5-Methoxybenzothiophene-2-Carboxamides as Inhibitors of Clk1/4: Optimization of Selectivity and Cellular Potency

Clks have been shown by recent studies to be promising targets for cancer therapy, as they are considered key regulators in the process of pre-mRNA splicing, which in turn affects every aspect of tumor biology. In particular, Clk1 and -4 are overexpressed in several human tumors. Most of the potent Clk1 inhibitors reported in the literature are non-selective, mainly showing off-target activity towards Clk2, Dyrk1A and Dyrk1B. Herein, we present new 5-methoxybenzothiophene-2-carboxamide derivatives with unprecedented selectivity. In particular, the introduction of a 3,5-difluoro benzyl extension to the methylated amide led to the discovery of compound 10b (cell-free IC50 = 12.7 nM), which was four times more selective for Clk1 over Clk2 than the previously published flagship compound 1b. Moreover, 10b showed an improved growth inhibitory activity with T24 cells (GI50 = 0.43 µM). Furthermore, a new binding model in the ATP pocket of Clk1 was developed based on the structure-activity relationships derived from new rigidified analogues.


Introduction
Pre-mRNA splicing represents a critically important step for various processes such as development and differentiation. Recently, there has been increased interest in mutations of pre-mRNA splicing factors and their roles in oncogenesis of hematological as well as solid cancers [1]. High-frequency mutations of SRSF2 (serine/arginine-rich splicing factor 2) have been described in patients with myelodysplastic syndromes (MDS), chronic myelomonocytic leukemia and acute myeloid leukemia (AML), in addition to various mutations that have been found in splicing-related genes in lung, breast and pancreatic cancers. Therefore, pharmacological modulation of pre-mRNA splicing as a novel strategy in combating cancer has gained attention in the last decade [1].
Clks (cdc-like kinases) belong to the most essential regulators in the process of pre-mRNA splicing, which is achieved through their ability to phosphorylate and activate serine/arginine-rich (SR) proteins. Upon phosphorylation, SR proteins are released from nuclear speckles into the nucleoplasm, where they modulate the selection of splice sites during pre-mRNA processing [2]. Targeting Clks has great potential to serve as a novel  The lack of isotype selectivity can be explained on the basis of the high homology between the isoenzymes: Both Clk1 and -4 share a 78.4% sequence identity [18], with both kinases harboring fully identical amino acid residues in and around their ATP binding pockets [2]. Moreover, Prak et al. also described a high degree of similarity between the catalytic domain of Clk1 and Clk2; indeed, the inhibitors reported in this study [19] and those cited above were equipotent against both kinases. Hence, it appears a challenging task to develop inhibitors specific for a certain isoform among the Clk enzyme family.
Herein we report the optimization of our previously reported benzo[b]thiophen-2carboxamides [2] through systematic synthetic modifications of the amide linker and the benzyl extensions to the amide function, followed by a biological evaluation.

Results
To improve the potency of the previous 5-methoxybenzothiophene-2-carboxamide (1b, Figure 2) and its selectivity over the most common off-targets for Clk1 inhibitors [13], various structural modifications were applied, which included modifications at the amide linker as well as introduction of various structural extensions at the amide function (summarized in Figure 2). These aimed at the stabilization of the biologically active conformation of the benzyl moiety for Clk1 inhibition. In addition, the conformationally constrained analogues were also planned in order to verify and possibly refine the previously proposed binding mode. In parallel, many new mono-and di-substituted benzyl extensions were included, taking advantage of the favorable effect of the fluorine substituent in 1b, while introducing additional groups for the interaction with the receptor but also for establishing intramolecular H-bonds (e.g., in 13a and 14a (Scheme 1), between the amide and the 2-methoxy substituent).

