 inhibitory activities of a set of dihydroquinolines 1a-h differently substituted at C-6 and C-7 (EWG = COMe and R 3 = Me). To our delight, compound 1h displayed interesting three-digit nanomolar hDYRK1A and hCLK1 inhibitory activities (Entry 8) comparable to those of INDY and TG003, whereas compounds 1a-d showed no inhibition on either of the two kinases (Entries 1−4). This first series revealed that a methoxy group at C-7 is essential while an additional substituent at C-6 appears to be deleterious, resulting in a complete loss of the inhibitory activity (Entries 6,7). The influence of the EWG at C-3 position, which is crucial to ensure good stability of the 1,4dihydroquinoline, was then assessed by replacing the acetyl group in 1h by a variety of EWGs (nitrile, amides, esters). The resulting compounds 1i-m turned out to be inactive toward both hDYRK1A and hCLK1 kinases (Entries 9−13), with the sole exception of the ester derivatives 1j,k that exhibited micromolar hCLK1 inhibitory activities (Entries 10,11). After demonstrating that both an acetyl group at C-3 position and a methoxy group at C-7 position are required, we turned our attention to the role of the alkyl group on the nitrogen dihydroquinoline scaffold. Compared to N-methyl dihydroquinoline 1h, N-benzyl and N-phenethyl dihydroquinolines 1n,o exhibited lower hDYRK1A inhibitory activities, while displaying comparable three-digit nanomolar hCLK1 inhibitory activities (Entries 14,15). Gratifyingly, N-propyl dihydroquinoline 1p turned out to be 2.5-fold more potent than N-methyl dihydroquinoline 1h toward hDYRK1A and 8-fold more potent than N-methyl dihydroquinoline 1h toward hCLK1 (Entries 8,16). To complete this SAR analysis, we sought to determine whether or not the enamine function is a crucial element for this new class of dual DYRK1A/CLK1 inhibitors. To tackle this issue, quinolinium salts 2h and quinoline 6h were also evaluated for inhibition of hDYRK1A/hCLK1. None of these synthetic intermediates showed inhibitory activities on hDYRK1A/hCLK1 kinases (IC 50 > 10 mM). These observations led us to conclude that, in addition to the acetyl group at C-3 and the methoxy group at C-7, the enamine-like nitrogen atom is also a prerequisite for inhibition of hDYRK1A/hCLK1.

Evaluation of the Antioxidant and Radical Scavenging Activities of 1p
Oxidative stress is also an important hallmark of AD [36][37][38][39][40] and is closely linked with the formation of both the neurofibrillary tangles and the amyloid plaques.
A further interesting aspect of this novel class of hDYRK1A/hCLK1 inhibitors is related to the potential of 1,4-dihydroquinolines 1 to prevent oxidative stress by exhibiting in vivo antioxidant and radical scavenging properties during their conversion to the corresponding quinolinium salts 2 in the Central Nervous System (CNS) [35,41].
To evaluate the ability of 1,4-dihydroquinolines 1 to act as antioxidants, we studied their behavior by incubating them in different oxidative media [42] (NAD + , mouse brain homogenate, H 2 O 2 , riboflavin) ( Figure 2). First of all, in vitro stability of the most potent inhibitor 1p was evaluated in PBS and human plasma at 37 • C. After 3.5 h incubation in these non-oxidizing media, only 3% of the oxidation product 2p was observed in PBS and 4% in human plasma (8% after 24 h for both media). This finding led us to predict a good stability of 1p at the periphery. Then, in the presence of the oxidizing medium consisting of 0.2% NAD + in PBS, the oxidation rate of dihydroquinoline 1p was relatively low (5% after 3.5 h and 16% after 24 h). However, the formation of the oxidation product 2p was shown to be much faster when using 2% NAD + (24% after 3.5 h and 77% after 24 h). A similar result was observed when 1p was exposed to 20% fresh mouse brain homogenate (25% after 3.5 h and 72% after 24 h), while hydrogen peroxide 0.1% appeared to be slightly less effective (13% after 3.5 h and 53% after 24 h). Finally, in vitro oxidation with 0.1% riboflavin showed an extremely rapid oxidative conversion of 1p into the corresponding quaternary salt 2p (100% in less than 3.5 h).
Molecules 2022, 27, x FOR PEER REVIEW 5 of 20 and the methoxy group at C-7, the enamine-like nitrogen atom is also a prerequisite for inhibition of hDYRK1A/hCLK1.