Results
To improve the potency of the previous 5-methoxybenzothiophene-2-carboxamide (1b, Figure 2) and its selectivity over the most common off-targets for Clk1 inhibitors [13], various structural modifications were applied, which included modifications at the amide linker as well as introduction of various structural extensions at the amide function (summarized in Figure 2). These aimed at the stabilization of the biologically active conformation of the benzyl moiety for Clk1 inhibition. In addition, the conformationally constrained analogues were also planned in order to verify and possibly refine the previously proposed binding mode. In parallel, many new mono-and di-substituted benzyl extensions were included, taking advantage of the favorable effect of the fluorine substituent in 1b, while introducing additional groups for the interaction with the receptor but also for establishing intramolecular H-bonds (e.g., in 13a and 14a (Scheme 1), between the amide and the 2-methoxy substituent).

Chemistry
A three-step synthesis was employed to access the 5-methoxybenzothiophene-2carboxamides (Scheme 1). Ethyl thioglycolate was reacted with 2-fluoro-5methoxybenzaldehyde in the presence of potassium carbonate to produce the 5methoxybenzothiophene-2-carboxylic acid ethyl ester (I) in a good yield. (I) was subjected to alkaline ester hydrolysis to produce the 5-methoxybenzothiophene carboxylic acid (II),

In Vitro Clk1/Clk2 Inhibitory Activity
All the newly synthesized derivatives (compounds 1c, 1d, 3a-16a, 3b, 4b, and 7b-15b) were tested for their ability to inhibit Clk1 and Clk2 in vitro. With Clk1, the compounds were initially screened at a concentration of 100 nM in duplicates; with Clk2 the initial screening dose increased to 250 nM. IC50s were determined for compounds that displayed a percentage of inhibition higher than 50% in the initial screening, through testing a range of five concentrations with at least two replicates per concentration (Tables  1 and 2). The Clk1 and Clk2 inhibitory activities of compounds 1a, 2a, 1b, 2b from our previous study are included in Table 1 for comparison. Scheme 2. Synthesis of compounds 1c, 1d, 3b, 4b, 7b-15b. Reagents and conditions: (i) 3 eq. KHMDS, 0 • C, 1 h, 1.5 eq. CH 3 I, or CH 3 CH 2 I or allyl bromide, room temperature, 24 h, %yield = 14-79%.

Biological
Evaluation and Development of a Binding Model 2.2.1. In Vitro Clk1/Clk2 Inhibitory Activity All the newly synthesized derivatives (compounds 1c, 1d, 3a-16a, 3b, 4b, and 7b-15b) were tested for their ability to inhibit Clk1 and Clk2 in vitro. With Clk1, the compounds were initially screened at a concentration of 100 nM in duplicates; with Clk2 the initial screening dose increased to 250 nM. IC 50 s were determined for compounds that displayed a percentage of inhibition higher than 50% in the initial screening, through testing a range of five concentrations with at least two replicates per concentration (Tables 1 and 2). The Clk1 and Clk2 inhibitory activities of compounds 1a, 2a, 1b, 2b from our previous study are included in Table 1 for comparison.                   a Data shown are the mean of at least two independent experiments, SD ≤ 10%. b IC50 ± SD, n.d., not determined; n.i., no inhibition. c Compounds 1a, 1b, 2a and 2b were reported previously in Ref. [2]. a Data shown are the mean of at least two independent experiments, SD ≤ 10%. b IC50 ± SD, n.d., not determined; n.i., no inhibition. c Compounds 1a, 1b, 2a and 2b were reported previously in Ref. [2]. a Data shown are the mean of at least two independent experiments, SD ≤ 10%. b IC50 ± SD, n.d., not determined; n.i., no inhibition. c Compounds 1a, 1b, 2a and 2b were reported previously in Ref. [2]. a Data shown are the mean of at least two independent experiments, SD ≤ 10%. b IC50 ± SD, n.d., not determined; n.i., no inhibition. c Compounds 1a, 1b, 2a and 2b were reported previously in Ref. [2]. a Data shown are the mean of at least two independent experiments, SD ≤ 10%. b IC50 ± SD, n.d., not determined; n.i., no inhibition. c Compounds 1a, 1b, 2a and 2b were reported previously in Ref. [2]. a Data shown are the mean of at least two independent experiments, SD ≤ 10%. b IC50 ± SD, n.d., not determined; n.i., no inhibition. c Compounds 1a, 1b, 2a and 2b were reported previously in Ref. [2]. a Data shown are the mean of at least two independent experiments, SD ≤ 10%. b IC50 ± SD, n.d., not determined; n.i., no inhibition. c Compounds 1a, 1b, 2a and 2b were reported previously in Ref. [2]. a Data shown are the mean of at least two independent experiments, SD ≤ 10%. b IC50 ± SD, n.d., not determined; n.i., no inhibition. c Compounds 1a, 1b, 2a and 2b were reported previously in Ref. [2]. a Data shown are the mean of at least two independent experiments, SD ≤ 10%. b IC50 ± SD, n.d., not determined; n.i., no inhibition. c Compounds 1a, 1b, 2a and 2b were reported previously in Ref. [2]. a Data shown are the mean of at least two independent experiments, SD ≤ 10%. b IC50 ± SD, n.d., not determined; n.i., no inhibition. c Compounds 1a, 1b, 2a and 2b were reported previously in Ref. [2]. a Data shown are the mean of at least two independent experiments, SD ≤ 10%. b IC50 ± SD, n.d., not determined; n.i., no inhibition. c Compounds 1a, 1b, 2a and 2b were reported previously in Ref. [2].     a Data shown are the mean of at least two independent experiments, SD ≤ 10%. b IC50 ± SD, n.d., not determined; n.i., no inhibition.

Development of a Binding Model Using Rigidified Analogues
The strongly rigidified analogues 5a and 6a were synthesized to probe the biologically active conformation of our inhibitor class on the basis of our previously presented binding model of 1b [2]. The dihedral angle of the benzyl moiety was fixed at two different degrees, which would direct the benzene ring either to a small hydrophobic cavity in the lid of the ATP pocket (in 6a) or more planar toward the pocket exit (in 5a), where favorable CH-π interactions might additionally boost the binding affinity. However, to our surprise, 5a and 6a were both inactive or only weakly active, respectively. This result suggested that the binding orientation predicted for 1b, where the carbonyl interacted with Lys191, might not be preferred by all 5-methoxybenzothiophene-2-carboxamide analogues, because in such case either 5a or 6a should have exhibited a clear inhibitory potency. Already in previous docking runs, an inverted binding mode had occasionally been observed, where the ligand was flipped by 180 • and the methoxy group interacted with Lys191. Hence, we decided to dock the most active and some inactive compounds of the present series and evaluated the consistency with the SAR in both potential binding modes. Indeed, we found that the SAR observed with the new compounds were all consistent with the inverted binding mode, where the amide carbonyl interacts with the Leu244-NH in the hinge region. Assuming this to be the generally preferred orientation of our ligands, the lack of activity observed with 5a could eventually be explained by a predicted clash of the dihydro-isoquinoline moiety with the Leu167 side chain ( Figure S1, Supplementary Material). With respect to the racemic probe compound 6a, only the (S)-enantiomer could be placed in the ATP binding pocket at all; however, this was at the cost of several inter-and intramolecular steric clashes, explaining the very low activity ( Figure S2B, Supplementary Material). Apparently, the original binding mode, with H-bonding between the carbonyl and Lys191, has other energetic disadvantages, resulting in a low binding affinity and precluding inhibition. Furthermore, the originally published binding mode would not be in accordance with the loss of potency found with 1d, 3a (S), 9b and 13b, while this could be explained unambiguously by the flipped binding orientation. Altogether, the SAR obtained with the entire set of the new, challenging modifications made it possible to derive a new binding model for the compounds presented here (cf. Figures 3 and 4), which will be employed in the following as a basis for the discussion.