Evaluation of the Antioxidant and Radical Scavenging Activities of 1p
Oxidative stress is also an important hallmark of AD [36][37][38][39][40] and is closely linked with the formation of both the neurofibrillary tangles and the amyloid plaques.
A further interesting aspect of this novel class of hDYRK1A/hCLK1 inhibitors is related to the potential of 1,4-dihydroquinolines 1 to prevent oxidative stress by exhibiting in vivo antioxidant and radical scavenging properties during their conversion to the corresponding quinolinium salts 2 in the Central Nervous System (CNS) [35,41].
To evaluate the ability of 1,4-dihydroquinolines 1 to act as antioxidants, we studied their behavior by incubating them in different oxidative media [42] (NAD + , mouse brain homogenate, H2O2, riboflavin) ( Figure 2). First of all, in vitro stability of the most potent inhibitor 1p was evaluated in PBS and human plasma at 37 °C . After 3.5 h incubation in these non-oxidizing media, only 3% of the oxidation product 2p was observed in PBS and 4% in human plasma (8% after 24 h for both media). This finding led us to predict a good stability of 1p at the periphery. Then, in the presence of the oxidizing medium consisting of 0.2% NAD + in PBS, the oxidation rate of dihydroquinoline 1p was relatively low (5% after 3.5 h and 16% after 24 h). However, the formation of the oxidation product 2p was shown to be much faster when using 2% NAD + (24% after 3.5 h and 77% after 24 h). A similar result was observed when 1p was exposed to 20% fresh mouse brain homogenate (25% after 3.5 h and 72% after 24 h), while hydrogen peroxide 0.1% appeared to be slightly less effective (13% after 3.5 h and 53% after 24 h). Finally, in vitro oxidation with 0.1% riboflavin showed an extremely rapid oxidative conversion of 1p into the corresponding quaternary salt 2p (100% in less than 3.5 h). Finally, the aptitude of 1p to act as a free radical scavenger was assessed using the rapid, inexpensive and widely used DPPH • (2,2-diphenyl-1-picrylhydrazyl) scavenging method [43].  Finally, the aptitude of 1p to act as a free radical scavenger was assessed using the rapid, inexpensive and widely used DPPH • (2,2-diphenyl-1-picrylhydrazyl) scavenging method [43]. Dark purple DPPH • is a stable free radical that can accept an electron or hydrogen to become a stable diamagnetic molecule with the loss of violet color. Interestingly, we were able to determine that dihydroquinoline 1p exhibited a noticeable radical scavenging activity (EC 50 = 128 µM). Moreover, 1 H NMR experiments conducted in DMSOd 6 showed that dihydroquinoline 1p reacts with DPPH • almost instantly to furnish the corresponding inactive quinolinium salt 2p (see Supplementary Materials). Thus, this novel class of hDYRK1A/hCLK1 inhibitors are all the more interesting that they may also potentially display an additional antioxidant activity to combat oxidative stress in AD.

Drug-likeness Evaluation
Drug candidates should possess favorable ADME (Absorption, Distribution, Metabolism, and Excretion) properties. In the literature, Lipinski [44], based on a study of orally active drugs, established a set of simple rules for estimating permeability. These rules were later adjusted and others were added to design successful CNS drugs [45]. To predict the CNS druggability of 1p, its ability to cross the blood-brain barrier (BBB) by passive diffusion was examined using the free online server SwissADME [46] (Table 2). According to the results obtained (see Supplementary Materials) and as illustrated by the "BOILED-egg" (Brain Or IntestinaL EstimateD permeation) [47], 1,4-dihydroquinoline 1p may be viewed as a promising CNS drug candidate. The "BOILED-egg" method is based on a descriptive delineation of well-absorbed drugs. This robust and accurate predictive model of passive absorption is based on the calculation of physicochemical descriptors.  materials). Thus, this novel class of hDYRK1A/hCLK1 inhibitors are all the more interesting that they may also potentially display an additional antioxidant activity to combat oxidative stress in AD.

Drug-likeness Evaluation
Drug candidates should possess favorable ADME (Absorption, Distribution, Metabolism, and Excretion) properties. In the literature, Lipinski [44], based on a study of orally active drugs, established a set of simple rules for estimating permeability. These rules were later adjusted and others were added to design successful CNS drugs [45]. To predict the CNS druggability of 1p, its ability to cross the blood-brain barrier (BBB) by passive diffusion was examined using the free online server SwissADME [46] (Table 2). According to the results obtained (see supplementary materials) and as illustrated by the "BOILEDegg" (Brain Or IntestinaL EstimateD permeation) [47], 1,4-dihydroquinoline 1p may be viewed as a promising CNS drug candidate. The "BOILED-egg" method is based on a descriptive delineation of well-absorbed drugs. This robust and accurate predictive model of passive absorption is based on the calculation of physicochemical descriptors. To confirm these encouraging results, the ability of 1p to cross the BBB by passive diffusion was assessed through the parallel artificial membrane permeability assay of blood-brain barrier (PAMPA-BBB) [48]. The obtained permeability value suggests that 1p easily diffuses across the BBB (Pe = 20.0 ± 0.9 × 10 −6 cm.s −1 ).