Variation of the N-alkylation and Methylation of the Benzyl Position
In our previous study, it was reported that N-methylation of the amide greatly enhanced the inhibitory activity against Clk1 [2] (cf. 1a vs. 1b). This overall tendency was confirmed with all unmethylated/ N-methylated compound pairs of the present series, too. However, when the methyl was enlarged to ethyl and allyl (compounds 1c and 1d, respectively, Table 1), no further enhancement, but rather a marked drop of activity was noted, suggesting that the available space around the methyl group is rather limited. This observation is in accordance with the binding models for the potent compounds 7b and 10b, which were derived later (cf. Figures 3 and 4).
The p-fluoro analogues having the methyl group added to the benzylic carbon spacer seemed to be more potent than the unmethylated and N-methylated congeners (compare 2a and 2b with 4a, Table 1). However, this was only true for the (R) enantiomer (4a), since the lower activity of the corresponding racemate (3a) indicated that the (S) enantiomer was inactive, probably due to a steric collision of the methyl group with Leu244 (cf. binding model in Figure S2A). In contrast, there is more space for the methyl in the (R) enantiomer (4a), as it will point away from Leu244.
On the other hand, combining a methyl group at the benzylic carbon spacer with an Nmethylation completely abolished the inhibitory activity, irrespective of the stereochemistry (compounds 3b and 4b, Table 1). It can be assumed that the additional methyl group at the benzylic carbon spacer is forced to adopt an anti or gauche conformation relative to the N-methyl, thus strongly increasing the steric demands which would prevent binding due to similar reasons as with 3a (S) (cf. Figure 3).

Variation at the Benzyl Amide Function
In the next optimization step, various new moieties were tested as an extension to the amide function, including mono-and di-substituted benzyl derivatives. Compound 7b, having the bulkier and more lipophilic m-chloro substituent on the phenyl ring, was equipotent to compound 1b; however, 7b was more selective for Clk1 over Clk2 (selectivity factor = 16) than 1b (selectivity factor = 2.7). The high potency of 7b could be explained based on our new binding model, predicting that CH-π interactions were established with Leu167 and Leu246 via the m-chloro phenyl moiety, and that halogen bonds were formed involving the chlorine, Ser247 and Asp250 (Figure 3).
Molecules 2021, 26, 1001 10 of 23 conformation relative to the N-methyl, thus strongly increasing the steric demands which would prevent binding due to similar reasons as with 3a (S) (cf. Figure 3).

Variation at the Benzyl Amide Function
In the next optimization step, various new moieties were tested as an extension to the amide function, including mono-and di-substituted benzyl derivatives. Compound 7b, having the bulkier and more lipophilic m-chloro substituent on the phenyl ring, was equipotent to compound 1b; however, 7b was more selective for Clk1 over Clk2 (selectivity factor = 16) than 1b (selectivity factor = 2.7). The high potency of 7b could be explained based on our new binding model, predicting that CH-π interactions were established with Leu167 and Leu246 via the m-chloro phenyl moiety, and that halogen bonds were formed involving the chlorine, Ser247 and Asp250 (Figure 3). Figure 3. The predicted binding mode of compound 7b (magenta) docked in the ATP binding pocket of Clk1 (PDB code of the coordinates: 1Z57) using MOE. 7b was anchored inside the pocket through H-bonds with Leu244 and Lys191 (indicated in black), CH-π interactions with Leu167 and Val175 (red), an edge-to-face CH-π interaction with Phe241, hydrophobic interactions between the N-methyl, Leu167 and Val175 residues (violet), and halogen bonds (green) between the chlorine and the carbonyl groups of Ser247 as well as Asp250. Interactions are indicated by dashed lines, and distances between the heavy atoms are given in Å. Figure 3. The predicted binding mode of compound 7b (magenta) docked in the ATP binding pocket of Clk1 (PDB code of the coordinates: 1Z57) using MOE. 7b was anchored inside the pocket through H-bonds with Leu244 and Lys191 (indicated in black), CH-π interactions with Leu167 and Val175 (red), an edge-to-face CH-π interaction with Phe241, hydrophobic interactions between the N-methyl, Leu167 and Val175 residues (violet), and halogen bonds (green) between the chlorine and the carbonyl groups of Ser247 as well as Asp250. Interactions are indicated by dashed lines, and distances between the heavy atoms are given in Å.
In our previous study, it was reported that using an electron withdrawing fluoro substituent at the phenyl ring was more favorable for the potency than an electron donating methyl or methoxy substituent at the meta-position; however, a closer investigation in the present study revealed that the effects were dependent on the substitution positions: fluorine was only favorable in the meta-position of the phenyl ring, while in the para-position, a methoxy substituent was superior to fluorine (compare 2b with 8b). The latter finding was in accordance with our binding model, where the predicted CH-π interactions involving Leu167 and Leu246 basically preferred an electron-rich phenyl ring; it is, therefore, likely that fluorine only becomes favorable if it can undergo additional interactions, such as an H-bond acceptor to Ser247 (cf. Figure 4). Based on these considerations, we decided to test the effect of disubstituted benzyl extensions, by combining fluorine with an electron donating group but also by introducing a second fluorine at different positions. With respect to the di-fluoro substitutions, the presence of one fluorine at the meta-position was essential for Clk1 inhibitory activity (compare 9b with 10b and 11b, Table 2). Installing both fluorine atoms at the meta-position was optimum among the di-fluorinated rings regarding Clk1 inhibitory activity (10b). The importance of the m-fluoro substituent could be explained by comparing the binding models of 10b (Figure 4) with that of 9b (Supplementary Materials, Figure S3) in the Clk1 ATP binding pocket; despite forming a CH-π interaction with Leu167, 9b failed to establish an H-bond interaction with Leu244 in the hinge region, and it was evident that the H-bond with Ser247 can only form with fluorine in the meta-position, like in 10b (Figure 4). Of note, despite being slightly less potent (IC 50 = 12.7 nM) than 1b, 10b was more selective for Clk1 over Clk2 (selectivity factor = 10). Regarding the combinations of fluorine with an electron withdrawing group on the phenyl ring, the 3-fluoro-4-methoxy (12b) and the 3-fluoro-5-methyl (15b) disubstituted benzyl rings were most beneficial for the Clk1 inhibitory activity. 15b also exhibited a high selectivity for Clk1 over Clk2 (15-fold), although it was slightly less potent than 7b against purified Clk1 (IC 50 s = 10.9 nM and 6.2 nM, respectively).  Leu244, Lys191 as well as a weaker H-bond interaction between the fluorine and Ser247. In addition, CH-π interactions (red) were predicted between the benzyl moiety and Leu167/Leu264, as well as between the benzothiophene core and Val175/Val324. An edge-to-face CH-π interaction with Phe241 and hydrophobic interactions (violet) between the Nmethyl, Leu167 and Val175 were also predicted. All interactions are indicated by dashed lines, and distances between the heavy atoms are given in Å.

Inhibition of Tumor Cell Growth In Vitro
In our previous study, the bladder carcinoma cell line T24 was identified as a sensitive system for testing the cellular activity of Clk1/4 inhibitors, where 1b showed a GI50 value of 0.63 µM [2]. This was in accordance with the Oncomine data from several studies (Oncomine.org) which had revealed Clk1 mRNA overexpression in bladder cancer. Therefore, we decided to evaluate the cellular activity of the most potent and selective Clk1 inhibitors among the new series (7b, 10b, 12b and 15b) against the T24 cell line (Table 3). The predicted binding mode of compound 10b (cyan). 10b was docked in the ATP binding pocket of Clk1 (PDB code of the coordinates: 1Z57) using MOE. 10b was anchored inside the pocket through H-bonds (indicated in black) with Leu244, Lys191 as well as a weaker H-bond interaction between the fluorine and Ser247. In addition, CH-π interactions (red) were predicted between the benzyl moiety and Leu167/Leu264, as well as between the benzothiophene core and Val175/Val324. An edge-to-face CH-π interaction with Phe241 and hydrophobic interactions (violet) between the N-methyl, Leu167 and Val175 were also predicted. All interactions are indicated by dashed lines, and distances between the heavy atoms are given in Å.