Docking Studies
Visualization of the hDYRK1A structure co-crystallized with INDY (PDB ID: 3ANQ [30]), a DYRK1A benzothiazole derivative ATP-competitive inhibitor (IC50 = 0.24 µ M and Ki = 0.18 µ M), and hCLK1 co-crystallized with the INDY derivative inhibitor TG003 (PDB ID: 6YTE [49], Kd = 75 nM) showed that the two ligands bind in both kinases' active sites through a very similar interaction network ( Figure 3A, 3B). In both structures, the aromatic ring of the ligands was placed parallel to the beta sheet delimiting the ATP binding site and the ligands established two hydrogen bonds: i) a first one with a backbone NH of Leu241-hDYRK1A or Leu244-hCLK1, and ii) a second one with NH3 + of the Lys188-hDYRK1A or Lys191-hCLK1 side chain. Even though the lysine side chain is long and flexible, in both kinases its orientation is fixed through salt bridges with Glu203-hDYRK1A and Glu206-hCLK1. Therefore, for a ligand binding in the "INDY way" to the To confirm these encouraging results, the ability of 1p to cross the BBB by passive diffusion was assessed through the parallel artificial membrane permeability assay of blood-brain barrier (PAMPA-BBB) [48]. The obtained permeability value suggests that 1p easily diffuses across the BBB (P e = 20.0 ± 0.9 × 10 −6 cm·s −1 ).

Docking Studies
Visualization of the hDYRK1A structure co-crystallized with INDY (PDB ID: 3ANQ [30]), a DYRK1A benzothiazole derivative ATP-competitive inhibitor (IC 50 = 0.24 µM and K i = 0.18 µM), and hCLK1 co-crystallized with the INDY derivative inhibitor TG003 (PDB ID: 6YTE [49], K d = 75 nM) showed that the two ligands bind in both kinases' active sites through a very similar interaction network ( Figure 3A,B). In both structures, the aromatic ring of the ligands was placed parallel to the beta sheet delimiting the ATP binding site and the ligands established two hydrogen bonds: (i) a first one with a backbone NH of Leu241-hDYRK1A or Leu244-hCLK1, and (ii) a second one with NH 3 + of the Lys188-hDYRK1A or Lys191-hCLK1 side chain. Even though the lysine side chain is long and flexible, in both kinases its orientation is fixed through salt bridges with Glu203-hDYRK1A and Glu206-hCLK1. Therefore, for a ligand binding in the "INDY way" to the two kinases, it is necessary to have two hydrogen bond acceptors at a distance of about 8.5 Å. Among the synthetized ligands, only ligands with the EWG = COCH 3 and R 1 substituent = OCH 3 presented interesting activities. Indeed, the distance between their oxygen atoms is about 8.7 Å. To gain further insight into the binding mode of the dihydroquinoline derivatives, compound 1h docking studies using the GOLD program applying the ChemPLP scoring function were carried out. GOLD is an automated ligand docking program that uses a genetic algorithm. GOLD's evolutionary algorithm modifies the position, orientation and conformation of a ligand to fit into one or more low energy states of the protein's active site. Our docking studies confirmed that compound 1h reproduces well the INDY binding mode in hDYRK1A and in hCLK1. For both kinases, the proposed poses by GOLD converged to the same ligand orientation and the best scoring ones are represented on Figure 3C,D. The compound 1h ChemPLP fit score was 63.59 in hDYRK1A and 67.99 in hCLK1. As can be seen on Figure 3, compound 1h is able to establish the two crucial hydrogen bonds with both kinases, the first one with the Leu backbone NH and the second one with NH 3 + of Lys.
Molecules 2022, 27, x FOR PEER REVIEW 7 of 20 two kinases, it is necessary to have two hydrogen bond acceptors at a distance of about 8.5 Å . Among the synthetized ligands, only ligands with the EWG = COCH3 and R 1 substituent = OCH3 presented interesting activities. Indeed, the distance between their oxygen atoms is about 8.7 Å . To gain further insight into the binding mode of the dihydroquinoline derivatives, compound 1h docking studies using the GOLD program applying the ChemPLP scoring function were carried out. GOLD is an automated ligand docking program that uses a genetic algorithm. GOLD's evolutionary algorithm modifies the position, orientation and conformation of a ligand to fit into one or more low energy states of the protein's active site. Our docking studies confirmed that compound 1h reproduces well the INDY binding mode in hDYRK1A and in hCLK1. For both kinases, the proposed poses by GOLD converged to the same ligand orientation and the best scoring ones are represented on Figure 3C,D. The compound 1h ChemPLP fit score was 63.59 in hDYRK1A and 67.99 in hCLK1. As can be seen on Figure 3, compound 1h is able to establish the two crucial hydrogen bonds with both kinases, the first one with the Leu backbone NH and the second one with NH3 + of Lys.