Inhibition of Tumor Cell Growth In Vitro
In our previous study, the bladder carcinoma cell line T24 was identified as a sensitive system for testing the cellular activity of Clk1/4 inhibitors, where 1b showed a GI 50 value of 0.63 µM [2]. This was in accordance with the Oncomine data from several studies (Oncomine.org) which had revealed Clk1 mRNA overexpression in bladder cancer. Therefore, we decided to evaluate the cellular activity of the most potent and selective Clk1 inhibitors among the new series (7b, 10b, 12b and 15b) against the T24 cell line (Table 3). In this assay, 10b emerged as the most potent inhibitor of T24 cell growth, exhibiting higher potency than 1b, despite being somewhat less potent in the cell-free assay. There was no strict correlation between the cell-free vs. cellular potencies, suggesting that pharmacokinetic properties, such as cellular uptake, played a major role for the efficacy in cells. Obviously, the mono-and difluoro-substitution in 1b and 10b strongly improved the physicochemical properties of the 5-methoxybenzothiophene-2-carboxamide scaffold so that the compounds became more available to the cellular Clk1. Since the selectivity of 10b had increased compared with the reference compound 1b (cf. Table 4), it was less likely that the inhibition of additional kinases was responsible for the higher cellular activity, though it cannot be totally ruled out.

Kinase Selectivity Profiling
Having identified 10b as the most potent inhibitor in T24 cells, it was important to further analyze the selectivity of this compound against a larger panel of kinases that were frequently reported as off-targets for Clk1 inhibitors [13,[20][21][22][23][24], due to the highly similar ATP binding pockets. The results are shown in Table 4.
The co-inhibition of Clk4 by 10b was expected, since Clk1 and Clk4 share the highest sequence identity among the Clk family members. In view of their fully identical ATP binding pocket [2], selective inhibition of Clk1 vs. Clk4 and vice versa might not be achievable by ATP competitive compounds. Among the tested non-Clk kinases, only Dyrk1A and Dyrk1B were appreciably inhibited; however, the IC 50 determination for Dyrk1A revealed that 10b was 28 times more potent for Clk1; this selectivity factor was higher than that reported for many highly potent Clk1 inhibitors in the literature: e.g., the quinazoline derivatives 34 and 35 described in [25], with selectivity factors of 2.8 and 2.7 respectively; the aminopyrimidine derivatives 15 and 17 reported in Ref. [26], with selectivity factors of 3.25 and 2, respectively; the azaindole derivative 10c reported in Ref. [27], with a selectivity factor of 3.2; TG003 reported in [28], which showed higher potency against Dyrk1A than against Clk1 (IC 50 s of 12 and 20 nM, respectively); and KH-CB19 reported in [29], with a selectivity factor of 2.8.

Chemistry
Solvents and reagents were obtained from commercial suppliers and used as received. Melting points were determined on a Stuart SMP3 melting point apparatus. A Bruker DRX 500 spectrometer was used to obtain the 1 H-NMR and 13 C-NMR spectra. The chemical shifts are referenced to the residual protonated solvent signals. All final compounds had a percentage purity of at least 95%, and this could be verified using HPLC coupled with mass spectrometry. Mass spectra (HPLC−ESIMS) were obtained using a TSQ quantum (Thermo Electron Corp., Waltham, MA, USA) instrument prepared with a triple quadrupole mass detector (Thermo Finnigan, Waltham, MA, USA) and an ESI source. All samples were injected using an autosampler (Surveyor, Thermo Finnigan) by an injection volume of 10 µL. The MS detection was determined using a source CID of 10 V and carried out at a spray voltage of 4.2 kV, a nitrogen sheath gas pressure of 4.0 × 105 Pa, a capillary temperature of 400 • C, a capillary voltage of 35 V, and an auxiliary gas pressure of 1.0 × 105 Pa. The stationary phase used was an RP C18 NUCLEODUR 100-3 (125 mm × 3 mm) column (Macherey & Nagel, Düren, Germany). The solvent system consisted of water containing 0.1% TFA (A) and 0.1% TFA in acetonitrile (B). The HPLC method used a flow rate of 400 µL/min. The percentage of B started at 5%, increased up to 100% during 7 min, was kept at 100% for 2 min, and was flushed back to 5% in 2 min and was kept at 5% for 2 min. Melting points were determined using a BUCHI B-540 melting point apparatus and are uncorrected. The high-resolution mass analyses were performed using a Orbitrap Q exactive mass spectrometer, equipped with a heated ESI source and an quadrupole-orbitrap coupled mass detector and an Ultimate3000 HPLC (Thermo Finnigan, San Jose, CA, USA). The MS detection was carried out at a spray voltage of 3.5 kV, a nitrogen sheath gas pressure of 4.0 × 10 5 Pa, an auxiliary gas pressure of 1.0 × 10 5 Pa and a capillary temperature of 300 • C. All samples were injected by autosampler with an injection volume of 15 µL. A RP Nucleoshell Phenyl-hexyl ® (100-3, 2.7 µm) column (Macherey-Nagel GmbH, Dühren, Germany) was used as stationary phase. The solvent system consisted of 0.1% formic acid (A) and 0.1% formic acid in acetonitrile (B). HPLC-Method: Flow rate 550 µL/min. The percentage of B started at an initial of 5%, increased up to 40% during 10 min, then to 99% during the next 4 min, kept at 99% for 2 min and flushed back to the initial 5%. Xcalibur software was used for data acquisition and plotting. To an ice-cooled suspension of (7.2 g, 2 eq.) of K 2 CO 3 in DMF (20 mL) was added ethyl thioglycolate (31.2 mmol, 1.2 eq.). The reaction mixture was stirred for 20 min at 0 • C under nitrogen, this was followed by the gradual addition of 2-fluoro-5-methoxybenzaldehyde (26 mmol, 1 eq.) in DMF (10 mL). The mixture was heated to reflux at 70 • C for 4 h. Afterwards, the suspension was cooled to room temperature and poured over 100 mL of 2 M HCl, the aqueous layer was extracted by DCM (5 × 10 mL), and the combined organic layers were thoroughly washed with water and brine, dried over anhydrous MgSO 4 , evaporated under reduced pressure and the resulting residue was purified by column chromatography (CC) using a solvent system of (petroleum ether/ DCM 3:2), giving the pure carboxylic acid ethyl ester as a yellow solid in a yield of 2.9 g (47%); 1  The ethyl ester (I) was suspended in a mixture of 50 mL of ethanol and 25 mL of water; this was followed by the addition of 4 eq. KOH. The mixture was heated to reflux at 80 • C for 3 h. Afterwards, the solvent was removed under reduced pressure, and the resulting aqueous layer was cooled to 0 • C, followed by the gradual addition of conc. HCl until the pH was adjusted to 1. The aqueous layer was then extracted using DCM (5 × 50 mL), the combined organic extracts were filtered over anhydrous MgSO 4 , and evaporated under reduced pressure giving the carboxylic acid derivative as a white solid in a yield of 2.3 g (90%). 1  thiophene-2-carboxylic acid (II) (0.1 g, 0.5 mmol) was dissolved in DCM (30 mL), then 0.3 g of HBTU and 2 mL of TEA were added. The reaction mixture was stirred for 30 min, following this, the appropriate amine (4 eq.) was added dropwise, and the reaction mixture was stirred at room temperature overnight. The solvent was evaporated under reduced pressure. The residue was partitioned between 50 mL of ethyl acetate and 20 mL of water then the aqueous layer was extracted with three 20 mL portions of ethyl acetate. The combined organic extracts were filtered over anhydrous MgSO 4 , the solvent was removed under reduced pressure, and the product was purified by CC to give the 5-methoxybenzo[b]thiophene-2-carboxylic acid amide derivatives in different yields.
Procedure D, General Procedure for the Synthesis of 5-Methoxybenzo[b]thiophene-2-carboxylic Acid Alkyl Amide Derivatives The respective amide derivative was dissolved in dry THF and stirred at 0 • C, this was followed by the gradual addition of 3 eq. of potassium bis(trimethylsilyl)amide (0.5 M solution in THF), the reaction mixture was left to stir for 1 h, after which methyliodide or ethyliodide or allylbromide (1.5 eq.) were added. The reaction mixture was left to stir for 24 h at room temperature, afterwards, the solvent was removed under reduced pressure, then a small amount of water was added, and extraction was carried out using DCM (5 × 30 mL). The organic layers were combined, dried over anhydrous MgSO 4 , the solvent was removed under reduced pressure and the product was purified by CC or by salting out to give the 5-methoxybenzo[b]thiophene-2-carboxylic acid alkyl amide derivatives in different yields.
MO, USA). T24 cells were seeded in a 96-well plate (5000 cells per well) already containing DMSO as a control (0.2% final concentration) or the test compounds dissolved in DMSO. The cells were grown for five days at 37 • C in a humidified incubator containing 5% CO 2 , without further change of medium, before the detection was carried out as described in [30].