Docking Studies
The initial model of compound 1h was built using BIOVIA Discovery Studio v19.1 (BIOVIA, San Diego, CA, USA) and its preferential protonation state at pH 7.4 was checked using standard tools of the ChemAxon Package (ChemAxon Ltd., Budapest, Hungary) [50]; the majority nonprotonated microspecies was used for the docking studies.
The crystallographic coordinates of human CLK1 used in this study were obtained from the X-ray structure of CLK1 bound with the benzothiazole TG003 inhibitor (PDB ID

Docking Studies
The initial model of compound 1h was built using BIOVIA Discovery Studio v19.1 (BIOVIA, San Diego, CA, USA) and its preferential protonation state at pH 7.4 was checked using standard tools of the ChemAxon Package (ChemAxon Ltd., Budapest, Hungary) [50]; the majority nonprotonated microspecies was used for the docking studies.
The crystallographic coordinates of human CLK1 used in this study were obtained from the X-ray structure of CLK1 bound with the benzothiazole TG003 inhibitor (PDB ID 6YTE [49], a structure refined to 2.30 Å with an R factor of 18.0%) and those of human DYRK1A from the X-ray structure of DYRK1A co-crystallized with the INDY inhibitor (PDB ID 3ANQ [30], a structure refined to 2.60 Å with an R factor of 23.6%).
The docking of compound 1h into the hCLK1 and hDYRK1 was carried out with the GOLD program v5.3 (The Cambridge Crystallographic Data Centre CCDC, Cambridge, United Kingdom) using the default parameters [51,52]. This program applies a genetic algorithm to explore conformational spaces and ligand binding modes. To evaluate the proposed ligand positions, the ChemPLP fitness function was applied. The binding site in both kinases was defined as a 6 Å sphere from the co-crystallized ligand.

Chemistry
All commercial reagents were used without further purification. The solvents were dried with appropriate desiccants and distilled prior to use or were obtained anhydrous from commercial suppliers. Silica gel (60, 230-400 mesh or 70-230 mesh) was used for column chromatography. Reactions were monitored by thin layer chromatography on silica gel precoated aluminum plates. UV light at 254 nm or KMnO 4 stains were used to visualize TLC plates. 1 H, 13 C, 19 F NMR spectra were recorded using a spectrometer operating at 300, 75 and 282 MHz, respectively. Abbreviations used for peak multiplicities are s: singlet, d: doublet, t: triplet, q: quadruplet dd = doublet of doublet, br = broad and m: multiplet. Coupling constants J are in Hz and chemical shifts are given in ppm and calibrated with DMSO-d 6 or CDCl 3 (residual solvent signals). 1 H NMR spectra obtained in CDCl 3 were referenced to 7.26 ppm. 13 C NMR spectra obtained in CDCl 3 were referenced to 77.16 ppm and in DMSO-d 6 were referenced to 39.52 ppm. 19 F NMR chemical shifts (δ) were determined relative to CFCl 3 as an internal standard ( 19 F, δ = 0.0 ppm). Nitrobenzaldehyde derivatives 3b-e, g, h and 4,4-dimethoxybutan-2-one 4 were obtained from Sigma-Aldrich, while 3f was commercially available at Fisher Scientific. 7-methoxy quinoline carboxylic acid 5 was prepared as previously reported in the literature [33,35,41]. 3-acetylquinoline 6a was purchased from Fisher Scientific. 7-methoxyquinoline-3-carbonitrile 6i was prepared using a synthesis previously described [33]. Elemental analyses were performed by the microanalysis service of the University of Rouen and were recorded with a Thermo Scientific™ FLASH 2000 analyzer (Thermo Fisher Scientific, Waltham, MA, USA).
General procedure A for the synthesis of compounds 1a-p The corresponding quinolinium salt 2a-p and BNAH were placed in dry and degassed dichloromethane in a round-bottomed flask and under inert atmosphere. The resulting suspension was stirred in the dark at 20 • C for 4 h. Thereafter, degassed dichloromethane (20 mL) was added to the reaction mixture and the solution was washed with degassed water (2 × 20 mL) and brine. The organic phase was dried on MgSO 4 and evaporated to dryness to give the corresponding dihydroquinoline. General procedure B for the synthesis of compounds 2a-p In a glass tube, to a solution of the corresponding quinoline 6a-m in dry acetonitrile was added a large excess of alkyl halide at 20 • C. The tube was sealed and the resulting mixture was heated at 85 • C until reaction completion (18 h for MeI and up to 4 days for the rest of the alkylating agents). After cooling at room temperature, dichloromethane (5 mL) and diethyl ether (15 mL) were added to the suspension and the mixture was stirred for 10 min. The formed precipitate was filtered, rinsed twice with diethyl ether, and dried under vacuum to afford the corresponding quinolinium salt.