Molecular Modeling
All procedures were performed using the Molecular Operating Environment (MOE) software package (version 2016, Chemical Computing Group, Montreal, Canada). For the docking simulations, PDB entry 1Z57 (Clk1 co-crystallized with hymenialdisine) [18] was used. Molecular docking simulations with the 3a (S), 5a, 6a, 7b, 9b and 10b ligands were performed using the MMFF94x force field, and the Amber10:EHT force field was used for compound 7b; the placement method was set to "Alpha PMI" (number of return poses set to 2000), and refinement was set to "induced fit" to enable free side chain movements. Using these settings, docking runs were performed in two different ways: (i) the binding pocket was defined by selecting nearby residues to the co-crystallized ligand (hymenialdisine); (ii) after an initial run using condition i, the MOE LigX routine was applied (settings: receptor strength = 5, ligand strength = 5000) for energy minimization, and a pharmacophore was defined based on a suitable pose by selecting both the methoxy oxygen close to the NH of the leucine and the carbonyl oxygen near the conserved lysine as acceptor points (this binding orientation prevailed among the poses with the compound docked deep in the pocket). In the pharmacophore definition window, the radius of the pharmacophore points was raised to 1.2 Å. The subsequent docking was performed using the pharmacophoresupported placement. The number of retained poses was set to 500 each time. Only the poses with the top 10 scoring values were further evaluated for plausibility. The final selected poses were optimized again using the LigX routine, using the same settings as described above.

Conclusions
In the present study, various strategies were employed to increase the potency and/or selectivity of our previously reported N-methylbenzothiophene-2-carboxamides as inhibitors of Clk1 [2]. While keeping the methylated amide function, introduction of the m-chlorobenzyl as well as di-substituted benzyl extensions led to an enhanced selectivity for Clk1 over Clk2 without loss of the high potency toward Clk1, when compared with the former flagship compound 1b. To the best of our knowledge, 10b, 7b, 12b and 15b show the highest selectivity for Clk1/4 over Clk2, as reported previously for Clk1 inhibitors. In addition, compound 10b was more potent in cells than 1b, exhibiting the highest antiproliferative activity against T24 bladder carcinoma cells. The overall selectivity profile of 10b against the most common off-targets for Clk1 inhibitors was comparable to that of 1b, with the important difference that 10b was four times more selective for Clk2 inhibition. The latter could be considered a significant achievement, because more selective Clk inhibitors might reduce the side effects that can be expected following the application of pan Clk inhibitors; in such cases, the alternative splicing in all tissues might be strongly compromised due to the simultaneous blocking of all Clk activities. In terms of a targeted tumor therapy, it seems more appropriate to inhibit only the Clk isoforms that are overexpressed in the respective tumor type. An analysis of the Oncomine mRNA expression database (oncomine.org) revealed that for dual Clk1/Clk4 inhibitors, the most promising oncologic indications are glioblastoma and lymphoma, in which mRNA overexpression compared to healthy tissue was detected for both isoforms.
Supplementary Materials: The following are available online, Figure S1: Molecular docking of compound 5a in the ATP binding pocket of Clk1, Figure S2: Molecular docking of compounds 3a (S) and 6a in the ATP binding pocket of Clk1, Figure S3: Molecular docking of compound 9b in the ATP binding pocket of Clk1.